Chromosomal Position of Ribosomal Protein Genes Affects Long-Term Evolution of Vibrio cholerae

ABSTRACT It is unclear how gene order within the chromosome influences genome evolution. Bacteria cluster transcription and translation genes close to the replication origin (oriC). In Vibrio cholerae, relocation of s10-spc-α locus (S10), the major locus of ribosomal protein genes, to ectopic genomic positions shows that its relative distance to the oriC correlates to a reduction in growth rate, fitness, and infectivity. To test the long-term impact of this trait, we evolved 12 populations of V. cholerae strains bearing S10 at an oriC-proximal or an oriC-distal location for 1,000 generations. During the first 250 generations, positive selection was the main force driving mutation. After 1,000 generations, we observed more nonadaptative mutations and hypermutator genotypes. Populations fixed inactivating mutations at many genes linked to virulence: flagellum, chemotaxis, biofilm, and quorum sensing. Throughout the experiment, all populations increased their growth rates. However, those bearing S10 close to oriC remained the fittest, indicating that suppressor mutations cannot compensate for the genomic position of the main ribosomal protein locus. Selection and sequencing of the fastest-growing clones allowed us to characterize mutations inactivating, among other sites, flagellum master regulators. Reintroduction of these mutations into the wild-type context led to a ≈10% growth improvement. In conclusion, the genomic location of ribosomal protein genes conditions the evolutionary trajectory of V. cholerae. While genomic content is highly plastic in prokaryotes, gene order is an underestimated factor that conditions cellular physiology and evolution. A lack of suppression enables artificial gene relocation as a tool for genetic circuit reprogramming.

IMPORTANCE The bacterial chromosome harbors several entangled processes such as replication, transcription, DNA repair, and segregation. Replication begins bidirectionally at the replication origin (oriC) until the terminal region (ter) organizing the genome along the ori-ter axis gene order along this axis could link genome structure to cell physiology. Fast-growing bacteria cluster translation genes near oriC. In Vibrio cholerae, moving them away was feasible but at the cost of losing fitness and infectivity. Here, we evolved strains harboring ribosomal genes close or far from oriC. Growth rate differences persisted after 1,000 generations. No mutation was able to compensate for the growth defect, showing that ribosomal gene location conditions their evolutionary trajectory. Despite the high plasticity of bacterial genomes, evolution has sculpted gene order to optimize the ecological strategy of the microorganism. We observed growth rate improvement throughout the evolution experiment that occurred at expense of energetically costly processes such the flagellum biosynthesis and virulence-related functions. From the biotechnological point of view, manipulation of gene order enables altering bacterial growth with no escape events.
replication, while Chr2 fires its replication origin (ori2) only after two-thirds of Chr1 is duplicated, and both chromosomes complete DNA synthesis synchronously (48). Nevertheless, replication-dependent gene-dosage effects exist in both replicons (30,49,50), while their relative timing modulates the gene dosage difference between both replicons. In Vibrionaceae, transcription and translation genes map close to the oriC of Chr1 (ori1) (31,49).
We have previously studied the chromosomal position of ribosomal protein (RP) genes to uncover the link between gene order and cell physiology (24,51). For this, we manipulated the genomic location of V. cholerae major ribosomal protein locus s10spc-a (S10), a 13-kbp array of essential genes highly conserved among prokaryotes and also in eukaryotic organelles (52,53). By relocating the S10 locus throughout both chromosomes (49), we generated a S10 movant strain set (i.e., isogenic strains in which the genomic position of the locus of interest is altered) whose growth rate, infectivity, and fitness decreased as a function of the distance between S10 and ori1. Relocation to the Chr2 showed a detrimental effect accordingly. These phenotypes were caused by the reduction of S10 genome-wide copy number that were mostly consequence of macromolecular crowding alterations rather than a deficit in the capacity to synthesize proteins (49,50,54). Further increasing S10 copy number was either deleterious or neutral. Overexpression of nonribosomal protein genes within S10 could not rescue the growth rate, indicating that these alterations were exclusively due to ribosomal protein genes (54). Genomic position of S10 maximizes cell physiology and fitness throughout both parts of V. cholerae life cycle. Meanwhile, S10 is close to ori1 among all Vibronaceae. Since this clade diverged from enterobacteria some 500 million years ago (55), such genome positioning is likely to be the result of selection in the long run rather than a trait acquired through genetic drift. In the last years, long-term evolution experiments emerged as the approach of choice to directly assess evolutionary trajectories of (micro)organisms (56)(57)(58)(59).
In this work, we showed that the genomic position of the major ribosomal protein locus conditions V. cholerae evolution in the long term. Toward this aim, we experimentally evolved strains in which the S10 locus was either placed at its original position or at the terminal region of the chromosome. We showed that the growth rate reduction observed in the latter case persisted over 1,000 generations of evolution, providing experimental support to the biased genomic position of these genes. In particular, we observed that, among other mutations, increased biofilm formation and growth rate occur at the expense of motility during in vitro environment evolution. However, no mutation could suppress S10 genome position, a finding that provides key evidence showing the importance of the long-term impact of gene order on genome evolution. We also provide novel evidence and observations on how V. cholerae evolves in vitro.

RESULTS
V. cholerae evolution experiments. Since we aimed at understanding how the genomic location of S10 could condition long-term evolution, we conducted the experimental evolution of six populations in which the locus is close to ori1 (blue) and six populations in which S10 resides at termini of the chromosomes (red) for 1,000 generations ( Fig. 1,  upper panel). For the six populations harboring S10 close to ori1, three populations were derived from a strain in which S10 was not transposed (49) (Fig. 1; Table S1, populations P1 to P3). The other three populations were evolved from S10Tnp-35, a movant in which the S10 locus was slightly relocated ( Fig. 1; Table S1, populations P3 to P6) that displays no fitness or growth rate defect (Fig. 1). It is isogenic with respect to the rest of the movants and shows only eight genes that are differentially expressed with respect to the parental strain (54). It is a control that any possible genetic change associated with S10 relocation per se could affect the evolution of V. cholerae.
In parallel, we evolved six populations where S10 was relocated far from ori1. We used three isolated clones of S10Tnp-1120 (Populations P7 to P9) and three clones of S10TnpC2 1 479 (Table S1, populations P10 to P12) in which S10 was moved to the ter region of Chr1 and Chr2, respectively (Fig. 1, upper panel, red). These are the most phenotypically affected movants, displaying an ;15% loss of maximal growth rate and lower fitness than the parental strain ( Fig. 1, lower panel, red) (50).
Due to the fast growth of V. cholerae, we passaged them twice a day in lysogeny broth (LB) supplemented with streptomycin (Str) to avoid contamination. The populations were regularly frozen. We also isolated single clones at 250 and 1,000 generations for later analysis.
Evolved V. cholerae populations increase their biofilm formation capacity. First, we noticed the emergence of strong biofilms in the Erlenmeyer flasks as early as generation 81 (G81). By generation G100, most populations displayed strong biofilms at the air-liquid interface and/or dense cell aggregates in the medium (Fig. S1A). By G250, biofilms were present in all populations. Accompanying this phenomenon, colonies of abnormal morphology appeared, particularly rugose colonies (Fig. S1B). The subculture of isolated smooth (S) and rugose (R) colonies showed that only the latter formed strong biofilms in the flasks (Fig. S1C). To quantitatively address the biofilm-forming capacity of the populations throughout experimental evolution, we cultured them statically in 96-well polystyrene plates, washed them, and then stained the wells with crystal violet at G0, G100, G250, G500, G700, and G1000 (Fig. 2). As positive and negative controls, we analyzed R1, a rugose clone isolated from P1 at G1000 (see above), and the ancestral parental strain, respectively. At the beginning of the experiment (G0), the levels of crystal violet stain were minimal. Then, biofilm-formation capacity increased rapidly in all populations. On average, the biofilm formation capacity reached the maximum at ;G250 (Fig. 2), although some specific populations displayed an earlier peak FIG 1 Genotype and associated phenotype of the evolved strains at the start of the experiment. Genome structure of the parental, the movant strains subjected to long-term evolution for 1,000 generations. Chromosome I and II are depicted as ovals. The orange arrow represents S10 at its genomic position with respect to ori1 (green dot) or ori2 (yellow dot). Colors are strain specific, and we indicate the population numbering assigned. Strains in which S10 is close to ori1 are colored with shades of blue, while those having the locus in a distal position are colored with shades of red. The maximum growth rate (m, gray boxes, right axis) and the relative fitness (W rel , white boxes, left axis) with respect to the parental strain obtained in previous works were plotted for each strain. The m obtained from the slope of the growth curve during exponential phase in fast-growing conditions (118). The fitness of each strain was measured by pairwise competition against a parental strain. The results are shown in standard box-and-whisker plots showing the medians, minima, and maxima. The detailed calculations of these parameters can be found in references 49 and 50. Statistical significance was analyzed using nonparametric tests and the Dunn test for multiple comparisons of the means obtained for each strain. ****, P , 0.0001. Experimental Evolution on Ribosomal Protein Gene Location mBio near ;G100 (Fig. S2). After G250, biofilm formation capacity decreased in all evolved populations without going back to initial G0 baseline levels. The emergence of rugose colonies paralleled the evolution of biofilm-forming capacity (Fig. S1D). All populations displayed a similar trend, suggesting that it was selected throughout the experiment independently of S10 genomic location ( Fig. 2; Fig. S2), being a common trait for V. cholerae in vitro evolution. Evolved populations display a growth rate increase. Previous studies showed that bacterial populations increased their fitness over time compared to the ancestral clone (58,60). We employed maximum growth rate (m) for fitness estimation, since in previous works, fitness measured by pairwise competitions and m changed similarly as a function of S10 genomic location (50). The evolved populations were subjected to automated growth curves. For this, at each time point, the 12 populations were diluted 1 in 1,000 and cultured in LB at 37°C measuring optical density at 620 nm (OD 620 ) in triplicate at least twice. In each experiment, the four ancestral strains (parental, S10Tnp 1 35, S10Tnp-1120, and S10TnpC2 1 479; Table S1) were used as a control. The value of m was calculated from the growth curves and relativized to the parental strains to remove variations between experiments. Then, we plotted the maximum growth rate as a function of elapsed generations ( Fig. 3A; Fig. S3A). During the first 150 generations, m remained stable among populations. As expected (49,50), populations derived from strains in which S10 is close to ori1 (P1 to P6, blue) grew ;15% faster than those originated from the most affected movants (P7 to P12, red). After this period, between G150 and G250, populations show a sudden m increment. All populations increased their growth rate throughout experimental evolution, but FIG 2 Evolution of biofilm formation capacity throughout the evolution experiment. Biofilm formation phenotype of 12 evolved populations from strains in which S10 is close to ori1 (shades of blue) or at a distal position from replication origin (shades of red) is shown. (A) Representative wells in which cells were stained with crystal violet are shown for P1 to P12 at different time points. Par, parental strain; R1, rugose clone used as negative and positive controls for biofilm-forming capacity. (B) The crystal violet stain was quantified for all populations throughout generations. The parental (light blue, P1 to P3), S10Tnp-35 (dark blue, P4 to P6), S10Tnp-1120 (light red, P7 to P9), and S10TnpC2 1 479 (red, P10 to P12) movants populations values were averaged for simplicity ( Fig. S2 shows the disaggregated data). The data are averages of the means from three independent experiments done in quadruplicate. Statistical significance was analyzed by ordinary two-way analysis of variance (ANOVA) two-tailed test with main effects only. Then, Tukey's test was done to compare the mean Abs 570 values. The factors analyzed were strains and generations. There were no significant differences in Abs 570 values between strains. Letters denote statistically significant differences in the means between generations (P , 0.05).
Experimental Evolution on Ribosomal Protein Gene Location mBio this rise occurred in a stepwise manner, with the most notable occurring at G250. After G500, the alterations in growth rate become noisier, since populations originating from the same strain tend to diverge ( Fig. S3A and C). Despite the general m increment through generations, the differences between populations evolving from faster strains in which S10 is close to ori1 (P1 to P6) and slower ones in which S10 was relocated to chromosome termini (P7 to P12) persisted throughout the experiment ( Fig. 3B; Fig. S3A to C). The evolutionary trajectory of this trait was similar for all of them (Fig. S3D). Importantly, we did not notice suppressor mutants compensating growth rate deficit in P7 to P12. Hence, mutations able to compensate genomic position of S10 locus are very difficult to achieve, at least in this time frame. This strongly suggests that the S10 genomic location conditions growth rate throughout experimental evolution, strongly indicating that the position of the main ribosomal protein locus determines fitness in the long run.
Positive selection drives genome dynamics in the first 250 generations of experimental evolution. To better characterize mutations occurring through the experiment, we deep-sequenced each ancestral strain as a reference and then each population at G250 and G1000. We speculated that mutations at G250 would correlate to the earlier increase in growth rate and the emergence of biofilm-formation capacity.
Deep sequencing of the 12 populations at G250 revealed 128 mutations observed in more than 25% of the reads obtained from each culture (Data Set S1A). Among them, we found 45 different mutations among all populations. The number of fixed DNA changes varied from 2 to 21 between each individual population (Fig. 4A). On average, each fixed mutation occurred in two or three populations, strongly indicating that they were under positive selection. Most of them (26 of 45) corresponded to nonsynonymous mutations or frameshifts of coding sequences (Fig. 4B). These changes are likely to alter or interrupt their encoded functions. Meanwhile, 17 changes (42%) occurred a priori in noncoding regions such as intergenic spaces and pseudogenes (Fig. 4B). We detected two (4.4%) synonymous mutations. When the mutations were grouped by coding sequences (CDS), we found that at least 17 genes suffered alterations (Data Set S1). Among them, six genes presented more than one mutation across multiple populations: aroG. ompT, flrA, flrB, flrC, and tagE2 (Table 1). First, ompT (61) is a highly expressed gene encoding one of the main porins of V. cholerae. tagE2 (VCA1043) was found to be altered in most of the populations. There is little FIG 3 S10 genomic location affects growth rate evolution in the long-term. (A) V. cholerae increases its growth rate throughout experimental evolution. At every time point, the populations from the frozen "fossil record" were subjected to automated growth curves and relativized to their ancestral strains. The strains in which S10 is close to ori1 are depicted in shades of blue, while those were bearing the locus at an ori1 distal location are shown in shades of red. We show the averages of parentals (light blue), S10Tnp-35 (blue), S10Tnp-1120 (light red), and S10TnpC2 1 479 (red) strains. The graph shows the average of at least two experiments done by triplicate of m 6 standard error of the mean (SEM) as a function of the generations elapsed. The average m from the ancestral strains is shown as dotted lines for reference. Individual populations are plotted in Fig. S3A. Statistical significance was assessed using ordinary two-way ANOVA two-tailed test with main effects only. Then, Tukey's test was done to compare the mean of m. Letters denote groups being statistically different. (B) Growth rate defect of movants persists through 1,000 generations of evolution. The obtained m for each population (P4 to P12) was relativized to growth rate of the Parental populations (P1 to P3) for each time point. Then, the mean with SEM of percentage of m variation was plotted as a function of generations. The dotted lines represent the average deviation from growth of the parental for each ancestral strain already shown in Fig. 1 (0% for parental, 0.79% for S10Tnp-35 and 215.5% for S10Tnp-1120 and 217.2% for S10TnpC2 1 479). Statistical significance against parental strain was assessed using ordinary two-way ANOVA two-tailed test with main effects only. Then, Tukey's test was done to compare the means. Across all generations, the percentage of m is significantly lower in strains in which S10 locus is far from ori1, and S10Tnp-35 is not significantly different than the parental strain. ****, P , 0.0001. Experimental Evolution on Ribosomal Protein Gene Location mBio evidence of its encoded function, although a recent study shows that it is an endopeptidase necessary for cell growth (62). The aroG gene is required for the synthesis of aromatic amino acids, although little work has been done in V. cholerae. Finally, we found single nucleotide polymorphisms (SNPs) at flr genes, the main regulators of flagellum biosynthesis (39,63,64). FlrA is the master regulator of flagellum biosynthesis that controls the expression of flrB and flrC, which in turn are downstream of the flagellar transcription hierarchy (39). We reasoned that this multiplicity of independent mutational events within the same regulation pathway suggests a fitness benefit from motility inactivation. Meanwhile, the gene order seems unaltered among all populations ( Fig. S4 and S5).  As a general scenario, we found that at G250 mutations occur several times independently, suggesting a positive selection of most of them rather than genetic drift. In particular, we observed that a majority of the populations displayed mutations at the main flagellar regulators (Data Set S1A; Table 1).
Evolutionary forces change beyond G250: adaptative, nonadaptative mutations, and hypermutators after 1,000 generations of experimental evolution. We speculated that the mutations observed at G1000 may reflect those that are fixed in the long run, since after ;G500, biofilm formation decreases and growth rate remains roughly constant. After 1,000 generations of experimental evolution, we observed 512 fixed changes in at least 25% of the population at 396 sites of the genome of V. cholerae (Fig. 4A, right bars). In contrast to the G250 scenario, few mutational events occurred in more than one population (1.3 events per population on average). We observed a lower proportion of mutated pseudogenes (4.5% in G1000 versus 26.6% in G250) and an increase in the proportion of synonymous mutations (4.4% versus 19.4%) (Fig. 4B). We also noticed mutational events at several genes involved in DNA repair, such as mutT (VC1342), recBCD (VC2319-VC2320), mutS (VC0535), and recJ (VC2417). In particular, mutL (VC0345) was mutated in 8 of 12 populations (P1, P2, P3, P4, P5, P6, P7, and P11). Populations with intact DNA repair mechanisms (i.e., P8, P10, and P12) were those showing the lowest number of mutations (Fig. 4A). These data suggest that in most populations at G1000, many of the identified mutations may result from a hypermutator phenotype rather than selection. In all populations in which S10 is close to ori1 (P1 to P6), we detected one or more mutations in DNA repair systems (Fig. 4A, blue). In contrast, half of the populations bearing S10 at an ori1 distal position (P7 to P12) displayed an intact DNA repair pathway (Fig. 4A, red).
Although not the most frequent scenario, we also detected many DNA changes in the same gene or operon occurring many times independently or emerging in parallel in more than one population, likely indicating positive fitness selection throughout the evolution experiment (Table 2). On top of aroG, ompT, flrABC, and tagE2 already acquired at G250, we found many mutations in interlinked cellular processes (38,43,44) such as rugosity, biofilm structure, and regulation (vpsA, vpsB, vpsP, vpsR, capK, and rbmB) (65-68), quorum sensing (luxP, luxO, and luxU) (69), flagellum biosynthesis (flgA, flgI, flgK, fliG, fliO, flag, and flhA) (39), and chemotaxis (cheY and cheR2) (39,70). We also identified several altered functions such as extracellular macromolecular structures linked to virulence (toxin coregulated pilus [tcp], mannose-sensitive hemagglutinin [MSHA] pilus, type 6 secretion system [T6SS]), iron incorporation and metabolism (particularly vibriobactin pathway genes and iscR), lipopolysaccharide biosynthesis (rfbA, rfbB, rfbV, and pgi), and phosphate sensing and incorporation (phoR, phoU, and phoH). Interestingly, the latter ones have been linked to biofilm downregulation (71). Most of them were nonsynonymous or led to frameshifts likely to be detrimental for their biological function (Data Set S1B). We interpret these mutations in light of a change from a host-environmental life cycle to the relatively constant culture condition of the evolution experiment. Functions such as virulence, swimming, chemotaxis, pili production, T6SS, and micronutrient scavenging are energetically costly structures under strong selection in V. cholerae natural environment that can become a burden during in vitro evolution.
Mutations linked to rugose phenotype. Throughout the experimental evolution, we observed the emergence of biofilms and aggregates linked to a rugose phenotype described earlier ( Fig. 2; Fig. S1 and S2). To uncover genomic changes associated with biofilm emergence, we fully sequenced the genome of a rugose and a smooth clone from a fast-growing (P1) and a slow-growing population (P12). After 1,000 generations, the clones originated from P1 and P12 fixed 9 and 6 mutations, respectively (Fig. 4C), independently of their smooth (S) or rugose (R) phenotype (1 mutation fixed every 133 6 30 generations). Although some DNA changes were common among R and S colonies (Data Set S1C), we found some differential mutations potentially involved in biofilm and rugose phenotype emergence (Table 3). In particular, mutations in vpsR, vpsB, cdgB, and VCA1031 were previously linked to rugose phenotype by Yildiz and coworkers (46,72,73). vpsR is a positive regulator of the vps cluster that suffered a nonsynonymous mutation (D40Y) in the S colony derived from P1. The vpsB gene codes for an enzyme that participates in the synthesis of nucleotide sugar precursor necessary for polysaccharide biosynthesis. We detected a premature stop codon in the S colony collected from P12 (P12S). The P12R displayed a nonsynonymous mutation (D101G) at the cdgB, a GGDEF protein upregulated in rugose clones (73,74). It also showed a nonsynonymous mutation in the iscR gene that codes for a transcription factor regulating genes related to iron starvation, oxidative stress, and oxygen limitation  in Escherichia coli and linked to virulence in other bacteria (75). Meanwhile, the P1S colony displayed a mutation at VCA1031, encoding a putative methyl-accepting chemotaxis protein previously reported to be overexpressed in rugose strains having a genetic DcdgC background (46). In parallel, P1R showed an intergenic mutation also found at G250, at G1000, and on fast-growing clones (see above). Finally, P12S displayed a synonymous mutation in the rpoA gene, a main component of the RNA polymerase. We found two other differential mutations unlikely to be causal rugose phenotypes but that might have been coselected. In summary, the S and R clones showed alterations in genes previously linked to biofilm generation and c-di-GMP regulation. However, we detected other changes in R clones such as in iscR and at the intergenic region between VC0448 and VC0449.
An increase in growth rate is often associated with motility loss. To uncover mutations associated with fast growth, we isolated four smooth colonies and a rugose clone (S1 to S4 and R, respectively) from P1 to P12 at G250, the moment in experimental evolution when we noticed a sudden growth rate increase in all populations. Next, we performed automated growth curves of these clones using as a reference the parental strain (Table S1). We calculated the growth rate for each strain (Fig. 5A), and we selected the fastest growing clones, from P1 to P6 and from P7 to P12. We called these clones fast-growing clones followed by population of origin and its smooth or rugose status. Thus, we isolated the clones FG-2-S2, FG-3-S2, FG-7-S1, FG-10-S1, and FG-11-S1. Interestingly, the three former clones grew 7% to 10% faster than the parental strain. Meanwhile, FG-G250-7-S1, FG-G250-10-S1, and FG-G250-11-S1 improved their growth rate compared to their ancestors; they still displayed m reductions of ;7% with respect to the parental strain (Fig. 5B). This is another line of evidence indicating that S10 genomic location cannot be suppressed in strains in which this locus is located far from the replication origin.
To identify the mutations responsible of m increase, we fully sequenced FG-G250-2-S2, FG-G250-3-S2, FG-G250-7-S1, FG-G250-10-S1, and FG-G250-11-S1 genomes and compared them to the ancestral strains. Overall, strains displayed 1 to 3 changes in the DNA sequence (1 mutation fixed every ;140 generations) (Fig. 4C). In the 5 strains, we found 8 different mutations (Table 4). All isolated strains harbored alterations either at flrA (4 of 5) or at flrB (1 strain), the main flagellum regulators, which were previously observed at the population level at G250 (Table 1). Three other mutations accompanied the latter: a nonsilent mutation at a magnesium transporter (mgtE), a point mutation at an intergenic base between genes VC0448 and VC0449, and a frameshift mutation at ompT. Except for the frameshift mutation, we found the exact same mutations at G250 and most of them at G1000 (Data Set S1D). Analysis of gene order among the FG clones showed that genome structure did not suffer large inversions or translocations affecting S10 genomic location (Fig. S5). Therefore,  (Table S2), obtaining several independent mutant clones that we tested on automated growth curves (Fig. 6A). This mutation increased the growth rate with respect to the wild-type strain. Similarly, the introduction of a frameshift in flrB resulted in a growth rate improvement. Also, the introduction of a nonsynonymous mutation at mgtE displayed a similar phenotype (Fig. 6A). To quantify the effect of individual mutations, we determined the percentage of generation time variation (%GT). Mutants bearing the flrA frameshift displayed the strongest reduction with a 13.41% 6 4.7% reduction of their generation time (95% CI, 14.9 to 11.9; P , 0.0001). Meanwhile, flrB and mgtE mutants displayed a 6% generation time reduction (95% CI, 6.9 to 5.4 and 7.5 to 4.6, respectively; P , 0.0001). These results coincide with the observed increase in growth rate of FG clones. Simultaneously, frameshifts in flrA and flrB caused a reduction of motility as shown by swimming assays (Fig. 6B), demonstrating that growth rate increase is associated with the loss of flrA and flrB function.
In summary, these results demonstrate that the mutations observed in the FG clones increased the growth rate in V. cholerae. They also suggest that some of the mutations detected at G250 and maybe at G1000 also contributed to the growth optimization of this microorganism.

DISCUSSION
Increasing evidence suggests that gene order within the bacterial chromosome contributes to cellular homeostasis by coordinating key entangled processes such as chromosome structuration, cell cycle, replication, and the expression of genetic information. While genome reshuffling experiments have not been yet performed in bacterial systems (79), approaches such as comparative genomics (17,18,21,23,31,22), systems biology (20), large DNA inversions (80)(81)(82), and relocation of individual gene sets (16,19,(83)(84)(85) have provided support to this notion. Genes encoding the genetic information flow are interesting models to test the role of genomic position on cellular physiology. In this work, we took this "positional genetics" approach (in which genes of interest are systematically relocated within the genome) a step further, by observing how movants bearing S10 at different genomic locations evolve in vitro for 1,000 generations in fast-growing conditions, those in which phenotypic effects were stronger (Fig. 1).
a The details are described in Data Set S1D. FS, frame-shift mutation; ND, not determined.

Experimental Evolution on Ribosomal Protein Gene Location mBio
Pioneer work by Richard Lenski and coworkers on long-term evolution was done for more than 75,000 generations (and counting!) in E. coli. However, the strongest fitness improvement occurred during the first 2,000 generations (58,60). Therefore, although we performed a shorter experiment, we believe that we observed a representative part of the evolutionary trajectory. Meanwhile, we developed a different approach: we used rich medium and a model organism that grows almost twice as fast as E. coli. Also, our system used 1 log larger bottlenecks, improving the chance of mutant selection and lowering the effects of genetic drift (i.e., we transferred 10 9 instead of 10 8 CFU). Finally, we evolved populations bearing the key locus S10 at different genomic locations.
Along with this experiment, we found the emergence of abundant biofilms and aggregates in all populations. Other experimental evolution approaches also led to biofilms and/or aggregates. In Acinetobacter baylyii ADP1 (86), it was the consequence of insertion sequences (IS) action that led to evolution of aggregating clones. The studies on Burkholderia cenocepacia (87) and Bacillus subtilis (88) actively selected for biofilm emergence with a wide variety of clones with clear fitness advantages such as labor division and heterogenicity emergence within the structure. A limitation of our study is that we sequenced rugose clones at G1000 instead of G250, when most biofilms were observed. At this time point, many other unrelated mutations could be coselected with the DNA changes causing rugose phenotypes. We described biofilm  (Table 4) were introduced into wild-type V. cholerae using MuGENT. The growth rate of these mutants was determined in automated growth curves performed by triplicate (n = 11). The left panel shows a representative growth curve. The right panel represents the median 6 CI of the percentage of generation time (GT) variation of the indicated mutants with respect to the wild-type strain (WT). Statistical significance was determined using one-way ANOVA. Then, Dunnet's test was applied for multiple comparisons. (B) Mutations in flrA and flrB reducing generation time reduce V. cholerae swimming capacity. Left panel show a representative swimming assay of the WT, the parental strain bearing the selection cassette (parental), and the clone bearing the indicated mutation (flrA* and flrB*). The right panel shows the analyzed data for all the assays performed per triplicate (n = 9). Statistical significance was analyzed by one-way ANOVA. Then, Tukey's test was done to compare the mean values. **, P , 0.01; ****, P , 0.0001. OD, optical density. Experimental Evolution on Ribosomal Protein Gene Location mBio emergence throughout experimental evolution, and we observed mutations co-occurring with this phenotype that are in agreement with previous papers (38, 66-68, 72, 74). During the experiment, the growth rate improved for all V. cholerae populations. This trait was not initially noticed in the long-term evolution experiment performed in E. coli, although fitness and cell size increment were observed since the initial steps of the experiment (60,89,90). Recently, an increase in growth rate could also be detected (91) in line with "bacterial growth laws" that correlate growth rate to cell size (92). In V. cholerae experimental evolution, growth rate improvement was readily detected. This could be the result of employing a richer broth, the evolution strategy, and/or the faster growth nature of V. cholerae. Since growth rate is a good proxy for fitness (50,93,94), we propose that the increment in growth rate observed reflects, at least in part, the expected fitness improvement. The growth stabilization after G500 could be the result of high divergence between populations or the consequence of terminating the experiment relatively early. Further generations of evolution should lead to growth increases in a stepwise manner.
Deep sequencing of the ancestral clones, the populations at G250, and at the end of the experiment suggested different selection forces at each time point. At G250, positive selection seems to be the main force driving mutation. This scenario changes at G1000, when most of them seemed not to be under strong selection. This is reinforced by the fact that many populations showed inactivation of DNA repair systems that in turn resulted on average in more DNA changes (Fig. 4A). Interestingly, populations where S10 is close to ori1 (P1-6) tend to have more alterations in repair systems and concurrently, more mutations at G1000. Some DNA changes arose in more than one flask, in particular inactivating genes encoding functions that might become unnecessarily costly during in vitro continuous culture such as virulence, motility, and nutrient scavenging. Notably, many functions associated with virulence were found inactivated (MSHA, tcp, T6SS, etc.) throughout the evolution experiment. We also noticed several mutations at quorum-sensing (QS) and chemotaxis-related genes at G1000 (see above). The former seems to be a common feature of laboratory domesticated V. cholerae strains (95). Both chemotaxis and QS are important factors during host infection (96,97). Taken together, these observations suggest that these mutations are likely to reduce V. cholerae virulence. Meanwhile, the increased growth and the higher biofilm-forming capacity may increase their infection capacity. Therefore, possible virulence reduction of the strains throughout experimental evolution is worth of investigation. In that case, understanding how resilient their encoding genes are in the absence of selective forces maintaining virulence should be studied.
Interestingly, mutations acquired at G250 were not lost, suggesting their continuous selection throughout the experiment. Moreover, in the case of flagellum biosynthetic pathway, we found, on top of those on the master regulators flrAB, mutations in transcriptional regulators of lower hierarchy such as flhA or components of the machinery such as flgA, flgI, or flgK.
At both G250 and G1000, we detected many mutations occurring in a priori noncoding regions such as intergenic regions and pseudogenes. We cannot discard biological effects interfering with promoter or regulatory regions, potential uncharacterized small RNAs, or pseudogenes expressing truncated but functional proteins. This might be the case with those occurring in several populations independently, such as the point mutations occurring at VCA0168, a pseudogene located on the secondary chromosome, mutated in 9 of 12 populations (Data Set S1).
Interestingly, S10 relocation to chromosome termini leads to a relatively strong fitness loss that is easily outcompeted by cells having an ori1-proximal location for the main ribosomal protein locus (50). This contributes to explain the S10 proximity to ori1 conserved among Vibronaceae (49). However, this change in chromosomal architecture is not lethal nor does it result in cells with lower viability or a sick phenotype. An easy way to regain fitness upon S10 relocation to the ter region would be large-scale chromosomal rearrangements such as inversions, translocations, or S10 duplication. However, we did not observe them in the populations (Fig. S4) nor in the isolated clones sequenced (Fig. S5). Indeed, these rearrangements were a common feature of long-term evolved E. coli populations. In some cases, such as the ribose utilization operon, they were linked to a fitness gain (98). However, reverting S10 ori1distal location could disturb the conserved structure of the bacterial chromosome. For example, essential genes are usually found in the replication leading strand to avoid deleterious conflicts between replication and transcription (7,99). Large chromosome inversions able to relocate S10 close to ori1 would encompass many highly transcribed essential genes that would not be viable or would be rapidly purified from populations. Similarly, large inversions or translocations could lead to replichore length imbalance. Meanwhile, chromosomal rearrangements were possible since we detected some events in sequenced clones (e.g., a large translocation in G250-FG-2; Fig. S5). However, these alterations are unlikely to be associated with the observed phenotypes (a faster growth), since reconstructing point mutations in flagellum regulators and in the mgtE gene fully reproduced the phenotype in a wild-type context. Also, point mutations suppressing S10 location change were not detected in our experiment after 1,000 generations. Therefore, we think that suppressing S10 either by altering chromosome architecture or by suppressor mutations is hard to achieve at least in this range of time. Meanwhile, under circumstances under which S10 oriC-proximal location becomes essential, these changes might be able to emerge. For instance, during infections in the Drosophila melanogaster model (96,97,100), S10 relocation to chromosomal termini leads to reduction of CFU/fly by orders of magnitude (49). Thus, experimental evolution in the fly gut or in conditions mimicking that of the strains S10Tnp-1120 and S10TnpC2 1 479 may lead to large chromosomal rearrangements or suppressor mutations that counterbalance S10 relocation far from ori1.
Importantly, we were able to distinguish mutations that improved bacterial growth from those with increasing biofilm-forming capacity. In particular, mutations inactivating the main flagellum regulators emerged in 7 and 11 of 12 populations at G250 and G1000, respectively. Their reintroduction into the wild-type context increased V. cholerae growth, demonstrating its specific effect and indicating that the rest of the detected mutations contributed growth enhancement. flrAB mutations also caused a reduction in motility that provide an ecological and physiological interpretation under the light of recent studies (101)(102)(103). These works show a trade-off between motility/chemotaxis and growth rate. In changing environmental conditions, cells deal with several simultaneous energetically costly processes such as growth, motility chemotaxis, and metabolism. Energy allocation to flagellar genes production, rotation of the flagellum, and, to a lesser extent, chemotaxis significantly reduces growth in E. coli (102,103). It is advantageous to dedicate energy to motility and chemotaxis in poor and heterogenous environments. Consistently, at G1000, we also noticed many mutations related to other swimming and chemotaxis genes (e.g., fliO, flrC, cheY, cheR2, flgA). In summary, many of the mutations detected throughout our experiment altered the motility/growth interplay in all the studied populations. However, many other mutations that contribute to improving fitness remain to be better characterized. For instance, mutations on mgtE, which are not clearly related to motility, also improved the growth rate. This opens the possibility of finding other mutations on genes not related to motility capable of improving the growth of an already very fast growth bacterium. In particular, magnesium might be a limiting micronutrient in fast-growing conditions, a factor to take into account when optimizing bacterial growth. Such a factor is worth testing in the V. cholerae relative Vibrio natriegens (GT = 10 min), since at the time it is a biotechnologically useful organism and would help to further push the growth on the fastest growing known organism.
After 1,000 generations of experimental evolution, although all populations shortened their generation time, those harboring S10 locus close to ori1 remained the fastest. This evidences the great potential of positional genetics approach (49) as a synthetic biology tool to condition cell physiology to permanently avoid the escape of suppressor variants. The latter conditions the design of stable genetic circuits. For instance, recent work by 1Izard et al. (104) created a growth switch in E. coli by controlling RNA polymerase expression from an inducible promoter that required three copies of the main repressor to avoid regulation escape. Despite the powerful nature of this synthetic circuit, escape could still be detected. Positional genetics offers a new way to cope with this problem to generate robust genetic circuits immune to suppressor mutations.
Conclusions. By applying the positional genetics approach to ribosomal protein genes, we contributed to understanding the role of gene order in bacterial cell physiology. The present work incorporates the evolutionary aspect of studies on positional genetics, expanding our knowledge on the long-term implication of gene order on genome evolution. The role of the genomic position of other key genes remains to be explored, such as ribosomal RNAs (105), tRNA UBI (22,24), RNA polymerase (51), and ATP synthase (17), among others. This is central to understanding how the genome primary structure affects cell physiology, in particular growth rate and genome evolution. Such knowledge will enable bacterial growth repurposing while allowing understanding of the behavior of more complex biological systems, thus promising a deep impact in genome design, bioengineering, and biotechnology.

MATERIALS AND METHODS
Bacterial strains and plasmids and culture conditions. All strains were derived from V. cholerae serotype O1 biotype El Tor strain N16961 with Str R (106). Bacterial growth was done in lysogeny broth (LB) at 37°C agitating at 200 rpm. The strains and plasmids used in this study are listed in Table S1.
General procedures. Genomic DNA was extracted using the GeneJET genomic DNA purification kit, while plasmid DNA was extracted using the GeneJET plasmid miniprep kit (Thermo Scientific) or ADN Puriprep B and P (InbioHighway, Argentina). PCR assays were performed using DreamTaq, Phusion High-Fidelity PCR Master Mix (Thermo Scientific), or MINT 2X (InbioHighway, Argentina). The oligonucleotides described in Table S2 were purchased from Macrogen (Seoul, Korea).
Experimental evolution setting. We studied four different strains bearing the S10 locus either close to their wild-type location (Par-1120 and S10Tnp-35; Table S1) or far from ori1 (S10Tnp-1120 and S10TnpC2 1 479; Table S1). Overnight cultures of each strain were used to obtain isolated colonies in LB agar Str plates. For each one, we picked three different individual colonies to start 5-mL independent cultures on LB Str that were cultured overnight at 37°C agitating at 180 rpm. Then, these cultures were frozen at 280°C on LB 10% dimethyl sulfoxide (DMSO; Merck). These stocks constitute the starting point of the experiment. The first inoculation was done with 50 mL of each initial culture (10e9 CFU/mL as assessed by plating) in 250 mL and cultured overnight (ca. 14 generations). Then, each population was propagated in 250-mL (7 generations) or 500-mL (8 generations) Erlenmeyer flasks. Throughout the experiment, the cells were cultured at 37°C agitating at 200 rpm on a dry incubator (Infors HT). For each passage, 1 mL of the previous flask is transferred to the next step. When clumps or biofilms appeared, we did not let the culture still for long time to avoid counterselection of aggregates and biofilms. Also, we used 1-mL pipette tips that allowed us to pipet up and down several times in case cell clumps were too big to entirely pass through the tip. Populations were regularly checked for contamination by plating in LB Str. Colonies with unusual morphologies were checked phenotypically and by PCR amplifying the V. cholerae specific gene rctB (3,398 to 3,407; Table S2) or a specific region of the secondary chromosome (Tgt4_1 to Tgt4_4; Table S2). During the first 400 generations, the populations were frozen by centrifuging 10 mL of each culture in 14-mL Falcon tubes at 5,000 rpm for 7 to 10 min, discarding the supernatant, and resuspending the cells on 1 mL of LB supplemented with 10% of DMSO (Merck) frozen at 280°C until later use. After G400, the populations were frozen every 100 generations. Then, the cultures were frozen at 280°C until later use.
Automated growth curve measurements. Overnight cultures of the indicated microorganism were diluted 1/1,000 in LB. Bacterial preparations were distributed by triplicate for evolved populations and sextuplicate for the ancestral clones in 96-well microplates. Growth curve experiments were performed using a TECAN Infinite Sunrise microplate reader, with absorbance (620 nm) taken at 5-min intervals for 5 h with agitation at 37°C. The experiments were done at least twice independently. The slopes during the exponential phase were directly obtained using a home-made Python script coupled to a growth rates program (73). To filter noise due to growth variations among experiments, the obtained m was relativized to the ancestral clones, and the relativized values were multiplied by the average of the ancestral clones throughout all the experiments. This normalization did not affect the results since raw data provided a similar picture as depicted in Fig. S3D. The percentage of GT was calculated as follows: %GT = (GT/GT WT 2 1) Â 100.
Biofilm formation assays. Nine-well PVC microtiter plates (BD Falcon) were used to monitor biofilm formation as described previously (107). Briefly, LB was inoculated with a 1/100 dilution directly obtained from the liquid overnight cultures of each population in rich medium. After inoculation, microtiter plates were incubated at 37°C for 24 h and rinsed, and 150 mL of a 0.1% solution of crystal violet was added to each well. The plates were incubated at room temperature for 30 min and rinsed, and biofilm formation was tested as follows: crystal violet was solubilized by addition of 150 mL of ethanol-acetone (80:20), and the OD 570 was determined. The results are presented as the means of four replicate wells in three independent experiments.
Deep sequencing analysis. The quality of Illumina reads was visualized by FastQC v0.10.1 Babraham Bioinformatics (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). All reads of 32 samples were aligned against the two chromosomes of V. cholerae (GenBank NC_002505 and NC_002506, 17 December 2014), using paired end mapping mode of BWA "mem" v0.7.4 (108) with the option "2M." The reference chromosome I was modified since the reference genome sequence (AE003852) showed an inversion around ori1 flanked by two rRNA operons (rrnB and rrnG-H) (48). Output SAM files were converted to BAM files using Experimental Evolution on Ribosomal Protein Gene Location mBio SAMtools v0.1.19 (109). The "Add Read Groups" step was made by Picard v1.131 (http://picard.sourceforge .net/). The aligned reads in BAM files were realigned with the command "IndelRealigner" implemented in GATK v2.2.16 (110). "MarkDuplicates" step from Picard flagged duplicates reads. We kept only uniquely mapped reads, using SAMtools (option "view -bq 20"). A summary of the sequencing depth for each population is summarized in Table S3. Then, mpileup files were generated using SAMtools without BAQ adjustments (option "mpileup -B"). SNPs and indels were called by the option "pileup2cns" of Varscan2 v2.3.2 (111) with a minimum depth of 10 reads, a threshold of 20 for minimum quality, a threshold of 0.25, and 0.8 for the minimum variant allele frequency for heterogeneous population strains and subclone isolated strains, respectively. The coverage around the mutation was checked with a Python script with a minimum depth of 7 reads. The annotation of identified variants was processed by SnpEff v4.1 (112). Common mutations between all sequenced strains were ignored, indicating differences with the reference sequence. We also looked at possible inversions or alterations in gene order in evolved populations. For this we employed Spades 3.6.2 (113) for assembly of trimmed reads with the kmer options "2k 21,51,71" and "2careful." Contigs were ordered and oriented using CONTIGuator 2.7.4 (114) using the two chromosomes of V. cholera as references. The Artemis comparison tool (ACT) (115) was used for synteny visualization and for the search for dynamics of large rearrangements such as translocations and inversions. All data generated and analyzed during this study is included in this published article, its supplementary information files, and publicly available repositories. The Illumina Deep Sequencing data sets are deposited at the GenBank SRA as BioProject PRJNA816505. Multiplex Genome Editing by Natural Transformation (MuGENT). To introduce the mutations observed throughout the evolution experiment, we employed MuGENT (76,77). Briefly, the wild-type strain (Table S1) was grown overnight from a single colony at 37°C in LB with agitation. This overnight culture was diluted 10 23 into fresh medium, and the strain was grown to an OD 600 ;0.5. The cells were washed by centrifugation and resuspended at the original volume in 1Â instant ocean sea salts (7 g/liter) (Blacksburg) and autoclaved chitin powder (Sigma-Aldrich) and incubated overnight without agitation at 30°C to induce competence. In parallel, DNA fragments harboring the mutations were synthesized using PCR assembly, cloned using a pEASY-Blunt Zero Cloning kit (Transbiotech, China), and verified using Sanger sequencing (Macrogen). Next, we cotransformed the targeted loci containing the desired chromosomal alterations and a spectinomycin resistance cassette directed to a neutral locus at the intergenic space between VC1903 and VC1902 regions (116) using 3 mg of each DNA molecule. After overnight incubation at 30°C without shaking, the sample was vortexed and plated in LB supplemented with spectinomycin. Then, several colonies were screened and confirmed by MASC-PCR (77).
Motility assay. The effect of mutations in flrA and flrB on the motility of V. cholerae was tested on LB soft agar plates (0.3s% bacteriological agar). Next, 2 mL of bacterial suspension (OD 600 = 1) was inoculated into the agar of each plate by triplicate to test bacterial swimming. The plates were incubated for 12 h at 37°C, and then the diameter of the motility halos was measured using ImageJ (117). The results are reported in mm of bacterial extension on the media.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.1 MB.