CRISPR-Cas immune systems adapt to new threats by acquiring new spacers from invading nucleic acids such as phage genomes. However, some CRISPR-Cas loci lack genes necessary for spacer acquisition despite variation in spacer content between microbial strains. It has been suggested that such loci may use acquisition machinery from cooccurring CRISPR-Cas systems within the same strain. Here, following infection by a virulent phage with a double-stranded DNA (dsDNA) genome, we observed spacer acquisition in the native host _Flavobacterium columnare_ that carries an acquisition-deficient CRISPR-Cas subtype VI-B system and a complete subtype II-C system. We show that the VI-B locus acquires spacers from both the bacterial and phage genomes, while the newly acquired II-C spacers mainly target the viral genome. Both loci preferably target the terminal end of the phage genome, with priming-like patterns around a preexisting II-C protospacer. Through gene deletion, we show that the RNA-cleaving VI-B system acquires spacers in _trans_ using acquisition machinery from the DNA-cleaving II-C system. Our observations support the concept of cross talk between CRISPR-Cas systems and raise further questions regarding the plasticity of adaptation modules.

Cooperation between Different CRISPR-Cas Types Enables Adaptation in an RNA-Targeting System

Materials and Methods

Regulation of CRISPR-Cas9 functions in vivo is conducive to developing precise therapeutic genome editing. Here, we report a CRISPR-Cas9 prodrug nanosystem (termed NanoProCas9), which combines the targeted delivery and the conditional activation of CRISPR-Cas9 for the precision therapy of inflammatory bowel disease. NanoProCas9 is composed of (i) cationic poly(β-amino ester) (PBAE) capable of complexing plasmid DNA encoding destabilized Cas9 (dsCas9) nuclease, (ii) a layer of biomimetic cell membrane coated on PBAE/plasmid nanocomplexes for the targeted delivery of PBAE/dsCas9 complexes, and (iii) the stimuli-responsive precursory molecules anchored on the exofacial membrane. The systemic administration of NanoProCas9 enables the targeted delivery of dsCas9 plasmid into inflammatory lesions, where the precursory small molecule can be activated by ROS signals to stabilize expressed dsCas9, thereby activating Cas9 function for inflammatory genome editing. The proposed “genome-editing prodrug” presents a proof-of-concept example to precisely regulate CRISPR-Cas9 functions by virtue of particular pathological stimuli in vivo.

Genome-editing prodrug: Targeted delivery and conditional stabilization of CRISPR-Cas9 for precision therapy of inflammatory disease

### Materials

All reagents used in the experiments were purchased from commercial sources without further purification. Anhydrous dichloromethane (DCM), anhydrous _N_,_N_-dimethylformamide (DMF), HBr [48% (w/w) in H2O], DCC, 4-dimethylaminopyridine (DMAP), rhodamine B, dimethyl sulfoxide (DMSO), and TMP were purchased from Energy Chemical Co. Ltd. (Shanghai, China). Branched polyethyleneimine [PEI25K; weight-average molecular weight (_M_w): 25 kDa] was supplied by Meilunbio Co. Ltd. (Dalian, China). Dulbecco’s modified Eagle’s medium, RPMI 1640, and FBS were purchased from Sigma-Aldrich (USA). Lipofectamine 2000 and Lipofectamine 3000 transfection reagents were purchased from Thermo Fisher Scientific (Germany). The transfection reagent jetPEI was supplied by PolyPlus. The CCK-8 and LDH release assay kit were obtained from Beyotime Co. Ltd. (Shanghai, China). The dialysis bag [molecular weight cutoff (MWCO): 1000 Da] was obtained from Yuanye Bio-Technology Co. Ltd. (Shanghai, China). T7E1 enzyme was purchased from New England Biolabs (USA). Ultrapure water was obtained from a Milli-Q system. The plasmid encoding CMV-DHFR-Cas9-DHFR-U6-sgRNA (no. 85447) was supplied by Addgene.org, which was examined by Sanger sequencing and enzyme digestion. CMVIE94-3XFLAG-NLS-Cas9-NLS-P2A-EGFP (CMV-Cas9-P2A-EGFP) and CMV-3XFLAG-NLS-Cas9-NLS (wild-type) were constructed in our laboratory. Target sgRNA (sg_PHD2_) was designed by online tools (http://crispr.mit.edu/ and http://chopchop.cbu.uib.no/). All primers used in this work were listed in tables S1 to S5. The plasmids used in this work and corresponding nanocomplex information were listed in table S6 and in the “Plasmid information” section in Supplementary Materials. Primary antibodies used in this project included the following: Anti–TNF-α (1:1000) antibody was obtained from Beyotime (Shanghai, China). Anti-PHD2 (1:1000) antibody and cell/tissue lysis buffer [radioimmunoprecipitation assay (RIPA) buffer] were obtained from Solarbio Science and Technology (Beijing, China). Anti-FLAG antibody was obtained from HuaAn Biotechnology (HUABIO, Hangzhou, China). Anti-GAPDH (glyceraldehyde-3-phosphate dehydrogenase) antibody was obtained from Abcam (Shanghai, China). All enzyme-linked immunosorbent assay (ELISA) kits used in this project were supplied by Multisciences (Hangzhou, China). Cationic PBAE was synthesized according to the previous literature (scheme S1 and fig. S1) (_16_, _17_).

### Characterization

The 1H NMR spectra and heteronuclear singular quantum correlation spectrum of PBAE were carried out on a Bruker 500-MHz NMR spectrometer and were reported as chemical shifts (δ) in ppm relative to tetramethylsilane (δ = 0). Proton spin multiplicities are reported as a singlet (s), doublet (d), triplet (t), quartet (q), and quintet (quint), with coupling constants (J) given in hertz, or multiplet (m). Mass spectra were recorded on a Shimadzu liquid chromatography–mass spectrometry (LC-MS) mass spectrometer (LC-MS 2020). TEM was carried out on a TEM (HT7700, Hitachi, Japan). Sizes and zeta potentials of the nanoparticles were characterized by Malvern Nano ZS90. Mass analysis of OE-PEG, BAM-TK-COOH, and BAM-TK-TMP was performed on MALDI-TOF (Ultraflextreme, Bruker, Germany). The molecular weight analysis of PBAE was performed on a Waters 1525/2414 GPC system. Confocal microscopy was performed on LSM-880 (Carl Zeiss, Germany). The DiI intensity was evaluated with an in vivo imaging system (IVIS Spectrum, PerkinElmer).

### Synthesis of 4-[((2,4-diaminopyrimidin- 5-yl) methyl)]-2,6-dimethoxyphenol (TMP-OH)

TMP-OH was synthesized according to the previous report (_33_). Briefly, TMP (1.00 g, 3.4 mmol) was dissolved in HBr (12.5 ml, 48% in H2O) and stirred at 95°C. After 20 min, the mixture was cooled down to room temperature and 2.97 ml of 50% (w/w) NaOH aqueous was added dropwise during stirring. After that, the solution was kept at 4°C overnight. The white precipitate was filtered and washed with double-distilled water (ddH2O). Residues were redissolved in boiling water, and 1 N NaOH was added to adjust pH to ~7.0 for recrystallization. After being cooled down at 4°C for a while, a pink solid, TMP-OH, was precipitated and washed by water (630 mg, 66%). 1H NMR (DMSO-_d6_) δ 8.06 (s, -OH), 7.45 (s, 1H), 6.48 (s, 2H), 5.99 (s, -NH2), 5.63 (s, -NH2), 3.71 (s, 6H), and 3.47 (s, 2H). MS mass/charge ratio (_m/z_) found: 277.15, calculated: 277.12 for C13H16N4O3 [M+H]+.

### Synthesis of polyethylene glycol monooleyl ether-thioketal (BAM-TK-COOH)

To a solution of polyethylene glycol monooleyl ether (_n_ = ~50, 250 mg, 0.1 mmol), DMAP (12 mg, 0.1 mmol) and thioketal (252 mg, 1 mmol) in 20 ml of anhydrous DCM at 0°C were added dropwise with 10 ml of DCC (206 mg, 1 mmol)/DCM solution. After completion of addition of DCC, the reaction mixture was transferred to room temperature and stirred for 72 hours under a nitrogen atmosphere. After that, the insoluble by-product was filtered and the solvents were removed by rotary evaporation. The residue was dissolved in water and dialyzed (MWCO, 1000 Da) against water/methanol (1:1, v/v) for the first 24 hours and water for 48 hours. The product, BAM-TK, was collected and lyophilized for further reaction (193 mg, 70%).

### Synthesis of polyethylene glycol monooleyl ether-thioketal (BAM-TK-TMP)

TMP-OH (19 mg, 0.07 mmol) was added to a solution of BAM-TK (190 mg, 0.07 mmol) and DMAP (2 mg, 0.02 mmol) in 20 ml of anhydrous DMF. After stirring for 10 min under a nitrogen atmosphere, DCC (29 mg, 0.14 mmol), which was dissolved in 5 ml of DMF, was added dropwise into reaction mixture. After addition, the mixture was stirred at 30°C for 72 hours. The solvent of crude product was removed by rotary evaporation, and the product was redissolved in DCM. The by-product was filtered, and the filtrate was collected. After removal of DCM by rotary evaporation, the solid was redissolved in methanol. The product, BAM-TK-TMP, was obtained after dialysis against methanol for 24 hours and water for 48 hours (MWCO, 1000 Da) and further lyophilization.

### Deep sequencing assay

The schematic illustration of deep sequencing assay was shown in scheme. S3. First, the off-target loci were evaluated and designed at CasOFFinder website (www.rgenome.net/cas-offinder/). Subsequently, the fragments were amplified with corresponding primers, termed first PCR product. After purification by PCR/Gel Extraction and Purification kits (Vazyme Biotech Co., Ltd), the first PCR product was further amplified to obtain PCR fragments less than 250 base pairs (bp) containing corresponding gene loci. Then, the product, termed second PCR product, was further extracted and purified. After that, the second PCR product was amplified with primers containing index sequence and then purified by PCR/Gel Extraction and Purification kits. The products were sent to company for sequencing. The results were analyzed by CRISPResso2 according to the instruction (_51_).

### Statistical analysis

All data and figures in this paper were analyzed and plotted by GraphPad Prism 8.0. The obtained data are expressed as means ± SD. Biological replicates were used in all experiments unless otherwise stated. The statistical significance was analyzed using one-way analysis of variance (ANOVA) with a Tukey post hoc test. _P_ value less than 0.05 was considered significant (*_P_ < 0.05, **_P_ < 0.01, \*\**_P_ < 0.001, and \*\*\**_P_ < 0.0001). ns means no significance. N.D. refers to not detectable.

CRISPR‐cas technology is being incorporated into undergraduate biology curriculum through lab experiences to immerse students in modern technology that is rapidly changing the landscape of science, medicine and agriculture. We developed and implemented an educational module that introduces students to CRISPR‐cas technology in a Genetic course and an Advanced Genetics course. Our primary teaching objective was to immerse students in the design, strategy, conceptual modeling, and application of CRISPR‐cas technology using the current research claim of the modification of theCCR5gene in twin girls. This also allowed us to engage students in an open conversation about the bioethical implications of heritable germline and non‐heritable somatic genomic editing. We assessed student‐learning outcomes and conclude that this learning module is an effective strategy for teaching undergraduates the fundamentals and application of CRISPR‐cas gene editing technology and can be adapted to other genes and diseases that are currently being treated with CRISPR‐cas technology.

“Designer babies?!” ACRISPR‐based learning module for undergraduates built around theCCR5gene

### 2.1. Intended audience and learning time

This work is intended for sophomore‐ and junior‐level undergraduate Biology students enrolled in a Genetics or Biotechnology course, but can be integrated and modified for Introductory Biology course or upper‐level Molecular Biology course. The main activity was implemented during a 2‐hour 45‐minute laboratory period. The activity can be adapted to fit two traditional 90‐minute class/lecture periods. Subsequent post‐activity student work where students created a research proposal was carried out over the next two laboratory periods and during students' own time.

#### 2.1.1. _Prerequisite student knowledge_ 

Student should have prior exposure to basic molecular genetics principles that include DNA/RNA structure, codons, reading frames, mutation. However, this lesson can be adapted to introduce and apply the concepts of codons/mutations.

### 2.2. Materials

Each student will require a computer/laptop with access to the internet and the software program, SnapGene Viewer (full version, https://www.snapgene.com). A free 30‐day trial to SnapGene Viewer can be downloaded by each student with their email address. Alternative molecular biology software could be used with modifications to the below resources.

The student handouts and accompanying instructor guide/key are provided in Supporting Information.

### 2.3. Instructions for faculty and students

#### 2.3.1. _Comparative analysis of_ _CCR5_ _gene and modified allele variants on paper_ 

The activity utilizes the gene encoding human CCR5; the intended target gene used in Dr. He's study. _CCR5_ gene encodes for a chemokine GPCR in T cells, and was targeted because of the known role of the CCR5 as a co‐receptor in HIV infection of humans. 28 _CCR5_ delta 32 (Δ32) is a naturally occurring allele lacking 32 nucleotides that correspond to a sequence that normally codes for part of the co‐receptors second extracellular loop. 28 Removing these 32 nucleotides results in a premature stop codon due to frameshift, and the resultant truncated protein product can no longer be exocytosed to the cell membrane. 29 Individuals homozygous for the _CCR5_ Δ32 variant are resistant to HIV infection, as CCR5 is required for membrane fusion during HIV infection. The HIV life cycle and the key molecular components of a HIV infection cycle in immune T cell can be introduced to students using this animation by Janet Iwasa. 30

Dr. He's goal of editing human embryos was to create a similar non‐functional HIV‐resistant CCR5 variants using CRISPR‐cas technology. 31 However, the results of these experiments have purportedly generated three new _CCR5_ allelic variants named after the twin girls, Lulu and Nana, with Lulu being heterozygotic for a novel _CCR5_ variant (the Lulu _CCR5_ allele) and wild‐type allele, and Nana being heterozygotic at the _CCR5_ locus (Nana _CCR5_ alleles 1 and 2 respectively). 4 , 31 He also claimed that a third child has been genetically modified in a similar manner.

To begin their comparative analysis of the different alleles, students were given a worksheet (Supporting Information) that contains the partial coding DNA sequences of an unmodified _CCR5_ allele, the HIV resistant allele Δ32 _CCR5_, and three novel alleles generated via CRISPR‐cas9‐based modification. They were tasked to determine the nucleotide differences between them and the effect on the reading frame, codon and amino acid (Figure ​(Figure2a,2a, example of student work; Table ​Table2,2, summary of CRISPR alleles). Students worked on their own initially and then in a group to compare answers.

  FIGURE 2

Student work aligning unmodified _CCR5_ allele to HIV resistant Δ32 and CRISPR‐cas modified alleles (a) on paper, and (b) using computer software program SnapGene. B) shows all alleles aligned to the unmodified allele. Panel (c) shows an expanded alignment view with the change in reading frame and early stop codon (red asterisks). Expanded alignment views for other modified alleles are shown in Figure S1

### TABLE 2

The effect of _CCR5_ mutant alleles on reading frame and protein structure

**AlleleNucleotide change in coding sequenceEffect on reading frame (RF)Effect on amino acidsProtein lengthProtein structure**Unmodified _CCR5_Normal (RF +1)3527 TM domains with extracellular N′ and intracellular C′Delta 32 (Δ32) _CCR5_Deletion of 32 nucleotides (911..942)Frameshift (RF +3) leading to early stop codonNew amino acids on C terminus (31 different amino acids from AA185)215Lacks 3 TM domains, altered C′ protein sequence after TM 3Lulu allele (Delta 15)Deletion of 15 nucleotides (900..914)In‐frame deletionDeletion of 5 amino acids from AA181 in extracellular loop 2347Extracellular loop 2 missing 5 AA

Nana allele 1

(Delta 4)

Deletion of 4 nucleotides (913..916)Frameshift leading to early stop codon (RF +2)9 different amino acids on C terminus from AA186194Lacks 3 TM domains, altered C′ protein sequence after TM 3Nana allele 2 (Plus 1)Addition of 1 nucleotide at bp 911Frameshift (RF +3) leading to early stop codon42 different amino acids on C terminus from AA185(31 of the amino acids are the same as Δ32)226Lacks 3 TM domains, altered C′ protein sequence after TM 3Open in a separate window

#### 2.3.2. _Comparative analysis of_ _CCR5_ _gene and modified allele variants on computer_ 

Students were guided to obtain the full reference coding sequence of _CCR5_ from NCBI (Accession number NM_000579). They created a “new sequence” in the DNA analysis software SnapGene. Students were tasked to create “new features” that represent the three exons to show where intron/exon boundaries would exist in genomic DNA. Using the open reading frame feature, the students determine the translated protein sequence and normal length of protein. They were given the three other new allele variants (Δ32, Lulu, Nana allele 1 and Nana allele 2) that had been created by the instructor based on reported CRISPR modified sequences. 32 Students were guided to align these sequences together and determine the nucleotide, reading frame, amino acid sequence and protein length changes that were created. The use of the full version SnapGene is key in visualizing the alignment. Although NCBI and free DNA softwares ApE and Benchling enables alignment of sequences together, the alignment in Snapgene is presented in an effective scheme that enables students to still visualize the double‐stranded DNA molecule with polarity of each DNA strain, the resulting translation, and the features labeled in one window (Figure 2b,c, Figure S1). Students compared their analysis on SnapGene to their worksheet. Using SnapGene allows students to determine whether premature stop codons are formed further downstream from the nucleotide change.

#### 2.3.3. _Predicting the effect of the mutations on protein structure_ 

The nucleotide sequence alignment allows students to visualize the effect on the linear primary sequence of the protein, but it is important that students move beyond the linear protein sequence and establish the effects of changes in the primary sequence to the tertiary structure. Students were shown published tertiary structures of the CCR5 protein 33 , 34 where the unmodified protein consists of seven transmembrane helices, characteristic of all G protein‐coupled receptors. Additionally, the web‐based 3D structure viewer iCn3D can be used by students to visualize the tertiary structure of the protein. The students were tasked to discuss and predict how the new alleles affect protein structure and draw the various predicted protein structures out on paper. Students of the Advanced Genetics class utilized the software BioRender (https://biorender.com), a free biology drawing software to create their schematics. Examples of student work are shown in Figure 3a,b.

  FIGURE 3

Student work depicting protein structure of CCR5 and predicted structures of its various modified versions (a) on paper and (b) using Biorender. Nucleotide sequences that were inserted or deleted are indicated, and nucleotides shared between Δ32 and Nana+1 alleles are highlighted

#### 2.3.4. _CRISPR‐cas9 editing design and strategy of_ _CCR5_ _gene_ 

Students were guided to design the strategy to program the cas9 nuclease to target and cut _CCR5_ gene at the appropriate region in order to create a loss‐of‐function CCR5 protein. Specifically, students identified and designed 3 key components: (a) the cas9 nuclear target sequence in _CCR5_, (b) the protospacer adjacent motif (PAM) sequence in _CCR5_ gene utilized by cas9, and (c) the complementary guide RNA that is guided to the target sequence by cas9 (Figure ​(Figure44).

  FIGURE 4

Necessary components for CRISPR‐cas9‐based genetic modifications. (a) cas9 and gRNA form a complex, then gRNA base pairs with complementary sequence in CCR5 gene with PAM aiding in bringing cas9 nuclease in, resulting in cas9 double‐stranded cleavage 3 nucleotides from PAM sequence. (b) A comparison of endogenous DNA repair mechanisms, non‐homologous end joining (NHEJ) and homology directed repair (HDR), which utilizes a repair template containing the desired nucleotide change. Nana alleles Δ4 and +1 are shown as outcomes of NHEJ, and HIV resistant Δ32 is shown as an outcome of HDR with the use of an appropriate repair template

Using the reference unmodified _CCR5_ gene, students utilized the “features” tool on SnapGene to identify and annotate the cas9 specific PAM sequence (5′NGG3′) chosen by Dr. He that was presented in his oral presentation at the Second International Summit on Human Genome Editing meeting (Start at: 1:17:57). 8 The guanine dinucleotides of the PAM sequence that is on the non‐target strand of the gRNA interacts with arginine amino acids of cas9 to assist the unwinding of double‐stranded DNA for subsequence nuclease activity (as reviewed in Chen and Doudna 3 ). Next, students were guided to label the target sequence of the cas9 nuclease, which are the 20 nucleotides upstream of the 5′TGG3′ PAM sequence (Figure ​(Figure5a).5a). The cas9 nuclease is brought to the target sequence by the single guide RNA, which is composed of an RNA sequence complementary to the target strand and an 80‐mer universal scaffold region that aids in cas9 binding. To model formation of gRNA‐cas9‐target DNA complexes, students were tasked to hand draw how all these components interact, labeling important components and sites including their PAM sequence, the location of the cas9 cut site, the annealing of the gRNA to its specific target nucleotide sequence. We intentionally had students draw this out based on our prior teaching experiences where assessment revealed students found it challenging to conceptualize the action and polarity of gRNA and cas9 complex with two complementary DNA strands (target and non‐target strands) of the target gene. 19

  FIGURE 5

Key components utilized by CRISPR‐cas9 system. View in SnapGene of (b) PAM sequence (green) and gRNA‐cas9 target sequence (light green) are annotated in unmodified CCR5 allele in top half. Potential PAM sequences are highlighted in yellow. Bottom half shows the alignment with Δ32 allelic variant to allow visualization of which PAM sequence is most appropriate. (b) An example HDR repair template lacking the 32 nucleotides missing in the Δ32 allelic version, and containing two homology arms. The blue line indicates where the 32 nucleotides in Δ32 allelic version has been deleted. (c) Student work showing a schematic of gRNA and cas9 complexing with target sequence in _optix_ gene

Students were prompted to discuss why this specific PAM associated with the target sequence was chosen of all possible PAM sequences of (5′NGG3′) as a quick “Find” of “GG” highlights in yellow all the GGs present in the sequence (yellow boxes, Figure ​Figure5a),5a), and students observe that there are many potential PAM sequences present throughout the gene. The discussion led the class to conclude that this specific 5′NGG3′ was chosen of many potential 5′NGG3′ because of the necessity for cas9 cut the _CCR5_ genomic DNA in close proximity to the physical location of the desired nucleotide change that could potentially give rise to Δ32 allele variant.

#### 2.3.5. _Interpreting_ _DNA_ _repair mechanisms that created Nana and Lulu alleles; designing a homology directed repair (HDR) repair template_ 

Identification of the likely _CCR5_ PAM sequence and cas9 cut site was followed by discussion of the outcomes for the _CCR5_ gene after cas9 cleavage, and how different endogenous repair mechanisms are used to achieve different research goals. Animation are shown to illustrate the two processes. 35 , 36 The process of non‐homologous end joining (NHEJ) is used after cas9 makes the double stranded cut to produce indels and potential loss‐of‐function alleles due to frameshift mutations. Homology Directed Repair (HDR) is typically utilized when a specific nucleotide change is desired. HDR requires the addition of a repair template with homology arms that enables the cell to utilize HDR to incorporate a desired nucleotide sequence after the cas9‐mediated dsDNA cut. As a class, we discussed the features that the HDR must contain: (a) the desired nucleotide change, (b) regions of DNA that were similar to the original allele (homology arms), and (c) a PAM sequence that was altered such that the repair template would not be cut by cas9. Students were guided to create an HDR template in SnapGene (Figure ​(Figure5b).5b). Students were then asked to work in groups to discuss which DNA repair strategy was likely utilized by Dr. He to produce the three new variant alleles found in Nana and Lulu. Most students report that NHEJ occurred in the embryos to create the three new variants and that a repair template was likely not used as the HIV resistant Δ32 allele was not found in Lulu and Nana.

#### 2.3.6. _Putting knowledge into practice:_ _CRISPR‐cas9 gene editing strategy with butterfly wing patterning gene optix and human oncogenic Ras gene_ 

Students worked on two follow‐up hypothetical scenarios to practice the CRISPR design and strategy of sgRNA and target sequence. The first scenario was that students were investigating the function of the gene _optix_ in _Lepidoptera_ butterflies, and were tasked to design a CRISPR‐cas9 strategy to create a loss‐of‐function _optix_ gene. The students were shown published results of a CRISPR‐cas9 strategy, where _optix_ CRISPR‐ed butterflies displayed color and patterning defects. 37 An example of student work is shown in Figure ​Figure5c.5c. The second scenario was that they had cultured cancer cells growing in lab with Ras gene that has the oncogenic mutation that causes abnormal cell proliferation. 38 , 39 The students' goal was to design a strategy to take this mutant Ras and genetically modify it to wild‐type Ras. They were given the coding sequence for the Ras gene in SnapGene with the oncogenic mutation causing 12th amino acid Glycine to Valine change (G12V). Students were asked to discuss in pairs what they would predict the phenotypic change would be after successful gene edit. The work was submitted to be checked by the instructor.

#### 2.3.7. _Bioethics discussion on implications of_ _CRISPR‐cas9 technology and_ _CCR5_ _editing in human society_ 

Following the exercise, we engaged the students in a robust bioethical discussion on the bioethical implications of CRISPR‐cas9 technology and the modification of the CCR5 gene. Questions that were posed to students for discussion are included in the Tables S2 and S3. One topic was the discussion of the novel and Δ32 mutations on CCR5's normal function in human health and disease. Whether Dr. He's novel _CCR5_ mutations in the twins actually confer HIV protection or has an effect on normal immune function has not been published yet. We asked students whether Lulu and Nana should be monitored by a specialized group of doctors and researchers for unintended consequences of their genetic edits since they have novel modified alleles that have not been studied before. Since Dr. He's made his claim, studies show that humans who are carriers for the Δ32 _CCR5_ allele have a faster recovery from strokes. 40 Furthermore, _CCR5_ plays a role in brain cognition, as mice that lack CCR5 have improved memory. 41

The use of CRISPR-Cas9 has transformed our ability to edit genome sequences from bacteria to humans. Here, we describe a species-specific version of CRISPR-Cas9 that has been modified from that developed for the relatedCandidaclade speciesCandida albicans(23,24). Our results highlight the necessity of modifying this system and the use of species-specific promoters to drive expression of CRISPR-Cas9 components for achieving high transformation efficiencies inC. lusitaniae. We show that CRISPR can be used to target multiple genes and that transformation efficiencies for template-directed repair are further increased upon deletion of the NHEJ pathway inC. lusitaniae. As an alternative approach, Hogan and colleagues have utilized the transformation of RNA-protein complexes that allow species-independent CRISPR-Cas9 editing ofCandidagenomes (see accompanying paper by Grahl et al. [34]). There are therefore now two distinct CRISPR-Cas9 methodologies that support efficient genome editing inC. lusitaniae.

Development of a CRISPR-Cas9 System for Efficient Genome Editing ofCandida lusitaniae

### Strains and media. 

_C. lusitaniae_ strains RSY284 (haploid **a**, _ura3_), RSY286 (haploid α, _leu2_), RSY147 (haploid **a**, _chx_r), RSY432 (diploid **a**/α, _ARG4/arg4_::_FRT ADE2/ade2_::_FRT URA3/ura3 chx_S/_chx_r) and RSY411 (haploid α, _arg4_::_FRT ade2_::_FRT chx_r) were used in this study (see Data Set S1 in the supplemental material). Strains RSY284 and RSY286 were mated to produce CAY5019 (diploid **a**/α, _URA3/ura3 LEU2/leu2_). All strains were grown in YPD (2% Bacto peptone, 2% glucose, 1% yeast extract, 25 µg/ml uridine) at 30°C with shaking, and transformants were selected on YPD supplemented with 200 µg/ml nourseothricin (YPD plus NAT).

### NHEJ plasmids and strain construction. 

_C. lusitaniae_ strain RSY147 was used to generate a _SAT1_-recycled _leu2_− strain (RSY281, haploid **a**, _leu2_), using homologous recombination of a _leu2_ deletion plasmid (pRB310) linearized with ApaI and SacII restriction enzymes (see Data Set S1 for primer sequences). _KU70_ and _LIG4_ deletion constructs were generated by PCR amplifying and cloning ~950-bp _C. lusitaniae_ 5′ and 3′ homologous flanks into the pSFS2a plasmid (26) containing the nourseothricin resistance marker (plasmids pRB254 [_KU70_] and pRB247 [_LIG4_]; see Data Set S1 for primer sequences). Plasmids were linearized with KpnI and SacII restriction enzymes prior to transformation. _C. lusitaniae_ strain RSY281 was used in electroporation transformations to delete _KU70_ (strain RSY376) and subsequently delete _LIG4_ (strain RSY426).

### CRISPR plasmids and DNA constructs. 

The CTG clade codon-optimized Cas9 (CaCas9) and guide RNA were PCR amplified from plasmid pV1093 (23) (see primers in Data Set S1). To optimize Cas9 and guide RNA expression in _C. lusitaniae_, the _C. albicans_ promoters were replaced with _C. lusitaniae_ constitutive _TDH3_ and _SNR52_ promoters, respectively (plasmids pRB732 to -734). The _TDH3_ locus for _C. lusitaniae_, _CLUG_03499_, is annotated in the Candida Genome Database (CGD) (35). The _C. lusitaniae SNR52_ locus was identified using the BLASTN tool on CGD to obtain the region of the _C. lusitaniae_ genome with the greatest homology to the _C. albicans SNR52_ gene. For _TDH3_ and _SNR52_, the upstream ~1-kb regions were PCR amplified from _C. lusitaniae_ genomic DNA, and fusion PCR was used to attach these fragments to the Cas9 and sgRNA scaffolds (see primers in Data Set S1). sgRNAs were designed to target _C. lusitaniae ADE2_, _UME6_, _MTLalpha1_, and _REC8_ genes by using 20-bp protospacers immediately followed by a PAM sequence of the structure NGG (or preceded by CCN if located on the reverse strand). Protospacers were identified based on previously established criteria (23), but we considered only candidates located within the first half of the genes so that mutations would more likely generate nonfunctional proteins. We also selected sgRNAs that had minimal off-target effects identified by cross-referencing the 12 nucleotides (nt) proximal to the PAM sequence against the _C. lusitaniae_ genome using NCBI download _Clavispora lusitaniae_ ATCC 42720 (assembly ASM383v1). By using oligonucleotide primers containing the unique 20-bp protospacer sequence plus a 20- to 30-bp overlap with the upstream or downstream sequence, we amplified the sgRNA in two fragments; these fragments were then stitched together by fusion PCR to integrate the new protospacer (see primers in Data Set S1). Deletion constructs were generated with either long (~1-kb) or short (80-bp) homologous flanks. Long-flank deletion constructs were generated using fusion PCR of 1-kb upstream and downstream flanking regions plus ~1.8 kb of the _SAT1_ marker, which were then cloned into a pCR-Blunt II-TOPO vector (_ADE2_, pRB620; _UME6_, pRB744; _MTLalpha1_, pRB768; _REC8_, pRB748). Short-flank deletion constructs were generated by PCR amplifying the ~4.3-kb _SAT1_ marker and flipper cassette with long oligonucleotides that contained ~80 bp of the upstream and downstream flanking regions of the target gene. PCRs (50-µl mixtures) were conducted with Phusion enzyme (Thermo Scientific) to generate all DNA constructs (1× Phusion HF buffer, 200 µM deoxynucleoside triphosphates [dNTPs], 0.2 µM [each] primer, ~20 to 50 ng template DNA, and 1 unit Phusion DNA polymerase; run according to the manufacturer’s recommendations). To stitch constructs together by fusion PCR, the products from prior Phusion PCRs were combined and diluted 10-fold, and then 1 µl was used as a template in the fusion PCR.

### _C. lusitaniae_ transformations. 

An electroporation transformation protocol used for both haploid and diploid strains was adapted from reference 21. (i) Overnight cultures were diluted to an optical density at 600 nm (OD600) of 0.4 in liquid YPD and then grown until they reached a final OD600 of 1.5 to 1.7. (ii) Cells were pelleted and resuspended in 10 ml of transformation buffer composed of 0.1 M lithium acetate (LiOAc), Tris-EDTA, pH 8.0, and 0.01 M dithiothreitol. (iii) Cells were incubated on a shaker at 22°C for 1 h, washed with ice-cold water, and resuspended in ice-cold 1 M sorbitol to acquire a concentration of 150 OD units/ml. (iv) Forty microliters of cells was combined with 3 µg deletion construct, 1 µg Cas9 construct, and/or 1 µg sgRNA construct (Table 1) in 0.2-cm electroporation cuvettes (Bio-Rad, Hercules, CA). (v) Cells were electroporated at 1.8 kV, 200 Ω, and 25 µF for 4.5 to 5 ms (Bio-Rad MicroPulser); immediately resuspended in 1 ml YPD; and allowed to recover overnight at 30°C. (vi) Cells were plated onto selective medium (YPD plus NAT) to identify successful transformants.

Systematic evaluation of genome-wide Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) off-target profiles is a fundamental step for the successful application of the CRISPR system to clinical therapies. Many experimental techniques and _in silico_ tools have been proposed for detecting and predicting genome-wide CRISPR off-target profiles. These techniques and tools, however, have not been systematically benchmarked. A comprehensive benchmark study and an integrated strategy that takes advantage of the currently available tools to improve predictions of genome-wide CRISPR off-target profiles are needed. We focused on the specificity of the traditional CRISPR SpCas9 system for gene knockout. First, we benchmarked 10 available genome-wide off-target cleavage site (OTS) detection techniques with the published OTS detection datasets. Second, taking the datasets generated from OTS detection techniques as the benchmark datasets, we benchmarked 17 available _in silico_ genome-wide OTS prediction tools to evaluate their genome-wide CRISPR off-target prediction performances. Finally, we present the first one-stop **i**ntegrated **G**enome-**W**ide **O**ff-target cleavage **S**earch platform (_iGWOS_) that was specifically designed for the optimal genome-wide OTS prediction by integrating the available OTS prediction algorithms with an AdaBoost ensemble framework.




Benchmarking and integrating genome-wide CRISPR off-target detection and prediction

To overcome CRISPR-Cas defense systems, many phages and mobile genetic elements encode CRISPR-Cas inhibitors called anti-CRISPRs (Acrs). Nearly all characterized Acrs directly bind Cas proteins to inactivate CRISPR immunity. Here, using functional metagenomic selection, we describe AcrIIA22, an unconventional Acr found in hypervariable genomic regions of clostridial bacteria and their prophages from human gut microbiomes. AcrIIA22 does not bind strongly to SpyCas9 but nonetheless potently inhibits its activity against plasmids. To gain insight into its mechanism, we obtained an X-ray crystal structure of AcrIIA22, which revealed homology to PC4-like nucleic-acid binding proteins. Based on mutational analyses and functional assays, we deduced that _acrIIA22_ encodes a DNA nickase that relieves torsional stress in supercoiled plasmids. This may render them less susceptible to SpyCas9, which uses free energy from negative supercoils to form stable R-loops. Modifying DNA topology may provide an additional route to CRISPR-Cas resistance in phages and mobile genetic elements.

The novel anti-CRISPR AcrIIA22 relieves DNA torsion in target plasmids and impairs SpyCas9 activity

Methods

**Background** Rhinoviruses (RVs) cause more than half of common cold and, in some cases, more severe diseases. Functional genomics analyses of RVs using siRNA or genome-wide CRISPR screen uncovered a limited set of host factors, few of which has proven clinical relevance.

**Results** Herein, we systematically compared genome-wide CRISPR screen and surface protein-focused CRISPR screen, referred to as surfaceome CRISPR screen, for their efficiencies in identifying RV host factors. It was found that surfaceome screen outperformed genome-wide screen in the success rate of hit identification. Importantly, using surfaceome screen we have identified olfactomedin like 3 (OLFML3) as a novel host factor of RV serotypes A and B including a clinical isolate. We found that OLFML3 was a RV-inducible suppressor of the innate immune response and that OLFML3 antagonized type I interferon (IFN) signaling in a SOCS3-dependent manner.

**Conclusion** Our study has suggested that RV-induced OLFML3expression is an important mechanism for RV to hijack the immune system and underscored surfaceome CRISPR screen in identifying viral host factors.

Surfaceome CRISPR Screen Identifies OLFML3 as a Rhinovirus-inducible IFN Antagonist

### **Design and construction of genome-wide and surfaceome CRISPR libraries**

sgRNAs were designed to target to protein coding regions (NCBI CCDS data, released on 8-Sep-2016) [**44**] and optimized by two steps. First, off-target scores were calculated according to an established algorithm [**30**]. Second, on-target scores were calculated using Rule Set 2 [**29**]. sgRNAs were ranked by on-target scores and the top 12 sgRNAs with off-target scores of less than 20 were selected for each gene. If less than 12 sgRNAs were obtained, the cutoff of off-target scores was increased sequentially to 40, 60, 80 and 100 until 12 sgRNAs were obtained. The genome-wide CRISPR library was divided into three sub-libraries A, B and C according to sgRNA rank. Non-targeting sgRNAs were included in each library.

Pooled sgRNA oligonucleotides were synthesized as 76-mers by Custom Array (Bothell, WA, USA) and were amplified by PCR with NEBNext High Fidelity PCR Master Mix (New England BioLabs, NEB, Ipswich, MA, USA) using customized primers (Additional file 2: **Table S1**). The PCR products were purified using MinElute PCR purification kit (Qiagen, Dusseldorf, Germany). Lentiviral vector LentiCRISPR-v2 was digested with Esp3I (Thermo Fisher Scientific, Waltham, MA, USA) at 37 °C for 3 h and gel-purified using Gel Extraction kit (Omega, Norcross, GA, USA). Purified digestion products were ligated to Esp3I-treated LentiCRISPR-v2 using Gibson assembly kit (NEB) following manufacturer’s instructions. The ligation product was purified by isopropanol precipitation and then transformed into electrocompetent _Escherichia coli_ (Lucigen, Middleton, WI, USA). Transformed cells were plated on to 15 cm Luria-Bertani (LB) agar plates supplemented with 50 μg/mL ampicillin (Sangon Biotech, Shanghai, China). Approximately 1-3 × 107 colonies were collected for each library to ensure 500-fold coverage. Plasmid DNA was extracted as pooled libraries using NucleoBond Xtra Maxi EF kit (Macherey-Nagel, Duere, Germany) and stored at −20 ℃.

### **Cell culture**

H1-Hela cells were obtained from the American Type Culture Collection (ATCC). HEK293T cells were obtained from the Cell Bank of Shanghai Institutes for Biological Science (SIBS) and were validated by VivaCell Biosciences (Shanghai, China). H1-Hela and HEK293T were grown in Dulbecco’s modified Eagle’s medium (DMEM, Thermo) supplemented with 10% fetal bovine serum (FBS, Thermo) and 1% penicillin-streptomycin (Thermo) and maintained at 37 ℃ in a fully humidified incubator containing 5% CO2. All cells were confirmed by PCR to be free of mycoplasma contamination.

### **LV production and transduction**

To produce LVs with high titers, HEK293T cells were seeded on to 6-well plates with 2 × 106 cells per well for single sgRNA, or 10 cm petri dishes with 107 cells for library construction. At 24 h after seeding, HEK293T cells at a confluence of 70%-90% were transfected with LV packaging plasmid pMD2.G, envelope plasmid psPAX and transfer plasmid pLentiCRISPR-v2 that carries single or pooled sgRNAs using Lipofectamine 3000 (Thermo). At 6 h after transfection, the medium was replaced with fresh medium. The medium supernatant containing LVs was harvested at 48 h post transfection by centrifugation at 2,000 rpm for 10 min, filtrated through a 0.45-μm filter and stored at −80 ℃.

H1-Hela cells were transduced with LVs at an MOI of 0.3 using spinfection. Briefly, H1-Hela cells were washed with phosphate buffered saline (PBS) and incubated with LVs in serum-free DMEM under centrifugation at 2,000 rpm for 2 h. Upon completion of spinfection, LV-containing medium was removed and cells were incubated in DMEM supplemented with 10% FBS and 2 μg/mL puromycin (Thermo) for 2 to 3 days to purge empty cells containing no LVs or sgRNAs. For each library, cells of more than 500-fold coverage of the library size were collected.

### **RV production and infection**

The full-length cDNA clones of RV-A16 (pR16.11, Cat. No. VRMC-8) and RV-B14 (pWR3.26, Cat. No. VRMC-7) were obtained from ATCC. To produce infectious viral RNA, RV-A16 and RV-B14 plasmids were linearized by SacI (NEB) digestion and then _in vitro_ transcribed using HiScribe T7 Transcription Kit (NEB). The RNA transcripts were extracted using Trizol (Thermo) and chloroform (Titan, Shanghai, China), followed by isopropanol precipitation. Viral RNA was transfected into H1-HeLa cells using Lipofectamine 3000 (Thermo) to generate infectious RV-A16 or RV-B14 particles. At 48 h post transfection, the supernatant containing RVs were collected for further infection on H1-Hela cells to produce RVs with higher titers. The aliquots of purified RVs were stored at −80 ℃. Virus titers were determined by the 50% tissue culture infectious dose (TCID50) assay.

For RV infection, H1-Hela cells were seeded on to 96- or 12-well plates with a density of 2×104 or 1.5×104 cells per well, respectively. Unless noted otherwise, at 24 h after seeding, cells were infected with RVs at an MOI of 2 for 1.5 h, washed with PBS for three times and then cultured in DMEM (Thermo) supplemented with 10% fetal bovine serum (FBS, Thermo) and 1% penicillin-streptomycin (Thermo) for 24 h.

### **CRISPR screen using RV-B14**

H1-Hela cells carrying genome-wide sub-library A or surfaceome library were seeded on to 15 cm petri dishes. Approximately 1.5 ×107 cells were seeded to ensure more than 200-fold coverage of sgRNA. These cells were infected with RV-14 at an MOI of 1 for 48 h. At the end point of RV-14 challenge, the cells were washed with PBS and attached cells were collected from the plates. Two biological replicates were performed for mock and test groups. Genomic DNA of the harvested cells was extracted using phenol: chloroform: isoamyl alcohol (v/v/v, 25:24:1) and then purified using ethanol precipitation.

### **Next-generation sequencing (NGS) analyses of sgRNA enrichment**

Genome-integrated sgRNAs were amplified from extracted genomic DNA by PCR using the primers containing Illumina adaptor (Additional file 2: **Table S1**). The PCR product was gel-purified and then analyzed on Illumina HiSeq 3000 platform by Genewiz (Suzhou, Jiangsu, China). After removing the adaptors, the 20 bp sgRNA was mapped to the reference sgRNA libraries with one nucleotide mismatch allowed for each sgRNA. Gini index was calculated to analyze the distribution of sgRNAs. The raw read counts were subjected to MAGeCK analyses [**45**] to determine the enriched sgRNA and genes. Enrichment of sgRNAs and genes was analyzed using MAGeCK (v0.5.7) by comparing the read counts from the cells infected with RV-B14 with those from uninfected cells. A false discovery rate (FDR) of less than 0.01 was applied to identify significantly enriched sgRNAs or gene knockout.

### **Generation of CRISPR-Cas9 knockout cells**

The transfer plasmid pLentiCRISPR-v2 carrying single sgRNA for targeted gene knockout was constructed as described above. Briefly, forward and reverse oligonucleotides encoding the 20 bp sgRNA (Additional file 2: **Table S2**) were annealed to generate double-stranded DNA with overhang that matched the sticky ends of Esp3I (Thermo)-treated pLeniCRISPR-v2 vector. Annealed sgRNA sequence was ligated into digested pLeniCRISPR-v2 and then transformed into DH5α _E. coli_ (Tsingke, Beijing, China).

The LVs carrying single sgRNA were packaged and transduced on to cells as described above. To evaluate the knockout efficiency, the genomic DNA of edited cells were extracted using Quick Extraction kit (Lucigen). Target sites carrying gene-edited sequences were PCR amplified using gene-specific primers (Additional file 2: **Table S3**). The knockout efficiency of each sgRNA was determined using T7E1 analysis. Single clones were obtained by limited dilution and genotyped by Sanger sequencing to determine the mutations at each allele. Knockout of target proteins was verified by western blotting (WB).

### **Cell viability assay**

H1-Hela cells were seeded into 96- or 24-well plates with a density of 5,000 or 50,000 cells per well respectively. At 24 h after seeding, cells were infected with RVs at an MOI of 1 for 1.5 h, washed with PBS for three times and cultured in fresh medium for 24 h unless noted otherwise. Cell counting Kit-8 (CCK-8, Dojindo, Kumamoto, Japan) was applied to determine cell viability according to manufacturer’s instructions. The absorbance at 450 nm was determined by Enspire multimode plate reader (PerkinElmer, Waltham, MA, USA).

### **Immunofluorescence (IF) and fluorescence** **_in situ_** **hybridization (FISH)**

IF staining of RV-B14 was performed using mouse anti-RV VP3 antibody (1:50, clone G47A, Thermo) and Alexa Fluor Plus 488 goat anti-mouse IgG (H+L) (1:1,000, A32723, Thermo). IF images were acquired and analyzed using Operatta high-content analysis system (PerkinElmer). At least 2,000 fluorescent cells were imaged and quantified for each replicate.

FISH was performed using an RNAscope Multiplex Fluorescent V2 Assay kit (Advanced Cell Diagnostics, Newark, USA) according to the manufacturer’s instructions. After fixation and pretreatment, RV-B14 RNA was detected using an RVB RNA probe (Advanced Cell Diagnostics, Cat. No. 447141) and TSA Plus Fluorescein (PerkinElmer, Cat. No. NEL741E001KT). FISH images were acquired and analyzed using a TissueFAXS 200 flow-type tissue quantitative analyser (TissueGnostics GmbH, Vienna, Austria). At least 5,000 cells in each replicate were included in analyses.

### **Real-time quantitative PCR (RT-qPCR)**

To determine viral loads, the medium supernatant or cell lysate containing RVs were harvested at 24 h after infection with RVs at an MOI of 2 unless noted otherwise. The total RNA from supernatant or cell lysate was purified using Trizol (Thermo) and chloroform (Titan), followed by purification using isopropanol precipitation. Purified viral RNA (vRNA) was reverse transcribed into cDNA using PrimeScript RT reagent Kit with gDNA Eraser (Takara Bio Inc., Shiga, Japan). The number of RV genome copy in the medium supernatant was determined using RT-qPCR with general or serotype-specific Taqman probe and primers (Additional file 2: **Table S4**) on Applied Biosystems Q6 Real-Time PCR cycler. The absolute viral titers were calculated based on a standard curve of RV genome with known TCID50 and the _R_ square of curve-fitting was guaranteed to be more than 0.99. The mRNA levels of RVs and IFN-stimulating genes (ISGs) in cell lysate were determined using RT-qPCR with SYBR green dye (Thermo) and specific primers (Additional file 2: **Table S5**) on Applied Biosystems Q6 Real-Time PCR cycler. All SYBR Green primers were validated with dissociation curves. The expression of vRNA and host genes in cell lysate is normalized to ribosomal gene RPLP0 (36b4).

### **Gene knockdown using siRNA**

H1-Hela cells were seeded on to 6-well plates with a density of 5×105 cells per well. At 24 h after seeding, cells were transfected with 100 pmol SOCS3 siRNA (Genepharma, Shanghai, China) (Additional file 2: **Table S6**) using 7.5 μL Lipofectamine 2000 (Thermo) for 6 h, washed with PBS and then cultured in fresh DMEM (Thermo) supplemented with 10% fetal bovine serum (FBS, Thermo). At 48 h post transfection, cells were infected with RV-B14 at an MOI of 2 for 1.5 h, washed with PBS for three times and then cultured in fresh medium for 24 h. The medium supernatant or cell lysate containing RVs were harvested and lysed for total RNA extraction, and then the mRNA levels of RVs, SOCS3, STAT1/2 and IFN-stimulating genes (ISGs) in cell lysate were determined using RT-qPCR as described above.

### **Virus attachment and entry assays**

H1-Hela cells were seeded in 12-well plate with a density of 200,000 cells per well and incubated overnight. For virus attachment assay, cells were incubated with RV-B14 or RV-A16 at an MOI of 20 in cold medium on ice for 60 min, washed by PBS for three times and then the total RNA was extracted using Trizol (Thermo). For virus entry assay, cells were incubated with RV-B14 or RV-A16 at an MOI of 20 in cold medium on ice for 60 min, washed by PBS for three times and then treated with pre-warmed medium for 40 min at 37 ℃. Then cells were washed with PBS for three times and treated with 0.25% trypsin for 2 min (Thermo, Cat. No. 25200072) to remove surface-bound viral particles. The internalized viral RNA was extracted using Trizol (Thermo) and viral loads were determined using RT-qPCR and FISH.

### **Rescue experiments by overexpression**

ICAM1, RAB5C and OLFML3 genes were codon-optimized for expression in human cells and synthesized by Genewiz. The 20 bp sgRNA-targeting sites and PAM sequences were mutated with silent mutations. Myc and FLAG tags were added to the C-terminus of these genes for WB detection. These genes were cloned into the EcoRI and XhoI sites of pCAGG plasmid that carries a separate mScarlet fluorescent protein as a transfection reporter.

### **WB analysis**

For WB analysis, cells were lysed with RIPA buffer (Beyotime Biotechnology, Beijing, China) on ice for 10 min. The total protein concentration in cell lysate was determined using the BCA Protein Assay Kit (Thermo). Cell lysate was mixed with SDS-PAGE loading buffer (Takara) containing 200 mM DTT, incubated at 95 °C for 10 min and resolved on NuPAGE 4-12% Bis-Tris gels (Thermo). Protein samples were transferred onto nitrocellulose membranes using an iBlot gel transfer system (Thermo). The following primary and secondary antibodies were used in WB including anti-ICAM1 rabbit antibody (Cell Signaling Technology, Cat. No. 4915S, Danvers, USA), anti-RAB5C rabbit antibody (Thermo, Cat. No. PA551932), anti-OLFML3 rabbit antibody (Thermo, Cat. No. PA531581), HRP-conjugated anti-rabbit IgG (CST, Cat. No. 7074S). Anti-β actin antibody conjugated with HRP (Abcam, Cat. No. ab49900, Cambridge, UK) was used as an internal control.

### **RNA-Seq analysis of mock and OLFML3-/-** **cells**

Non-targeting sgRNA-treated mock cells and OLFML3-/- cells were seeded on to 10 cm plates at a density of 1×106 cells per plate. At 24 h after seeding, the cells were infected with RV-B14 at an MOI of 1 or with PBS as mock infection. After 24 h of RV challenge, the remaining cells were washed with PBS and collected by Trizol (Thermo) treatment. Three biological triplicates were prepared for each group. The whole-transcriptome RNA sequencing was performed by Genewiz.

RNA-Seq short reads were aligned to the human genome (GRCh37) using Hisat2 (v2.0.1) [**46**]. Gene expression was counted as the number of short reads fully or partially aligned to the annotated gene model using HTSeq (v0.6.1) [**47**]. Genotype (non-targeting sgRNA-transduced mock cells and OLFML3-/- cells) and treatment conditions (mock and RV-B14 infection) were the factor variables in our RNA-Seq data. DESeq2 (v1.26.0) [**48**] was used to examine the difference between the response of mock and knockout cells to RV infection, which was captured by the interaction term. RV infection-induced gene upregulation and downregulation were calculated and the differentially upregulated or downregulated genes in mock and knockout cells were defined as DEGs. Significant DEGs were filtrated with an adjusted _P_ value of less than 0.05 and a fold change value of more than 2.

GO enrichment analysis was performed using cluster Profiler (v3.14.3) by comparing DEGs to a list of all human genes [**49**]. Adjusted _P_ value of less than 0.001 or 0.05 was set as the filter for biological process and molecular function terms respectively.

### **Isolation of clinical RV strain**

Nasopharyngeal swab samples were collected from the hospitalized patients bearing respiratory infection symptoms such as fever, cough, pharyngalgia and others. Nasopharyngeal swab samples were maintained in viral-transport medium and all samples were conserved on −80°C until analyses. The swab samples were diagnosed for respiratory viruses and the remaining samples were used for isolation of RVs. All procedures were complied with the Measures for the Ethical Review of Biomedical Research Involving Human Subjects issued by the National Health and Family Planning Commission of the People’s Republic of China. The Ruijin Hospital Ethics Committee, Shanghai Jiaotong University School of Medicine, approved the sample collection protocol with a permit number of Ruijin Hospital Ethics Committee 2018-48.

For isolation of RVs, the specimens were centrifuged at 2,000 rpm for 10 min and the supernatant was collected for RT-qPCR analysis using universal RV primers (**Supplementary Table 4**). The samples with high RV loads were selected for subsequent RV isolation in H1-Hela cells. H1-Hela cells were seeded on to 12-well plates at a cell density of 100,000 cells per well. Specimen supernatant (700 μL) was mixed with 300 μL of fresh medium and incubated with cells under centrifugation at 2,000 rpm for 2 h. Upon completion of spinfection, the supernatant was replaced with fresh medium and cells were incubated for another 2 days. Thereafter, medium supernatant and attached cells were collected and subjected to quick freeze-thaw cycles for three times to release viral particles. The medium supernatant or cell lysate from above was added to cells on 12-well plates. At 24 h after incubation, the supernatant was replaced with fresh medium and cells were incubated for 2 to 3 days before reaching 100% confluence. Isolation of RV strain was confirmed by CPEs and sequencing results.

### **Statistical analyses**

All data are the results from at least three biological replicates and are shown as mean ± SD unless noted otherwise. Statistical analyses and graphing were performed with GraphPad Prism 7.0. The _P_ values were determined using two-tailed unpaired Student’s _t_-test unless otherwise noted.

The advance of CRISPR/Cas9 technology has enabled us easily to generate gene knockout cell lines by introducing insertion/deletion mutations (indels) at the target site via the error-prone non-homologous end joining repair system. Frameshift-promoting indels can disrupt gene functions by generation of a premature stop codon. However, there is growing evidence that targeted genes are not always knocked-out by the indel-based gene disruption. In this study, we optimized CRISPR-del, which induces a large chromosomal deletion by cutting two different target sites, to perform “complete” gene knockout in non-transformed human diploid RPE1 cells. By improving several procedures, the optimized CRISPR-del allowed us to generate knockout cell lines harboring bi-allelic large chromosomal deletions in a high-throughput manner. Quantitative analyses show that the frequency of gene deletion with this approach is much higher than that of conventional CRISPR-del methods. The lengths of the deleted genomic regions demonstrated in this study are longer than those of 95% of the human protein-coding genes. Furthermore, the ability of this method to introduce a large chromosomal deletion enables the generation of a model cell line having a bi-allelic cancer-associated chromosomal deletion. Overall, these data lead us to propose that the optimized CRISPR-del is a high-throughput method for performing “complete” gene knockout in RPE1 cells.

Quantitative analysis of CRISPR-del for complete gene knockout in human diploid cells

### **Cell culture**

RPE1 cells were grown in Dulbecco’s Modified Eagle’s Medium F-12 (DMEM/F-12) with 10% FBS and 1% penicillin/streptomycin. All cell lines were cultured at 37°C in a humidified 5% CO2 incubator.

### **sgRNA synthesis**

sgRNAs were transcribed _in vitro_ from PCR-generated DNA templates according to a previously published method (**DeWitt et al., 2016**) with slight modifications. Briefly, template DNA was assembled by PCR from five different primers: 1) a variable forward primer containing T7 promoter and desired guide sequence, 2) a variable reverse primer containing the reverse complement of the guide sequence and the first 15 nt of the non-variable region of the sgRNA, 3) a forward primer containing the entire invariant region of the sgRNA, and 4), 5) two amplification primers. The assembled template was purified and subjected to _in vitro_ transcription by T7 RNA polymerase using the Hiscribe T7 High Yield RNA Synthesis Kit (New England Biolabs). The reaction product was treated with DNAse I, and the synthesized sgRNA was purified using the RNA Clean & Concentrator Kit (ZYMO RESEARCH). All sgRNA and primer sequences used in this study are listed in Supplementary Table1 and 2, respectively.

### **CRISPR-del-mediated gene knockout**

For large chromosomal deletions, the CRISPR-del method was performed with Cas9 protein and two synthesized sgRNAs. Cas9/sgRNA RNPs were electroporated into RPE1 cells using the Neon Transfection System (Thermo Fischer Scientific) according to the manufacturer’s protocol. Briefly, HiFi Cas9 protein (1.55 μM) from Integrated DNA Technologies (IDT) and two sgRNAs (0.92 μM each) were pre-incubated in Resuspension buffer R and mixed with cells (0.125 x105/μl) and Cas9 electroporation enhancer (1.8 μM, IDT). After resuspension, electroporation was immediately conducted using a 10 μl Neon tip at a voltage of 1300 V with two 20 ms pulses. The transfected cells were seeded into a 24-well plate. After recovery from the electroporation, single cells were isolated into 96-well plates using cellenONE (cellenion) according to the manufacturer’s protocol. After cell expansion, each 96-well plate was duplicated for genotyping and preparation of a frozen stock. Briefly, cells in 96-well plates were washed with PBS and treated with 25 μl of Trypsin/EDTA solution (nacalai tesque). After brief incubation at 37°C, the detached cells were resuspended with 75 μl of Cell Reservoir One (nacalai tesque), a DMSO-free cryopreservation medium. 25 μl of the cell mixture were transferred into a well of another 96-well plate filled with 175 μl of growth medium for cell expansion followed by genotyping analysis. The 96-well plate with the remaining cell suspension was placed in a deep freezer at −80°C.

### **Genotyping**

For high-throughput genotyping, genomic DNA was directly extracted from single cell clones in 96-well plates using DNAzol Direct (Molecular Research Center) and subjected to PCR for the detection of both WT and the deleted alleles using appropriate primers. Briefly, after removal of culture medium, 20 μl of DNAzol Direct was added to each well and the 96-well plate was shaken at 800 rpm for 20 min at room temperature. 1μl of the lysate containing genomic DNA was used for 10 μl of PCR reaction using KOD One PCR Master Mix (TOYOBO). The PCR products were analyzed by the automated microchip electrophoresis system MCE-202 MultiNa (Shimadzu). To confirm the genotype of homozygous KO clones, their genomic DNA was purified using NucleoSpin DNA RapidLyse kit (Macherey-Nagel) and subjected again to genotyping PCR. The PCR products were analyzed by agarose gel electrophoresis.

### **CRISPR/Cpf1-mediated gene knock-in**

Endogenous mNG tagging of HNRNPA1 by CRISPR-Cpf1 system was performed with the electroporation of Cpf1/crRNA RNP and dsDNA repair template. crRNA was designed to target the site immediately downstream of the stop codon of _HNRNPA1_ and transcribed _in vitro_ as described above. The DNA template for crRNA synthesis was assembled by PCR using a forward primer containing T7 promoter and the target sequence, and a reverse primer containing the reverse complement of the target sequence and the non-variable region of crRNA. The dsDNA repair template was amplified by PCR from a plasmid encoding the mNG sequence using two primers containing 90 bp left and right homology arm sequence, respectively. Electroporation of Cpf1/crRNA and the repair template was conducted similarly to the Cas9/sgRNA condition described above, with a modification in the electroporation solution. A.s.Cpf1 Ultra (1 μM, IDT) and crRNA (1 μM) were pre-incubated in buffer R and mixed with cells (0.125 ×105 /μL), Cpf1 electroporation enhancer (1.8 μM, IDT) and the repair template (33 nM). mNG-positive cells were sorted using FACS Aria III (BD Biosciences), equipped with 355/405/488/561/633 nm lasers. After cell expansion, cell sorting was repeated to re-enrich mNG-positive cells. The knock-in cell pool was subjected to CRISPR-del experiments. The crRNA sequence is listed in Supplementary Table1.

### **CRISPR-del efficiency assessed by Flow cytometry**

HNRNPA1-mNG knock-in cells were electroporated with Cas9 and the indicated sgRNA pairs for CRISPR-del application. After 10 days in culture, the cells were harvested with trypsin/EDTA solution, washed in cold PBS, and analyzed by FACS Aria III for mNG expression. Data for approximately 5,000 gated events were collected. Due to the presence of a very small portion of mNG-negative cells in control samples, the mean percentage (1.47 %) was then subtracted from that of each CRISPR-del sample before comparison.

### **Western blotting**

Cells were lysed on ice in lysis buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1% Triton X-100, 5 mM EGTA, 1 mM DTT, 2 mM MgCl2, and 1:1,000 protease inhibitor cocktail [Nakarai Tesque]). After centrifugation, the supernatant was added to Laemmli sample buffer, boiled and subjected to SDS-PAGE. Separated proteins were transferred onto Immobilon-P membrane (Merck) using Trans-Blot SD Semi-Dry Electrophoretic Transfer Cell (Bio-Rad Laboratories). The membranes were probed with the primary antibodies, followed by incubation with their respective HRP-conjugated secondary antibodies (Promega). The membrane was soaked with Chemi-Lumi Super (Nakarai Tesque) for the signal detection using ChemiDoc XRS+ (Bio-Rad Laboratories).

### **Immunofluorescence**

For the immunofluorescence analyses, cells cultured on coverslips (Matsunami) were fixed with −20 °C methanol for 7 min, or 4 % PFA at room temperature for 30min. Fixed cells were incubated with blocking buffer (1 % bovine serum albumin in PBS containing 0.05 % Triton X-100) for 30 min at room temperature. The cells were then incubated with primary antibodies in the blocking buffer for 1 hour in a humid chamber. After washing with PBS, the cells were incubated with secondary antibodies and Hoechst 33258 (DOJINDO, 1:3000–1:5000) in the blocking buffer for 30 min, followed by final wash with PBS. The coverslips were mounted onto glass slides (Matsunami) using ProLong Gold (Molecular Probes), with the cell side down.

### **Antibodies**

The following primary antibodies were used in this study: anti-CEP128 (abcam, ab118797; IF 1:500, WB 1:1000), anti-γ-tubulin (Merck, GTU88; IF 1:1,000), anti-p15 (Santa Cruz Biotechnology, sc-271791; WB 1:1000), anti-HSP90 (BD Biosciences, 610419; WB 1:1000) and anti-_β_-actin (Santa Cruz Biotechnology, sc-47778; WB 1:1000). The following secondary antibodies were used: anti-mouse IgG Alexa Fluor 488 (Molecular Probes, 1:1000), anti-rabbit IgG Alexa Fluor 555 (Molecular Probes, 1:1000), anti-mouse IgG HRP (Promega, WB 1:10000) and anti-rabbit IgG HRP (Promega, WB 1:10000).

### **Statistics analysis**

Statistical comparison between the data from different groups was performed in PRISM v.8 software (GraphPad) using either a Mann–Whitney U test or a Tukey–Kramer test as indicated in the legend. P values <0.05 were considered statistically significant. All data shown are mean ± S.D. The sample size is indicated in the figure legends.

Treg cell therapy is a promising curative approach for a variety of immune-mediated conditions. CRISPR-based genome editing allows precise insertion of transgenes through homology-directed repair, but use in human Tregs has been limited. We report an optimized protocol for CRISPR-mediated gene knock-in in human Tregs with high-yield expansion. To establish a benchmark of human Treg dysfunction, we targeted the master transcription factor FOXP3 in naive and memory Tregs. Although FOXP3-knockout Tregs upregulated cytokine expression, effects on suppressive capacity manifested slowly and primarily in memory Tregs. Moreover, FOXP3-knockout Tregs retained their characteristic phenotype and had few changes in their DNA methylation landscape, with FOXP3 maintaining methylation at regions enriched for AP-1 binding sites. Thus, while FOXP3 is important for human Treg development, it has a limited role in maintaining mature Treg identity. Optimized gene knock-in with human Tregs will enable mechanistic studies and the development of tailored, next-generation Treg cell therapies.

### **Competing Interest Statement**

MKL received research funding from Sangamo Therapeutics, Bristol-Myers Squibb, Pfizer, Takeda, and CRISPR Therapeutics for work unrelated to this study. All other authors declare no competing interests.

Optimized CRISPR-mediated gene knock-in reveals FOXP3-independent control of human Treg identity

### **CRISPR and HDR template design and assembly**

_CD226_- and _FOXP3_-targeting gRNAs were synthesized as chemically modified crRNAs (IDT) and duplexed at a 1:1 molar ratio with tracrRNA (IDT) by denaturation for 5 min at 95°C followed by gradual cooling to room temperature. The resulting gRNA was complexed with Cas9 (QB3 Macrolab) at a 2:1 molar ratio for 10 min at room temperature. gRNA target sequences: CD226 (5′-GTTAAGAGGTCGATCTGACG-3′), FOXP3 CR1 (5′-AGGACCCGATGCCCAACCCC-3′), FOXP3 CR2 (5′-GGGCCGAGATCTTCGAGGCG-3′), FOXP3 CR3 (5′-GCAGCTGCGATGGTGGCATG-3′), FOXP3 CR4 (5′-TGCCCCCCAGCTCTCAACGG-3′), FOXP3 CR5 (5′-CCCACCCACAGGGATCAACG-3′). HDR designs contained homology arms of at least 800 bp flanking a promoterless or EF-1α-driven transgene marker (sfGFP or ΔNGFR). Fragments were PCR-amplified from genomic DNA, plasmids, or synthesized (IDT gBlocks or GenScript GenParts), then cloned by Gibson assembly into an AAV transfer plasmid between two ITRs.

### **AAV6 production, purification, and titration**

For recombinant AAV6 production, HEK-293T/17 cells (ATCC CRL-11268) were co-transfected by calcium phosphate with an adenoviral pHelper plasmid, a pAAV6-Rep-Cap plasmid (both Cell Biolabs), and an ITR-flanked transfer plasmid at a 1:1:1 molar ratio. Cells were harvested after 3 d and AAV6 was purified with a filtration-based kit (Takara AAVpro Purification Kit) per the manufacturer’s protocol. AAV6 vector genomes were titrated by ITR-specific quantitative PCR (Takara AAVpro Titration Kit Ver.2) per the manufacturer’s protocol.

### **Cell isolation and culture**

Human blood sample collection from healthy adults was performed in accordance with protocols approved by the University of British Columbia Clinical Research Ethics Board and Canadian Blood Services. CD3+ T cells or CD4+ T cells were isolated via Lymphoprep and RosetteSep (both STEMCELL Technologies). For Treg isolation, CD4+ T cells were enriched with CD25 MicroBeads II (Miltenyi Biotec) before flow sorting on a MoFlow Astrios (Beckman Coulter) or FACSAria IIu (BD Biosciences). Sorting strategies: total Tregs (CD4+CD25hiCD127lo), naive Tregs (nTreg; CD4+CD25hiCD45RA+CD127lo), memory Tregs (mTreg; CD4+CD25hiCD45RA−CD127lo), and conventional T cells (Tconv; CD4+CD25loCD127hi). Unless otherwise indicated, all cells were cultured at 37°C, 5% CO2 in X-VIVO 15 (Lonza) supplemented with 5% (v/v) human serum (WISENT), 1% (v/v) penicillin-streptomycin (Gibco), 2 mM GlutaMAX (Gibco), and 15.97 mg/L phenol red (Sigma-Aldrich). During optimization, in some cases as indicated, CTS OpTmizer T Cell Expansion SFM (Gibco) was used in place of X-VIVO 15, human serum, and phenol red.

### **Human Treg expansion with CRISPR/HDR-mediated gene knock-in**

In all cases during pre-activation and expansion, media was supplemented with IL-2 (Proleukin; 1000 IU/ml for Tregs, 100 IU/ml for Tconvs or CD4+ T cells). Tregs, Treg subsets, Tconvs, and CD4+ T cells were pre-activated with either CD3/CD28/CD2 tetrameric antibody complexes (ImmunoCult by STEMCELL Technologies) or 1:1 aAPCs for 5 d. aAPCs were L cells (ATCC CRL-2648) expressing CD32, CD58, and CD80, gamma-irradiated (75 Gy), and loaded with and anti-CD3 (OKT3, 0.1 μg/ml; University of British Columbia Antibody Lab) (de Waal **Malefyt et al., 1993**). Media and IL-2 were replenished every 2–3 d.

Cells were electroporated with the Neon Transfection 10 μL Kit (Invitrogen) per the manufacturer’s protocol. Briefly, cells were washed twice with PBS, resuspended in Buffer T (≤20×106 cells/mL) with Cas9 or RNP (40 pmol gRNA + 20 pmol Cas9 per transfection), electroporated at 1400 V / 30 ms / 1 pulse, then immediately transferred into prewarmed antibiotic-free media containing AAV6 at the specified vg/cell and simultaneously expanded with aAPCs and IL-2 for 7 d as above. After 7 d expansion (total 12 d), cells were flow-sorted as sfGFP+, then rested overnight in reduced IL-2 (100 IU/ml for Tregs, none for Tconvs) before functional assays; in some cases, flow-sorted cells were further expanded for 7 d with aAPCs and IL-2 as above before use in assays. Cell counts at 5 d were determined by trypan blue (Gibco); cell counts at 12 d and 19 d were determined by ViaStain Acridine Orange/Propdium Iodide Staining Solution (Nexcelom).

For experiments investigating the role of FOXP3 in Treg biology, we pre-activated cells with CD3/CD28/CD2 tetramers followed by aAPC-based expansions, for two reasons. First, Treg yield by day 19 of expansion was comparable to a fully aAPC-based expansion (**Figure 1D**). Second, we found that repetitive (3×) aAPC-based stimulation occasionally resulted in spontaneous FOXP3 loss (data not shown), which could be circumvented by using a weaker reagent to pre-activate Tregs, i.e., CD3/CD28/CD2 tetramers.

During optimization, cells were alternatively pre-activated with IL-2 as above and one of: 1:1 anti-CD3/anti-CD28-coated beads (Gibco Dynabeads Human T-Expander or Human Treg Expander), CD3/CD28 tetrameric antibody complexes (ImmunoCult by STEMCELL Technologies), or a combination of plate-bound anti-CD3 (OKT3, 10 μg/ml in 0.1 M Trizma HCl; Sigma-Aldrich) with soluble anti-CD28 (CD28.2, 4 μg/ml; BD Biosciences). Dynabead-activated cells were magnetically de-beaded immediately prior to electroporation. Divergent T cell pre-activation conditions and Neon electroporation parameters are specified as applicable. In some cases, HDR Enhancer (IDT, v1) was added to AAV6-containing media prior to electroporation at the indicated concentrations.

### **Evaluation of genome editing**

To assess indel formation after CRISPR editing, genomic DNA was extracted from cells 3 d after electroporation using a QIAamp DNA Mini Kit or Micro Kit (QIAGEN) per the manufacturer’s protocols. The region flanking the gRNA cut site was PCR-amplified, Sanger-sequenced, and chromatograms uploaded to the ICE analysis tool (Synthego; https://ice.synthego.com/) (**Hsiau et al., 2019**). Primer sequences for FOXP3 CR1 and FOXP3 CR3: FWD (5′-CTAGAGCTGGGGTGCAACTATG-3′), REV* (5′-TCTTCTCTTGTCACATGGGGATG-3′). Primer sequences for FOXP3 CR5: FWD* (5′-GAATGGCCGTCTTTAAGCTTCTC-3′), REV (5′-TTATTGGGATGAAGCCTGAGCTG-3′). An asterisk (*) denotes the primer used for Sanger sequencing.

### **T cell suppression assay**

Expanded Tregs and ex vivo-isolated allogeneic CD3+ T cells (responder cells) were labelled with Cell Proliferation Dye (Invitrogen) eFluor 670 and eFluor 450, respectively, then cocultured at the indicated ratios and activated with anti-CD3/anti-CD28-coated beads (Gibco Dynabeads T-Expander; 1:16 bead:responder cells) for 4 d. Proliferation of CD4+ and CD8+ cells within the responder-cell fraction was determined by dilution of Cell Proliferation Dye; responder cells activated with beads served as positive controls. Percent suppression of proliferation was calculated as: (1 − (division index of sample / division index of positive control)) * 100.

### **Assessment of cytokine production**

Expanded Tregs or Tconvs were activated with IL-2 (100 IU/mL) and 1:1 anti-CD3/anti-CD28-coated beads (Gibco Dynabeads T-Expander) for 4 d. For secreted cytokines, supernatants were collected, cytokines measured by cytometric bead array (BioLegend LEGENDplex Human Th Panel 13-plex), and data analysed with Qognit software (BioLegend) per the manufacturer’s protocols. For intracellular cytokine expression, cells were restimulated with PMA (10 ng/ml), ionomycin (500 μg/ml), and brefeldin A (10 μg/ml; all Sigma-Aldrich) for an additional 4 h.

### **Flow cytometry**

Antibodies are listed in **Table S2**. Cells were stained for surface proteins in PBS (Gibco) or Brilliant Stain Buffer (BD Biosciences) for 20 min at room temperature; Fixable Viability Dye (Invitrogen) or 7-AAD (BioLegend) was used to exclude dead cells. For detection of intracellular proteins, cells were fixed and permeabilized with the eBioscience Foxp3 / Transcription Factor Staining Buffer Set (Invitrogen) for 40 min at room temperature, then stained for intracellular proteins for 40 min at room temperature. As applicable, Brilliant Stain Buffer Plus (BD Biosciences) was included during intracellular staining. Samples were acquired on an LSRFortessa X-20, FACSymphony A5 (both BD Biosciences), or Cytoflex (Beckman Coulter), and data were analysed with FlowJo software (BD Biosciences; v10.7). All live single cells were first gated as CD4+, except for the T cell suppression assay in which live single cells were gated as CD4+ or CD8+. Percent protein KO was calculated as: (1 – (% marker+ of RNP-edited sample / % marker+ of negative control)) * 100.

### **Genome-wide DNA methylation profiling**

FOXP3HDR-KO nTregs, Cas9 nTregs, and Tconvs expanded for 20 d from 5 individuals were used for genome-wide DNA methylation profiling. Genomic DNA was isolated with an AllPrep DNA/RNA Mini Kit (QIAGEN), bisulfite-converted with an EZ DNA Methylation Kit (Zymo Research), hybridized to Infinium MethylationEPIC BeadChips (Illumina), and imaged using an iScan System (Illumina) per the manufacturers’ protocols. Raw red and green intensity data (IDAT) were imported into the R statistical environment (v4.0.2) with RStudio (v1.3.1093), converted into methylated and unmethylated signals, normalized using noob (**Triche et al., 2013**) and functional normalization (**Fortin et al., 2014**) by preprocessFunnorm in minfi (v1.34.0) and by BMIQ (**Teschendorff et al., 2013**) in wateRmelon (v1.32.0), and corrected for batch effects (BeadChip ID and position) using ComBat in sva (v3.36.0). Probes with a detection p-value > 0.01 or a bead count < 3 in at least one sample were discarded. Probes previously characterized to have low-quality genomic mapping, cross-reactive or polymorphic sequences, or type I probes targeting a common SNP that could cause a colour channel switch were discarded (**Zhou, Laird and Shen, 2017**) (https://zwdzwd.github.io/InfiniumAnnotation). Because the cohort contained cells from males and females, probes targeting X/Y chromosomes were discarded. The total remaining CpG sites after preprocessing was 702,481. Results were visualized as beta values, which range from 0 (unmethylated) to 1 (fully methylated).

Principal component analysis was performed using prcomp. Differential methylation analysis for all paired comparisons was performed with logit-transformed beta values (M values), considered more statistically robust (**Du et al., 2010**), using limma (v3.44.3). CpG sites with a mean methylation (beta) difference > 0.05 and Benjamini-Hochberg-corrected p-value < 0.05 were considered differentially methylated. DMRs were then identified by DMRcate (v2.2.3) (**Peters et al., 2015**) using a bandwidth of 1000 nt (lambda = 1000) and a scaling factor of 2 (C = 2). Regions with contiguous, differentially methylated CpG sites within lambda nucleotides, a region mean methylation (beta) difference > 0.05, and Fisher’s multiple-comparison statistic < 0.05 were considered significantly differentially methylated.

To identify differentially methylated regions that overlapped a FOXP3-binding region, a publicly available human FOXP3 whole-genome ChIP-chip tiled array dataset was used (**Sadlon et al., 2010**). The genomic coordinates of FOXP3-binding sites identified by model-based analysis of tiling-arrays were lifted over from hg18 to hg19, then compared to the DMRs identified above. A DMR was considered overlapping a FOXP3-binding region if there was an overlap of at least 1 nt.

HOMER (**Heinz et al., 2010**) was used for transcription factor binding motif enrichment analysis. The coordinates of demethylated DMRs were scanned using findMotifsGenome.pl with default parameters, exact (given) fragment size, and “-mask” to mask repeat sequences. Background sequences with matching GC content were randomly selected as controls. Demethylated DMRs were searched for overrepresented motifs with lengths of 8, 10, and 12 bp relative to the background sequences. De novo motifs with p-value < 1×10−12 were considered enriched.

### **Statistical analysis**

Normality was assumed. Statistical significance between more than two groups was determined by matched 1-way ANOVA or matched 2-way mixed-effects model and Dunnett’s or Tukey’s multiple comparisons tests, as appropriate. Significance between two groups was determined by paired t-test; for lognormal distributions, ratio paired t-test or Mann-Whitney test was used, as appropriate. For AAV6 dose response experiments, areas under the curve were used to determine significance. For suppression of proliferation, areas under the curve were used to determine significance, taking into account all degrees of freedom, as previously described (**Akimova et al., 2016**). The n values used to calculate statistics are defined in the figure legends; significance (p < 0.05 was considered significant) is indicated within the figures. Analysis was performed using Prism software (GraphPad; v9.0.0). Statistical testing of DNA methylation patterns was performed in R, as described in the preceding section.

The type V-K CRISPR-Cas system, featured by Cas12k effector with a naturally inactivated RuvC domain and associated with Tn7-like transposon for RNA-guided DNA transposition, is a promising tool for precise DNA insertion. To reveal the mechanism underlying target DNA recognition, we determined a cryo-EM structure of Cas12k from cyanobacteria _Scytonema hofmanni_ in complex with a single guide RNA (sgRNA) and a double-stranded target DNA. Coupled with mutagenesis and _in vitro_ DNA transposition assay, our results revealed mechanisms for the recognition of the GGTT PAM sequence and the structural elements of Cas12k critical for RNA-guided DNA transposition. These structural and mechanistic insights should aid in the development of type V-K CRISPR-transposon systems as tools for genome editing.

Structural basis of target DNA recognition by CRISPR-Cas12k for RNA-guided DNA transposition

### **Protein expression and purification**

Gene fragments of TnsB, TnsC, and TniQ were ordered from Integrated DNA Technologies (IDT) and cloned into bacterial expression plasmid pET28-MKH8SUMO (Addgene: #79526). The gene fragment for Cas12k was cloned into the bacterial expression plasmid pET-His6-StrepII-TEV LIC (Addgene: #29718). Cas12k, TnsB, and TniQ were expressed in _Escherichia coli_ BL21(DE3) while TnsC was expressed in Rosetta™(DE3)pLysS (Novagen: #70956) containing a pLysS-tRNA plasmid. Cells were grown to OD600 =0.6 in Terrific Broth (TB) and protein expression was induced by adding 0.3 mM of IPTG followed by overnight incubation at 16°C. The cells were collected and resuspended in lysis buffer (50 mM Tris-HCl, pH□7.6, 500 mM NaCl, 5% glycerol) supplemented with 1 mM PMSF and 5 mM β-mercaptoethanol, and then disrupted by sonication. Cell lysate was clarified by centrifugation. The supernatant was loaded onto Ni-NTA resin. After extensive washing with lysis buffer supplemented with 30 mM imidazole, target proteins were eluted with lysis buffer supplemented with 250 mM imidazole. The His-SUMO tag of TnsB, TnsC, and TniQ and His-StrepII tag of Cas12k were removed by overnight digestion with TEV protease at 4°C. The protein was diluted with buffer containing 50 mM Tris-HCl pH 7.6, 200 mM NaCl, and 5% glycerol and loaded onto a Heparin column (GE Healthcare), eluted with a linear NaCl gradient (0.1 to 1M). After concentration, the proteins were further purified by size exclusion chromatography (SEC) over a Superdex 200 increase 10/300 GL column (Cytiva) in buffer containing 25 mM Tris-HCl (pH□7.6), 500 mM NaCl, 10% glycerol, and 1 mM DTT (0.5 mM EDTA was added to the buffer for TnsC). Fractions were concentrated and stored at −80°C.

To assemble the Cas12k–sgRNA binary complex, Cas12k proteins were incubated with sgRNA **(Table S1)** at a ratio of 1:1.15 at 37°C for 30 min in buffer A (25 mM Tris-HCl, pH□7.6, 150 mM NaCl, 2 mM DTT and 1 mM MgCl2). To reconstitute the Cas12k–sgRNA–target DNA ternary complex, Cas12k protein was incubated with sgRNA at 37°C for 30 min followed by the addition of target DNA synthesized from IDT **(Table S1)** at a ratio of 1:1.1.5:1.3. After 30 min, the mixture was subjected to SEC over a Superdex 200 column (Cytiva) equilibrated with buffer A for further purification.

### **sgRNA preparation**

sgRNAs were produced by _in vitro_ transcription using the HiScribe T7 High Yield RNA synthesis kit (NEB) with PCR amplified gBlocks (IDT) as templates. sgRNAs were purified over a Resource-Q column (Cytiva) and eluted with a linear NaCl gradient (50 mM–1000 mM) in 25 mM Tris-HCl, pH 8.0. The eluted sgRNAs were concentrated and stored at −80°C

### **Mutagenesis**

Single amino acid mutations were introduced by the QuikChange site-directed mutagenesis method. Mutations with multiple amino acids were introduced by ligating inverse PCR-amplified backbone with mutations bearing DNA oligonucleotides via the In-Fusion Cloning Kit (ClonTech). All mutants were confirmed by Sanger sequencing.

### **_In vitro_** **transposition assay**

Donor plasmid (pDonor) and target plasmid (pTarget) were gifts from Feng Zhang (Addgene #127924 and #127926, respectively). _In vitro_ transposition reaction was conducted as previously described unless otherwise stated. All proteins were diluted to 2 μM with 25 mM Tris-HCl, pH 8.0, 500 mM NaCl, 1 mM EDTA, 1 mM DTT, and 25% glycerol. 50 nM of each proteins, 600 nM sgRNA, 20 ng pTarget, and 100 ng pDonor were added sequentially to the reaction buffer containing 26 mM HEPES pH 7.5, 4.2 mM Tris-HCl pH 8.0, 2.1 mM DTT, 0.05 mM EDTA, 0.2 mM MgCl2, 28 mM NaCl, 21 mM KCl, 1.35% glycerol, 50 μg/mL BSA, and 2 mM ATP (final pH 7.5) to a total volume of 20 μL. Reactions were incubated at 30°C for 40 min before being supplemented with 25 mM MgOAc2 and incubated at 37°C for another 2 hours. 1 μL of the final products was taken out for direct PCR readout. The remaining sample was digested with 1 μL of Proteinase K (Thermo Fisher Scientific) at 37°C for 15 min before transformation into Stellar competent cells. Colonies were grown on kanamycin and chloramphenicol plates. Single colonies were randomly picked for plasmid preparation. After extraction, plasmids were analyzed by PCR, restriction enzyme (BamHI) digestion and sanger sequencing.

### **Polymerase Chain Reaction (PCR)**

Forward primer pTarget_F, reverse primer pDonor_R (**Table S1**), and _in vitro_ transposition reaction product were mixed to a final volume of 25 μL for PCR reactions. Cycling conditions were as follows: 1 cycle, 94°C, 3 min; 35 cycles, 98°C, 10 s, 66.9°C, 15 s, 72°C, 8 s; 1 cycle, 72°C, 10 min. Plasmids extracted from single colonies were analyzed by PCR under cycling conditions as follows: 1 cycle, 98°C, 3 min; 35 cycles, 98°C, 10 s, 69.9°C, 15 s, 72°C, 12 s; 1 cycle, 72°C, 10 min.

### **Electron Microscopy**

Aliquots of 4 μL Cas12k–sgRNA binary complex (1 mg/mL) and Cas12k–sgRNA–dsDNA ternary complex (1 mg/mL) were applied to glow-discharged UltrAuFoil holey gold grids (R1.2/1.3, 300 mesh). The grids were blotted for 2 seconds and plunged into liquid ethane using a Vitrobot Mark IV. Cryo-EM data were collected with a Titan Krios microscope operated at 300 kV and images were collected using Leginon (**Suloway et al., 2005**) at a nominal magnification of 81,000x (resulting in a calibrated physical pixel size of 1.05 Å/pixel) with a defocus range of 0.8–2.0 μm. The images were recorded on a K3 electron direct detector in super-resolution mode at the end of a GIF-Quantum energy filter operated with a slit width of 20 eV. A dose rate of 20 electrons per pixel per second and an exposure time of 3.12 seconds were used, generating 40 movie frames with a total dose of ~ 54 electrons per Å2. Statistics for cryo-EM data are listed in **Table 1**.

### **Image Processing**

Movie frames were aligned using MotionCor2 (**Zheng et al., 2017**) with a binning factor of 2. The motion-corrected micrographs were imported into cryoSPARC (**Punjani et al., 2017**). Contrast transfer function (CTF) parameters were estimated using CTFFIND4 (**Rohou and Grigorieff, 2015**). A few thousand particles were auto-picked without template to generate 2D averages for subsequent template-based auto-picking. The auto-picked and extracted particles were processed for 2D classifications, which were used to exclude false and bad particles that fell into 2D averages with poor features. An initial reconstruction was done in cryoSPARC using 100,000 particles (**Punjani et al., 2017**). Heterogenous refinement was further performed to sort out different conformational heterogeneity. To further screen homogenous particles, 3D variance analysis (**Punjani and Fleet, 2021**) was performed and the resulting maps with different conformations (frame_000.mrc and frame_019.mrc) are used for supervised heterogenous refinement. The homogeneous dataset was used for final 3D refinement with C1 symmetry, resulting in 3.65 Å resolution from 183,870 particles.

The Cas12k–sgRNA binary complex dataset were processed in a similar way as the ternary complex. 114,383 particles were selected for a final reconstruction at 3.80 Å resolution. Cryo-EM image processing is summarized in **Table 1**.

### **Model building, refinement, and validation**

_De novo_ model building of the Cas12k–sgRNA–target DNA structure was performed manually in COOT (**Emsley et al., 2010**) guided by secondary structure predictions from PSIPRED (**Jones, 1999**) of Cas12k protein and structure prediction of sgRNA by RNAComposer (**Biesiada et al., 2016**). Refinement of the structure models against corresponding maps were performed using the _phenix.real_space_refine_ tool in Phenix (version 1.19.2) (**Afonine et al., 2018**). For the Cas12k–sgRNA complex, the structure model of the Cas12k–sgRNA–target-DNA complex was fitted into the cryo-EM map with models for target DNA deleted. The model is adjusted by all-atom refinement in COOT with self-restrains. The resultant model was refined against the corresponding cryo-EM map using the _phenix.real_space_refine_ tool in Phenix.

### **Structure-based sequence alignment**

PROMALS3D program (**Pei et al., 2008**) was used to align the sequences of Cas12k and Cas12f based on structure. The alignment diagram was plotted using ESPript (**Robert and Gouet, 2014**). Sequence identities and similarities were calculated using Sequence Manipulation Suite (**Stothard, 2000**). Root-mean-square deviation (RMSD) of the Cα atomic was calculated using the _cealign_ command in PyMOL.

### **Structural visualization**

Figures were generated using PyMOL and UCSF Chimera (**Pettersen et al., 2004**).

SpacePHARER (CRISPR Spacer Phage-Host Pair Finder) is a sensitive and fast tool for _de novo_ prediction of phage-host relationships via identifying phage genomes that match CRISPR spacers in genomic or metagenomic data. SpacePHARER gains sensitivity by comparing spacers and phages at the protein level, optimizing its scores for matching very short sequences, and combining evidence from multiple matches, while controlling for false positives. We demonstrate SpacePHARER by searching a comprehensive spacer list against all complete phage genomes.

SpacePHARER: Sensitive identification of phages from CRISPR spacers in prokaryotic hosts

### **Input**

SpacePHARER accepts spacer sequences as multiple FASTA files each containing spacers from a single prokaryotic genome or as multiple output files from the CRISPR detection tools PILER-CR [**7**], CRT [**5**], MinCED [**13**] or CRISPRDetect [**4**]. Phage genomes are supplied as separate FASTA files or can be downloaded by SpacePHARER from NCBI GenBank [**2**]. Optionally, additional taxonomic labels can be provided for spacers or phages to be included in the final report.

### **Algorithm**

SpacePHARER is divided into five steps (**Figure 1A, Supp. Materials**). (0) Preprocess input: scan the phage genome and CRISPR spacers in six reading frames, extract and translate all putative coding fragments of at least 27 nt, with user-definable translation tables. Each query set _Q_ consists of the translated ORFs _q_ of CRISPR spacers extracted from one prokaryotic genome, and each target set _T_ comprises the putative protein sequences _t_ from a single phage. We refer to similar _q_ and _t_ as _hit_, and an identified host-phage relationship _Q_ − _T_ as _match_. (1) Search all _q_’s against all _t_’s using the fast, sensitive MMseqs2 protein search [**14**], with VTML40 substitution matrix [**10**], gap open cost of 16 and extension cost of 2 (Figure S1). We optimized a short, spaced k-mer pattern for the prefilter stage (10111011) with six informative (‘1’) positions. In addition, align all _q_ − _t_ hits reported in previous search on nucleotide level and prioritize near-perfect nucleotide hits (**Supp. Materials**). (2) For each _q_ − _T_ pair, compute the P-value for the best hit _p_bh from first-order statistics. (3) Compute a combined score _S_comb from best-hit P-values of multiple hits between _Q_ and _T_ using a modified truncated-product method (**Supp. Materials**). (4) Compute the false discovery rate (FDR = FP /(TP + FP)) and only retain matches with FDR _<_ 0.05. For that purpose, SpacePHARER is run on a null model database and the fraction of null matches with _S_comb below a cut-off (empirical P-value) is used to estimate the FDR. (5) Scan 10 nt upstream and downstream of the phage’s protospacer for a possible PAM.

**FIG 1.**

(**A**) SpacePHARER algorithm. A query set _Q_ consists of 6-frame translated ORFs (_q_) from CRISPR spacers, and a target set _T_ consists of 6-frame translated ORFs (_t_) of phage proteins. (1) Search all _q_s against all _t_s using MMseqs2. Align the _q_ − _t_ hits on nucleotide level and prioritize near-perfect nucleotide hits. (2) For each _q_ − _T_ pair, compute the P-value for the best hit from first-order statistics. (3) Compute score _S_comb by combining the best-hit P-values from multiple hits between _Q_ and _T_ using a modified truncated-product method. (4) Estimate the FDR by searching a null database. (5) Scan for possible protospacer adjacent motif (PAM). (**B**) Performance comparison between SpacePHARER (blue) and BLASTN (red) using inverted phage sequences (solid lines) or eukaryotic viral ORFs as null set (dashed lines) demonstrated by expected number of true positive (TP) predictions at different false discovery rates (FDRs). (**C**) Performance comparison between BLASTN (left), SpacePHARER using the weighted lowest common ancestor procedure (LCA, right) at FDR = 0.02, evaluated by the number of correct (blue) and incorrect (red) predictions, for all the host predictions made at each taxonomic rank or below.

### **Output**

is a tab-separated text file. Each host-phage match spans two or more lines. The first starts with ‘#’: prokaryote accession, phage accession, _S_comb, number of hits in the match. Each following line describes an individual hit: spacer accession, phage accession, _p_bh, spacer start and end, phage start and end, possible 5’ PAM|3’ PAM, possible 5’ PAM|3’ PAM on the reverse strand. If requested, the spacer–phage alignments are included.

If taxonomic labels are provided, taxonomic reports based on the weighted lowest common ancestor (LCA) procedure described in [**9**] are created for host LCAs of each phage genome or phage LCAs of each spacer as additional tab-separated text files.

Genome-wide association studies have identified 161 genetic variants associated with coronary artery disease (CAD), but the causal genes and biological pathways remain unknown at most loci. Here, we used CRISPR knockout, inhibition and activation to target 1998 variants at 83 CAD loci to assess their effect on six vascular endothelial cell phenotypes (E-selectin, ICAM1, VCAM1, nitric oxide, reactive oxygen species, calcium signalling). We identified 42 significant variants located within 26 CAD loci. Detailed characterization of the RNA helicase _DHX38_ and CRISPR activation at the _FURIN/FES, CCDC92/ZNF664_ and _CNNM2_ loci revealed a strong effect on vascular endothelial cell senescence.

CRISPR perturbations at many coronary artery disease loci impair vascular endothelial cell functions

### **Design of the sgRNA library**

We retrieved 92 sentinel genetic variants associated with coronary artery disease (CAD) at genome-wide significant levels (P-value ≤5×10−8) from four GWAS meta-analyses available at the time of the design of this experiment**5**,**6**,**18**,**19**. For the design of the sgRNA library, we included all sentinel variants as well as variants in strong LD (_r_2 >0.8 in the 1000 Genomes Project European-ancestry populations). For each variant - sentinel and LD proxy - we identified all possible sgRNA in a 100-bp window centered on the variant itself. We prioritized sgRNA with the highest predicted quality using the CRISPR OffTarget Tool (version 2.0.3)**43** with a Targeting_guide_score ≥ 20 and the “matches with 0 mismatches” = 1 and “matches with 1 mismatch” = 0 settings. We discarded sgRNA that overlapped heterozygous variants, indels and/or multi-allelic variants in the teloHAEC genome (build hg19). We selected sgRNA targeting essential genes from a previously published study**44**. For potential positive control genes (_SELE, SELP, ICAM1, VCAM1, PECAM1, NOS3, VWF, SOD2, SOD3, GPX3, CAT, ITPR1, ITPR2, ITPR3, ATP2A2, ATP2A3, PLN, CAV1,_ and _TRPV4),_ we selected sgRNA from the Human GeCKOv2 CRISPR knockout pooled library**45**. We also selected sgRNA that targeted the promoter (300-bp window before the transcriptional start site) of positive control genes for the CRISPRa (dCas9-VP64) experiments. For all selected loci (variants, coding sequences, gene promoters), we retained the five top scoring sgRNA for the library design. Finally, we added two sgRNA for the _SELE_ locus (SELE_g1, SELE_g2) that we frequently use to validate TNFα stimulation. This resulted in a final library of 8051 sgRNA (**Supplementary Table 2**).

The sgRNA were synthesized in duplicates by Agilent Technologies (Cat-#: G7555B) to accommodate the specific requirements of the Cas9/dCas9-KRAB and dCas9-VP64 (specific MS2 tracrRNA) experiments. We amplified each specific pool of oligonucleotides as previously described**16**, with the following small modifications: we performed two PCR using NebNext High fidelity Master mix (Cat-#: M0541L). The first PCR was used to amplify each pool separately using 2.5ng of pooled oligonucleotides and 500nM of each primer (for the Cas9/dCas9-KRAB library, we used U6_subpool_fwd and Guide_CM_barcode1_rev; for the dCas9-VP64 library, we used U6_subpool_fwd and Guide_MS2_Barcode2_rev). Cycling conditions for PCR1 were 98°C for 30 sec, then 15 cycles of 98°C for 10 sec; 55°C for 10 sec; 72°C for 15 sec and a final step of 72°C for 2 min and a 10°C hold. We performed the second PCR to add homologous sequences, using the U6_screen_fwd and Tracr_rev oligonucleotides for the Cas9/dCas9-KRAB library, and the U6_screen_fwd and Tracr_MS2_rev oligonucleotides for the dCas9-VP64 library, in both cases using ⅕ of PCR1 as template. Cycling conditions for PCR2 were 98°C for 30 sec, then 10 cycles of 98°C for 10 sec; 55C for 10 sec; 72C for 15 sec and a final cycle of 72C for 2 min and 10C hold. See table **Supplementary Table 9** for primer details.

After gel extraction and PCR purification, we performed Gibson assembly in both respective vectors (pHKO9-Neo and lentisgRNA(MS2)-zeo backbone addgene 61427). For pHKO9-Neo, we replaced the Crimson fluorescent gene in the pHKO9-Crimson-CM vector (gift from Dan Bauer’s lab) by a neomycin resistance (NeoR) sequence. Briefly, we amplified the NeoR gene from our pCas9-Neo vector**46** using BsiWI-Neo-Fwd and MluI-Neo_rev primer (**Supplementary Table 9**). After digestion by BsiWI and MluI, we cloned the segment in pHKO9-Crimson_CM, which had been digested with BsiWI and MluI. We amplified each library using ten independent maxi-preparations (Macherey-Nagel cat# 740424). To control the quality of both libraries, we sequenced them on an Illumina HiSeq4000 instrument and calculated the Gini index, which summarizes read distribution across sgRNA in a given pool. For a good-quality sgRNA library, the expected Gini index is ≤0.2, and we obtained Gini indexes of 0.050 and 0.052 for the Cas9/dCas9-KRAB and dCas9-VP64 library, respectively.

### **Engineering of teloHAEC cell lines to stably express Cas9 variants**

TeloHAEC are immortalized human aortic endothelial cells obtained by over-expressing telomerase (ATCC CRL-4052). These cells have a normal female karyotype (46;XX) and exhibit many of the properties and functions of human vascular endothelial cells**9**. We generated our teloHAEC cells models expressing either Cas9, dCas9-KRAB or dCas9-VP64 + MPHv2 using Addgene vectors #52962, #46911, #61425 and #89308, and lentiviral infection as previously described**46**.

### **Pooled CRISPR screen experiments**

We produced four batches of lentiviruses for each sgRNA library pool (Cas9/dCas9-KRAB, dCas9-VP64). We infected each teloHAEC cell line (Cas9, dCas9-KRAB, dCas9-VP64) at a multiplicity of infection of 0.3 using each batch of viruses separately. Following viral infection, we selected cells using zeocin (teloHAEC-dCas9-VP64) or G418 (teloHAEC-Cas9/-dCas9-KRAB) for five (teloHAEC-dCas9-VP64) or seven days (teloHAEC-Cas9/-dCas9-KRAB) in vascular cell basal medium (ATCC PCS-100-030) to remove any cells that did not incorporate a vector. After selection, we stimulated cells expressing Cas9 or dCas9-KRAB using TNFα (10ng/μl) for four hours to induce a pro-inflammatory response; we did not stimulate cells expressing dCas9-VP64, reasoning that the VP64 transcriptional domain should activate gene expression. Following TNFα stimulation, we immunostained cells (around 50M cells) with antibodies linked to phycoerythrin for adhesion molecules (E-selectin (BD BIOSCIENCES Cat-#: 551145), VCAM-1 (Cat-#: 12-1069-42), ICAM-1 (Cat-#: 12-0549-42)) or we incubated with fluorescent dye-based reagents for endothelial signaling markers: (nitric oxide (NO) (DAF-FM Diacetate, Cat-#: D23844)), reactive oxygen species (ROS) (CM-H2DCFDA, Cat-#: C6827), calcium signaling (Fura Red, Cat-#: F3021)). We calibrated the FACS assays with positive control treatments to make sure that we could robustly detect changes in the measured phenotypes. Antibodies and fluorescent dye-based reagents were titrated to use optimal concentrations. We also quantified how teloHAEC were responding to ionomycin for calcium signaling, to sodium nitroprusside for NO and to TNFα for reactive oxygen species. For adhesion molecules, we utilized sgRNA targeting coding exons and promoter regions of _SELE, ICAM1_ and _VCAM1_ as positive controls. Unless otherwise stated, we purchased all antibodies and dyes from ThermoFisher Scientific. Subsequently, we sorted stained cells by flow cytometry on a BD FACSARIA FUSION flow cytometer to collect the top and bottom 10% of fluorescently labeled cells. FACS traces were generated with FlowJo (BD Biosciences). We extracted genomic DNA from both top and bottom 10% cell fractions separately (around 5M cells in each fraction) using the QIAGEN DNeasy Blood and Tissue kit (Cat No. 69504) according to manufacturer’s instructions.

### **Amplification and sequencing of pooled CRISPR experiments**

We amplified sgRNA sequences from genomic DNA via PCR, followed by a cleanup step using the QIAGEN QIAquick PCR purification kit (Cat-#: 28104) according to the manufacturer’s instructions. We used the primer sequences and PCR settings as previously described in ref. **16**. We created sequencing libraries using Illumina TruSeq adapters according to the manufacturer’s protocols. We sequenced the libraries on an Illumina Hiseq4000 instrument at the McGill University and Genome Quebec Innovation Centre (MUGQIC). Generally 6 samples were multiplexed per sequencing lane for a target read coverage of ~500 reads per sgRNA per sample (**Supplementary Figure 4B**).

### **Computational analysis of pooled CRISPR screen data**

We processed raw sequencing data from the BCL to the FASTQ format using bcl2fastq at MUGQIC. Raw FASTQ reads were quality-controlled using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and MultiQC**47**. We performed downstream analysis of sgRNA sequencing data using MAGECK (v.0.5.9)**48**. We quantified sgRNA sequences using MAGECK count against the list of sgRNA sequences in the library (**Supplementary Table 3**), allowing for no mismatches in the sgRNA sequence. We then tested the difference in sgRNA counts between cell fractions or conditions using MAGECK maximum likelihood estimation (_mle_) method with median normalization**49**.

### **UMAP representation of sample-level count data**

We normalized raw sgRNA counts using variance stabilizing transformation (vst) in DESeq2**50**. To account for baseline differences between plasmid preparations, we further normalized samples to their respective vector library by dividing the vst normalized sgRNA count by the vst normalized count of the Cas9/dCas9-KRAB or dCas9/VP64 library, respectively. We calculated principal components using the top 10% most variable sgRNA (805 sgRNA) across all cell sorted samples based on normalized counts. Next, we used the loadings from the first three principal components in Uniform Manifold Approximation and Projection (UMAP)**51** to create a two-dimensional embedding of the normalized sgRNA count data. Each dot in the UMAP plot represents one sequenced sample (top or bottom 10% of stained cells).

### **Analysis of sgRNAs targeting essential genes**

To test for potential effects of sgRNA on endothelial cell death and proliferation, we compared sgRNA counts of all samples across the same cellular model (Cas9, dCas9-KRAB, dCas9-VP64) against the respective baseline vector library sgRNA count using MAGECK _mle._ We used sgRNAs targeting essential genes in the teloHAEC Cas9 cellular model as positive controls.

### **Single sgRNA validation**

We individually cloned each sgRNA for validation as previously described**52**. We produced lentiviruses, infected cells, performed antibiotic selection, and stained cells as for the pooled CRISPR screen. We analyzed cells using flow cytometry (BD FACSCelesta (BD Biosciences, San Jose, CA, USA) equipped with a 20 mW blue laser (488 nm), a 40 mW red laser (640 nm), and a 50 mW violet laser (405 nm). For each experiment, we measured the mean fluorescent intensity (MFI) obtained for sgRNA of interest and compared it with the MFI for control sgRNA (safe-harbor and/or scrambled sgRNA). We performed each experiment at least three times. For statistical analyses, we used Student’s _t_-test and determined that a sgRNA had a significant effect on the measured phenotype when a one-tailed P-value ≤0.05.

For _DHX38_ Crimson experiments, we individually cloned each sgRNA in pHKO9-Crimson-CM vector (gift from Dan’s Bauer lab). We produced lentiviruses, infected cells and performed flow cytometry on a BD FACSARIA FUSION flow cytometer at day 2, day 4 and day 7 post-infection. We analyzed the percentage of Crimson positive cells and we sorted Crimson positive and negative cells to extract RNA in each fraction. We extracted total RNA using RNeasy Plus Mini Kit (Qiagen cat #: 74136). We measured RNA integrity and concentration using Agilent RNA 6000 Nano II assays (Agilent Technologies) on an Agilent 2100 Bioanalyzer and Take3 on Cytation V (Biotek). We reverse transcribed 750ng of total RNA using random primers and 1 U of the MultiScribe Reverse Transcriptase (Applied Biosystems) in a 20 μL reaction volume at 100 mM dNTPS and 20 U of RNase inhibitor with these three steps: 10 min at 25 °C, 120 min at 37 °C and 5 min at 85 °C. We followed the MIQE guidelines to assess quality and reproducibility of our qPCR results **53**. We performed qPCR in triplicates for all samples using: 1.25 μL of cDNA (1/50 dilution), 5 μL of Platinum SYBR Green qPCR SuperMix-UDG (Life Technologies) and 3.75 μL of primer pair mix at 0.8 μM on a CFX384 from Biorad. We used the following thermal profile: 10 min at 95 °C, and 40 cycles of 30 s at 95 °C, 30 s at 55 °C and 45 s at 72 °C. We carried out melting curve analyses after the amplification process to ensure the specificity of the amplified products. We also simultaneously performed qPCR reactions with no template controls for each gene to test the absence of non-specific products. Cq values were determined with the CFX Manager 3.1 (Bio-Rad) software and expression levels were normalized on the expression levels of the house-keeping genes TATA-box binding protein (_TBP_), hypoxanthine-guanine phosphoribosyltransferase (_HPRT_), and glyceraldehyde 3-phosphate dehydrogenase (_GAPDH)_ using the ΔΔCt method. The primer sequences are in **Supplementary Table 8**.

### **PCR for determination of CRISPR-Cas9-induced indels**

We isolated gDNA using QuickExtract DNA Extraction Solution (Epicentre, QE0905) from 1×105 cells. We used 100 or 200 ng of gDNA as a template for PCR reaction with the corresponding primers (see **Supplementary Table 9**). gDNA from parental teloHAEC cells was used as control. Obtained PCR products were analysed by electrophoresis on a 1% agarose gel prior to Sanger sequencing. We used TIDE (tracking of indels deconvolution) software for analysis**54**.

### **Assays for assessment of cell senescence**

Using the same experimental design (DHX38-Crimson), we performed beta-galactosidase staining using the CellEvent Senescence Green Flow Cytometry Assay Kit from Invitrogen at day 2, day 4 and day 7 following the manufacturer’s protocol. Briefly, we trypsinized and we fixed the cells with 2% paraformaldehyde solution during 10 minutes at room temperature, washed them in 1%BSA/PBS and incubated for 1h30 in 1/500 working solution. After incubation, we washed the cells with 1%BSA/PBS and analyzed them by flow cytometry. We measured β-galactosidase fluorescence signal in positive and negative Crimson cells independently. As positive control, non-infected cells were treated with 20 μM of Etoposide (Sigma, E1383-25) for 2, 4 and 7 days.

### **Transcriptome data analysis**

For RNA-seq analysis, we extracted RNA using RNeasy plus mini kit from Qiagen (cat #: 74136). RNA-seq experiments were carried out by the Centre d’Expertise et de Services Genome Quebec using rRNA-depleted TruSeq stranded (HMR) libraries (Illumina) on an Illumina Hiseq 4000 instrument (paired-ends, 100-bp reads) and by The Center for applied Genomics (Toronto) using rRNA-deletion library prep on an Illumina NovaSeq-SP flow cell. We quality-controlled raw fastq files with FastQC and multiQC**47**. We used kallisto (v. 0.46.0) to quantify transcript abundances**55** against ENSEMBL reference transcripts (release 94) followed by tximport to calculate gene-level counts**56**. We utilized regularized log-transformation (rlog) in DESeq2**50** as input for principal component analysis (PCA). DESeq2**50** was further used to identify differentially-expressed genes between teloHAEC cell models (Cas9, dCas9-VP64) infected by lentiviruses with safe-harbor sgRNA or sgRNA identified in the pooled CRISPR screens. We excluded genes expressed with less than 10 reads across all samples from the analysis. We performed shrinkage for effect size estimates using apeglm using the lfcShrink method**57**. Genes differentially expressed with a Benjamini-Hochberg adjusted p-value ≤ 0.05 were considered significant. Gene set enrichment analysis was performed using the R package fgsea using 100,000 permutations against the Hallmark gene sets from msigdbr (https://igordot.github.io/msigdbr/)**58**,**59**. We quantified short indels in the RNA-seq data of _DHX38_ (sgRNA_10966) and _MAT2A_ (sgRNA_02249) using the tools transIndel and Genesis-Indel, which are specifically designed to identify indels in the unmapped read fraction of samples**60**,**61**.

### **Analysis of scRNA-seq data from human coronary arteries**

Single-cell gene expression matrix from human right atherosclerotic coronary arteries (three male and one female donors), was downloaded from NCBI GEO (GSE131780, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE131780). The data was re-analyzed using the Seurat package in R with a standard single-cell clustering pipeline. Gene expression data was normalized using the SCTransform function from Seurat (v.3.2.3), regressing out the percentage of mitochondrial gene expression. Principal components analysis was performed, followed by dimensional reduction with Uniform Manifold Approximation and Projection (UMAP) using the first 20 principal components as input. Gene expression was visualized on the first two UMAP dimensions using the kernel density function (plot_density) from the Nebulosa package (v.0.99.92)**62** for endothelial cell marker and candidate genes.

### **Statistics and data analysis**

Unless noted otherwise, we performed all data and statistical analyses in R (v.3.6.0) using Rstudio. We ran our analyses on a high performance computing cluster (Beluga) from Calcul Quebec/Compute Canada. For MAGECK variant-level analyses, permutation-based FDR of ≤10% were considered significant. For RNA-seq analysis, genes with a Benjamini-Hochberg adjusted P-value in DESeq2 ≤0.05 were considered significant**50**.

Online methods

Rapid assessment of CRISPR/Cas (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated Cas protein)-based genome editing (GE) tools and their components are critical aspects for successful applications in different organisms. In many bacteria, double-stranded breaks (DSBs) generated by CRISPR/Cas tool generally cause cell death due to the lack of an efficient non-homologous end-joining pathway and restricts its use. CRISPR-based DSB-free base editors (BEs) have been applied for precise nucleotide editing in bacteria, which does not need to make DSBs. However, optimization of newer BE tools in bacteria is challenging owing to the toxic effects of BE reagents expressed using strong promoters. Improved variants of two main BEs capable of converting C-to-T (CBE) and A-to-G (ABE) have been recently developed but yet to be tested in bacteria. Here, we report a platform for _in vivo_ rapid investigation of CRISPR-BE components in _Escherichia coli_ (IRI-CCE) comprising different combinations of promoters/terminators. We demonstrate the use of IRI-CEE to characterize different variants of CBEs (PmCDA1, evoCDA1, APOBEC3A) and ABEs (ABE8e, ABE9e), exhibiting that each independent BE has its specific editing pattern for a given target site and promoter type. Additionally, the IRI-CCE platform offers a rapid way to screen functional gRNAs. In summary, CRISPR-BE components expressed by promoters of different strengths in the IRI-CCE allow an analysis of various BE tools, including cloned biopart modules and gRNAs.

IRI-CCE: A Platform for Evaluating CRISPR-based Base Editing Tools and Its Components

### **2.1** **_E. coli_** **strains**

The four _E. coli_ strains were used: 10-beta, DH5α, DB3.1, and BL21(DE3). The information about strain genotype provided in **Supplementary Table 1**. The primers and plasmids used in this study are listed in **Supplementary Tables 2 and 3**, respectively. The _E. coli_ strains were cultured in Luria-Bertani (LB) broth with appropriate antibiotics.

### **2.2 Plasmid construction and cloning**

The required bioparts were amplified by conventional polymerase chain reaction (PCR) using a Phusion high-fidelity DNA polymerase (Thermo Fisher Scientific, Waltham, MA). The cloning of different plasmid vectors was designed and performed by following the principle of MoClo (**Weber et al., 2011**) and Golden Gate assembly (**Engler et al., 2014**) protocols using BpiI/BsaI Type IIS enzyme digestion-ligation. The DNA recognition sites of restriction enzymes BsaI and BpiI were removed from internal sequences of bioparts during PCR amplification to make them suitable for Golden Gate cloning (a procedure also termed domestication). The DNA oligonucleotide pairs for the pEc1 (SJM901+RBSTL2) promoter and TerL3S2P21 terminator were annealed and ligated into BpiI-digested acceptor plasmids. Various promoters including pGlpT, p35S(Long), p35S(Short), pRbc, and pRPS5a were used from the pYTK001 (Addgene #65108) (**Lee et al., 2015**), pICH51266 (Addgene #50267) (**Engler et al., 2014**), pICH51277 (Addgene #50268) (**Engler et al., 2014**), pICH45195 (Addgene #50275) (**Engler et al., 2014**), and pKI1.1R (Addgene #85808) (**Tsutsui and Higashiyama, 2017**), respectively. The promoter DNA sequences and their evaluation provided in **Supplementary Figures 1, 2, 3 and 4**. Similarly, DNA sequences of Terminators, Ter35S, and TerHSP were adopted from the pICH41414 (Addgene #50337) (**Engler et al., 2014**) and pKI1.1R (Addgene #85808), respectively.

Three cytidine deaminases: PmCDA1 (Target-AID) (**Nishida et al., 2016**), evoCDA1 (**Thuronyi et al., 2019**), and APOBEC3A (**Zong et al., 2018**), were PCR amplified from PmCDA1-1x uracil-DNA glycosylase inhibitor protein (UGI) (Addgene #79620), evoCDA1 pBT277 (Addgene #122608), and A3A-PBE-ΔUGI (Addgene #119770), respectively. For all three deaminases, the same linker regions were used as reported in the previous studies. The ABE8e was synthesized from Bioneer Co. (Daejon, Korea), and the ABE9e variant (containing additional V82S/Q154R mutations) was cloned using ABE9e module as a template by site-directed mutagenesis PCR. PmCDA1-1xUGI was fused to the C terminus of nCas9. The evoCDA1, A3A, and ABE8e were fused to the N terminus of nCas9 with the XTEN linker. The 2xUGI module (template: Addgene #122608) was fused to the C terminus of evoCDA1-nCas9 and A3A-nCas9. The sgRNA expression was driven by bacterial pJ23119 (synthetically cloned) or plant AtU6 promoter (pICSL01009, Addgene #46968). The nCas9(D10A) was generated by the PCR method using previously optimized Cas9 as a template (Level 1 hCas9 module, Addgene #49771). Desired sgRNA sequence was PCR amplified using plasmid pICH86966::AtU6p::sgRNA_PDS (Addgene #46966) as a template and cloned together with either pJ23119 or pAtU6 for sgRNA expression.

For cloning of target region of the desired sgRNA sequence, a pair of oligonucleotide DNAs that contains the target sequence with PAM was annealed and ligated into BsmBI-digested universal target-acceptor plasmid L1 or L2 having optimized superfolder green fluorescent protein (sfGFP) (**Lee et al., 2015**) at downstream side (**Supplementary Figure 5 and 6**).

### **2.3 Bacterial transformation, plasmid isolation, and Sanger sequencing**

Ligation products of all the steps during the cloning were transformed into competent cells of _E. coli_ 10-beta strain by a heat-shock method. The bacterial culture was spread on the LB media containing desired antibiotic and incubated at 37°C for 18-24 h. The cloned plasmids and BE activities were confirmed by Sanger sequencing at Solgent Ltd. (Daejeon, Korea) or Cosmogentech Ltd. (Seoul, Korea). For mutagenesis assay, the different _E. coli_ strains were transformed with the appropriate plasmids using the heat-shock method and were pre-cultured for 1 h with 1 ml of LB medium. After incubation for 1 h at 37°C, the fraction of cell cultures were spread on LB agar (1.5%) plates containing selection antibiotics with needed concentrations. The next day, individual colonies from the plate were inoculated in 3 ml LB broth with appropriate antibiotics and cultured at 37°C. The culture time was varied according to the experimental parameters and mentioned in appropriate sections. The plasmid isolation was done using Plasmid Mini-Prep Kit from BioFact Co. Ltd. (Daejeon, Korea) for Sanger sequencing analysis.

### **2.4 Promoter activity analysis**

As shown in **Figure 1B**, all the tested promoters were cloned with sfGFP sequence at the downstream side, followed by a termination signal. Bacterial cultures were grown from the transformed single colonies in LB broth, shaking at 37°C for 24 h. The grown cultures were incubated for 24 h at 4°C, and then OD600 values were normalized to 1. The cells were harvested at 10000 rpm for 1 min and washed with 1x PBS (Phosphate buffered saline). Images were captured in blue light with a FluoroBox from CELLGENTEK Co., Ltd. (Deajeon, South Korea). The fluorescence intensity level was quantified using the ImageJ software (**Schneider et al., 2012**), and corrected total cell fluorescence (CTCF) was calculated using the following formula: CTCF = Integrated Density - (Area of selected cell x Mean fluorescence of background readings).

For qRT-PCR, RNA was extracted using the following method. _E. coli_-carrying plasmids (**Figure 1B**) were grown in LB medium. Cells were harvested during the exponential growth phase (12 h with OD600 value 0.5). Total RNA was extracted using RNeasy Protect Bacteria Mini Kit from Qiagen. For all samples, 600 ng of total RNA was used for complementary DNA (cDNA) synthesis using a QuantiTect Reverse Transcription Kit from Qiagen following the manufacturer’s instructions. To estimate the relative _sfGFP_ transcript, the quantitative real-time PCR (qRT-PCR) reactions were carried out using the KAPA SYBR FAST qPCR kit from Kapa Biosystems (MA, USA) with _sfGFP_-specific primer sets (**Supplementary Table 4**). Cycling of PCR consisted of pre-denaturation at 95°C for 5 min followed by 40 cycles of a denaturation step at 95°C for 10 min, an annealing step at 60°C for 15 s, and final extension step at 72 °C for 20 s using the CFX384 Real-Time System from Bio-Rad, Hercules (CA, USA). The qRT-PCR reactions were performed with independent biological replicates. Relative _sfGFP_ transcript values normalized against internal control 16S ribosomal RNA (_rrsA_) gene. Data analyses were performed by the 2−ΔΔCt method (**Livak and Schmittgen, 2001**).

### **2.5 Evaluation of editing activities**

The single colonies were cultured, and the plasmid vectors containing synthetic targets were purified using Plasmid Mini-Prep Kit from BioFact Co. Ltd. (Daejeon, Korea) for further sequencing analysis. For sequencing analysis of genomic loci, the target fragments were PCR amplified using target-specific primers from the randomly picked colonies and then analyzed by Sanger sequencing. Sanger sequencing data analysis was performed using SnapGene software (GSL Biotech; available at snapgene.com). The editing efficiency was determined either by the ratio of non-edited to edited colonies from the randomly picked cells or by the nucleotide conversion rate using the online tool EditR (**Kluesner et al., 2018**) from C-to-T and A-to-G for CBE and ABE, respectively. The data were statistically analyzed and plotted in GraphPad Prism 9.0.0 (www.graphpad.com).

Deep mutational scanning is a powerful method to explore the mutational fitness landscape of proteins. Its adaptation to anti-CRISPR proteins, which are natural CRISPR-Cas inhibitors and key players in the co-evolution of microbes and phages, would facilitate their in-depth characterization and optimization. Here, we developed a robust anti-CRISPR deep mutational scanning pipeline in _Escherichia coli_ combining synthetic gene circuits based on CRISPR interference with flow cytometry-coupled sequencing and mathematical modeling. Using this pipeline, we created and characterized comprehensive single point mutation libraries for AcrIIA4 and AcrIIA5, two potent inhibitors of _Streptococcus pyogenes_ Cas9. The resulting mutational fitness landscapes revealed that both Acrs possess a considerable mutational tolerance as well as an intrinsic redundancy with respect to Cas9 inhibitory features, suggesting evolutionary pressure towards high plasticity and robustness. Finally, to demonstrate that our pipeline can inform the optimization and fine-tuning of Acrs for genome editing applications, we cross-validated a subset of AcrIIA4 mutants via gene editing assays in mammalian cells and _in vitro_ affinity measurements. Together, our work establishes deep mutational scanning as powerful method for anti-CRISPR protein characterization and optimization.

A deep mutational scanning platform to characterize the fitness landscape of anti-CRISPR proteins

### **Molecular cloning**

All plasmids used in this study are listed in Supplementary Table 1, sgRNA target sequences are shown in Supplementary Table 2. Sequences and maps for the CRISRPi selection vectors are available as Supplementary data 1 (Genbank files). Oligonucleotides and synthetic double-stranded DNA fragments were obtained from Integrated DNA Technologies. PCRs were performed with Q5 Hot Start High Fidelity Polymerase (New England Biolabs) or Invitrogen Platinum Superfi II DNA Polymerase (Thermo Fisher Scientific). Expression vectors were created using classical restriction enzyme cloning or Golden-Gate assembly (**40**). Restriction enzymes were obtained from New England Biolabs, T4 DNA ligase was obtained from Jena Biosciences. Agarose gel electrophoresis was used to analyze PCR and restriction products. Bands of the expected size were cut out and DNA was extracted with a ZymoClean Gel DNA recovery kit (Zymo Research). For all cloning steps, chemically competent _E. coli_ K12 DH5α cells (New England Biolabs) were used. Antibiotics were used at the following concentrations: carbenicillin, 50 µg mL−1; chloramphenicol, 25 µg mL−1; kanamycin, 50 µg mL−1. Plasmid DNA was purified with the ZR Plasmid Miniprep or ZymoPure II Midiprep kit (both Zymo Research). The integrity of all plasmids was verified by Sanger Sequencing (GENEWIZ Europe).

AcrIIA4 (_Listeria monocytogenes)_ (**19**) and AcrIIA5 (Phage D4276) (**14**) coding sequences were codon optimized for expression in _E. coli_, obtained as double stranded gene fragments and cloned into pBAD24 (pBAD24-sfGFPx1 was a gift from Sankar Adhya & Francisco Malagon, Addgene plasmid #51558) (**41**) via unique EcoRI and HindIII restriction sites, resulting in the plasmids pBAD24-AcrIIA4 and pBAD24-AcrIIA5.

To construct plasmids co-expressing d_Spy_Cas9 and sgRNA scaffolds, a second multiple cloning site was introduced into pBbA5 (pBbA5c-RFP was a gift from Jay Keasling, Addgene plasmid #35281) (**42**) by PCR, resulting in pBbA5C_sgMCS. A fragment encoding an _E. coli_ codon optimized d_Spy_Cas9 was PCR amplified from vector pdCas9 (gift from Luciano Marraffini, Addgene plasmid #46569) (**43**) and cloned into pBbA5C_sgMCS using EcoRI and BamHI restriction sites, thereby generating pBbA5C_sgMCS_d_Spy_Cas9. A sgRNA expression cassette with an RFP reporter-targeting spacer sequence was ordered as overlapping DNA oligonucleotides, extended by PCR and cloned into pBbA5c-spCas9 using NotI and SalI restriction sites, resulting in pBbA5C_sgMCS_d_Spy_Cas9_RFP_guide. To generate the RFP reporter plasmid, a bacterial codon optimized RFP encoding sequence was PCR amplified from pBbA5c-RFP; the Anderson Promoter J23102 (http://parts.igem.org/Promoters/Catalog/Anderson) was included in the 5’-extension of the forward primer. A sequence encoding a SsrA degradation tag (**44**) (AANDENYADAS; corresponds to part BBa_M0052 in iGEM parts registry - parts.igem.org) was included as 5’- extension into the reverse primer to reduce the half-live of RFP in bacterial cells. The resulting PCR fragments were cloned into pJUMP27 (pJUMP27-1AsfGFP was a gift from Chris French, Addgene plasmid #126974) (**45**) via XbaI and PstI restriction sites, thereby resulting in pJUMP27_J23102_RFP_M0052.

The single codon mutational libraries of AcrIIA4 and AcrIIA5 were generated by back-to-back PCR on pBAD24-AcrIIA4 and -AcrIIA5 with forward primers containing NNB overhangs as previously described (**46**). PCRs were performed for each position individually to avoid PCR bias, followed by gel extraction. The purified fragments were treated with KLD enzyme mix (New England Biolabs) and transformed into _E. coli_ DH5a chemically competent cells. Cells were grown in LB supplemented with carbenicillin. Note that sub-libraries corresponding to a single codon were grown individually and to stationary phase in a 96 deep-well plate. Cultures were then combined at an equal volume and further grown in 50 ml LB carbenicillin until saturation, followed by extraction of plasmid DNA.

Single mutant Acr variants for bacterial validation experiments were created via back-to-back PCR on template vectors pBAD24-AcrIIA4 and -5 and by incorporating the mutations into the 5’-extension of one primer, followed by KLD treatment.

The vectors co-encoding firefly luciferase, a firefly luciferase gene targeting sgRNA and _Renilla_ luciferase (for normalization purposes) used for the dual luciferase assay experiments in mammalian cells were previously reported by us (**32**). Vector pCMV-AcrIIA4 was previously reported by us (**32**). Single codon substitutions were introduced into the Acr expression constructs via back-to-back PCR by incorporating the mutations into the 5’- extension of one primer, followed by KLD treatment.

### **Transformation of** **_E. coli_** **with Acr libraries**

For library generation, chemically competent bacterial cells were first co-transformed with the d_Spy_Cas9 plasmid and the RFP reporter plasmid and selected on LB containing kanamycin and chloramphenicol. An overnight culture of a single bacterial colony was inoculated in Super Optimal Broth (SOB) medium supplemented with kanamycin and chloramphenicol, grown to an optical density (OD)600 of 0.5 and chemical competent cells were prepared according to the Inoue transformation protocol (**47**). AcrIIA4 and AcrIIA5 libraries were transformed into chemical competent cells by heat-shock at 42°C for 1 min. 0.1% of the total transformation volume was plated on LB Agar plates containing carbenicillin, chloramphenicol and kanamycin to estimate the transformation efficiency and corresponding library complexity (Supplementary Tables 3 and 4). The remaining cells were grown in liquid LB medium containing carbenicillin, chloramphenicol and kanamycin until stationary phase, followed by cryopreservation in aliquots.

### **Fluorescence-activated cell sorting**

Cryopreserved cells were thawed and grown in LB medium containing carbenicillin, chloramphenicol, kanamycin until stationary phase, followed by induction with 1 mM IPTG and 4 mM arabinose at a starting OD600 of 0.05. Following a 12-hour incubation period, cells were collected by centrifugation at 5000×g for 10 minutes, the supernatant was removed, and cells were resuspended in 10 volumes 1×PBS buffer.

Fluorescence-activated cell sorting (FACS) experiments were performed on a BD FACSAria™ _Fusion_ flow cytometer at the ZMBH flow cytometry core facility (Heidelberg University). _E. coli c_ells were first gated using the forward-scatter area (FSC-A) and site-scatter area (SSC-A). Bacterial cells were then sorted into eight fractions according to their RFP intensity using 150,000 total cells per fraction. Collected cells were either frozen for DNA extraction followed by deep amplicon sequencing or recovered by growing the cells in LB medium containing the necessary antibiotics for subsequent characterization of fluorescence enrichments. In the latter case, cells were washed in PBS twice and grown in LB medium containing carbenicillin, chloramphenicol, kanamycin overnight. The cultures form the individual fractions were then induced with 1 mM IPTG and 4 mM arabinose at a starting OD600 of 0.05 and analyzed on a BD FACSAria™ _Fusion_ flow cytometer 12 hours later as described above.

### **Amplicon Deep Sequencing**

FACS-sorted _E. coli_ cell fractions were lysed and DNA extracted for subsequent 1st stage PCR amplification of AcrIIA4 or AcrIIA5 genes with primers containing Nextera XT index overhangs. PCR amplicons were purified by gel extraction and the concentrations of each sample was adjusted to 25 ng/µl for the 2nd stage barcoding PCR using TG Nextera XT Index Kit v2 Set A (Illumina) according to the manufacturer’s protocol. Libraries were sequenced at the EMBL Heidelberg GeneCore facility on the Illumina MiSeq system using 2×250 paired-end sequencing reagents (MiSeq Reagent Kit v2, 500-cycles).

### **NGS data analysis**

Paired-end reads for each sorted fraction were assembled by their overlap and filtered to remove reads corresponding to sequences not contained in the original library. Only reads corresponding the wild-type Acr and reads containing mutations in _exactly one_ codon relative to wild-type were kept for further analysis. Read counts of the remaining reads were augmented with pseudocounts and normalized within each sorting fraction. Then, for each single mutant, read counts were normalized across fractions, resulting in a distribution of read counts over fractions for each mutant.

### **Acr mutant activity regression**

As a proxy for Acr activity, linear regression of the binned distribution and mean of log fluorescence intensities was performed for a set of single Acr mutants (benchmarks). To this end, log fluorescence intensity data from flow cytometry for the benchmark mutant set were binned (bin width 0.5, range from 0 to 12) and bins normalized to sum to one. An affine regression model relating the read distribution across sorting fractions to the distribution of log fluorescence intensity was zero initialized and fitted using gradient descent to minimize squared error under L2 regularization. Model fits were evaluated using mean square error calculated via leave-one-out cross-validation on the set of benchmark mutants.

Log fluorescence intensity distributions were predicted for all single Acr mutants. As a measure of confidence for the predicted distribution, the relative entropy of the underlying NGS read distribution was computed relative to a uniform distribution. Large relative entropy indicates a distribution with a majority of probability mass focused on one sequenced fraction (high confidence), whereas a low relative entropy indicates a distribution close to the uniform distribution (low confidence).

### **Mammalian cell culture**

HEK293T (human embryonic kidney) cells were cultured at 5% CO2 and 37°C in a humidified incubator and passaged every 2-3 days. Cells were maintained in DMEM (Thermo Fisher Scientific) supplemented with 10% (v/v) fetal calf serum (Thermo Fisher Scientific), 100 U mL-1 penicillin, 100 µg mL-1 streptomycin (Thermo Fisher Scientific). Prior to the assays, cells were checked for mycoplasma contamination by qPCR (Mycoplasma Check, GATC Eurofins).

12,500 cells were seeded in a 96 well culture plate and transfected the next day with Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s protocol. Cells were co-transfected with 33 ng of (i) a plasmid co-expressing _Renilla_ and firefly luciferase as well as a sgRNA targeting the firefly reporter gene, (ii) 33 ng of a plasmid encoding _Spy_Cas9 with C- and N-terminal nuclear localization signals (NLS) and an N-terminal 3xFlag tag, (iii) and either 11 ng, 33 ng or 99 ng of a vector encoding human codon-optimized Acr mutant variants. A stuffer plasmid (pUC19) was used to top up plasmid levels per sample to 165 ng, thereby keeping the total amount of DNA constant across all samples. 72 hours post-transfection, cells were washed with PBS and firefly and _Renilla_ luciferase activities were measured using the Dual-Glo luciferase assay system (Promega) according to manufacturer’s instructions. First, one volume Dual-Glo reagent was added to each well. Following a ten-minute incubation time, lysates were transferred to a white 96-well plate and firefly luciferase photon counts were measured using a FLUOstar Omega multimode reader (BMG Labtech). Subsequently, one volume Dual-Glo Stop & Glo was added to quench the firefly signal and activate _Renilla_ luciferase, samples were incubated for ten minutes, and _Renilla_ luciferase photon counts were measured. To calculate the reported luciferase activity values, firefly luciferase photon counts were normalized to _Renilla_ luciferase photon counts in each sample.

### **Protein production for Acr single point mutants**

Bacterial codon-optimized sequences of AcrIIA4 mutants were PCR amplified and cloned into pET-28b(+) using BsaI sites. Constructs were expressed with an N-terminal His-tag. Plasmids were transformed into BL21(DE3) competent cells (Thermo Fisher Scientific). Protein expression was conducted in LB medium supplemented with kanamycin, and cells were induced with 0.5 mM IPTG at an OD600 of 0.6, followed by incubation at 18°C for 16 hours. Next, cells were harvested by centrifugation, resuspended in lysis buffer (50 mM Tris-HCL pH 7.5, 200 mM NaCl, 1 mM DTT, 1 mM PMSF, and 0.3 mg ml-1 lysozyme) and sonicated. Subsequently, samples were centrifuged, and the cleared lysates were incubated with 1 mL HisPur Ni-NTA Resin (Thermo Fisher Scientific), washed and eluted. Eluates were desalted using 7K MWCO Zeba Spin Desalting columns, stored in phosphate-buffered saline solution (PBS). The flow-through was collected and concentrated using 3K MWCO Pierce Protein Concentrators PES (Thermo Fisher Scientific), followed by flash freezing.

### **Affinity measurements**

Biolayer interferometry binding data were collected with a Gator Bioanalysis System and processed using the instrument’s integrated software. For the binding assay, His tagged designs were loaded onto anti-NTA biosensors (Gator probes) at 5 µg mL-1 in binding buffer (PBS pH 7.4) for 120 s. Homemade CRISPR-Cas9 RNP was diluted from 200 nM to 6.25 nM in binding buffer. After baseline measurement in the binding buffer alone, the binding kinetics were monitored by dipping the biosensors in wells containing the Cas9 RNP at the indicated concentration (association step) and then dipping the sensors back into baseline/buffer (dissociation).

Site-specific genetic and epigenetic targeting of distinct cell populations is a central goal in molecular neuroscience and is crucial to understand the gene regulatory mechanisms that underlie complex phenotypes and behaviors. While recent technological advances have enabled unprecedented control over gene expression, many of these approaches are focused on selected model organisms and/or require labor-intensive customizations for different applications. The simplicity and modularity of CRISPR-based systems have transformed this aspect of genome editing, providing a variety of possible applications and targets. However, there are currently few available tools for cell-selective CRISPR regulation in neurons. Here, we designed, validated, and optimized CRISPR activation (CRISPRa) and CRISPR interference (CRISPRi) systems for Cre recombinase-dependent gene regulation. Unexpectedly, CRISPRa systems based on a traditional double-floxed inverted open reading frame (DIO) strategy exhibited leaky target gene induction in the absence of Cre. Therefore, we developed an intron-containing Cre-dependent CRISPRa system (SVI-DIO-dCas9-VPR) that alleviated leaky gene induction and outperformed the traditional DIO system at endogenous genes in both HEK293T cells and rat primary neuron cultures. Using gene-specific CRISPR sgRNAs, we demonstrate that SVI-DIO-dCas9-VPR can activate highly inducible genes (_GRM2, Tent5b_, and _Fos_) as well as moderately inducible genes (_Sstr2_ and _Gadd45b_) in a Cre-specific manner. Furthermore, to illustrate the versatility of this tool, we created a parallel CRISPRi construct that successfully inhibited expression from of a luciferase reporter in HEK293T cells only in the presence of Cre. These results provide a robust framework for Cre-dependent CRISPR-dCas9 approaches across different model systems, and will enable cell-specific targeting when combined with common Cre driver lines or Cre delivery via viral vectors.

A Cre-dependent CRISPR/dCas9 system for gene expression regulation in neurons

#### _Cultured neuron experiments_

Primary rat neuronal cultures were generated from embryonic day 18 rat striatal tissue as described previously (**Savell et al., 2019a**). Briefly, cell culture wells were coated overnight at 37° C with poly-L-lysine (0.05 mg/ml for culture wells supplemented with up to 0.05 mg/ml Laminin) and rinsed with diH2O. Dissected tissues were incubated with papain for 25 min at 37°C. After rinsing in Hank’s Balanced Salt Solution (HBSS), a single cell suspension of the tissue was re-suspended in Neurobasal media (Invitrogen) by trituration through a series of large to small fire-polished Pasteur pipets. Primary neuronal cells were passed through a 100 μM cell strainer, spun and re-suspended in fresh media. Cells were then counted and plated to a density of 125,000 cells per well on 24-well culture plate with or without glass coverslips (60,000 cells/cm). Cells were grown in Neurobasal media plus B-27 and L-glutamine supplement (complete Neurobasal media) for 11 DIV in a humidified CO2 (5%) incubator at 37° C.

For viral transduction, cells were transduced with lentiviruses on DIV 4 or 5. All viruses had a minimum titer of 1×109 GC/ml, with a target multiplicity of infection (MOIs) of at least 1000. After an 8-16 hr incubation period, virus-containing media was replaced with conditioned media to minimize toxicity. A regular half-media change followed on DIV 8. On DIV 11, transduced cells were imaged and virus expression was verified prior to RNA extraction. EGFP and mCherry expression was also used to visualize successful transduction using a Nikon TiS inverted epifluorescence microscope.

#### _RNA extraction and RT-qPCR_

Total RNA was extracted (RNAeasy kit, Qiagen) with DNase treatment (RNase free DNAse, Qiagen), and reverse-transcribed (iScript cDNA Synthesis Kit, Bio-Rad). cDNA was subject to qPCR for genes of interest, as described previously (**Savell et al., 2016**). A list of PCR primer sequences is provided in **Supplementary Data Table 1**.

- **View inline**
- **View popup**

**Supplemental Data Table 1.**

Sequences of primers and sgRNAs

#### _CRISPR-dCas9 construct design_

To achieve transcriptional activation or inactivation, lentivirus-compatible plasmids were engineered to express dCas9 fused to either VPR or KRAB-MeCP2, based on existing published plasmids (Addgene plasmid # 114196 (**Savell et al., 2019a**); Addgene plasmid # 155365 (**Duke et al., 2020**)). A Cre-dependent DIO version of the dCas9-VPR construct was generated by insertion of LoxP and Lox2272 sequences flanking the dCas9-VPR cassette. dCas9-VPR was PCR-amplified to insert additional restriction sites (KpnI, BmtI, and BspDI) to allow for LoxP and Lox2272 insertion and was subsequently inserted in reverse orientation to create an intermediate construct via sequential digest and ligation (AgeI, EcoRI). LoxP and Lox2272 sites were amplified from a DIO construct (Addgene plasmid # 113685 (**Don et al., 2017**)) to create restriction sites (KpnI and BmtI around one set; BspDI and KpnI around another set) and were inserted into the intermediate construct via sequential restriction digest and ligation (KpnI, BmtI, BspDI and EcoRI). Intron-containing dCas9-VPR was constructed by the insertion of a gBlock containing the SV40 intron into the original dCas9-VPR plasmid via Gibson assembly (Gibson Assembly kit, New England BioLabs). The SVI-FLEX construct was built via Gibson assembly of the original dCas9-VPR backbone and two gBlocks encoding the SV40 intron sequence, LoxP and Lox2272 sites (sequences based on DIO construct described above) and dCas9-part1 in inverted orientation. The intron-containing CRISPRi construct (SVI-dCas9-KRAB-MeCP2) was built via restriction digest and Gibson assembly (SfiI, EcoRI, and Xhol, Gibson Assembly kit, New England BioLabs) of the KRAB-MeCP2 and SVI-dCas9-VPR constructs. A Cre-encoding construct (Addgene plasmid # 49056 ((**Kaspar et al., 2002**)) was used to amplify and insert a Cre transgene into a lentivirus compatible backbone that contained the hSYN promoter and expressed mCherry for visualization via Gibson assembly. To create an additional Cre construct, mCherry was replaced with GFP via sequential digest and ligation (EcoRI and XhoI). dCas9-VPR-expressing constructs were co-transduced with sgRNA-containing constructs. Gene-specific sgRNAs were designed using an online sgRNA tool, provided by the Zhang Lab at MIT (crispr.mit. edu) and inserted in a previously described lentivirus compatible sgRNA scaffold construct (Addgene plasmid # 114199 (**Savell et al., 2019b**)). To ensure specificity, all CRISPR crRNA sequences were analyzed with the National Center for Biotechnology Information’s (NCBI) Basic Local Alignment Search Tool (BLAST) and Cas-OFFinder (http://www.rgenome.net/cas-offinder/). sgRNAs were designed to target _GRM2_, _Tent5b, Sstr2_, _Gadd45b_, and _Fos_ respectively. A list of the target sequences is provided in **Supplementary Data Table 1**. crRNA sequences were annealed and ligated into the sgRNA scaffold using the BbsI or BsmBI cut site. Plasmids were sequence-verified with Sanger sequencing; final crRNA insertion was verified using PCR. Lentivirus-compatible SYN-SVI-DIO-dCas9-VPR and SYN-SVI-DIO-dCas9-KRAB-MeCP2 plasmids will be made available on Addgene (plasmid #164576 and # 170378).

#### _Lentivirus production_

Viruses were produced in a sterile environment subject to BSL-2 safety by transfecting HEK293T cells with specified CRISPR-dCas9 plasmids, the psPAX2 packaging plasmid, and the pCMV-VSV-G envelope plasmid (Addgene plasmids #12260 & #8454) with FuGene HD (Promega) for 40-48 hrs as previously described (**Savell et al., 2019b**). Viruses were purified using filter (0.45 μm) and ultracentrifugation (25,000 rpm, 1 hr 45 min) steps. Viral titer was determined using a qPCR Lentivirus Titration Kit (Lenti-X, qRT-PCR Titration Kit, Takara). For smaller scale virus preparation, each sgRNA plasmid was transfected in a 12-well culture plate as described above. After 40-48 hr, lentiviruses were concentrated with Lenti-X concentrator (Takara), resuspended in sterile PBS, and used immediately. Viruses were stored in sterile PBS at −80°C in single-use aliquots.

#### _HEK293T cell culturing and transfection_

HEK293T cells were obtained from American Type Culture Collection (ATCC catalog# CRL-3216, RRID:CVCL_0063) and cultured in standard HEK Media: DMEM (DMEM High glucose, pyruvate; Gibco 11995081) supplemented with 10% bovine serum (Qualified US Origin; BioFluid 200-500-Q) and 1U Penicillin-Streptomycin (Gibco 15140122). Cells were maintained in T75 or T225 tissue flasks. At each passage, cells were trypsinized for 1-3 min (0.25% trypsin and 1 mM EDTA in PBS pH 7.4) at room temperature. For transfection experiment cells were plated in 24 well plates and transfected with FuGene HD (Promega).

#### _Luciferase assay_

Bidirectional regulation by SVI-DIO CRISPRa and CRISPRi machinery was examined using a previously described _Fos_ luciferase reporter plasmid (**Duke et al., 2020**). 80,000 HEK293T cells were plated in 500 μl HEK Media. After cells reached 40-50% confluence, 500 ng total plasmid DNA was transfected with 1.5 μl FuGene HD (Promega) as follows: 50 ng of luciferase plasmid, 450 ng in 1:2 molar ratio of total sgRNA:CRISPRa or CRISPRi plasmid. A luciferase glow assay was performed according to manufacturer’s instructions 40 hrs following transfection (Thermo Scientific Pierce Firefly Glow Assay; Thermo Scientific 16177). Cells were lysed in 100 μl 1x Luciferase Cell Lysis Buffer while shaking at low speed and protected from light for 30 min. 20 μl of cell lysate was then added to an opaque 96-well microplate (Corning 353296) and combined with 50 μl 1x D-Luciferin Working Solution supplemented with 1x Firefly Signal Enhancer (Thermo Scientific Pierce Firefly Signal Enhancer; Thermo Scientific 16180). Following a 10 min dark incubation period to allow for signal stabilization, luminescence was recorded using a Synergy 2 Multi-Detection Microplate Reader (BioTek). Luminescence in dCas9-KRAB-MeCP2 and dCas9-VPR experiments was recorded with a read height of 1 mm, 1 s integration time, and 135 or 100 ms delay, respectively. Representative images of luciferase reporter activity presented in **Figure 4** were captured using an Azure c600 imager (Azure Biosystems).

Materials and methods

CRISPR systems are prokaryotic adaptive immune systems that use RNA-guided Cas nucleases to recognize and destroy foreign genetic elements, like bacteriophages and plasmids. To overcome CRISPR immunity, phages have evolved diverse families of anti-CRISPR proteins (Acrs), each of which inhibits the nucleic acid binding or cleavage activities of specific Cas protein families. Recently, **Lin et al. (2020)** described the discovery and characterization of 7 different Acr families (AcrVIA1-7) that inhibit type VI-A CRISPR systems, which use the nuclease Cas13a to perform RNA-guided RNA cleavage. In this Matters Arising article, we detail several inconsistencies that question the results reported in the **Lin et al. (2020)** study. These include inaccurate bioinformatics analyses, as well as reported experiments involving bacterial strains that are impossible to construct. The authors were unable to provide their published strains with which we might reproduce their experiments. We independently tested the Acr sequences described in **Lin et al. (2020)** in two different Cas13 inhibition assays, but could not detect anti-CRISPR activity. Taken together, our data and analyses prompt us to question the claim that AcrVIA1-7 reported in Lin et al. are bona fide type VI anti-CRISPR proteins.

Lack of Cas13a inhibition by anti-CRISPR proteins from Leptotrichia prophages

### **Bacterial strains and culture conditions**

Plasmid culture and extraction, as well as plasmid transformation assays were performed with New England Biolabs Turbo Competent E. coli. Bacteria were grown in Lysogeny Broth (LB) supplemented with antibiotics: kanamycin (50 µg/mL), chloramphenicol (25 µg/ml), and anhydrotetracyline (100 ng/mL). 0.2% arabinose was added to induce expression of Cas13 from Para promoters.

### **Plasmid construction**

Cas13 plasmids for transformation assay were cloned into the vector pAM38 containing p15A origin of replication, chloramphenicol resistance marker, and an arabinose-inducible Para promoter. Coding sequences for LwaCas13a, LbuCas13a, or LseCas13a were inserted downstream of the Para promoter, and crRNA constructs were inserted downstream of the cas13 coding sequence driven by a synthetic promoter (J23119). These plasmids were constructed by three piece Gibson assembly as described below and validated by Sanger sequencing. Lists of plasmids, oligonucleotide primers, and crRNAs can be found in Supplementary Tables.

LwaCas13a plasmids: Three piece Gibson assembly using (1) SalI/HindIII-digested pAM38, (2) LwaCas13a coding sequence and Shine-Dalgarno sequence amplified from Addgene #91865 using primers oAM1496 and oAM1497, and (3) J23119-driven crRNA gBlock fragments synthesized by IDT was used to generate pAM491 (LwaCas13a kan crRNA) and pAM494 (LwaCas13a non-targeting crRNA).

LbuCas13a plasmids: Three piece Gibson assembly using (1) SalI/HindIII-digested pAM38, (2) LbuCas13a coding sequence codon-optimized for E. coli (synthesized by Genewiz) and Shine-Dalgarno sequence amplified using primers oAM211 and oAM1477, and (3) J23119-driven crRNA gBlock fragments synthesized by IDT was used to generate pAM492 (LbuCas13a kan crRNA) and pAM475 (LbuCas13a non-targeting crRNA).

LseCas13a plasmids: Three piece Gibson assembly using (1) SalI/HindIII-digested pAM38, (2) LseCas13a coding sequence codon-optimized for E. coli (synthesized by Genewiz) and Shine-Dalgarno sequence amplified using primers oAM207 and oAM994, and (3) J23119-driven crRNA gBlock fragments synthesized by IDT was used to generate pAM526 (LseCas13a kan crRNA) and pAM319 (LseCas13a non-targeting crRNA).

Acr plasmids: Acrs were expressed from pAM326 (carrying kanamycin resistance marker and pWV01 origin of replication) and driven by an anhydrotetracycline-inducible Ptet promoter. Acr coding sequences were synthesized by Genewiz, amplified using primers oAM1529 and oAM1530, and inserted with Ptet fragment into HindIII/EagI-digested pAM326 via three piece Gibson assembly to make pAM495 (AcrVIA1), pAM496 (AcrVIA2), pAM497 (AcrVIA3), pAM498 (AcrVIA4), pAM499 (AcrVIA5), pAM500 (AcrVIA6), pAM501 (AcrVIA7). The AcrVIA1Lse plasmid pAM383 was generated previously (**Meeske et al. 2020**).

### **Plasmid transformation assay**

NEB Turbo competent E. coli were transformed with either empty vector or plasmids harboring LwaCas13a, LbuCas13a, or LseCas13a and expressing non-targeting or targeting crRNAs, and selected on LB agar containing chloramphenicol and 0.2% glucose to repress Cas13a expression. Each transformed strain was made chemically competent by resuspension of exponentially growing cells in ⅕ culture volume of ice cold TFB I (10 mM CaCl2,30 mM potassium acetate, pH 5.8, 100 mM RbCl, 50mM MnCl2, 15% glycerol), incubation on ice for 15 minutes, then pelleting and resuspension in 1/25 culture volume of ice cold TFB II (10 mM MOPS pH 6.5, 10 mM RbCl, 75 mM CaCl2, 15% glycerol). Chemically competent cells were transformed by heat shock at 42°C with 100 ng Acr plasmids or empty vector control, then recovered for 1 hour in LB. Tenfold serial dilutions of recovered transformants were made and 5 µL of each dilution was spotted onto LB agar supplemented with kanamycin, chloramphenicol, arabinose, and anhydrotetracycline. Plates were photographed after 1 day of incubation at 37°C.

### **Human cell culture and transfections**

Human HEK 293T cells (ATCC) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated FBS (HI-FBS) and 1% penicillin/streptomycin. Supernatants from cell cultures were tested monthly for mycoplasma using MycoAlert PLUS (Lonza). For transfections, approximately 2 × 104 cells per well were seeded in 96-well plates. Approximately 18-22 hours after seeding, transfections were performed using TransIT-X2 (Mirus) transfection reagent with 60 ng of LwaCas13a plasmid (Addgene ID 91924), 20 ng of gRNA plasmid (generated by cloning oligos into an LwaCas13a crRNA entry plasmid, LTH151, Addgene ID 171129), and with 0, 2, 12, or 72 ng of Acr plasmid. The human codon optimized Acr constructs for AcrVIA4 and AcrVIA5 were synthesized by Twist Biosciences and cloned into a pCMV backbone (plasmid IDs LTH956 and LTH957 respectively; see **Supplementary Sequences**), similar to how we previously generated the human cell expression constructs for AcrIIA4 and AcrIIA5 (Addgene IDs 133801 and 133802, respectively) (see **Supplementary Sequences**). The total amount of DNA in each transfection remained constant (152 ng), with a pCMV-null plasmid (Addgene ID 171128) used to balance DNA amounts when titrating the Acr plasmid. The plasmid mixtures were combined with 0.54 ul of TransIT-X2 and volume of Opti-MEM (ThermoFisher) to a final volume of 15 ul, incubated at room temperature for 15 minutes, and then applied to the HEK 293T cells. All experiments were performed with 3 independent biological replicates.

### **RNA extraction and RT-qPCR**

At 48 hours post-transfection, total RNA was extracted from transfected cells using an RNeasy Plus Mini Kit (Qiagen; cat. no. 74136). Purified RNA was reverse transcribed using a High-Capacity RNA-to-cDNA kit (ThermoFisher; cat. no. 4388950) using up to 250 ng of RNA as input. Prior to qPCR, the cDNA library for each sample was diluted 1:20. Samples for qPCR were prepared in technical triplicate with 5 ul of Fast SYBR Green Master Mix (ThermoFisher; cat. no. 4385610), 3 ul diluted cDNA, and 2 ul qPCR primer pairs specific to the target (IDT; **Supplementary Table 1**). Control reactions amplifying ACTB were set up in parallel for each cDNA library. Reaction cycling was performed using a Roche LightCycler480. To determine the fold change of each sample, the expression levels were normalized to a negative control transfection containing LwaCas13a plasmid and an empty pUC19 backbone plasmid (Addgene ID 133961). Each qPCR reaction was analyzed for purity by melting curve analysis to confirm a single PCR product.

CRISPR-Cas biology and technologies have been largely shaped to-date by the characterization and use of single-effector nucleases. In contrast, multi-subunit effectors dominate natural systems, represent emerging technologies, and were recently associated with RNA-guided DNA transposition. This disconnect stems from the challenge of working with multiple protein subunits _in vitro_ and _in vivo_. Here, we apply cell-free transcription-translation (TXTL) to radically accelerate the characterization of multi-subunit CRISPR effectors and transposons. Numerous DNA constructs can be combined in one TXTL reaction, yielding defined biomolecular readouts in hours. Using TXTL, we mined phylogenetically diverse I-E effectors, interrogated extensively self-targeting I-C and I-F systems, and elucidated targeting rules for I-B and I-F CRISPR transposons using only DNA-binding components. We further recapitulated DNA transposition in TXTL, which helped reveal a distinct branch of I-B CRISPR transposons. These capabilities will facilitate the study and exploitation of the broad yet underexplored diversity of CRISPR-Cas systems and transposons.

- PAM-DETECT for rapid determination of PAMs for Type I CRISPR-Cas systems in TXTL
- Mining of Type I orthologs and characterization of extensively self-targeting systems
- TXTL-based assessment of DNA target recognition and transposition by CRISPR transposons
- Identification of a distinct branch of Type I-B CRISPR transposons

Rapid cell-free characterization of multi-subunit CRISPR effectors and transposons

STAR Methods

Although performing adaptive immunity, CRISPR-Cas systems are present in only 40% of bacterial genomes. Here, we observed an abrupt transition of bacterial CRISPR-Cas abundance at around 45°C. Phylogenetic comparative analyses confirmed that the abundance correlates with growth temperature only at the temperature range around 45°C. Meanwhile, we noticed that the diversities of cellular predators have a precipitous decline at this temperature range. The grazing risk faced by bacteria reduces substantially at around 45°C and almost disappears above 60°C. So viral lysis would become the dominating factor of bacterial mortality, and antivirus immunity has a higher priority. In temperature ranges where the abundance of cellular predators does not change with temperature, temperatures would have negligible effects on CRISPR-Cas abundance. The hypothesis predicts that bacteria should also be rich in CRISPR-Cas systems if they live in other extreme conditions inaccessible to grazing predators.

Bacterial CRISPR-Cas Abundance Increases Precipitously at Around 45°C: Linking Antivirus Immunity to Grazing Risk

From the growth temperature database TEMPURA and the Genome Taxonomy Database,**22**, **23** we retrieved 682 bacteria and 156 archaea with growth temperatures and phylogenetic information. The sequences of these genomes were downloaded from ftp://ftp.ncbi.nlm.nih.gov/genomes/. We annotated their CRISPR-Cas systems using CRISPRCasFinder v1.3 program.**24** The raw data of these 682 bacteria and 156 archaea were deposited as Table S1.

According to Couvin et al.,**24** the annotated CRISPR arrays were classified into four categories, 1 to 4, according to their evidence levels. The CRISPR arrays with evidence levels 3 and 4 are highly likely candidates, and those with evidence levels 1 and 2 are potentially invalid. Therefore, only the CRISPR arrays with evidence levels 3 and 4 were counted in calculating CRISPR array abundance. We counted the putative CRISPR arrays with evidence levels 1 and 2 as zero and presented the results in the main text and Table S2. We also replicated our analyses by regarding the putative CRISPR arrays with evidence levels 1 and 2 as controversial CRISPR arrays and discarding the species having only CRISPR arrays with evidence levels 1 and 2 in calculating CRISPR array abundance. That is, these species were excluded from both numerator and denominator in the calculation of CRISPR-Cas abundance. Similar results (Table S3) were obtained, and the same conclusion was supported.

The CRISPR array abundance in a bacterial or archaeal group was defined as the total number of CRISPR arrays annotated in their genomes divided by the genome numbers of the group. The abundances of CRISPR spacers, _cas_ genes, and _cas_ gene clusters were defined similarly.

The phylogenetic signals (λ) of CRISPR array abundance, CRISPR spacer abundance, _cas_ gene abundance, _cas_ gene cluster abundance, and growth temperatures were estimated using the phylosig function of the R (Version 4.0.3) package phytools (Version 0.7-70)**25**. The phylogenetic generalized least squares (PGLS) regression was performed using the R (Version 4.0.3) package phylolm (version 2.6.2)**26**. Pagel’s lambda model has been applied in the analyses.

This study describes the strong anti-CRISPR activity of the bacterial AcrIIA4 protein in _Nicotiana benthamiana_, a model plant used as molecular farming platform. The results demonstrate that AcrIIA4 abolishes site-directed mutagenesis in leaves when transiently co-expressed with CRISPR/Cas9. We also show that AcrIIA4 represses CRISPR/dCas9-based transcriptional activation (CRISPRa) of both reporter and endogenous genes in a highly efficient, dose-dependent manner. Furthermore, the fusion of an auxin degron to AcrIIA4 results in auxin-regulated activation of a downstream reporter gene. The strong anti-Cas9 activity of AcrIIA4 reported here opens new possibilities for customized control of gene editing and gene expression in plants.

Strong and tunable anti-CRISPR/Cas9 activity of AcrIIA4 in plants

### **GoldenBraid Cloning**

DNA constructs used in this work were assembled using GoldenBraid**25**. AcrIIA4 protein sequence, as originally reported by Rauch et al.**26** fused to SV40 NLS and mAID47 sequence, as reported by Brosh et al.**21** (modified as N-tag or C-tag), were codon optimized for _N. benthamiana_ using IDT Codon Optimization Tool (https://eu.idtdna.com/CodonOpt) and subsequently domesticated at https://gbcloning.upv.es/do/synthesis/. Additionally, a _N. benthamiana_ codon optimized AcrIIA4 protein version without SV40-NLS, B3-B4 version, was domesticated using GoldenBraid GB Domesticator tool (https://gbcloning.upv.es/do/domestication/). Level 1 assemblies of transcriptional units from individual Level 0 parts were performed through Golden Gate-like multipartite BsaI reactions. All Level 0 and Level 1 assemblies were performed as previously reported **19**. Level 0 assemblies were confirmed by restriction analysis and Sanger sequencing and Level 1 assemblies were verified by restriction analysis. An exhaustive list of all plasmids used in this work are listed in Supplementary Table 2.

### **N. benthamiana transient expression**

For transient expression assays, plasmids were transferred to _Agrobacterium tumefaciens_ strain C58. Five to six weeks old _N. benthamiana_ plants grown at 24°C and 16h (light)/ 20°C and 8h (darkness) conditions were used. Agroinfiltration was carried out as previously reported**27**. Briefly, overnight _A. tumefaciens_ cultures were pelleted and resuspended in agroinfiltration solution (10 mM MES, pH 5.6, 10 mM MgCl2 and 200 μM acetosyringone) to an optical density of 0.1 at 600nm (OD600). Bacterial suspensions were incubated for 2h at room temperature on a horizontal rolling mixer, and then mixed for co-expression experiments, in which more than one GB element was used. Finally, agroinfiltrations were carried out through the abaxial surface of the three youngest fully expanded leaves of each plant with a 1ml needle-free syringe. For some experiments, agroinfiltrations were carried out with small modifications from the general procedure described above. For dose-response assays, SV40-AcrIIA4 TU (GB3344) was resuspended in agroinfiltration solution to an OD600 of 0.1 and subsequently diluted to 0.05, 0.01, 0.005, and 0.001 using a culture carrying an empty vector to maintain the final OD600 at 0.1. For the time-course assay, the strain carrying the SV40-AcrIIA4 TU (GB3344) was agroinfiltrated at OD600 of 0.05 at 0, 24, 48 and 72 h after infiltration of the dCasEV2.1 with the DFR gRNA (GB2513) and the pSlDFR reporter construct (GB1160). Detailed information of the experimental design can be found in the Supplementary Materials and Methods section.

### **Luciferase activity and determination of relative transcriptional activity**

Leaf samples were collected at 4 days post infiltration (dpi). For the determination of the FLuc/RLuc activity, one 0.8 cm diameter disc per agroinfiltrated leaf was excised. Leaf discs were frozen in liquid nitrogen and subsequently homogenized with 180 μl of Passive Lysis buffer, followed by 15 min of centrifugation at 14000 x g at 4°C. 10 μl of crude extract were mixed with 40 μl of LARII and Firefly luciferase (FLuc) activity was determined using a GloMax 96 microplate luminometer (Promega) with a 2-s delay and a 10-s measurement time. After the measurements, 40 μl of Stop&Glo Reagent were added per sample and Renilla luciferase (RLuc) activity was determined using the same protocol. Sample FLuc/RLuc ratios were calculated as the mean value of the three independent agroinfiltrated leaves. Relative transcriptional activities (RTAs) were calculated as the Fluc/Rluc ratios of the pSlDFR reporter in each sample normalized with the Fluc/Rluc ratios produced by a pNos reporter (GB1116) assayed in parallel and expressed in relative promoter units (rpu)**19**.

### **Determination of Cas9-mediated editing activity**

Genomic DNA was extracted from leaf samples 5 dpi following the CTAB protocol**28**. The DNA was used as template for PCR amplification of the targeted sites with primers listed on Supplementary Table 3 and using MyTaq™ DNA Polymerase (Bioline). Subsequently, PCR products were analyzed in 1% agarose gel electrophoresis, purified with the ExoSAP-IT™ PCR Product Cleanup Reagent (Applied Biosystems™) following manufacturer instructions and Sanger-sequenced. Finally, sequencing results were analyzed using Synthego CRISPR Performance Analysis (https://ice.synthego.com/#/) to determine the ICE score.

### **RNA isolation and reverse transcription-quantitative PCR (RT-qPCR)**

Leaf samples from infiltrated plants were harvested at 4 dpi and 100 mg of tissue were used for total RNA isolation using the Thermo Scientific™ GeneJET RNA Purification Kit. Total RNA was treated with Recombinant DNase I (RNase-free) (Takara) following manufacturer’s instructions. Aliquots of 1 μg of the treated RNA were used for cDNA synthesis with oligo dT using PrimeScript™ RT-PCR kit (Takara). cDNAs (0.4 μl) were used to determine the expression levels for each gene in triplicated 25 μl reactions with the SYBR® Premix Ex Taq (Takara) using the Applied biosystem 7500 Fast Real Time PCR system. _N. benthamiana_ F-BOX gene was used as internal reference**29**. Calculations of each sample were carried out according the comparative ΔΔCT method**30**. Primers used for qRT-PCR reactions are listed in Supplementary Table 3.

The acquisition of antibiotic resistance genes via horizontal gene transfer is a key driver of the rise in multidrug resistance amongst bacterial pathogens. Bacterial defence systems per definition restrict the influx of foreign genetic material, and may therefore limit the acquisition of antibiotic resistance. CRISPR-Cas adaptive immune systems are one of the most prevalent defences in bacteria, found in roughly half of bacterial genomes, but it has remained unclear if and how much they contribute to restricting the spread of antibiotic resistance. We analysed ~40,000 whole genomes comprising the full RefSeq dataset for 11 species of clinically important genera of human pathogens including _Enterococcus_, _Staphylococcus_, _Acinetobacter_ and _Pseudomonas_. We modelled the association between CRISPR-Cas and indicators of horizontal gene transfer, and found that pathogens with a CRISPR-Cas system were less likely to carry antibiotic resistance genes than those lacking this defence system. Analysis of the mobile genetic elements targeted by CRISPR-Cas supports a model where this host defence system blocks important vectors of antibiotic resistance. These results suggest a potential “immunocompromised” state for multidrug-resistant strains that may be exploited in tailored interventions that rely on mobile genetic elements, such as phage or phagemids, to treat infections caused by bacterial pathogens.

CRISPR-Cas is associated with fewer antibiotic resistance genes in bacterial pathogens

### **Genomes**

RefSeq genomes for _P. aeruginosa_, _A. baumannii_, _Neisseria meningitidis_, _Staphylococcus epidermidis_, _Streptococcus pyogenes_, _Francisella tularensis_, _Mycobacterium tuberculosis_, _Neisseria gonorrhoeae_, _E. faecium_ and _E. faecalis_ were retrieved from NCBI between December 2020 and January 2021 in nucleotide fasta format using ncbi-genome-download v0.3.0 (https://github.com/kblin/ncbi-genome-download). The strains were selected to include a variety of CRISPR-Cas system types, pathogenic lifestyles (i.e. obligate pathogens such as _F. tularensis_, _M. tuberculosis_ and _N. gonorrhoeae_ as well as opportunistic pathogens like _P. aeruginosa_, _S. aureus and S. pyogenes_), and species of significance in the spread of ABR such as _A. baumannii_ and _E. faecium_. The species used are summarised in **Table 1**.

- **View inline**
- **View popup**

**Table 1 –**

Summary of species included

### **_Identification of CRISPR-Cas systems, ABR genes, plasmids, ICEs, and_** **intI1**

CRISPR-Cas systems were detected using CRISPRCasTyper [**20**], and orphan CRISPR arrays were not included in any spacer analysis. CRISPR-Cas systems with “Unknown” predictions were also removed from the dataset. Acquired ABR genes and plasmid replicons were identified using abricate (https://github.com/tseemann/abricate) with the NCBI AMRFinderPlus [**21**] and PlasmidFinder [**22**] databases, both from 19/04/2019 and containing 5386 and 460 sequences, respectively. To detect ICEs and _intI1_, custom BLAST databases were created based on the ICEberg database (retrieved January 2021; Liu et al. 2019), and the NCBI gene entry for class 1 integron integrase _intI1_ [NC_019069.1 (99852.100865, complement)], and used with abricate.

### **Identification of spacer targets and MGEs carrying ABR genes**

Spacer sequences were searched against BLAST databases constructed from the aforementioned ABR, ICE and prophage databases as well as a curated plasmid database [**24**], and the complete RefSeq viral release (retrieved February 2021). Only hits with 100% identity were used for downstream analyses. The total number of spacers per genome was approximated by subtracting 1 from the number of repeats for each array and summing these values, and any spacers with no hits in any database were assigned as “unknown”.

MGEs carrying ABR genes were identified by running a blastn search with default parameters, querying all sequences in search databases used for ICE, plasmid and virus detection against a custom BLAST database built from the aforementioned NCBI AMRFinderPlus database. Subsequently, a percent identity cutoff of 90% was applied to hits. The presence of ABR genes on MGEs was classified in a binomial way, in which ≥ 1 ABR gene was counted as the presence of ABR on the element. In cases where multiple MGEs were targeted by a single spacer, the spacer was classified as targeting an MGE with ABR if at least one of these MGEs carried at least one ABR gene.

### **Data analysis and statistical modelling**

The University of Exeter’s Advanced Research Computing facilities were used to carry out this work. Data analyses were conducted in R v4.0.2 using the tidyverse suite of packages [**25**]. Unless otherwise specified, linear or generalized linear models were fitted using the _lm_ or _glm_ functions in base R. Linear models allow the inclusion of multiple predictors in one model, termed fixed effects, to estimate the relationship between these predictors and a response variable. These models generate a formula that calculates the mean of a response variable based on the value of each fixed effect (predictor). Regression coefficients are fitted, which give the slope and the intercept (value of the response variable when the predictor = 0) for each fixed effect. For categorical predictors, a different intercept is fitted for each level of the predictor (e.g. each species in a multispecies model). To fit different slopes for levels of categorical predictor, interaction terms must be fitted between them and a continuous predictor. For example, an interaction between species and the count of ABR genes would allow varying slopes and intercepts for each individual species.

For these analyses, a maximal model was generated and all possible candidate models were compared using the AIC method using _dredge_ from the MuMIn package [**26**]. AIC values assess the fit of a model by looking at the likelihood of a model given the data, penalising for increased number of parameters (as increased complexity of the model increases parameter uncertainty). The model with the lowest AIC value was selected, and no alternative models were within 2 AICs of the best fitting model.

Prediction data frames with 95% confidence intervals were generated using _ggpredict_ from the package ggeffects [**27**], model dispersion was tested and scaled residuals were examined using DHARMa residual diagnostics [**28**], and the final predictions were visualised with ggplot2, ghibli [**29**] and cowplot (https://wilkelab.org/cowplot/index.html). Prediction plots were produced only for biologically realistic combinations of fixed effects and response, i.e. those where sufficient data was present in the data set.

Associations between genome length and CRISPR-Cas presence/absence were tested using a linear model with genome length as the response variable and CRISPR-Cas presence/absence and species as fixed effects, with CRISPR-Cas presence/absence × species as an interaction term.

The effect of various MGEs on the distribution of CRISPR-Cas was modelled using a binomial GLM with CRISPR-Cas presence/absence as the response variable and counts of ICEs, plasmid replicons, ABR genes and _intI1_ copies, as well as species, as fixed effects. In addition, we fitted an interaction between species and every other fixed effect. Prediction data frames were created up to the maximum number of each MGE detected in each species (unless this value was an outlier, in which case a more biologically informative upper limit was set).

The relationship between spacers targeting MGEs with and without ABR and the count of ABR genes was modelled using a Bayesian Poisson GLM in MCMCglmm. The default weakly informative priors were used, and models were run for 20,000 iterations with a burn in period of 10,000 iterations and a thinning interval of 5. Two models were run, with the number of ABR genes as the response variable, species as a fixed effect, and either the counts of spacers targeting MGEs with ABR or targeting MGEs without ABR as an additional fixed effect. Prediction data frames were created up to the maximum number of spacers detected targeting MGEs with and without ABR.

### **Construction of trees and genetic distance-controlled single-species models**

The package mashtree [**30**] with mash v2.0.0 [**31**] was used to create neighbour-joining trees for each species based on whole genomes. Genomes were sketched using mash with default parameters prior to using mashtree. Mash distance was calculated for all isolates relative to the first genome entry for the group, in order to remove outliers with a distance > 0.1, which are likely to have had their species misclassified. Trees were subsequently rooted to outgroups and ultrametric trees were produced with the ape package in R [**32**] using the chronos function with smoothing parameters (λ) of 1 and 10. The “correlated”, “discrete” and “relaxed” models were used for each value of λ, which vary in the extent to which they permit adjacent parts of the tree to evolve at different rates. The tree with the highest log-likelihood for each species was used. Due to the large number of genomes and low prevalence of CRISPR-Cas, we did not perform this analysis for _S. aureus_.

The MCMCglmm package [**33**] was used to run Bayesian GLMs incorporating trees. The default weakly informative priors were used, and all models were run for 30,000 iterations with a burn in period of 20,000 iterations and a thinning interval of 5. Poisson GLMs were run for each species, with count of ABR genes as the response variable and CRISPR-Cas type or spacer count (in which case genomes with no CRISPR-Cas were removed) as a fixed effect. In cases where more than one CRISPR-Cas type was present, CRISPR-Cas type × total spacers was fit as an interaction. The small sample size for _S. epidermidis_ with CRISPR-Cas systems (n = 54), meant we chose not to model the association between spacer count and ABR genes for this species. Prediction data frames for spacer counts were created up to the maximum count of spacers detected for that species, with the exception of _S. pyogenes_, for which the top value was an outlier at 33 so the next highest value of 17 was taken.

### **Code availability and reproducibility**

The snakemake pipeline used to calculate genome lengths and detect CRISPR-Cas systems, ABR genes, _intI1_, ICEs and plasmid replicons, as well as R scripts used for producing trees and statistical analysis for this work are available at https://github.com/elliekpursey/crispr-pathogens. Snakemake v5.18.1 [**34**] and Python v3.8.3 with Biopython v1.78 [**35**] were used for this work.

CRISPR/Cas9 can be used to inactivate or modify genes by inducing double-stranded DNA breaks1–3. As a protective cellular response, DNA breaks result in p53-mediated cell cycle arrest and activation of cell death programs4,5. Inactivating _p53_ mutations are the most commonly found genetic alterations in cancer, highlighting the important role of the gene6–8. Here, we show that cells deficient in p53, as well as in genes of a core CRISPR-p53 tumor suppressor interactome, are enriched in a cell population when CRISPR is applied. Such enrichment could pose a challenge for clinical CRISPR use. Importantly, we identify that transient p53 inhibition suppresses the enrichment of cells with these mutations. Furthermore, in a data set of >800 human cancer cell lines, we identify parameters influencing the enrichment of _p53_ mutated cells, including strong baseline _CDKN1A_ expression as a predictor for an active CRISPR-p53 axis. Taken together, our data identify strategies enabling safe CRISPR use.

CRISPR enriches for cells with mutations in a p53-related interactome, and this can be inhibited

### **Cells**

The Hox bone marrow cell line was generated by transducing bone marrow cells of C57BL/6 Cas9+ GFP+ mice (The Jackson Laboratory, #026179) with an estrogen inducible retroviral construct expressing HoxB8 (ER-Hoxb8, a kind gift from Mark P. Kamps, University of California, San Diego) as described in**1**,**2**. Cells were cultured in 1 μM β-estradiol (BE, Sigma-Aldrich #E2758) and 25 nM mouse Stem Cell Factor (SCF, Peprotech #250-03) for several weeks with HoxB8 expression turned on to establish a cell line like population. Hox cells were cultured in RPMI-1640 (Sigma-Aldrich #R0883) with 10% heat-inactivated fetal bovine serum, 1% penicillin-streptomycin-glutamine (Gibco #10378016), 1 μM BE, and 25 nM SCF.

B16-F10 cell is a mouse melanoma cell line, purchased from ATCC (#CRL6475) and used at a low passage number. Cas9 expressing cells were generated by transducing B16-F10 cells with lentiCas9-Blast (Addgene #52962) lentiviral particles. B16 cells were cultured in RPMI-1640 (Sigma-Aldrich #R0883) with 10% heat-inactivated fetal bovine serum, and 1% penicillin-streptomycin-glutamine (Gibco #10378016)

### **Viral preparation and transduction**

Lentiviral particles were generated by seeding 2×106 HEK293T cells in 10 cm plate in 10 ml of DMEM (Sigma-Aldrich #D6546) with 10% heat-inactivated fetal bovine serum and 1% L-glutamine (Gibco #A2916801). After ~24 h of culture, the cells reached ~60-70% confluency, and the medium was replaced by 5 ml of prewarmed fresh media. Transfer plasmids (lentiCas9-Blast, Addgene #52962; or LentiGuide-Puro-P2A-EGFP_mRFPstuf, Addgene #137730), pMD2.G (Addgene #12259), and psPAX2 (Addgene #12260) were mixed at 4:5:1 ratio (10 μg : 12.5 μg : 2.5 μg for 10 cm plate), and transfected into HEK293T cells using Lyovec (Invivogen #lyec) according to the manufacturer’s protocol. After 12 h, the medium was replaced by 8 ml of DMEM with 30% heat-inactivated fetal bovine serum and 1% L-glutamine. After another 36 h, the supernatant containing the virus was collected and centrifuged to remove the cell debris and used to spin infect cells.

For ER-Hoxb8 retrovirus preparation, plasmids including ER-Hoxb8 and the EcoPac gag-pol-env plasmid (both are kind gifts from Mark P. Kamps, University of California, San Diego) were mixed at 1:1 ratio (12 μg:12 μg for 10 cm plate) for transfection into HEK293T following the same approach as for generating lentiviral particles.

To transduce Hox or B16 cells with LentiGuide-Puro-P2A-EGFP_mRFPstuf, the multiplicity of infection (MOI) of the viral particles was tested by infection with serial dilutions of virus particles, to find a dilution resulting in a suitable MOI described below. Virus supernatant was added to each well of 6-well plate containing cells (4×105 for Hox cells with SCF and BE, and 1×105 for B16 cells) with 8 μg/ml polybrene (Sigma-Aldrich #H9268). The plate was centrifuged at 37°C, 1200g (120 min for Hox cells, and 45 min for B16 cells). After 24h, the virus-containing medium was replaced with fresh medium, and the infection rate was measured by the percentage of GFP+ cells if the vector contains GFP. Puromycin (Invivogen #ant-pr) selection (10 μg/ml for HoxB8 cells, and 5ug/ml for B16 cells) or Blasticidin (Invivogen #ant-bl) selection (10 μg/ml for B16 cells) was performed for 24 h to remove the non-infected cells.

### **Electroporation and transfection**

sgRNAs were designed using the Green Listed software**3**,**4** using sgRNA design from the Doench mouse library**5**. 2’-O-methyl and phosphorothioate stabilized sgRNAs (**Supplementary Table 8**) were ordered from Sigma-Aldrich. Off-target activity was calculated using Cas-OFFinder (http://www.rgenome.net/cas-offinder/)**6**, and On-target activity was extracted from CHOPCHOP (https://chopchop.cbu.uib.no/)**7** by searching for Ccr1, or entering the EGFP FASTA sequence from the LentiGuide-Puro-P2A-EGFP_mRFPstuf (Addgene #137730).

For Hox cells, Neon Transfection System (Invitrogen #MPK5000) was used to perform the electroporation following the manufacturer’s instructions (Pulse voltage: 1700, Pulse width: 20 ms, Pulse number: 1, for Hox cells). 100 pmol of sgRNA were electroporated into 2×105 Hox cells for each electroporation experiment using Neon Transfection System 10 μL Kit (Invitrogen #MPK1096). For B16 cells, Lipofectamine 2000 Transfection Reagent (Invitrogen, #11668019) was used following the recommended protocol. 100pmol of sgRNA were transfected into 1×105 B16 cells for each transfection experiment.

Trp53 ON-TARGETplus siRNA SMARTPool was ordered from Horizon. For Hox cells, 100 pmol of siRNA was electroporated into 2×105 Hox cells for each electroporation experiment. For B16 cells, 100 pmol of siRNA was transfected into 1×105 B16 cells for each transfection experiment. The Trp53 siRNA was typically delivered in the same reaction as the sgRNAs.

### **CRISPR KO genotyping**

1×105 cells were collected for genomic DNA extraction using DNeasy Blood & Tissue Kit (Qiagen #69504) following the recommended protocol. Primers (Sigma-Aldrich) were designed using Primer-BLAST (**Supplementary Table 8**), aiming for a 400-1000 bp amplicon with the sgRNA target in the middle. Amplicons were gel purified and recovered using Zymoclean Gel DNA Recovery Kit (Zymo Research #D4007/D4008). The PCR products were quantified using Nanodrop and sequenced by Eurofins Genomics. The Sanger sequencing data was subsequently analyzed by ICE (Synthego, https://ice.synthego.com).

### **Growth curve characterization**

Hox cells were cultured with the following interventions: electroporation with a GFP targeting sgRNA; 0.5 μg/ml Etoposide (Sigma-Aldrich #E1383); 3 μM AMG232 (Axon Medchem #2639). The Etoposide and AMG232 were removed after 8 h by removing the medium and adding fresh medium without Etoposide or AMG232. The cells were cultured in 6 well plates with 4ml medium and passaged every day at the ratio of 1:3. 80 ul of cells were taken every day for flow cytometry (BD Accuri) with the existence of CountBright Absolute Counting Beads (Invitrogen #C36950). The absolute viable cell number was calculated by comparing the events number of viable cells and counting beads in the flow cytometry data.

### **Real-Time PCR**

TRIzol Reagent (Invitrogen #15596026) and Direct-zol RNA MiniPrep Kit (Zymo Research #R2051) was used to extract RNA. RNA was then converted into cDNA using High Capacity RNA-to cDNA kit (Applied Biosystem #4388950). The expression of Cdkn1a was quantified with a CFX 384 Real-Time PCR machine (Bio-Rad) using TaqMan gene expression FAM assays for Cdkn1a (Mm00432448_m1) with the TaqMan Gene Expression Master Mix (Applied Biosystem #4369542) as suggested by the manufacturer. Expression was normalized by TaqMan gene expression VIC assays for β-actin (Mm00607939) and Gene expression was quantified using the ddCT method.

### **Apoptosis TUNEL assay**

Cells were collected and fixed by PFA at different time point, and FlowTAC Apoptosis Detection Kit (R&D Systems #4817-60-K) was used to stain apoptotic cells following the manufacturer’s instructions. Stained cells were analyzed by flow cytometry (BD Accuri).

### **Cloning of sgRNAs into lentiviral transfer plasmid, and CRISPR screens**

sgRNAs with overhangs for the plasmids (**Supplementary Table 1**) were designed using the Green Listed software**3**,**4** using sgRNA design from the Doench mouse library**5** and, for intergenic controls, the Wang mouse library**8**. Individual sgRNAs were ordered from Sigma-Aldrich, and the sgRNA library was ordered from CustomArray as a DNA oligo pool. Cloning was performed using 150 ng of BsmBI (New England Biolabs #R0739) cleaved lentiGuide-Puro-P2A-EGFP_mRFPstuf plasmid (Addgene #137730), and 10 ng of the library oligo pool, using NEBuilder HiFi DNA assembly master mix (New England Biolabs #E2621). Endura ElectroCompetent Cells (Lucigen #60242), were subsequently transformed with the cloned plasmid pool using electroporation (1.0 mm cuvette, 10 μF, 600 Ohms, 1800 Volts) following the suggested protocol. The electroporated cells were combined and seeded on ten 20 cm LB agar plates with 100 μg/ml carbenicillin and grown at 37°C overnight. The lentiGuide-Puro-P2A-EGFP_mRFPstuf plasmid (Addgene #137730) makes bacteria red if the stuffer has not been exchanged by a sgRNA, and a few red clones were removed before collecting all other white clones. Then the plasmids were purified using EndoFree Plasmid Maxi Kit (Qiagen #12362).

The sgRNA cloned lentiGuide-Puro-P2A-EGFP_mRFPstuf was used as transfer plasmid for lentiviral preparation and transduction. The total amount of transduced cells was calculated based on MOI (0.25 for B16 cells, and 0.05-0.1 for Hox cells, which were difficult to transduce), based on the % GFP+ cells before puromycin selection, aiming for 1000 transduced cells for each sgRNA.

The CRISPR library transduced cells were exposed to GFP targeting sgRNA electroporation with or without Trp53 siRNA, 0.5 μg/ml Etoposide (Sigma-Aldrich #E1383) 8 h pulse stimulation, or 3 μM AMG232 (Axon Medchem #2639) 8 h pulse stimulation.

Cells were collected for genomic DNA extraction using DNeasy Blood & Tissue Kit (Qiagen #69504). Genomic DNA was then amplified using Q5 High-Fidelity DNA Polymerase (New England Biolabs #M0491) while introduced sample-specific barcodes and adapters for Illumina Sequencing similar as described in Joung J et al.**9** using primers specified in **Supplementary Table 9**. The final PCR products were gel purified and recovered using Zymoclean Gel DNA Recovery Kit (Zymo Research #D4007/D4008), and quantified with Qubit 4 Fluorometer (Invitrogen #Q33238) using Qubit dsDNA HS Assay Kit (Invitrogen #Q32851) and pooled for next-generation sequencing (Illumina MiSeq v3 run, 2×75bp reads). The raw FASTQ data were analyzed by MAGeCK**10**. Read counts from CRISPR screens found in **Supplementary Table 10-11**.

### **JAK1/STAT1 signaling assay**

Hox cells were cultured with or without mouse Interferon Beta (IFNβ, Nordic BioSite #12405-1) and the Jak1 inhibitor Solcitinib (Selleckchem #S5917) for 7 days. Cells were passaged at the ratio of 1:8 every other day. 100 ul of cells were taken on day 7 for flow cytometry (BD Accuri) with the existence of CountBright Absolute Counting Beads (Invitrogen #C36950). The absolute viable cell number was calculated by comparing the events number of viable cells and counting beads in the flow cytometry data.

### **Competitive co-culture assay**

Trp53 KO and WT cells were mixed at 1:4 ratio, and were electroporated with sgRNA or transduced with lentivirus (lentiGuide-Puro-P2A-EGFP_mRFPstuf) and cultured for 7 days; or cultured with Etoposide (Sigma-Aldrich, #E1383, 0.05 μg/ml) or AMG232 (Axon Medchem #2639, 0.5 μM for Hox cells, 4 μM for B16 cells) for 7 days; or cultured with Cobalt(II) chloride (CoCl2, Sigma-Aldrich #232696, 10-20 μg/ml) for 7 days. Different p53 related inhibitors were added during culture: Trp53 ON-TARGETplus siRNA SMARTPool (Horizon), KU55933 (Sigma-Aldrich #SML1109, 100 ng/ml for B16, 10 ng/ml for Hox), VE821 (Sigma-Aldrich #SML1415, 50 ng/ml), Pifithrin-μ (Sigma-Aldrich #P0122, 2 μg/ml for B16, 200 ng/ml for Hox), Cyclic Pifithrin-α (Sigma-Aldrich #P4236, 700 ng/ml for B16, 70 ng/ml for Hox), C646 (Sigma-Aldrich, #SML0002, 1 μg/ml), AZD2461 (Selleckchem #S7029, 250 ng/ml), LJI308 (Sigma-Aldrich #SML1788, 2 ng/ml), Z-VAD-FMK (Selleckchem #S7023, 50 μM). siRNA was delivered to cells 1 day before CRISPR/Etoposide/AMG232 exposure (100 pmol of siRNA electroporated into 2×105 Hox cells, or transfected into 1×105 B16 cells), or together with sgRNA for the transfection/electroporation groups. Other inhibitors were added to cell culture media 1 day before CRISPR/Etoposide/AMG232 exposure and cultured for 7 days. Cells were then collected for Trp53 KO genotyping. **Supplementary Table 7** further describes how stock solutions of inhibitors were generated and stored.

### **Flow Cytometry Analysis**

Fresh bone marrow cells from C57BL/6 Cas9+ GFP+ mice and Hox cells were stained with the following antibodies: FITC Rat anti-Mouse CD34 (BD Biosciences #553733, 1:500), PE anti-mouse CD150 (BioLegend #115904, 1:200), PerCP/Cyanine5.5 anti-mouse Ly-6A/E (BioLegend #122524, 1:200), APC anti-mouse CD117 (BioLegend #105812, 1:500), APC/Cyanine7 anti-mouse CD16/32 (BioLegend #156612, 1:500), Biotin anti-mouse Lineage Panel (BioLegend #133307, 1:100), BV421 Streptavidin (BD Biosciences #563259, 1:1000), LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Invitrogen #L34957, 1:2000). After 30 min of staining, the cells were washed and analyzed by flow cytometry (BD FACSVerse). FACS FCS files were analyzed by FlowJo version 10 (FlowJo, LLC).

### **Analysis of data from the Depmap portal**

sgRNA enrichment (CRISPR (Avana) Public 20Q4 release), mutation profile (Mutation Public 20Q4 release), drug sensitivity (PRISM Repurposing Primary Screen 19Q4 release), and mRNA expression levels (Expression Public 20Q4 release) was extracted December 13th 2020 from the Depmap portal (https://depmap.org/portal/)**11**–**13**. Connectivity maps were generated using the geneMANIA plugin for Cytoscape**14**,**15**. tSNE plots were made with the Rtsne package (https://github.com/jkrijthe/Rtsne) to analyze the cluster and ggplot2 (https://github.com/tidyverse/ggplot2) to visualize the data, or tSNE-online (https://github.com/jefworks/tsne-online). The “ENCODE and ChEA Consensus TFs from ChIP-X” functionality of Enrichr (https://maayanlab.cloud/Enrichr/index.html)**16**,**17** was used to identify transcription factor binding to gene sets.

CRISPR-Cas systems present an evolutionary tradeoff: does defense against phages and other parasitic DNA also prevent cells from acquiring potentially helpful new genes? Genomic analyses of this conundrum have arrived at often contradictory conclusions. Meanwhile, experimental studies have focused mainly on phages, conjugation, or artificial transformation, but less work has examined natural competence, a major driver of evolution and antibiotic resistance. Here, we use _Acinetobacter baylyi_, which combines high natural competence with a functional CRISPR-Cas system, to experimentally probe the interactions between CRISPR-Cas and natural competence. In these bacteria, the endogenous CRISPR array largely allows natural transformation by targeted DNA. However, CRISPR-Cas then kills the newly autoimmune cells in a form of programmed cell death. CRISPR-Cas often allows self-targeting cells to form colonies, albeit with fitness costs. Thus CRISPR-Cas appears to block natural transformation in a process more akin to altruistic group defense than an individual immune system.

CRISPR-Cas Inhibits Natural Transformation Through Altruistic Group Defense and Self-Sacrifice

### **Cell culture and cloning**

_A. baylyi_ strain ADP1 was obtained from the American Type Culture Collection (ATCC #33305). Cells were propagated in standard LB media at 30 or 37 °C. DNA constructs were prepared using Gibson Assembly or Golden Gate cloning methods to generate linear DNA, which was used to transfom _A. baylyi_ directly, by incubating 20-100 μl cells plus DNA for 30 minutes to a few hours. Clones were picked from selective plates, genomic DNA was purified using a genomic DNA miniprep kit, regions of interest were amplified by PCR, and the transformants were verified by Sanger sequencing. For _PrecA_-_GFP_, clones were screened for low-level GFP fluorescence by placing the plate on a blue transilluminator, rather than antibiotic selection. Multiplex CRISPR arrays were assembled using a method described previously **19**.

Three genomic loci were used in the experiments described here. One was the natural CRISPR array located between ACIAD2484 and ACIAD2500. A second was the gene remnant ACIAD2826, which was replaced as a neutral locus for insertions, denoted _ntrl1_. A third was a neutral locus used previously, replacing ACIAD2186, 2187, and part of 2185, into which the ectopic CRISPR arrays were inserted**40**. For ectopic CRISPR arrays, the inserts included 684 bp upstream of the first repeat from the endogenous array. For _PrecA_-_GFP_, the insert included 214 bp of promoter from upstream of the _recA_ gene, as described previously**23**.

### **Transformation experiments**

_A. baylyi_ was grown overnight in LB broth at 30°C with shaking. Just before experiments, cultures were washed and resuspended in fresh LB. 20 μl of cells were mixed with 40 ng of purified DNA and incubated at 30°C with shaking for 2 hours. Samples were then serially 10-fold diluted and spotted on LB agar plates with and without selective antibiotics (35 μg/ml kanamycin or spectinomycin), with 3 technical replicates for each dilution level. Plates were incubated at 30°C overnight, and colonies were counted at the lowest dilution level that still had clearly separated colonies. Counts for the three technical replicate spots were averaged. Each experiment was repeated at least 3 times, on at least 2 separate days.

For pulse-chase experiments, cells were growth overnight at 30°C, washed and diluted 1/13x into fresh LB, and re-grown for 3 hours at 37°C. The cells were then washed into fresh LB again, divided into 20 μl aliquots, mixed with 40 ng DNA, and incubated at 37°C. After 30 minutes, DNase and DNase buffer were added, and incubation continued for another 15 minutes at 37°C to digest remaining donor DNA. The cultures were then diluted into 600 μl fresh LB, and incubated at 37°C, with aliquots taken for serial dilution and spotting after 15 minutes and 3-4 hours.

Relative fitness was calculated using the equation   , where _A_ is cells transformed by DNA containing a given target, _B_ is cells transformed by control DNA without a protospacer, and the subscripts indicate time points. Killing rates were calculated as   , where _γ_ is the growth rate of transformants with non-targeted control DNA without a protospacer, and _n_ is the number of transformants at each time point.

### **Statistics**

Serial dilutions produce data best compared and plotted on a log scale, so data were plotted and compared after taking base 10 logarithms. Hypothesis testing was performed in Matlab using single-tailed t-tests. Where multiple samples were tested at once, the standard significance threshold of p<.05 was divided by the number of samples (Bonferroni correction for multiple hypothesis testing). Where data points were below detection, they were manually set to the limit of detection as the most conservative way to include them in the statistical analysis.

### **Visualizing DNA damage**

The construct _PrecA_-_GFP_ was prepared containing 214 bp of the _A. baylyi recA_ promoter fused to GFP**23**, with approximately 1 kb homology arms surrounding a _vgrG_ operon, and transformed into _A. baylyi._ This is a Type VI secretion system toxin-antitoxin operon, so in the absence of cells with the toxin, this insertion is neutral. In addition, cells contained a spectinomycin resistance gene in another neutral locus. Donor DNA for visualization experiments consisted of 3 kb covering that spectinomycin resistance gene and surrounding DNA, with a 32 bp protospacer inserted before the spectinomycin gene. Thus, the donor DNA had a 32 bp insert embedded within 3 kb of total homology to recipient cells. Cells were grown overnight in LB with shaking at 30 °C, washed, diluted 1:20 into fresh LB, and mixed 1:1 with donor DNA. Agar pads were prepared by pouring molten LB containing 2% wt/vol agar into 35 mm glass-bottom dishes, allowed to solidify, scooped out with a spatula, and placed upside-down. Spots of 0.5 μl of the cell-DNA mixtures were deposited onto the raised pads created from the glass window cutout and allowed to dry, after which the pad was placed back into the glass-bottom dish. Each experimental run contained 6 spots, including a no-DNA control. Three experiments were run for each CRISPR target and two for each plasmid. An additional 5 spots were imaged for protospacer 1 to obtain higher statistical power. A total of 9 no-DNA control spots were imaged, one for each experimental run. Experiments using plasmid DNA also included integrating DNA spots as positive controls. Time-lapse movies were obtained on a Nikon TE microscope using a 20x objective, with a 30-minute time step. Time points for analysis were selected at the point when growth significantly slowed, generally after 36 hours. GFP images were analyzed using the following procedure:

- Preprocessing: Images were preprocessed by dividing by a blank background image, which was itself first smoothed with a Gaussian filter with a radius of 5 pixels.
- Excluding empty areas: Regions with no cells were outlined manually, refined using the Image Segmenter app in Matlab, and excluded from further analysis.
- Background autofluorescence: To compensate for increased background fluorescence in the centers of microcolonies, an image-specific background was constructed. First, a Gaussian curve was constructed using the mean and variance of the GFP pixel histogram. Pixels greater than the 99th percentile of that curve were suppressed down to the 99th percentile level. The resulting image was smoothed with a Gaussian filter of radius 20 pixels to create an autofluorescence background. This was subtracted from the pre-processed GFP image, yielding the final background-corrected image.
- Histogram normalization: A histogram of GFP pixel intensity was constructed, and a new Gaussian curve was fit to the lower 60% of pixels, representing an expected distribution in the absence of GFP expression. The histogram was shifted laterally to be centered at 0, its width was normalized to σ = 1, and its height was normalized such that the total area under the curve was 1, to allow visual histogram overlays (Supplemental Fig. S5).
- Bright pixels: Bright pixels were defined as those greater than the 99.5 percentile of the aligned Gaussian.

RNA-guided nucleases from clustered regularly interspaced short palindromic repeats (CRISPR) systems expand opportunities for precise, targeted genome modification. Endogenous CRISPR systems in many bacteria and archaea are particularly attractive to circumvent expression, functionality, and unintended activity hurdles posed by heterologous CRISPR effectors. However, each CRISPR system recognizes a unique set of PAM sequences, which requires extensive screening of randomized DNA libraries. This challenge makes it difficult to develop endogenous CRISPR systems, especially in organisms that are slow-growing or have transformation idiosyncrasies. To address this limitation, we present Spacer2PAM, an easy-to-use, easy-to-interpret R package built to identify potential PAM sequences for any CRISPR system given its corresponding CRISPR array as input. Spacer2PAM can be used in “Quick” mode to generate a single PAM prediction that is likely to be functional or in “Comprehensive” mode to inform targeted, unpooled PAM libraries small enough to screen in difficult to transform organisms. We demonstrate Spacer2PAM by predicting PAM sequences for industrially relevant organisms and experimentally identifying seven PAM sequences that mediate interference from the Spacer2PAM-predicted PAM library for the type I-B CRISPR system from _Clostridium autoethanogenum_. We anticipate that Spacer2PAM will facilitate the use of endogenous CRISPR systems for industrial biotechnology and synthetic biology.

Spacer2PAM: A computational framework for identification of functional PAM sequences for endogenous CRISPR systems

### **Prediction of PAM Sequences**

All CRISPR arrays were retrieved from CRISPRCasdb, part of CRISPR-Cas++, which can be found at https://crisprcas.i2bc.paris-saclay.fr/ (**28**). Alignment of CRISPR spacers to genomes was done via the NCBI BLAST web interface(**29**) using the BLASTn algorithm excluding Eukaryotes (taxid:2759) as well as the organism that encodes the CRISPR system. All other manipulations of sequence information and prediction of PAM sequences were completed using Spacer2PAM which is available at [INSERT URL TO GITHUB ONCE PUBLIC]. Spacer2PAM requires the following dependencies: dplyr, ggplot2, ggseqlogo(**30**), taxonomizr, HelpersMG, httr, jsonlite, spatstat.utils, and seqinr. Prophage prediction uses the Phaster API(**31**). More information about Spacer2PAM can be found in the program documentation.

### **Plasmid Construction**

All individual plasmids and libraries in this work were generated by two-piece Gibson assembly using the GeneArt Seamless Plus kit. Linear backbone was generated by PCR of pMTL82254 using Kapa DNA polymerase Master Mix and purification by gel electrophoresis and extraction with Zymoclean Gel DNA recovery Kit. Linear dsDNA gBlocks ordered from IDT containing the PAM sequence upstream of _C. autoethanogenum_ CRISPR array 1 spacer 19 were used as inserts. Gibson assembly products were transformed into chemically competent One Shot™ MAX Efficiency™ DH10B T1 Phage-Resistant Cells using standard procedures. DNA sequence was confirmed by Illumina MiSeq Sequencing V2 and V3 chemistry.

### **Spacer2PAM-informed PAM Prediction Screening**

Spacer2PAM was applied to the type I-B CRISPR system of C. autoethanogenum using the comprehensive method. The top 25% of high scoring PAM predictions were used to determine a set of 16 four nucleotide PAM sequences that are likely to be functional. The Spacer2PAM-informed, unpooled PAM library constructs were transformed into _E. coli_ HB101 carrying R702(**32**) (CA434(**33**)) in parallel. Conjugation of library members into _C. autoethanogenum_ DSM 19630, a derivate of type strain DSM 10061, was performed as described earlier(**33**, **34**) using erythromycin (250 µg/mL) and clarithromycin (5 µg/mL) for plasmid selection in _E. coli_ and _C. autoethanogenum_, respectively, and trimethoprim (10 µg/mL) as counter selection against _E. coli_ CA434. Optical density of donor _E. coli_ cultures were measured prior to addition to _C. autoethanogenum_ cells. Transconjugant colonies were counted following 4 days of incubation at 37°C under 1.7 × 105 Pa gas (55% CO, 10% N2, 30% CO2, and 5% H2) in gas-tight jars. This was performed in biological triplicate, with 3 separate cultures of donor _E. coli_ conjugated to aliquots of a single _C. autoethanogenum_ culture.

### **Randomized PAM Library Screening**

The randomized, pooled PAM library was transformed into NEBExpress® _E. coli_ and then purified by QIAprep Spin Miniprep Kit. An aliquot of this DNA was saved to determine PAM frequencies before exposure to the CRISPR system. Electroporation into _C. autoethanogenum_ was performed as described previously(**35**, **36**). Following recovery, cells were pelleted by centrifugation at 4000 X g for 10 minutes, 9.5 mL of supernatant was discarded, and cells were resuspended in 500 µL YTF. Resuspensions were split by volume and spread on YTF 1.5% agar supplemented with 5 µG/mL clarithromycin, allowed to dry for ∼30 minutes, and incubated at 37°C for 4 days under 1.7 × 105 Pa gas (55% CO, 10% N2, 30% CO2, and 5% H2) in gas-tight jars. 2.5 mL of Luria broth was added to each plate and plates were scraped. Total DNA from the cell suspension was purified using the MasterPure™ Gram Positive DNA Purification Kit. PCR across the PAM and spacer was performed using Kapa DNA polymerase Master Mix followed by purification by gel electrophoresis (1.5% agarose) and extraction with Zymoclean Gel DNA recovery Kit. Extracts were quantified by Quant-iT (Thermo Fisher Scientific), diluted to 1 ng/uL, and prepared for sequencing following the Illumina 16S amplicon protocol starting at the Index PCR step https://support.illumina.com/documents/documentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf. Ampure XP purified libraries were quantified by Quant-iT and sequenced using MiSeq Reagent Kit V3. Frequency of each PAM was determined by counting the occurrence of each PAM next to a correct protospacer sequence within the read. Briefly, all sequence reads are searched for the presence of the C. autoethanogenum Array 1 spacer 19 sequence and are binned as a forward read, reverse read, or does not contain the spacer. For all reads in the forward and reverse bins, the immediate 4 nucleotides upstream or downstream, respectively, are extracted. The sequences extracted from reverse reads are converted to their reverse complement to be compatible with the sequences extracted from forward reads and the two sets of sequences are combined. The frequency of each 4-nucleotide sequence in the combined list is then counted and recorded. The frequency of each PAM was converted to a relative frequency within the total library and the log2-fold change in relative frequency was calculated from exposure to the CRISPR system.

CRISPR-Cas systems are adaptive immune systems commonly found in prokaryotes that provide sequence-specific defense against invading mobile genetic elements (MGEs). The memory of these immunological encounters are stored in CRISPR arrays, where spacer sequences record the identity and history of past invaders. Analyzing such CRISPR arrays provide insights into the dynamics of CRISPR-Cas systems and the adaptation of their host bacteria to rapidly changing environments such as the human gut. In this study, we utilized 601 _Bacteroides fragilis_ genome isolates from 12 healthy individuals, 6 of which include longitudinal observations, and 222 available _B. fragilis_ reference genomes to update the understanding of _B. fragilis_ CRISPR-Cas dynamics and their differential activities. Analysis of longitudinal genomic data showed that some CRISPR array structures remained relatively stable over time whereas others involved radical spacer acquisition during some periods, and diverse CRISPR arrays (associated with multiple isolates) co-existed in the same individuals with some persisted over time. Furthermore, features of CRISPR adaptation, evolution, and microdynamics were highlighted through an analysis of host-MGE network, such as modules of multiple MGEs and hosts, reflecting complex interactions between _B. fragilis_ and its invaders mediated through the CRISPR-Cas systems. This work demonstrates the power of using culture-based population genomics to reveal the activities and evolution of the CRISPR-Cas systems associated with gut bacteria in human population. We made available of all annotated CRISPR-Cas systems and their target MGEs, and their interaction network as a web resource at https://omics.informatics.indiana.edu/CRISPRone/Bfragilis.

Diversity and dynamics of the CRISPR-Cas systems associated with Bacteroides fragilis in human population

### **Genomic data processing and assembly**

Reads from 601 _B. fragilis_ isolates from the Zhao et al. study [**40**] were downloaded from the NCBI BioProject Accession PRJNA524913, henceforth referred to as the ‘Zhao2019 dataset’. Raw shotgun sequencing reads were trimmed using Trimmomatic v0.39 [**41**] (parameters used: LEADING:5 TRAILING:5 SLIDINGWINDOW:4:10 MINLEN:60). Trimmed reads were then assembled using SPAdes v3.12 [**42**] with default settings. FragGeneScan [**43**] was then used to predict protein coding genes of metagenome assemblies.

A total of 222 _B. fragilis_ reference genomes, 16 complete and 202 draft genomes, were downloaded from the NCBI ftp website as of Jan 18, 2021. A list of genomes included in this analysis can be found at the CRISPRone website http://omics.informatics.indiana.edu/CRISPRone/Bfragilis.

### **Characterization of CRISPR-Cas systems**

To identify CRISPR-Cas systems in _B. fragilis_ genomes, we utilized CRISPRone [**44**] which predicts both CRISPR arrays and _cas_ genes within a given input genome sequence. Predicted CRISPR-Cas systems were then further refined through a reference based approach. Repeat sequences of CRISPRone predicted spacers were extracted and clustered to obtain consensus reference repeats using CD-HIT-EST [**45**] with 85% sequence identity. Consensus reference repeats were then used as input for CRISPRAlign [**46**], a reference based approach to identify CRISPR arrays. As the exact boundaries of CRISPR arrays predicted by _de novo_ approaches may sometimes be blurred due to small CRISPR arrays, sequencing errors, and mutations in repeat sequences, we utilize a reference based approach to redefine the repeat-spacer boundaries of CRISPR arrays predicted by CRISPRone.

To compare spacer sequences across different arrays, reduce spacer redundancy, and the eventual computation of spacer content heterogeneity, spacers were clustered with CD-HIT-EST [**45**] at 85% sequence identity. An 85% sequence identity was used to provide greater flexibility in spacer sequences, and allow for a small amount of sequence variation either due to sequencing error or real mutations found between individual spacers. Spacer sequences that clustered together were considered identical spacer sequences. Spacer clusters were reserved for downstream computation of spacer content heterogeneity (**Figure 1A**) and construction of compressed spacer graphs (**Figure 1B**).

- **Download figure**
- **Open in new tab**

**Figure 1:**

Approaches used for the identification and refinement of the CRISPR arrays and construction of spacer graphs. (A) CRISPR arrays are analyzed in groups such that each group shares identical or very similar repeats (repeats are shown as diamonds and spacers are shown as boxes). CRISPR arrays containing no spacer variance or no CRISPR spacer hetergeneity are discarded. (B) Example of spacer sharing CRISPR arrays can be represented as a simplified graphical structure (spacer graph), in which the edges record the ordering of the spacers in arrays.

In some cases, it may be difficult to differentiate between true CRISPR-Cas systems and false positive CRISPR-Cas systems (e.g., false CRISPR-arrays, inactive CRISPR-Cas systems). While manual curation can help filter out some of these issues, it becomes difficult to screen out hundreds to thousands of genomes. To help filter out false positive arrays and inactive CRISPR-Cas systems, we propose a metric of heterogeneity to measure the rate of change (i.e., growth and turnover of spacers) in CRISPR arrays with the assumption that CRISPR arrays of active CRISPR-Cas systems undergo active expansion and turnover of spacers. Here we define spacer content heterogeneity score as:  Where _n_ is defined as the number CRISPR arrays, with each CRISPR array containing _c_1, _c_2,…, _cn_ unique spacers (in some rare cases, CRISPR arrays may contain multiple copies of the same spacer, which will be considered as one spacer) and _m_ denotes the number of unique spacers found from all _n_ arrays combined. Spacer heterogeneity scores range from 0 to 1, where 0 indicates no spacer heterogeneity (i.e., two CRISPR arrays share all spacers), and 1 indicates the greatest possible extent of spacer content heterogeneity (i.e. two CRISPR arrays share no spacers).

Because spacer content heterogeneity alone is not enough to rule out false positive or inactive CRISPR-Cas systems, predicted CRISPR-Cas systems were further filtered out by coupling spacer content heterogeneity with gene content information. CRISPR groups that lack spacer content heterogeneity and had no adjacent _cas_ genes were considered inactive or false positive, and thus discarded from further analysis; all filtered arrays were also manually inspected prior to their removal.

### **Compressed spacer graph for summarizing the sharing of spacers among a group of CRISPR arrays**

Compressed spacer graphs [**47**] were constructed for each CRISPR-Cas type to summarize and illustrate CRISPR array dynamics. For every spacer in a given array, where each spacer was represented by a node of its representative spacer cluster, a directed edge was built between nodes of neighboring spacers in sequential order. Once all CRISPR arrays were represented in the graph structure, the spacer graph was then simplified by collapsing neighboring nodes if two neighboring nodes shared an “in-degree” and “out-degree” equal to or less than one (**Figure 1B**). Compressed spacer graphs highlight CRISPR array structure and dynamics (e.g. branching structures representing spacer gain and loss). Arrays that share no spacers result in disconnected components in the compressed spacer graph.

### **Mobile Genetic Element Databases**

A collection of mobile genetic element (MGE) databases were gathered, including phage and plasmid databases. The phage databases included the Gut Phage Database [**48**] (GPD), MicrobeVersusPhage [**49**] (MVP) database, and the reference viral database [**50**] (RVDB). The plasmid databases included the Comprehensive and Complete Plasmid Database [**51**] (COMPASS), and PLSDB [**52**]. The phage and plasmid databases included sequences from the NCBI reference database, NCBI nucleotide database, MGEs identified from metagenomic assemblies, and prophages identified in prokaryotic genomes. We collectively refer to these databases as the ‘MGE database’ for simplicity.

### **Identification of CRISPR Targets**

All unique spacer sequences extracted from _B. fragilis_’ CRISPR arrays were queried against the MGE database using BLASTN [**53**] to search for putative invaders that were targeted by _B. fragilis_. For this analysis, we used all unique spacers instead of 85%-similarity nonredundant set to increase the search sensitivity. Results were filtered to retain hits with a greater than 90% sequence identity, query coverage per hsp greater than 80%, and an e-value of less than 0.001. We noticed that even after dereplication by dRep [**54**] (with default parameters), there was still a large redundancy in the identified MGEs. Instead, we devised a greedy algorithm to select the minimum number of MGEs that collectively contain all protospacers matching the spacers. Similarly, we selected the minimum number of _B. fragilis_ isolates that contained all identified spacers and only included them in the network. Selected MGEs and isolates are then used for building spacer-MGE and host-MGE networks. In the spacer-MGE network, spacer sequence clusters (called spacers for simplicity) and MGEs are represented as nodes and an edge is added between a spacer node and MGE node if the MGE contains a segment that matches the spacer (i.e., protospacer). In the host-MGE network, an edge is added to a host and a MGE if the host and MGE pair contain at least one matching protospacer and spacer. For MGEs that are phages (or prophages), we applied PhaGCN [**55**] to assign their taxonomic groups (ICTV [**56**] families). All visualizations and manual inspection of the networks were performed using Cytoscape [**57**].

CRISPR-Cas are adaptive immune systems that protect their hosts against viruses and other parasitic mobile genetic elements. Consequently, selection from viruses and other genetic parasites is often assumed to drive the acquisition and maintenance of these immune systems in nature, but this remains untested. Here, we analyse the abundance of CRISPR arrays in natural environments using metagenomic datasets from 332 terrestrial, aquatic and host-associated ecosystems. For each metagenome we quantified viral abundance and levels of viral community diversity to test whether these variables can explain variation in CRISPR-Cas abundance across ecosystems. We find a strong positive correlation between CRISPR-Cas abundance and viral abundance. In addition, when controlling for differences in viral abundance, we found that the CRISPR-Cas systems are more abundant when viral diversity is low. We also found differences in relative CRISPR-Cas abundance among environments, with environmental classification explaining ∼24% of variation in CRISPR-Cas abundance. However, the correlations with viral abundance and diversity are broadly consistent across diverse natural environments. These results indicate that viral abundance and diversity are major ecological factors that drive the selection and maintenance of CRISPR-Cas in microbial ecosystems.

High viral abundance and low diversity are associated with increased CRISPR-Cas prevalence across microbial ecosystems

### **Computational pipeline**

An overview of the computational pipeline is provided in **Figure S1**.

### **Collection of samples and assembly**

Paired-end metagenomic libraries (sequenced with Illumina platform) were downloaded from the NCBI SRA database (**Table S1**). Libraries were processed with BBMap suite of tools for error correction [**12**] and for each metagenomic library a representative FASTA file was created by combining both the merged and unmerged reads. A total of 332 libraries (each library containing at least 1 million reads with a minimum length of 100 nucleotides (nt) and insert size >=150) were selected to represent a wide range of biome diversity (**table S1**). Libraries were assembled using MegaHit (version 1.1.3) [**13**] with default parameters and contigs with minimum length 200nt were retained. Contigs were classified as archaea, bacteria or virus using a purpose-built classification tool: MIUMS (Microbial Identification Using Marker Sequence, https://github.com/ambarishbiswas/miums_v1.0). MIUMS is designed to classify contigs based on a reference database containing protein sequence fragments highly specific to bacteria, archaea, viruses. Full details of MIUMS reference database construction and prediction process are given in the supplementary information. Each metagenome was then subsampled to 1 million randomly selected reads with a minimum length of 100 nt.

### **Generation of archaeal and bacterial abundance tables**

Subsampled reads were screened using metaxa2 [**14**] (GSU parameters: -g ssu -f f, LSU parameters: -g lsu -f f) to generate a table of bacterial and archaeal abundances. Both the LSU and SSU based methods were used. A reference sequence database was made from contigs classified as viral by MIUMS. Subsampled reads were then mapped to the assembled contigs using Magic-BLAST [**15**] (parameters: -no_unaligned -no_query_id_trim -perc_identity 95 -outfmt tabular). Reads with a minimum of 95% sequence identity and coverage were used.

### **CRISPR array identification**

Accurately identifying CRISPR arrays in metagenomic data is challenging for a number of reasons. Firstly, a large proportion of the direct repeats (DRs) identified from metagenomic contigs often show little sequence similarity to the CRISPR repeats found in published genomes and lack an isolated representative [**2**]. Secondly, CRISPR arrays found in metagenomic reads are generally short (i.e. < 3 DRs) and often missing one or both flanking regions. To overcome these issues we combined information on existing, published genomes and their CRISPR arrays as well _de novo_ extraction of putative CRISPR arrays from our assembled contigs (**Fig. S1**). A database of metagenomic CRISPR arrays was first constructed by processing all assembled contigs with CRISPRDetect version 3 (CRISPRDetect3, https://github.com/ambarishbiswas/CRISPRDetect_3.0). CRISPRDetect version 3 was also modified to allow prediction of shorter CRISPRs (e.g. partial/broken CRISPRs with as little as 1.5 repeats). A higher CRISPR likelihood score cut-off of 4.5 was used instead of the default score cut-off of 3 to reduce potential non-CRISPR arrays. CRISPRDetect [**16**] uses several CRISPR elements (e.g. repeats, spacers, cas genes, AT composition of flanking regions etc.) from published genomes to identify and separate true CRISPRs from other genomic repeats. In this study, a modified version of the CRISPRDetect tool was used, which uses a reference repeat database created using the cluster representative DRs from the metagenomic contigs as well as DRs found in published genomes. Predicted CRISPR arrays were checked to ensure that the total array degeneracy (i.e. number of insertion, deletion, mutation or presence of Ns in the array) was less than the total number of DRs in the array, which resulted in 51395 CRISPR arrays. Direct repeat sequences (23 to 60 nt) were extracted and clustered with cd-hit-est (parameters: -n 3 -c 0.90 -aL 0.90 -aS 0.90) [**17**], which resulted in 33745 repeat clusters. A vast majority of these clusters (i.e. 22808) consisted of a single DR. This low level of redundancy was important to ensure the successful subsequent mapping of reads to DRs.

Subsampled reads were then screened against this database using metaCRISPRDetect (https://github.com/ambarishbiswas/metaCRISPRDetect_1.0), which supports rapid identification of CRISPR arrays in short reads using user-provided reference repeat database as an extension of CRISPRDetect [**16**]. Arrays with a likelihood score > 3 were added to the existing CRISPR reference database. Subsampled reads were then mapped to the reference database using blastn [**18**] [parameters: -task blastn-short with default parameters]. Abundance tables were generated for all spacers and direct repeat sequences by tallying the number of reads that mapped to a given DR with 100% sequence identity and coverage.

### **Identifying potential false positive CRISPRs**

CRISPRs predicted from metagenomes are generally short, often incomplete and missing flanking regions which makes it hard to distinguish true CRISPRs from other genomic repeats. To measure how many of our identified arrays occurred in known prokaryotic genomes we compared the metagenomic CRISPR repeats against CRISPRs found in RefSeq and GenBank prokaryotic sequences (sequences published before September 9, 2019) using NCBI blast (parameters: -task blastn -word_size 11 -dust no -culling_limit 1 -num_alignments 1) [**19**]. Metagenomic DRs with >=90% identity and >=90% sequence coverage against RefSeq or GenBank DRs were considered as a positive match. Out of the 7562 repeat clusters 1649 were found in CRISPRs predicted from RefSeq or GenBank prokaryotic sequences. Similarly, the metagenomic repeats were screened against an _in silico_ generated set of eukaryotic reads from 6 eukaryotic reference genomes (**Table S2**). Ten thousand 250bp paired reads were generated from each reference genome using WGSIM (https://github.com/lh3/wgsim). Eukaryotic reads were subsampled to an equal depth to remove differences due to genome size. Out of the 7562 DR clusters a total of 168 repeat clusters were found in the eukaryotic reads suggesting a high level of eukaryotic sequence contamination may increase the false positive rate of our analysis. We therefore removed samples where > 10% of reads were classified as of eukaryotic origin.

### **Virome analysis**

To assess the diversity of viruses in the different environments, we analysed all MIUMS classified contigs from the 332 metagenomes. Across all datasets, we identified 2583 archaeal prophages, 375179 bacterial prophages, 1218279 bacteriophages, and 174792 archaeal viruses. Per metagenome, the sub-sampled reads used for the CRISPR quantification analysis were mapped to the total set of all viral contigs with bwa mem [**20**]. To quantify the genetic diversity of the viral community, we calculated the richness as the total number of detected viral contigs, plus the evenness and Shannon’s diversity index based on relative abundances calculated from the mean depth of coverage of all detected viral contigs. The intra-population diversity (micro-diversity) was calculated as the mean heterozygosity of viruses in the community by averaging Nei’s per-nucleotide diversity index across all detected contigs [**21**].  where pj is the frequency of allele _j_ at the position _i_ of a contig of the length L, N is the number of viruses in a dataset. Single nucleotide variability was assessed with VarScan software [**22**].

Viral abundance was calculated by first collecting the average depth values for all viral contigs in each sample using the ‘jgi_summarize_bam_contig_depths’ script from the Metabat package [**23**]. These depth scores were then summed per sample to give an approximation of overall viral abundance relative to bacterial abundance.

### **Statistical analysis**

For each sample, the number of reads that mapped to a direct repeat were counted to give a measure of CRISPR array abundance per sample. Earth Microbiome Project Ontology levels were assigned using the framework previously described [**24**](**Fig 1D**) based on the associated metadata from the NCBI short read archive (SRA). In the case of bioreactor samples, a literature search was conducted to identify the original material described in the associated study (see **Table S1**).

General linear models (GLM) were constructed for each of the reported correlations. In each case log10 transformations were applied to conform to model assumptions. Checks of model residuals were performed to assess model fit. Significance was determined using F-tests between null models and those containing the variable or interaction of interest.

### **Microbial community assessment**

CRISPR is more common in archaea. Therefore, in order to minimise any phylogenetic effects deriving from high archaeal abundances in samples we estimated the number of archaeal reads in the subsampled reads using metaxa2 [**14**] and converted this to relative abundance per sample. We then included this value as a fixed effect in additional GLMs to check the influence of archaeal abundance on our results. For additional phylogenetic assessments, reads classified as non-prokaryotic were removed and relative abundances were generated using the output from metaxa2 in order to assess effects of community composition. Permutational ANOVAs were performed on species abundance matrices using Bray-Curtis dissimilarity and CRISPR abundance as explanatory variable with 9999 permutations, using the function ‘adonis’ from ‘vegan’ package [**25**]. Visualisation of clustering analyses, at the class level, was performed using non-metric multidimensional scaling (NMDS) with Bray-Curtis dissimilarity through the ‘metaMDS’ function in ‘vegan’ [**25**]. Smooth surfaces were fit to these points via a generalized additive model (GAM), using either CRISPR abundance as the explanatory term, with the ‘ordisurf’ function in ‘vegan’[**25**].

Many prokaryotes encode CRISPR-Cas systems as immune protection against mobile genetic elements (MGEs), yet, a number of MGEs also harbor CRISPR-Cas components. With a few exceptions, CRISPR-Cas loci encoded on MGEs are uncharted and a comprehensive analysis of their distribution, prevalence, diversity, and function is lacking. Here, we systematically investigated CRISPR-Cas loci across the largest curated collection of natural bacterial and archaeal plasmids. CRISPR-Cas loci are widely but heterogeneously distributed across plasmids and, in comparison to host chromosomes, their mean prevalence per Mbp is higher and their distribution is markedly distinct. Furthermore, the spacer content of plasmid CRISPRs exhibits a strong targeting bias towards other plasmids, while chromosomal arrays are enriched with virus-targeting spacers. These contrasting targeting preferences dominate across the diversity of CRISPR-Cas subtypes and host taxa, highlighting the genetic independence of plasmids and suggesting a major role of CRISPR-Cas for mediating plasmid-plasmid conflicts. Altogether, CRISPR-Cas are frequent accessory components of many plasmids, which is an overlooked phenomenon that possibly facilitates their dissemination across microbiomes.

CRISPR-Cas systems are widespread accessory elements across bacterial and archaeal plasmids

### **Software and code availability**

Scripts for downloading data and reproducing all analyses are available at https://github.com/Russel88/CRISPRCas_on_Plasmids. Analyses were made with a combination of shell, python 3, and R 3.6.3 scripting. Plots were made with ggplot2, heatmaps with pheatmap, phylogenetic trees with iTOL (**Letunic and Bork, 2021**), and networks with gephi (**Bastian, Heymann and Jacomy, 2009**).

### **Dataset construction**

A total of 27,939 complete bacterial plasmid sequences were downloaded from PLSDB 2020_11_19 (https://ccb-microbe.cs.uni-saarland.de/plsdb) (**Galata** **_et al_., 2019**), together with their associated metadata (**Galata** **_et al_., 2019**). A total of 253 manually curated archaeal plasmids were downloaded from NCBI RefSeq on January 6th 2020. Plasmid-host chromosome associations were determined through the NCBI assembly information, for which only sequences annotated as “chromosome” were included as host sequences. Using this approach, we were able to assign a host for 21,974 of the plasmids. The number of archaeal plasmids selected is relatively low because few archaeal plasmids have been characterised and sequenced (Lipps, 2007). We used GTDBtk v1.4.1 (**Chaumeil** **_et al_., 2019**) to re-annotate the taxonomy of the host of each plasmid in a common phylogenomic framework. To filter out redundant plasmids, they were de-replicated using dRep version 3.1.0 (**Olm** **_et al_., 2017**) with the following parameters: 90% ANI cut-off for primary clustering, 95% ANI cut-off for secondary clustering and a total coverage of 90%, with fastANI (**Marçais** **_et al_., 2018**) as secondary clustering algorithm. Size was the only criterion used to choose the plasmid to include in each cluster, such that the largest plasmid (or random among these given ties) was picked among the clustered plasmids. Dereplication resulted in a total of 17,828 plasmids, out of which 13,265 could be associated with known prokaryotic hosts.

### **Identification of CRISPR loci**

Detection of CRISPR arrays was carried out by using CRISPRCasFinder 4.2.17 (**Couvin** **_et al_., 2018**), coupled to an optimized algorithm for false-positive array removal (**Supplementary Figure S12**) and an additional analysis for finding CRISPR loci that are commonly missed by this algorithm. Briefly, high confidence arrays predicted by CRISPRCasFinder (evidence level 4) were automatically kept. The remaining arrays were binned into a “quarantine list” if they were found to clear a series of conservative manually-curated parameter cutoffs: 1) calculated average CRISPR repeat conservation across the array > 70%, 2) spacer conservation < 50%, 3) standard error of the mean of the array’s spacer lengths < 3, and 4) array does not overlap with an open reading frame (ORF) with a prediction confidence of at least 90% (**Hyatt** **_et al_., 2010**). Putative arrays from the quarantined list were rescued for further analysis if they were found within 1 Kb to a predicted _cas_ gene or matched (95% coverage and 95% identity) with any previously defined high confidence CRISPR repeat: CRISPRCasFinder evidence level 4 or archived in CRISPRCasdb (**Pourcel** **_et al_., 2020**). This upgrade reduced the rate of detection of false positive CRISPRs, most of which constitute short repetitive genomic regions that are erroneously selected by CRISPRCasFinder (**Zhang and Ye, 2017**), and which are more common on plasmids (e.g. iterons and tandem transposon-associated repeats) (**Chattoraj, 2000**; **Giraldo and Fernández-Tresguerres, 2004**; **Oliveira** **_et al_., 2010**). High confidence CRISPR repeats (see above) were then BLASTed (task: blastn-short, 95% coverage and 95% identity) to a database in which the CRISPR loci that were already detected were masked and any matches within 100 bp were clustered into arrays. Arrays with less than 3 repeats were excluded from all analyses.

### **Identification and typing of** **_cas_** **loci**

The prediction and classification (at the subtype or variant level) of _cas_ operons was carried out by CRISPRCasTyper 1.2.4 (https://github.com/Russel88/CRISPRCasTyper) (**Russel** **_et al_., 2020**). CRISPR arrays closer than 10 Kbp to the nearest _cas_ operon were considered to be linked; the 10 Kbp cutoff was based on an analysis of the distribution of distances of CRISPR arrays to the closest _cas_ operon (**Supplementary Figure S13**). Furthermore, we used CRISPR-repeat similarity information to type arrays that were not found linked to _cas_ operons. These distant arrays (>10 Kbp from the nearest _cas_ operon) were considered associated with a _cas_ operon if the direct repeat sequence was at least 85% identical to the direct repeat sequence of an array adjacent to that _cas_ operon (**Supplementary Figure S14**). When possible, CRISPR-Cas systems annotated as “Ambiguous” were manually subtyped.

### **Indicator analysis**

Enrichment of certain CRISPR-Cas subtypes on either plasmids or host chromosomes was investigated with an indicator species analysis, using the indicspecies R package. For the comparison between all plasmids and chromosomes the IndVal.g statistic was used, which controls for difference in group sizes. For the direct comparison between plasmids and hosts chromosomes, where both carry CRISPR-Cas, the IndVal statistic was used. Statistical significance was determined by permutation (n=9999) and a Bonferroni adjusted p-value threshold of 0.05 was used.

### **Plasmid conjugative transfer and incompatibility group prediction**

The conjugative transfer functions and incompatibility (Inc) typing of all plasmids in PLSDB was predicted with MOB-suite v3.0.1 using _mob_typer_ function (**Robertson and Nash, 2018**) using default parameters.

### **Spacer-protospacer match analysis**

The genomic regions where CRISPR arrays were identified on plasmids (including CRISPR arrays with 2 repeats, which were otherwise excluded from the analyses) were masked in order to avoid false positive matches to spacers in arrays. Furthermore, for matches to plasmids only matches to high confidence ORFs were included, also to rule out any matches to possibly undetected CRISPR arrays. Spacers from orphan arrays whose consensus repeat could not be typed by repeatTyper from CRISPRCasTyper (https://typer.crispr.dk, model version 2021_03 (**Russel** **_et al_., 2020**)) were excluded from the spacer analysis to avoid any bias stemming from possible false positive arrays in this group.

Viral genomes were obtained from the IMG/VR v3 (2020-10-12_5.1, (**Roux** **_et al_., 2021**)) only including those annotated as “Reference”, which includes 39,296 viral genomes. Spacer sequences from plasmids and plasmid-associated host chromosomes were aligned against the masked dereplicated plasmid database and the virus database using FASTA 36.3.8e (**Pearson, 2014**). Alignments were filtered using an e-value cutoff of 0.05. To reduce redundancy bias, spacers were only counted once, no matter the absolute number of matches.

Networks were visualized in gephi with layout generated by a combination of OpenOrd and Noverlap algorithms. For calculating taxonomic confinement of spacer-protospacer matches between plasmids, each pair of plasmids connected by at least one spacer-protospacer match was counted as one matching pair. Cross-targeting plasmids were included as two separate plasmid pairs. Confinement was calculated as the number of matches found exclusively within a specific taxonomic rank, such that each plasmid-plasmid pair was only counted once. For estimating confinement of random spacer-protospacer matching, the taxonomic annotations were permuted among the plasmid-plasmid pairs with observed spacer-protospacer matches. This was repeated 100 times and the median number of matches was used as an estimate of confinement for hypothetically random matches. For estimating targeting bias towards conjugative vs. non-conjugative plasmids each unique spacer was counted with a weight of 1 with the targeting bias proportional to the number of matches to conjugative and non-conjugative plasmids, respectively. For example, a spacer matching 4 conjugative plasmids and 1 non-conjugative plasmids is counted as 0.8 for conjugative matches and 0.2 for non-conjugative matches.

Multiple advances have been made to increase the efficiency of CRISPR/Cas9 editing using the model genetic organism _Caenorhabditis elegans_ (_C. elegans_). Here we report on the use of co-CRISPR ‘marker’ genes: worms in which co-CRISPR events have occurred have overt, visible phenotypes which facilitates the selection of worms that harbour CRISPR events in the target gene. Mutation in the co-CRISPR gene is then removed by outcrossing to wild type but this can be challenging if the CRISPR and co-CRISPR gene are hard to segregate. However, outcrossing can be avoided by selecting worms of wild type appearance from a ‘jackpot’ brood. These are broods in which a high proportion of the progeny of a single injected worm display the co-CRISPR phenotype suggesting high CRISPR efficiency. This can deliver worms that harbour the desired mutation in the target gene locus without the co-CRISPR mutation. We have successfully generated a discrete mutation in the _C. elegans nlg-1_ gene using this method. However, in the process of sequencing to authenticate editing in the _nlg-1_ gene we discovered genomic rearrangements that arise at the co-CRISPR gene _unc-58_ that by visual observation were phenotypically silent but nonetheless resulted in a significant reduction in motility scored by thrashing behaviour. This highlights that careful consideration of the hidden consequences of co-CRISPR mediated genetic changes should be taken before downstream analysis of gene function. Given this, we suggest sequencing of co-CRISPR genes following CRISPR procedures that utilise phenotypic selection as part of the pipeline.

Confounds of using the unc-58 selection marker highlights the importance of genotyping co-CRISPR genes

### **_C. elegans_** **culturing and strains used**

All _C. elegans_ strains were maintained using standard conditions (**Brenner, 1974**). Strains used: Bristol N2 (wild-type), VC228 _nlg-1(ok259) X_ (x6 outcrossed), CB665 _unc-58(e665) X_, TN64 _dpy-10(cn64) II_, provided by _Caenorhabditis_ Genetics Center (CGC). XA3780 _nlg-1(qa3780) X_ (x2 outcrossed), XA3773 _unc-58(qa3788) X; nlg-1(qa3780) X_ (x1 outcrossed), XA3788 _unc-58(qa3788) X_ (x2 outcrossed), generated in this study. JIP1154 _unc-58(bln223) X_, provided by Thomas Boulin.

### **CRISPR/Cas9 genome editing**

The previously reported method by El Mouridi et al was followed for CRSIPR/Cas9 editing (**El Mouridi** **_et al_., 2017**). L4+1 day old hermaphrodites were microinjected with expression vectors for Cas9 and sgRNAs targeting _nlg-1_ and _unc-58_ or _dpy-10_ for the first and second round of CRISPR, respectively. The design of the sgRNAs and repair templates to edit _unc-58, dpy-10_ and _nlg-1_ genes were as previously reported (**Rawsthorne** **_et al_., 2020**; **Arribere** **_et al_., 2014**). sgRNAs and repair templates were purchased from Integrated DNA Technologies (IDT™ – Integrated DNA Technologies). All injection reagents were diluted in molecular grade water to a final concentration of 50ng/µl in the CRISPR mix.

### **Molecular Screening and sequencing**

_nlg-1_ primers, forward 5’-ATGAGTATACAGATTGGGAAAATCCC-3’ and reverse 5’-ACTGTTTGGTTGCTCTTGGCTCCAAG-3’, were used to amplify the CRISPR targeted region of _nlg-1_ using a single worm PCR protocol (**He, 2011**). A BanI site was used to screen PCR amplicons from individual worms for the incorporation or loss of the d10 sequence in _nlg-1_. The primers used for sequencing of the target regions of _unc-58_ and _dpy-10_ were: forward 5’-GACTCGGAGATATCGTTGTGACTG-3’, reverse 5’-CGCGGAGTTCGTTATCCAGGAAG-3’ and forward 5’-ACTAATTCAGAGTCATCATCTCGCC-3’, reverse 5’-CATCAATTCCCTTAAGTCCTGGTGG-3’ respectively.

### **Removal of** **_unc-58_** **background mutation**

_unc-58(qa3788); nlg-1(qa3780)_ double CRISPR mutant strain was backcrossed with N2 males once to generate heterozygous F1 progeny. These progeny were cloned by picking a single F1 onto individual plates before being left for 2-3 days to self-fertilise and F2 progeny were screened using a single worm PCR protocol (**He, 2011**) and restriction digest. Restriction digest of _unc-58 and nlg-1_ PCR amplicons were performed using restriction enzymes BsiWI and BstBI respectively to screen for the presence or absence of CRISPR mediated edits. Restriction enzymes were supplied by New England BioLabs (NEB) and used according to manufacturer instructions.

### **Food leaving assay**

Food leaving assays were carried out as previously described (**Rawsthorne** **_et al_., 2020**). Briefly, 50µl of OP50 _E. coli_ at OD600 of 0.8 was gently spotted on to the middle of an unseeded plate the day prior to the assay. Seven age synchronised L4+1 day old hermaphrodites were gently picked onto the centre of the bacterial lawn on the assay plate. At 2 and 24 hours food leaving events were counted visually using a Nikon SMZ800 microscope (X10 magnification) during 30 minute observations. A food leaving event was defined as when the whole of the worm’s body exited the bacterial lawn. N2 animals were used as a paired control, run in parallel with the strain under investigation and the investigator was blind to the genotypes being observed.

### **Thrashing assay**

Using a 24 well plate, a single worm was picked per well containing 500µl of M9 with 0.1% bovine serum albumin and left for 5 minutes before thrashing was measured. For each worm, thrashing events were visually counted under a Nikon SMZ800 microscope (X30 magnification) for a period of 30 seconds. This was repeated three consecutive times and the mean was calculated. Each thrash was defined as a complete movement through the midpoint of the worms body and back. N2 animals were used as a paired control, run in parallel with the strain under investigation and the investigator was blind to the genotypes.

### **_unc-58_** **protein sequence analysis**

UNC-58, isoform b protein sequence was downloaded from WormBase version WS278. The wild-type or CRISPR mutant protein sequence was entered into the membrane topology prediction tool TMHMM (v.2.0) (**Sonnhammer, von Heijne and Krogh, 1998**; **Krogh** **_et al_., 2001**) to determine the predicted transmembrane topology.

The adaptive CRISPR-Cas immune system stores sequences from past invaders as spacers in CRISPR arrays and thereby provides direct evidence that links invaders to hosts. Mapping CRISPR spacers has revealed many aspects of CRISPR biology, including target requirements such as the protospacer adjacent motif (PAM). However, studies have so far been limited by a low number of mapped spacers in the database. By using vast metagenomic sequence databases, we mapped one third (∼70,000) of more than 200,000 unique CRISPR spacers from a variety of microbes, and derived a catalog of more than one hundred unique PAM sequences associated with specific CRISPR subtypes. These PAMs were further used to correctly assign the orientation of CRISPR arrays, revealing conserved patterns between the last nucleotides of the CRISPR repeat and PAM. From the curated CRISPR arrays dataset we could also deduce CRISPR subtype specific preferences for targeting either template or coding strand of open reading frames. While some DNA-targeting systems (e.g. Type I-E and Type II systems) prefer the template strand and avoid mRNA, other DNA- and RNA-targeting systems (i.e. Type I-A, I-B and Type III systems) prefer the coding strand and mRNA. In addition, we found large scale evidence that both CRISPR adaptation machinery and CRISPR arrays are shared between different CRISPR-Cas systems. This could lead to simultaneous DNA- and RNA targeting of invaders, which may be effective at combating mobile genetic invaders.

Comprehensive PAM prediction for CRISPR-Cas systems reveals evidence for spacer sharing, preferred strand targeting and conserved links with CRISPR repeats

### **CRISPR spacers and sequence data**

221 089 spacers along with information on _cas_ gene presence, genome and repeat sequence were obtained from CRISPRCasDb (**Pourcel et al., 2020**) in February 2020 and the taxonomy of the genomes was obtained from NCBI Taxonomy database (**Federhen, 2012**). We created our own sequence database by combining all sequences from the NCBI nucleotide database (**Benson et al., 2018**; **Pruitt et al., 2005**), environmental nucleotide database (**Sayers et al., 2009**), PHASTER (**Arndt et al., 2016**), Mgnify (**Mitchell et al., 2020**), IMG/M (**I. M. A. Chen et al., 2017**), IMG/Vr (**Paez-Espino et al., 2019**), HuVirDb (**Soto-Perez et al., 2019**), HMP database (**Peterson et al., 2009**), and data from **Pasolli et al., 2019**. All databases were accessed in February 2020.

Subtypes were predicted based on the repeat sequences using the subtype predictions and method described by **Bernheim et al., 2020**, where the subtype of a spacer was inferred by the similarity of its repeat sequence to repeat sequences with known subtype (74% identity threshold to infer subtype).

### **Blast hits and filtering**

Hits between spacers and sequences from the aforementioned databases were obtained using the command line blastn program (**Altschul et al., 1990**) version 2.10.0, which was run with parameters word_size 10, gapopen 10, penalty 1 and an e-value cutoff of 1, to find as many potential targets as possible. These blast hits were then filtered to remove hits of spacers inside CRISPR arrays and false positive hits found by chance. Hits inside CRISPR arrays were detected by aligning the repeat sequence of the spacer to the flanking regions of the spacer hit (23 nucleotides on both sides). This alignment was done using the globalxs function from the Biopython pairwise2 package (**Cock et al., 2009**) with −3 gap open and −3 gap extend parameters. If more than 13 nucleotides were identical in the alignment of at least one flank, the hit was suspected to fall inside a CRISPR array and was filtered out.

To minimize the number of hits found by chance, we filtered hits based on the fraction of spacer nucleotides that hit the target sequence, as this metric considers both the sequence identity and the coverage of the spacer by the blast hit. In a first step, only hits with this fraction higher than 90% were kept. To find targets for even more spacers while keeping the number of false positives low, we included a second step where hits with a fraction higher than 80% were kept if another spacer from the same genus hit the same contig or genome in the first step. This second step did not introduce hits on any new contigs or genomes and was based on the assumption that multiple spacers from the same genus hitting the same contig or genome is unlikely to be caused by chance. Finally, we removed spacers that were shorter than 27 nucleotides (54 spacers) and removed 7 spacers that were hitting aspecifically, such as inside ribosomal RNAs or tRNAs. This left 72,099 unique spacers with target hits for downstream analysis.

### **Protospacer flank alignment for orientation and PAM predictions**

The PAM is known to occur on the 5’ end of the protospacer for Type I, Type IV and V CRISPR-Cas systems, and on the 3’ end for Type II systems (**Collias & Beisel, 2021**; **Jackson et al., 2017**). We used this property to predict the orientation of transcription of CRISPR arrays and sequence of crRNA. The PAM sides were compared to the nucleotide conservation in the flanking regions of the spacer hits and the spacer orientations were predicted such that the flank with the greater conservation matched the known PAM side.

To measure the nucleotide conservation in the flanking regions, data from multiple spacers was combined based on the subtype and repeat sequences of the spacers. Highly similar repeat sequences from the same subtype were clustered using CD-HIT (**Fu et al., 2012**) with a 90% identity threshold. We hypothesized that similar repeat sequences would be used in a similar orientation and would utilize the same PAM sequences, as coevolution of PAM, repeat and Cas1 and Cas2 sequences has been shown previously (**Alkhnbashi et al., 2014**; **Lange et al., 2013**). For each repeat cluster the flanking regions of the spacer hits were aligned. To equally weigh each spacer within the repeat cluster, irrespective of the number of blast hits, consensus flanks were obtained per spacer. These consensus flanks contained the most frequent nucleotide per position of the flanking regions. From the alignment of consensus flanks the nucleotide conservation, or information content, in each flank was calculated in bitscore (**Schneider & Stephens, 1990**) using the Sequence logo python package. We corrected for GC-content of the targeted sequences by calculating the expected occurrences of each nucleotide based on the GC-content of the spacer sequences. To minimize the number of orientation predictions based on little or noisy data, we only predicted the orientation for repeat clusters when the alignment of consensus flanks consisted of at least 10 unique protospacers. Furthermore, the information content of at least two positions was higher than 0.3 bitscore and higher than 5 times the median bitscore calculated from 23-nt flanks on both sides. These parameters were chosen as strictly as possible, while still yielding orientation predictions for the highest number of spacers.

Using the orientation predictions described above, we predicted the PAMs for each repeat cluster by checking which nucleotide positions were conserved. To minimize PAM predictions based on noise, we only predicted the PAM for repeat clusters where the alignment of consensus flanks consisted of at least 10 unique protospacers. A nucleotide position was predicted to be part of the PAM when higher than 0.5 bitscore and higher than 10 times the median bitscore. These parameters were chosen as strictly as possible, while maximizing the number of repeat clusters with PAM predictions and minimizing the number of unique PAMs predicted.

We subsequently categorized and counted multi-effector compatible spacers in the following ways. Firstly by an occurrence of multiple repeat clusters with different subtype classification that both contained the same PAM, for two DNA targeting clusters (category I) or a DNA and a RNA targeting cluster (category II). Secondly if multiple _cas_ gene clusters from different subtypes were in the vicinity of a single repeat cluster and their genomes did not further contain other arrays linked to these _cas_ gene clusters they were counted as a third category multi-effector compatible array.

### **Coding versus template strand targeting analysis**

For each spacer target inside an open reading frame (ORF), we determined if the spacer targets the coding (DNA and RNA) or template strand (DNA-only) during transcription. The ORFs and its orientation were predicted using Prodigal (**Hyatt et al., 2010**) for one target sequence per spacer. The target sequence of each spacer was selected as the longest hit sequence in the NCBI nucleotide database, excluding ‘other sequences’, or, if no such sequence was hit, the longest hit sequence in metagenomics database. Using our spacer orientation predictions for Type I, II and IV spacers, and the orientation predictions from CRISPRCasDb for the other spacers, we checked if the spacer target (blast hit orientation) was on the coding or template strand of the predicted ORF. To test for significant bias towards either the temperate or the coding strand, a two-sided tailed binomial test was performed with an expected probability of 0.5.

COVID-19 has forced rapid clinical translation of novel vaccine technologies, principally mRNA vaccines, that have resulted in meaningful efficacy and adequate safety in response to the global pandemic. Notwithstanding this success, there remains an opportunity for innovation in vaccine technology to address current limitations and meet the challenges of inevitable future pandemics. We describe a universal vaccine cell (UVC) rationally designed to mimic the natural physiologic immunity induced post viral infection of host cells. Induced pluripotent stem cells were CRISPR engineered to delete MHC-I expression and simultaneously overexpress a NK Ligand adjuvant to increase rapid cellular apoptosis which was hypothesized to enhance viral antigen presentation in the resulting immune microenvironment leading to a protective immune response. Cells were further engineered to express the parental variant WA1/2020 SARS-CoV-2 spike protein as a representative viral antigen prior to irradiation and cryopreservation. The cellular vaccine was then used to immunize non-human primates in a standard 2-dose, IM injected prime + boost vaccination with 1e8 cells per 1 ml dose resulting in robust neutralizing antibody responses (1e3 nAb titers) with decreasing levels at 6 months duration. Similar titers generated in this established NHP model have translated into protective human neutralizing antibody levels in SARS-Cov-2 vaccinated individuals. Animals vaccinated with WA1/2020 spike antigens were subsequently challenged with 1.0 × 105 TCID50 infectious Delta (B.1.617.2) SARS-CoV-2 in a heterologous challenge which resulted in an approximately 3-log order decrease in viral RNA load in the lungs. These heterologous viral challenge results reflect the ongoing real-world experience of original variant WA1/2020 spike antigen vaccinated populations exposed to rapidly emerging variants like Delta and now Omicron. This cellular vaccine is designed to be a rapidly scalable cell line with a modular poly-antigenic payload to allow for practical, large-scale clinical manufacturing and use in an evolving viral variant environment. Human clinical translation of the UVC is being actively explored for this and potential future pandemics.

A Novel CRISPR-Engineered, Stem Cell-Derived Cellular Vaccine

### **iPS cell culture, irradiation, and cryopreservation**

Human iPS cells (Thermo Fisher) were cultured on vitronectin-coated T225cm2 flasks using complete mTesSR Plus medium (StemCell Technologies) supplemented with 1% penicillin/streptomycin, Rock inhibitor (StemCell Technologies) at 1:1000 dilution. For drug selection, G148 was used at 500ug/ml and puromycin at 5ug/ml (Sigma-Aldrich). Cultures were maintained at 37 °C, 5% CO2 in a humidified incubator. Harvesting of engineered UVC was performed using accutase (StemCell Technologies) and cells were counted using a CellDrop cell counter (DeNovix). Cells were irradiated at a total single dose of 10 Gy, before centrifugation at 300 xg for 10 minutes followed by resuspension in 100 μl of CryoStor-CS10 freezing media (StemCell Technologies). The UVC preparations for use in non-human primate studies were analyzed for endotoxin levels (Wickham Laboratories Ltd) and absence of mycoplasma (Mycoplasma Experience Ltd).

### **CRISPR genetic engineering**

CRISPR sgRNAs targeting the human B2M gene, PPP1R12C (AAVS1), and the ROSAβgeo26 locus were designed and validated for indel formation at the selected genomic site. Up to 6 sgRNAs per target gene were tested and the most efficient sgRNA was selected containing 2’-O-methyl and 3’ phosphorothioate modifications to the first three 5’ and the last three 3’ nucleotides (Synthego). 2×106 UVC cells were electroporated using a Neon Nucleofector (Lonza) in Buffer P3 (Lonza) with Cas9 protein (IDT) precomplexed with sgRNA, in a total volume of 100 μl using electroporation program CM138. Gene targeting vectors carrying an expression cassette for expression of human MICA or the SARS-CoV-2 WA1/2020 spike gene, targeting the Rosa26 and AAVS1 locus respectively, were co-electroporated at 4 μg. Indels introduced by CRISPR editing were detected by PCR and Sanger sequence using DNA primers designed to amplify a 600-900 base pair region surrounding the sgRNA target site. A minimum of 24 hours after electroporation, genomic DNA was extracted using the DirectPCR Lysis solution (Viagen Biotech) containing Proteinase K and target regions were amplified by PCR using the GoTaq G2 PCR mastermix (Promega). Correct and unique amplification of the target regions was verified by agarose gel electrophoresis before purifying PCR products using the QIAquick PCR Purification Kit (Qiagen). For analysis by TIDE, PCR amplicons were Sanger sequenced (Eurofins or Genewiz) and paired .ab1 files of control versus edited samples were analyzed using Synthego’s ICE tool (https://ice.synthego.com).

### **Intracellular spike protein staining**

Engineered UVC were harvested and then fixed and permeabilized using BD Cytofix/Cytoperm Fixation/Permeabilization Solution (ThermoFisher). Cells were then stained for intracellular spike protein using an Anti-SARS-CoV-2 Spike Glycoprotein S1 antibody (Abcam, ab275759, 1:50) followed by Goat Anti-Rabbit IgG H&L (Alexa Fluor 488) (Abcam, ab150077, 1:500). Flow analysis was carried out on a Fortessa flow cytometer (BD Bioscience), and data analyzed, and flow cytometry figures generated using FlowJo 10 software (BD Biosciences).

### **Flow cytometry analysis of cell surface antigen expression**

For flow cytometric analysis of cell surface expression of MHC-I, MICA and SARS-CoV-2 spike protein, cells were harvested from culture plates and washed using PBS with 1% Bovine Serum Albumen (Thermo Scientific) and were then stained with PE anti-human MICA/MICB Antibody (6D4, Biolegend), Alexa Fluor 647 anti-human HLA-A,B,C (W6/32, Biolegend), and anti-SARS-CoV-2 Spike Glycoprotein S1 antibody (Abcam, ab275759, 1:50) followed by Goat Anti-Rabbit IgG H&L (Alexa Fluor 488) (Abcam, ab150077, 1:500). Live/Dead Fixable Dead Cell Stains (Invitrogen) were included in all experiments to exclude dead cells. After staining, cells were resuspended in PBS with 2% Human Heat Inactivated AB Serum (Sigma) and 0.1 M EDTA pH 8.0 (Invitrogen) before analysis on a Fortessa flow cytometer (BD Bioscience) and data analyzed using FlowJo 10 software (BD Biosciences).

### **Western blot**

The SARS-CoV-2 spike glycoprotein was detected in UVC lysates by western blotting. Briefly, cells were lysed by RIPA buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% NP40, 1x protease inhibitor cocktail). Samples were spun at 4°C for 10 mins at 12,000 xg and the pellet discarded. Protein content was measured using BCA Assay (ThermoFisher) using a PHERAstar plate reader (BMG Labtech) at 560 nm. LDS Sample Buffer was added to 30 ng of protein sample to make a 1x solution, with 0.5 μl of b-mercaptoethanol per well and heated at 70°C for 10 minutes before separation on a polyacrylamide gel (Bio-Rad Mini-PROTEAN TGX Gel 4-15%) and transferred to a PVDF membrane. Membranes were blocked in blocking buffer (5% non-fat powdered milk in TBST), before incubation with primary antibodies in blocking buffer (Rabbit polyclonal anti-SARS-Cov2, Sino Biological 40591-T62, 1:6000 dilution or Mouse b-actin, Abcam 8226, 1 μg/ml), detected with HRP conjugated secondaries in blocking buffer (Goat anti-Rabbit HRP, Sino Biological SSA003, 0.5 μg/ml or Goat anti-Mouse HRP, Abcam ab205719, 1: 4000 dilution) and visualised using the SuperSignal West Femto kit (ThermoFisher) as per kit instructions.

### **qPCR measurement of stem cell factors**

Total RNA was extracted from UVC cells using the ReliaPrep RNA miniprep (Promega) according to the manufacturer’s instructions (a DNase treatment was included for all samples), and RNA concentration and absorbance ratios were measured using a Nanodrop One Spectrophotometer (ThermoFisher). cDNA was synthesized using a High-Capacity cDNA Reverse Transcription Kit (the Applied Biosystems) in a total volume of 20 μl to produce DNA that was subsequently assessed by spectrophotometric analysis and diluted to 100 ng/μl. Individual master mixes with each of the DNA-primer combinations for detection of human SOX2, NANOG, OCT4, DNC, Vimentin, HES5 and GATA6 genes were made for 3 replicates using the Brilliant III Ultra-Fast SYBR green qPCR master mix (Agilent Technologies) and analyzed on a CFX Opus Real-Time PCR system (BioRad) using the following program: 95 °C for 15 minutes for 1 cycle; 95°C for 15 seconds for 40 cycles; 60°C for 30 seconds.

### **SARS-CoV-2 spike protein ELISA**

Cell pellets were harvested and lysed in 20 μl Cell Extraction Buffer (Invitrogen) containing protease inhibitors (Sigma) on ice for 30 minutes, with 3 brief vortexing every 10 minutes. Samples were centrifuged at 13,000 rpm for 10 minutes at 4°C to pellet insoluble contents. S1 Spike protein was detected using a Covid-19 S-protein ELISA kit (Abcam) specific to S1RBD. Samples were diluted to a range determined to be within the working range of the ELISA kit used and the assay procedure was follows as per manufacturer’s instructions. The resulting colorimetric signal was detected at 450 nm using a PHERAstar (BMG LABTECH) plate reader. GraphPad Prism was used to plot a standard curve and interpolate the sample values using a 4-parameter logistic fit.

### **UVC Proliferation and apoptosis assays**

To quantify apoptosis of UVC post-irradiation, cells were stained using a FITC Annexin V Apoptosis Detection Kit with 7-AAD (Biolegend). Proliferation of cells was measured staining of control and Irradiated UVC with either 2 μM Cell Trace Yellow (Abcam) according to kit protocol and analyzing the dilution of the dye at 24-hour periods over 3-days and measuring fluorescence intensity. Flow analysis was carried out on a Fortessa flow cytometer (BD Bioscience), and data analyzed, and flow cytometry figures generated using FlowJo 10 software (BD Biosciences).

### **CAM cytotoxicity assay**

Both MHC-I expressing and MHC-I deficient (B2M knockout) UVC were used as target cells for NK cell cytotoxicity assay. Trypsinized cells were stained with calcein acetoxymethyl ester (CAM, Invitrogen) at a 10 μM concentration for 1 hour at 37°C and then washed to remove excess dye. NK cells highly enriched from normal cynomolgus macaque (_Macaca fascicularis_) blood samples using a CD3 depletion kit (Miltenyi Biotec), were used as effector cells. NK cell effectors and stained target cells were co-cultured in 96 well round bottom plates at effector: target (E:T) ratios of 1:1 and 5:1. Control wells included – only target cells for spontaneous release of CAM and target cells treated with Triton-X100 for maximum release of CAM. At the end of 4-hour incubation, supernatant was collected for CAM measurement in a fluorescent plate reader at 530 nm. Percent-specific lysis = (test release - spontaneous release)/(maximum release - spontaneous release).

### **Nucleofection of NKG2D ligands in iPS cells**

UVC were cultured in EGM2 (Lonza) media supplemented with 20 ng/ml VEG-F (Peprotech) until 70-90% confluent, in tissue culture flasks pre-coated with sterile 0.1% gelatin in PBS for 1 hour at 37°C. The cells were removed from culture flasks using trypsin, washed, and transfected with plasmid DNA containing either MICA, MICB or ULBP-1 genes after optimizing nucleofection conditions using primary cell 4D nucleofector kit and 4D nucleofector system (Lonza). After 48 hours of culture, transfected cells were stained with aqua dye for live/dead discrimination and corresponding antibodies-MICA/MICB (Clone 6D4, PE, BioLegend) or ULBP-1 (clone 170818, PE, R & D Systems). Stained cells were fixed with 2% paraformaldehyde and acquired on LSRII flow cytometer. Transfection efficiency was calculated as % live cells expressing transfected protein.

### **NK cell intracellular cytokine staining assay**

NK cell effectors were enriched from normal cynomolgus macaque (Macaca fascicularis) blood samples using a CD3 depletion Kit (Miltenyi Biotec). Target and effector cells were plated at E:T ratio of 2:1 in a 96 well round bottom plate. Anti-CD107a antibody (clone H4A3, ECD conjugate, BD Biosciences), brefeldin A and monensin (BD Biosciences) were added to all the samples prior to incubation. After 6 hours of incubation at 37°C, the cells were washed and stained with aqua dye used for live and dead cell discrimination for 20 minutes at room temperature. The cells were then washed and stained for surface markers that included CD3 (SP34.2, BV421, BD Biosciences), CD14 (M5E2, BV650, BD Biosciences), CD16 (3G8, BUV496, BD Biosciences), CD20 (L27, BV570, BD Biosciences), CD56 (NCAM1.2, BV605, BD Biosciences), HLA-DR (G46-6, APC-H7, BD Biosciences) and NKG2A (Z199, PE-Cy7, BD Biosciences) to delineate NK effector cells. Following incubation for 20 minutes, cells were washed and permeabilized using fix & perm reagent (Thermofisher Scientific) as per manufacturer’s recommendation. Intracellular cytokine staining was performed for macrophage inflammatory protein 1β (MIP-1β; D21-1351, FITC, BD Biosciences) interferon-γ (IFN-γ; B27, BUV395, BD Biosciences), tumor necrosis factor-α (TNF-α; Mab11, BV650, BD Biosciences) at 4°C for 15 minutes.

Cells were washed, fixed, and acquired on LSRII flow cytometer. Unstimulated NK cells were used for background subtraction of percent positive cells. NK cells stimulated with leukocyte activation cocktail (BD Biosciences) were used as positive control for the assay.

### **Animals and study design**

Outbred adult male and female rhesus macaques (_M. mulatta_) and cynomolgus macaques (_M. fascicularis_), 6–12 years old, were randomly allocated to groups. All macaques were housed at Bioqual. Macaques were treated with irradiated UVC at doses of either 1×107 or 1×108 cells (_n_ = 3-6), and sham controls (_n_ = 3-6). Prior to immunization, the cryopreserved doses of irradiated UVC were thawed at 37°C, then 900 μl of 1xPBS was added to each vial of 100 μl UVC in CryoStore freezing media. Macaques received a prime immunization of 1ml of UVC by the intramuscular route without adjuvant at week 0. At weeks 4 or 6, macaques received a boost immunization of either 1×107 or 1×108 UVC. At week 10 all macaques were challenged with 1.0 × 105 TCID50 (1.2 × 108 RNA copies, 1.1 × 104 PFU) SARS-CoV-2, which was derived from B.1.617.2 (Delta). Viral particle titers were assessed by RT– PCR. Virus was administered as 1 ml by the intranasal route (0.5 ml in each nare) and 1 ml by the intratracheal route. All immunological and virological assays were performed blinded. All animal studies were conducted in compliance with all relevant local, state, and federal regulations and were approved by the Bioqual Institutional Animal Care and Use Committee (IACUC).

### **Subgenomic viral mRNA assay**

SARS-CoV-2 _E_ gene sgRNA was assessed by RT–PCR using primers and probes as previously described**49**, **50**, **51**, **52**. In brief, to generate a standard curve, the SARS-CoV-2 _E_ gene sgRNA was cloned into a pcDNA3.1 expression plasmid; this insert was transcribed using an AmpliCap-Max T7 High Yield Message Maker Kit (Cellscript) to obtain RNA for standards. Before RT–PCR, samples collected from challenged macaques or standards were reverse-transcribed using Superscript III VILO (Invitrogen) according to the manufacturer’s instructions. A Taqman custom gene expression assay (ThermoFisher Scientific) was designed using the sequences targeting the _E_ gene sgRNA. Reactions were carried out on a QuantStudio 6 and 7 Flex Real-Time PCR System (Applied Biosystems) according to the manufacturer’s specifications. Standard curves were used to calculate sgRNA in copies per ml or per swab; the quantitative assay sensitivity was 50 copies per ml or per swab.

### **Serum antibody ELISA**

RBD-specific binding antibodies were assessed by ELISA as previously described**9**,**10**. In brief, 96-well plates were coated with 1 μg ml-1 SARS-CoV-2 RBD protein (A. Schmidt, MassCPR) in 1× DPBS and incubated at 4 °C overnight. After incubation, plates were washed once with wash buffer (0.05% Tween 20 in 1× DPBS) and blocked with 350 μl casein block per well for 2–3 hour at room temperature. After incubation, block solution was discarded, and plates were blotted dry. Serial dilutions of heat-inactivated serum diluted in casein block were added to wells and plates were incubated for 1 hour at room temperature, before three further washes and a 1-hour incubation with a 1:1,000 dilution of anti-macaque IgG HRP (NIH NHP Reagent Program) at room temperature in the dark. Plates were then washed three times, and 100 μl of SeraCare KPL TMB SureBlue Start solution was added to each well; plate development was halted by the addition of 100 μl SeraCare KPL TMB Stop solution per well. The absorbance at 450 nm was recorded using a VersaMax or Omega microplate reader. ELISA endpoint titers were defined as the highest reciprocal serum dilution that yielded an absorbance >0.2. The log10(endpoint titers) are reported.

### **Pseudovirus neutralization assay**

The SARS-CoV-2 pseudovirus expressing a luciferase reporter gene were generated in a similar approach to that previously described**9**,**10**,**16**. In brief, the packaging construct psPAX2 (AIDS Resource and Reagent Program), luciferase reporter plasmid pLenti-CMV Puro-Luc (Addgene), and spike protein expressing pcDNA3.1-SARS-CoV-2 SΔCT were co-transfected into HEK293T cells with calcium phosphate. The supernatants containing the pseudotype viruses were collected 48 hours after transfection; pseudotype viruses were purified by filtration with 0.45-μm filter. To determine the neutralization activity of the antisera from vaccinated macaques, HEK293T-hACE2 cells were seeded in 96-well tissue culture plates at a density of 1.75 × 104 cells per well overnight. Twofold serial dilutions of heat-inactivated serum samples were prepared and mixed with 50 μl of pseudovirus. The mixture was incubated at 37 °C for 1 hour before adding to HEK293T-hACE2 cells. After 48 hours, cells were lysed in Steady-Glo Luciferase Assay (Promega) according to the manufacturer’s instructions. SARS-CoV-2 neutralization titers were defined as the sample dilution at which a 50% reduction in relative light units was observed relative to the average of the virus control wells.

### **Statistical Analyses**

Statistical differences between two sample groups, where appropriate, were analyzed by a standard Student’s two-tailed, non-paired, t-test and between three or more sample groups using two-way or three-way ANOVA using GraphPad Prism 9. Analysis of virological data was performed using two-sided Mann–Whitney tests. Correlations were assessed by two-sided Spearman rank-correlation tests. P values are included in the figures or referred to in the legends where statistical analyses have been carried out. P values of less than 0.05 were considered significant.

Nucleophilic amino acids are important in covalent drug development yet underutilized as antimicrobial targets. Over recent years, several chemoproteomic technologies have been developed to mine chemically-accessible residues via their intrinsic reactivity toward electrophilic probes. However, these approaches cannot discern which reactive sites contribute to protein function and should therefore be prioritized for drug discovery. To address this, we have developed a CRISPR-based Oligo Recombineering (CORe) platform to systematically prioritize reactive amino acids according to their contribution to protein function. Our approach directly couples protein sequence and function with biological fitness. Here, we profile the reactivity of >1,000 cysteines on ~700 proteins in the eukaryotic pathogen _Toxoplasma gondii_ and prioritize functional sites using CORe. We competitively compared the fitness effect of 370 codon switches at 74 cysteines and identify functional sites in a diverse range of proteins. In our proof of concept, CORe performed >800 times faster than a standard genetic workflow. Reactive cysteines decorating the ribosome were found to be critical for parasite growth, with subsequent target-based screening validating the apicomplexan translation machinery as a target for covalent ligand development. CORe is system-agnostic, and supports expedient identification, functional prioritization, and rational targeting of reactive sites in a wide range of organisms and diseases.

CRISPR

Skills & Expertise

Have a question?