Horizontal Gene Transfer Between Subspecies Affects Bacterial Genome Dynamics

Different bacterial subspecies live in close proximity in soil. Horizontal gene transfer enables them to exchange genetic information. Little is known about the efficiency of gene transfer and whether it occurs randomly across the genome. In the first part of this thesis, we designed an evolution experiment and analysis algorithm to character- ize genome dynamics in the presence of a different subspecies. Bacillus subtilis is naturally competent for transformation: it takes up DNA from its environment and integrates segments into its chromosome by recombination. Eight clonal populations of B. subtilis 168 (Bsu168) were evolved with genomic DNA from B. subtilis W23 (BsuW23) for 21 cycles. To minimize variability in transformation rates as a func- tion of time, we generated a strain with inducible competence, and competence was induced once per cycle. Evolved cultures had more than 100 homologous recombination events, per replicate, and the length of recombined segments was exponentially distributed with a characteristic length of 3500 bp −1 . Average recipient genome replacement occurred at a constant rate over the course of the experiment, averaging 0.47% replacement every cycle. In addition to homologous recombined-segments, de novo segments from the BsuW23 auxiliary genome and de novo variants were also detected. The de novo segments from the auxiliary genome had a mean length of 2.2 kbp. Bacteria evolving in the presence of BsuW23 DNA showed five times as many de novo variants in regions upstream of genes and five times as many missense indels, compared to control samples receiving either no DNA or self (Bsu168) DNA. Of those upstream mutations, 75% were inside a recombined segment. This hinted at the possibility of those mutations being compensatory mutations, as upstream and missense mutations are likely to affect gene expression levels. We conclude that the recipient genome was replaced by donor genome at a constant rate and constant segment-length distribution, up to a total of 10% genome replacement. Introduction of de novo variants is likely to affect the levels of gene expression. The probability of replacing a specific gene was in good agreement with a binomial distribution, suggesting that replacement occurred close to random across the genome. However, we found important deviations from random integration. At both the single-cell level and the population level, we obtained evidence that homologous recombination does not occur stochastically. At the single-cell level, imported segments had a higher average identity, 93.6 %, than the Bsu168 and BsuW23 inter-subspecies identity of 92.4%. Interestingly, recombined segments had one end of integration with a significantly higher identity than expected from simulations using the same length distribution. We found that the increased sequence identity extends to roughly 500 bp. The bias towards higher than average sequence identity is most likely caused by the recombination process. At the population level, we found further evidence that recombination did not occur randomly. On one hand, several genes, such as leu and eps, were presumptively selected for, as they were replaced in nearly all replicates. On the other hand, recombined donor segments were underrepresented in prophages and mobile-elements genes, most likely because they correspond to auxiliary genes in each subspecies. Essential genes were overrepresented, plausibly because essential genes have a higher average identity compared to nonessential genes. There was a preference to replace genes and operons fully. Two-thirds of all affected genes and operons were fully replaced. Full gene replacement is explained, in part, by the average import length being 1.9 kbp, ∼2 genes. The average operon length is also comparable, 3.2 kbp. To summarize, homologous replacement is biased to higher than average sequence-identity. Given the length distribution of recombined segments, the generation of hybrid genes is less likely than the replacement of full genes. Only for a few genes is there evidence of selection. In the second part of this thesis, we investigated the cost of competence for transformation in the stationary growth phase. We characterized the competition dynamics of strains with various probabilities of entering the K-state, to quantify the effect the K-state has on fitness in B. subtilis. Relative fitness was found to decrease with increasing probability of entering the K-state, during both the exponential and stationary growth phases. Using a microfluidic chamber, we were able to characterize generation times in the stationary phase for both K-state and non-K-state cells. We found a strong cost of competence due to growth inhibition, even in the stationary state. These findings emphasized that the stationary phase is dynamic. We conclude that gene transfer between subspecies of B. subtilis is highly efficient, with 10% of the chromosome being replaced in a total of 42 h of competence during the evolution experiment. DNA uptake in B. subtilis 168 from donor strain B. subtilis W23 occurred at a constant rate and fairly random. Recombined segments had higher identities than the between-subspecies average, pointing to the importance of sequence divergence in recombination. Several alleles were replaced in the majority of the replicates, indicative of a selection advantage. It has been proposed that recombination is beneficial because it can cure the recipient from prophages. In contrast, here we observed that replacement of essential genes was overrepresented, while prophage genes were hardly affected. This suggested that gene transfer between subspecies functions as genome maintenance. The appearance of variants, including indels, in possible regulatory regions calls for studies of the effects of recombination on transcription. We speculate that recombination may be suppressed between subspecies by other factors, such as quorum sensing