Phage and bacteria
Flavobacterium columnare strain B245 was isolated from the same fish farm in Central Finland in 2009 as its phage V15635. Conventional culturing of B245 was made using Shieh medium36 without glucose. V156 stocks were produced by harvesting confluent soft-agar layers from double-agar plates, adding 4ml of media, centrifuging (10.000 rpm, 10 minutes, Sorvall RC34 rotor) and filtering the supernatant through 0.22 µm filters. Aeromonas sp. B135 was isolated from a small natural brook in Central Finland (2008). E. coli DSM613 strain was obtained from DSMZ GmbH (Braunschweig, Germany).
Long-term culturing conditions and sampling
The effect of four different nutritional conditions on phage resistance mechanisms were studied: autoclaved lake water alone, lake water supplemented with 0.1% purified porcine mucin, 0.1x Shieh medium alone and 0.1x Shieh media supplemented with 0.1% purified porcine mucin. The lake water was collected from Lake Jyväsjärvi (Jyväskylä, Finland) on February 13th 2018 and autoclaved. The water was analyzed by Eurofins Scientific and contained N: 790 µg L-1, P: 18 µg L-1 and Fe: 540 µg L-1. Shieh media was diluted in ultrapure sterile water. Autoclaved 2% w:v solution of purified porcine mucin (Sigma-Aldrich, catalog no. M1778) was used as stock for preparing the simulated mucosal cultures.
The initial inoculum in each culture was 5×104 colony forming units (cfu) of F. columnare B245 and 5×103 plaque forming units (pfu) of phage V156 (multiplicity of infection of 0.1). Each condition was tested in triplicates, in a final volume of five milliliters, and incubated at 26 degrees under 120 rpm. As non-infected controls, one culture of each condition was made with only B245 without phage.
Every week after day zero, one milliliter of each culture was removed and replaced with one milliliter of the corresponding culturing media (autoclaved lake water or 0.1x Shieh, supplemented or not with 0.1% mucin). An overview of the experiment setup is shown in Figure 1.
Overview of the experimental setup. The 16-week experiment (denoted by the horizontal line) contained four culturing conditions, which were sampled and restocked with fresh media weekly. Bacterial isolates were characterized by their morphotype and CRISPR spacer content and used later for growth tests with or without the ancestral phage. Phage genomes (population level) were sequenced at week 16. Genomes from representative bacterial isolates were sequenced throughout the experiment. Figure made in ©BioRender - biorender.com
Bacteria and phage titrations
Immediately after sampling, each sample was serially diluted and plated on Shieh-agar plates for analysis of bacterial population size. Chloroform (10% v:v) was added to the remaining sample to kill bacterial cells. Serial dilutions of the chloroform-treated supernatants were used for titrating phages with the double-agar layer method37 using the ancestral B245 as host. Bacteria and phage plates were incubated at room temperature for three days, followed by the enumeration of bacterial colonies and viral plaques. In the first experiment phages were titrated every week during the experiment, while bacteria were titrated in weeks 1 to 8, 10, 12 and 15. The competition experiment was sampled at days 7, 14, 32 and 56 for phage and bacteria titers.
Detecting the acquisition of new CRISPR spacers (both loci)
We aimed at collecting an equal number of Rough and Rhizoid colonies per sample whenever possible. After counting, random colonies were picked and transferred to 50 microliters of Shieh on 96 well plates. Two microliters of each resuspended colony were used as template for PCR reactions designed to detect the insertion of new spacers on both CRISPR loci. Reactions were made with DreamTaq polymerase (Thermo Fisher), in 20 μL reactions containing 0.5mM of DNTPs and 0.5 μM of each primer using previously published primers for strain B24532. Cycling conditions were 95 degrees for 3 minutes followed by 30 cycles of 95 degrees for 30 seconds, 60 degrees for 30 seconds, 72 degrees for one minute and a final extension step of 72 degrees for 15 minutes. PCR reactions were resolved in 2% agarose gels and the addition of spacers verified by the size of each amplicon.
Bacterial growth characteristics
Representative bacterial isolates, considering colony morphology and CRISPR loci sizes, were chosen every week for further analysis. Following PCR, the remaining volume of resuspended colonies was transferred to 5mL of Shieh media. After overnight growth (120 rpm, 26 degrees) the individual colonies/isolates were frozen at -80 degrees for future use, and revived overnight for testing their immunity against the ancestral phage using Bioscreen C® (Growth curves Ltd, Helsinki, Finland). 1000 cfu mL-1 of each isolate were added to Bioscreen plates, in triplicates (200 microliters per well). Each isolate was tested in the presence and in the absence of the ancestral V156 phage (103 pfu mL-1, MOI 1). Ancestral bacterial strain B245 was included on every plate as control, also in triplicates. Optical density measurements were made every 10 minutes for 4 days. Plates were kept without agitation at 27 degrees for the whole time. Minor differences between plates were accounted for by including the plate ID as a random effect in statistical models when appropriate. However, differences between plates was generally minimal (Supplementary Figure 1). Testing the growth of bacterial isolates was not made for the competition experiment (see below).
Genomic DNA extraction and sequencing
Population-level phage DNA at week 16 and clonal DNA of selected bacterial isolates from different time points were sequenced with Illumina (BGI). For bacterial genomic DNA extraction, isolates were taken from the freezer and grown overnight. DNA of turbid cultures was extracted using the GeneJet Genomic DNA Purification Kit (Thermo Fisher). For phage DNA extraction, lysates from week 16 were used to infect B245. Confluent soft-agar bacterial lawns were collected, mixed with 4ml of Shieh media, centrifuged (10.000 rpm, 10 minutes, Sorvall RC34) and filtered. Phage precipitation was made with ZnCl2 followed by removal of host DNA with nucleases38. After Protease K treatment, the material was mixed with Guanidine:Ethanol (1 part 6M guanidine and 2 parts 99% ethanol, v.v.) and the extraction finished using the GeneJet Genomic DNA Purification Kit (Thermo Fisher). All samples were sequenced using 150PE BGISEQ platform at BGI Group. We were unable to obtain phage sequence data from the replicate b of the lake water with mucin condition due to technical problems.
Construction of the B245 reference genome
The ancestral B245 genome was assembled from Illumina reads using Spades 3.14.1 (--isolate mode)39. The resulting 487 contigs were combined to a single contig relying on a previously compiled complete F. columnare genome FCO-F2 (accession number CP051861) as reference using RagOO 1.140. The genome was annotated for the purpose of mutational analysis using dFast 1.2.341.
We used Breseq 0.35.142 to analyze mutations occurring in the phage and bacterial genomes. Since the phage samples represented mixed phage populations, Breseq was run in --polymorphism-prediction mode. Bacterial samples were run in default mode. As references we used the ancestral B245 bacterial genome or the previously published V156 phage genome31. To reduce noise in phage mutation, we used a value of 0.05 for --polymorphism-bias-cutoff. During analysis of the phage genomes, we discovered a large number of identical polymorphisms in some samples. As it is very unlikely that such mutations occur independently, we discarded any identical mutations that co-occur in two or more samples as false positives. When investigating mutations in unknown putative phage genes, we used HHpred to predict protein function43.
A second experiment was made using only the lake water supplemented with 0.1% mucin condition (LW + M). The experimental setup was similar to the growth experiment described above, but besides adding 5×104 cfu F. columnare B245 and 5×103 pfu phage V156, we also included 5×104 cfu of Aeromonas sp. B135 and of E. coli DSM613 as competitors. Cultures containing all three bacteria and V156, any combination of two bacteria and V156, and only F. columnare B245 host and V156 were made. Another control consisted of cultures containing only the V156 phage to follow its stability over time in the absence of its host. Each condition was tested in triplicates, in a final volume of five milliliters, and incubated at 25 degrees under 120 rpm. For this competition experiment we followed phage and bacterial titers, spacer acquisition in F. columnare and colony morphology.
All data analysis and statistic were done in R 3.5.3 using RStudio 1.1.463. Model comparisons and inclusion of random effects was aided by Akaike’s information criterion (AIC) comparison when applicable44. For details, see Supplementary Information.
Parasitism by bacteriophages has led to the evolution of a variety of defense mechanisms in their host bacteria. However, it is unclear what factors lead to specific defenses being deployed upon phage infection. To explore this question, we exposed the bacterial fish pathogen Flavobacterium columnare to its virulent phage V156 in the presence of a eukaryotic host signal (mucin). All tested conditions led to some level of innate immunity, but the presence of mucin led to a dramatic increase in CRISPR spacer acquisition, especially in low nutrient conditions where over 60% of colonies had obtained at least one new spacer. Additionally, we show that the presence of a competitor bacterium further increases CRISPR spacer acquisition in F. columnare. These results suggest that ecological factors are important in determining defense strategies against phages, and that the concentration of phages on metazoan surfaces may select for the diversification of bacterial immune systems.