Mucin induces CRISPR-Cas defense in an opportunistic pathogen

Presence of mucin stabilizes survival of both the bacterium and the phage during 16 weeks of co-culture

An overview of our main experimental setup is shown in Fig. 1. To avoid population bottlenecks, our sampling was based on the weekly collecting 20% of the cultures and replacing with the same volume of fresh medium. Long term co-existence of both F. columnare B245 and its phage V156 was observed in all treatments. In lake water with (LW + M) or without mucin (LW), the closest approximations of natural conditions for F. columnare, the phage titers remained similar until week 9, after which LW + M showed a significant decline in phage numbers compared to LW (LM, t_1,46 = −2.737, P = 0.0088) with roughly a ten-fold difference at week 16 (Fig. 2a, Supplementary Fig. 1a). Bacterial population densities in these treatments were opposite and more dramatic, with an average of 45-fold higher numbers in LW + M than in LM across all time points after an initial spike at week 1 (LM, t_1,77 = 4.836, P < 0.001) (Fig. 2b, Supplementary Fig. 1b). Surprisingly, bacteria in the no-phage control of LW became extinct after week 10, while no extinction occurred in the phage-containing cultures.

Cultures in Shieh medium (without mucin) had large variations between replicates in both bacterial and phage titers. Despite similar bacterial titers at week one, differences between replicates grew to more than 100-fold towards the end of the experiment (Fig. 2b). Phage titers declined in Shieh for the first 4–5 weeks, after which replicates b and c recovered while replicate a stabilized. Around week 10, phage titers declined sharply in replicates a and replicate c, while replicate b remained with high titer until the end (Fig. 2a). The presence of mucin in Shieh decreased variation between replicates compared to Shieh alone, with higher phage and bacterial titers (Fig. 2). No statistical analysis was performed on the Shieh cultures due to high divergence in replicates.

Presence of mucin enhances spacer acquisition in type II-C and VI-B CRISPR-Cas loci

Culture conditions, especially the presence of mucin, had a significant impact in the acquisition of new CRISPR spacers (Fig. 3a, b). The presence of mucin in lake water increased spacer acquisition 9-fold compared to plain lake water (GLMM, Z = 4.271, P < 0.001), and the efficiency of the acquisition was 5-fold when comparing lake water with mucin to Shieh with mucin (GLMM, Z = −3.367, P < 0.001) (Fig. 3b). Timewise, the maximum efficiency of the acquisition was reached around week 3 in the LW + M treatment, with over 60% of LW + M colonies having acquired new spacers (Fig. 3a). The number of spacers in one locus in a single isolate was up to six spacers in LW + M, two in LW and up to three in Shieh with mucin (Fig. 3b). The efficiency of spacer acquisition was roughly similar between the two CRISPR loci regardless of the treatment. Shieh without mucin did not show any new spacers.

Surprisingly, spacer acquisition was not exclusive to the Rhizoid morphotype. In fact, in Shieh + M treatment, isolates with the highest number of spacers were Rough, indicating an overlap of surface modification and spacer acquisition (Fig. 3b). However, after a peak in isolates with CRISPR spacers in this treatment around week 7, no more spacer mutants appeared. When only considering morphotype, LW had no Rhizoid colonies while other conditions had a minor bias towards them (Fig. 3c).

Co-culturing with phage leads to immunity

To detect the development of phage resistance and any associated costs during the 16-week co-culture experiment, we grew bacterial isolates obtained during the experiment in the presence or absence of the ancestral phage and compared the maximum OD (OD_MAX) reached to that of ancestral B245. In principle Rhizoid colonies are phage susceptible while Rough colonies are phage resistant due to SM. In the absence of phage, the OD_MAX of the four treatments (LW, LW + M, Shieh, Shieh+M) did not significantly differ from the ancestral B245 (Fig. 4a). In the presence of phage, however, the OD_MAX of the ancestral B245 decreased from 0.41 to 0.19 (GLMM, Z = −13.78, P < 0.001), while OD_MAX in the four treatments remained unchanged, indicating phage resistance (GLMM, Z = −0.281 to −0.599, P > 0.3 in all) (Fig. 4a). These results suggest that co-culturing F. columnare with phage caused phage resistance to evolve in all four treatments.

We also measured the growth of isolates from the control cultures from the 16-week experiment (same conditions but without phage). As expected, isolates from most treatments had lower OD_MAX in the presence of phage compared to the absence of phage (Fig. 4b). The only exception was the Shieh + M isolate, which had a significantly lower OD_MAX in the absence of phage compared to the ancestral bacterium (OD 0.41) with a predicted OD of 0.276 (LM, t_2,24 = −3.053, P = 0.0055). In the presence of phage, however, this control had a predicted OD of 0.287, which is significantly higher than that of the ancestor’s 0.199 (LM, T_2,24 = 3.565, P = 0.0016). These results suggest that prolonged incubation in Shieh + M in the absence of phage made the cells grow slower, but also made them phage resistant, perhaps incidentally through adaptation to the nutrient-rich mucin environment.

Fitness benefits of CRISPR adaptation depends on the environment

We next investigated how spacer acquisition in different morphotypes affected bacterial growth. This analysis was only done for the LW + M and Shieh+M treatments, which produced enough isolates with expanded CRISPR arrays for statistical analysis. In the absence of phage, Rhizoid and Rough isolates with native CRISPR arrays from the LW + M treatment reached predicted OD_MAX of 0.173 and 0.421, respectively (Fig. 5a). This difference was statistically significant (GLMM, Z = 6.34, P < 0.001). However, the acquisition of one or more spacers almost doubled the predicted OD_MAX of Rhizoid isolates from 0.173 to 0.337 (GLMM, Z = 5.56, P < 0.001). The effect of new spacers was opposite for rough isolates, whose OD_MAX was reduced by 20% to a predicted 0.340 (GLMM, Z = −5.326, P < 0.001). These effects were similar and significant also in the presence of phage (Fig. 5). Sanger sequencing of selected bacterial isolates revealed that all sequenced spacers in the type II-C CRISPR-Cas locus were targeting the phage (16/16), while in the VI-B locus roughly two-thirds (19/28) were targeting the phage and the rest (9/28) targeting the bacterial genome (Supplementary data 1).

In Shieh+M samples without phage, Rough isolates had a significantly higher OD_MAX than Rhizoid ones (0.400 vs 0.339, GLMM, Z = 2.240, P = 0.025) (Fig. 5b). Surprisingly, new spacers increased OD_MAX of Rough isolates by 40% (GLMM, Z = 2.589, P = 0.01) but did not affect Rhizoid OD_MAX. Again, results were similar when these isolates were subjected to phage (Fig. 5b).

Overall, CRISPR spacer acquisition has a positive impact on Rhizoid colonies and a negative impact on Rough colonies when the isolates originate from an environment that favors spacer acquisition (LW + M, Fig. 1a). However, when isolates originate from a nutrient-rich environment that does not favor CRISPR-Cas (Shieh + M, Fig. 1a), there is no or little benefit from the acquired spacers.

Competition in lake water supplemented with mucin enhances spacer acquisition

After detecting that mucin in lake water caused an increase in CRISPR spacer acquisition, we performed a follow-up experiment testing the effect of competing bacterial species on F. columnare spacer acquisition in this condition. Our aim was to see if the presence of a competing bacterial species would provoke F. columnare to maintain phage defenses that enable efficient nutrient intake (CRISPR-Cas) in a competitive setting. Initially, we chose Aeromonas sp. which responds to mucin similarly as F. columnare by forming biofilm and by becoming more susceptible to phage infections, and E. coli DSM613 for apparently not being affected by mucin exposure⁷. However, the E. coli populations were either extinguished quickly (no colonies were seen in the first samplings) or remained at very low levels (a few colonies were seen at the last time point) during the experiment (Supplementary fig. 2). This, allied to the fact that we could not measure which effect the initial input of E. coli had to mucin content in the cultures, led us to discard the E. coli -containing conditions from the final analysis. The presence of Aeromonas sp. significantly enhanced F. columnare spacer acquisition as measured by the total number of spacers acquired across CRISPR loci (Fig. 6). In the absence of Aeromonas sp., the expected total number of new spacers in F. columnare was 0.57 per colony, whereas in a co-culture each colony was expected to acquire 1.64 spacers (GLMM, Z = −3.381, P < 0.001). Interestingly, in this experiment the Rhizoid colony type bacteria acquired more spacers than the Rough types (Fig. 6).

Given the association of Rhizoid colony type with spacer acquisition in this competitive environment, we tested if morphotype affects F. columnare competetiveness. To test this, we grew Aeromonas sp. in the presence of either Rhizoid or Rough F. columnare. As controls, both species were grown alone. The presence of Rhizoid F. columnare dramatically reduced the concentration of Aeromonas sp. in all time points compared to the Aeromonas sp. control (GLM, t_2,23 = −8.185, P < 0.001), whereas rough F. columnare did not have a significant effect on Aeromonas (GLM, t_2,23 = −1.580, P = 0.128) (Fig. 7).

Genome analysis

We sequenced phage and bacterial genomes to search for genetic variations resulting from coevolution. We also sequenced the control bacterial genomes (evolved without phage) to differentiate between mutations caused by interaction with phage and those that may arise from different culturing conditions.

Phage genomes were investigated on the population level: at the end of the 16-week experiment, phage samples of each culture (representing the variety of phages present in that culture) was used to infect the ancestral bacterium and the resulting lysate was deep sequenced. We found an abundance of shared mutations across multiple replicates. Due to the low likelihood of the same mutations occurring convergently across multiple samples (suggesting common sequencing artefacts), we discarded most mutations as false positives (however, all mutations are listed in Supplementary data 2). We were confidently able to recover individual mutations in only two cultures: Replicate c from Shieh + M treatment had a non-synonymous mutation (A₃₅₇T) in a predicted phage baseplate protein and a nonsynonymous mutation (M₂₀₀V) in a putative protein with no predicted function (Table 1). Replicate b from Shieh treatment had a non-synonymous mutation in a protein with an unknown function near the aforementioned baseplate protein (V₃₀₀I), as well as a synonymous mutation (K₁₀K) in a predicted DNA helicase (Table 1). Common to both replicates is that they showed a drop in phage titer after the initial spike but recovered towards the end (Fig. 2a). In fact, replicate b in Shieh treatment was the only replicate in its treatment group in which the phage titer remained high at the end of the experiment.

Bacterial genomes were investigated on the isolate (clonal) level. Isolates were picked from different treatments at several time points during the experiment. Evidence of surface modification was found in almost all phage-exposed samples in the flavobacterial gliding motility genes which are associated with type IX secretion system and whose mutations are expected to cause colony morphotype change from Rhizoid to Rough³⁵. Most of these mutations caused premature stop codons or introduced frameshift mutations (Table 2). Most of the T9SS mutants were Rough (Table 2). Isolates from the control treatments without phage did not show mutations in gliding motility genes but had variation in other ORFs. It is therefore possible that extended growth in these conditions introduced other adaptive changes, although these mutations were not present in the phage-exposed samples. Indeed, most variation was detected in the Shieh + M control isolate, which seemed to have had developed resistance against phage even in the phage’s absence (Fig. 2b). This sample contained non-synonymous mutations in several metabolism related genes (e.g. Lon, rpoB and surE), and in a putative type VI secretion system-like gene (Supplementary data 3). Type VI secretion systems have previously been shown to be associated with host colonization and bacterial antagonism in Flavobacterium johnsoniae³⁶. Mutations in the gld and spr genes resulted in resistance against the ancestral V156 phage.

Source: Ecology - nature.com