Modulation of rpoS fitness by loss of cpdA activity during stationary-phase in Escherichia coli

Experimental evolution of Escherichia coli in one month long stationary-phase in lysogeny broth batch cultures repeatedly selected mutations in the genes for the stationary-phase sigma factor RpoS and the cAMP phosphodiesterase CpdA. The founder strain carried a previously identified allele of rpoS, referred to as rpoS819, a partially functional variant that confers growth advantage in stationary-phase (GASP). The 46 base duplication at the 3’ end of rpoS819 produces a longer protein present at very low levels compared to wild type RpoS. A new rpoS variant rpoS92, carrying a re-duplication of the original duplication in rpoS819, arose during the first week of our evolution experiment. In rpoS92, an in-frame stop codon truncated RpoS819 creating a shorter RpoS92 whose levels are restored to that of wild type RpoS. Transcription profiling of rpoS92 indicated a shift in gene-expression to that of wild-type rpoS, reversing some of the expression trends of rpoS819. Δ3cpdA, carrying an in-frame three base deletion, had arisen late in our evolution experiment. It is a loss of function mutation, which elevates cAMP levels. Using mixed culture competition experiments, we demonstrate that rpoS92 confers GASP, whereas Δ3cpdA confers relatively modest GASP in comparison to the ancestral rpoS819. Δ3cpdA mediates epistatic repression of rpoS92 GASP. The original survivor carrying both rpoS92 and Δ3cpdA besides other mutations displays robust GASP, highlighting the role of these additional mutations in reversing the epistatic interaction between Δ3cpdA and rpoS92. In 10- and 20-day old spent media, there is a reduction in the competitive fitness of rpoS92, which is arrested by Δ3cpdA. Thus the activity of RpoS fluctuates via genetic mutations in deep stationary phase, and additional mutations in CpdA helps modulate the competitive fitness of RpoS variants.


Bacterial Media and Growth Conditions.
Various E. coli strains were cultured in Lysogeny broth (LB) at 37 °C at 200 rpm. For determining viable counts, cultures were plated on LB agar medium supplemented with 35 μg of kanamycin (Kan)/ml, 100 μg of ampicillin (Amp)/ml, 10 μg of tetracycline (Tet)/ml, or 15 μg of chloramphenicol (Cam)/ml when required. Competition experiments were performed as described below.

P1 transduction
P1 phage transduction were performed by standard protocol described in Miller (40).

Competition Assays
For a standard stationary-phase mixed culture competition experiment, monocultures of the two competing strains were grown in 5 ml of LB at 37 °C with aeration for 12 hours. These overnight grown cultures were mixed in the desired proportion reciprocally (1:1000) and incubated under the same conditions without the addition of fresh medium. The two populations marked with two different antibiotic resistance markers were tracked by plating on LB agar plates containing the respective antibiotics every-day of the competition experiment duration. For competitions in conditioned medium, the parent strain ZK819 was grown for 21 days at 37 °C with periodic addition of sterile-distilled water to compensate for evaporation. Twenty-one day old culture was pelleted and supernatant was filtered twice through a filter with pore size 0.2 micron (Sartorius) to remove cells. To replicate the competitions in 21 day old conditioned medium, the overnight grown monocultures of competing

Construction of cpdA deletion strain.
Gene knockouts were generated by the procedure of Datsenko and Wanner (41). Hybrid primers with 36-bp extensions homologous to the N terminus and C terminus of the cpdA locus spanning positions +2 and +445 were designed for the PCR amplification of kanamycin resistant gene flanked by FLP  (44,45) . Readcounts per gene were generated from SAM files using a custom perl script. Subsequently the R package DESeq was used to call differentially expression of genes (46).
All subsequent data analyses on this set of mutations were performed using the statistical programming language R (v3.4.4).

Genetic parallelism at rpoS and cpdA during LTSP evolution
In a previous study, we had sequenced genomic DNA from multiple long term stationary phase (LTSP) populations of E. coli and found evidence of genetic parallelism in which the same gene acquired mutations in multiple independent lines. Among such genes were rpoS and cpdA (39). RpoS is a general stress response sigma factor and a frequent target of mutation and selection under complex environments involving prolonged periods of nutrient limitation (6,8,10). The founder E. coli strain ZK819 used in our evolution experiment carries an rpoS allele, referred to as rpoS819, which shows attenuated activity. rpoS819 was naturally selected during prolonged starvation in lysogeny broth batch culture and was shown to confer growth advantage in stationary-phase (10). We observed that all five independent replicate populations had diversified at the rpoS locus giving rise to four rpoS populations

rpoS92 confers GASP in stationary-phase
The rpoS92 allele arose early in our evolution experiment and steadily increased to nearly 40% (39). It had appeared in all the replicate populations by the end of the first quarter of our one month long experiment. Simultaneously, the ancestral rpoS819 allele frequency declined in all replicates except one wherein it roughly maintained its starting frequency (39). To determine the GASP phenotype of rpoS92, we transduced the rpoS92 allele to the ancestor ZK819 to create ZK92. To set up a stationary-phase competition, overnight grown monocultures of ZK92 and the ancestor ZK819 were mixed in 1:1000 ratio reciprocally. Growth trends of the strains in mixed culture over four days of competition showed that ZK92 minority counts increased by ~11 fold whereas the ancestor ZK819 majority declined by ~31 fold. This closes the gap in size between the populations, indicating that rpoS92 confers growth advantage in stationary-phase ( Fig. 1A; paired t-test comparing the slopes of the competing populations, N = 4, P = 0.003). In the reciprocal mix, the ZK92 majority decline was ~10 fold whereas the ancestor ZK819 minority counts dropped by ~50 fold (

RpoS92 is a truncated version of RpoS819
The  amino acids of wild-type RpoS are replaced with a stretch of 39 amino acids ( Fig.2A & B). Sequence analysis of the new allele rpoS92 revealed the reduplication of the original 46 base duplication ( Fig.   2A). In rpoS92, the second duplication introduces an early stop codon that truncates RpoS819 ( Fig. 2A & B). We probed the cellular levels of the RpoS polypeptide variants using anti-RpoS antibodies (Fig.   2C). RpoS819 indeed is the longest polypeptide, but present at very low levels (Fig. 2C, Lane 3).
RpoS92 is noticeably shorter than RpoS819 and shows higher expression than RpoS819, to a level comparable to (or even higher than) that of wild-type RpoS (Fig. 2B, Lane 4). Differential geneexpression analysis of RpoS targets in rpoS92 and rpoS819 allelic background using RNA-seq shows that rpoS92 has partially restored the rpoS regulon expression to that of wild-type rpoS (Fig. 2D). Thus the activity of RpoS fluctuates -through genotype alterations -in deep stationary phase.

The original cpdA survivor Sur_Δ3cpdA displays GASP
We tested the fitness of original Sur_Δ3cpdA against the ancestor ZK819 in mixed culture competition assay. By the end of three days of competition, Sur_Δ3cpdA minority counts had increased by ~32 folds while the ancestor majority counts dropped by ~806 folds (Fig. 3A ; paired t-test comparing slopes of competing populations, N = 4, P = 2.21 x 10 -06 ). This establishes the GASP phenotype of Sur_Δ3cpdA.
In the reciprocal mix, Sur_Δ3cpdA majority declined by ~7 fold whereas ancestor minority declined bỹ 3 fold ( Fig. 3B ; paired t-test comparing slopes of competing populations, N =4, P= 0.008). One order of magnitude decline of either majority population was observed in control competition between ancestor pair marked with kanamycin resistance (~9 fold) or tetracycline resistance (10 fold), and therefore not taken into account while drawing inferences.
To determine the contribution of Δ3cpdA in Sur_Δ3cpdA GASP, the Δ3cpdA allele in Sur_Δ3cpdA was replaced with the wild-type cpdA using P1 transduction. In Sur_cpdA wt versus ancestor ZK819 competition, Sur_cpdA wt minority counts increased by ~24 fold whereas ancestor majority declined bỹ 77 fold ( Fig. 3C ; paired t-test comparing slopes of competing populations, N = 4, P = 9.22 x 10 -06 ). In the reciprocal mix, Sur_cpdA wt majority counts dropped by ~18 fold while the ancestor minority counts showed little change ( Fig. 3D; paired t-test comparing slopes of competing populations, N = 4, P = 0.02). Thus the replacement of the Δ3cpdA allele by its wild type equivalent resulted in a decline in the ability of the survivor to out compete the ancestor ZK819.

Δ3cpdA is a null allele and confers modest GASP
The cAMP phosphodiesterase encoded by cpdA hydrolyzes cAMP (29). In Δ3cpdA, a three base  rpoS819 and rpoS92 (Fig. 4). Deletion of cpdA gene in ancestor ZK819 results transductant ZK819Δ3cpdA indicates that the genetic background does not influence the cAMP levels in these strains (Fig. 4).
In mixed culture competitions with the ancestor ZK819, ZK819cpdA null minority counts remained stable whereas ancestor majority declined by ~37 fold (Fig. 5A; paired t-test comparing slopes of the competing populations, N =4, P = 0.02). In the reciprocal mix, there was ~73 fold reduction in ancestor minority counts while ZK819cpdA null majority counts dropped by ~6 fold (Fig. 5B; paired t-test comparing slopes of the competing populations, N =4, P = 0.06). In Δ3cpdA transductant ZK819Δ3cpdA versus ZK819 competition, ZK819Δ3cpdA minority declined by ~9 fold whereas ancestor majority counts dropped by ~186 fold (Fig. 5C; paired t-test comparing slopes of the competing populations, N =4, P = 0.002). In the reciprocal mix, ZK819Δ3cpdA majority counts dropped by 59 fold and ancestor minority declined by ~235 fold (Fig. 5D; paired t-test comparing slopes of the competing populations, N =4, P = 0.03). These competition assays demonstrate that loss of cpdA function confers modest GASP.
Δ3cpdA is epistatic to rpoS92 in ancestor genetic background GASP of ZK92 indicating that Δ3cpdA is epistatic to rpoS92 in ancestor background (compare Fig. 1 with Fig. 6). The absence of this epistatic interaction in the original survivor Sur_Δ3cpdA suggests the role of the additional mutation/s present on its genome in neutralizing this epistatic interaction.

Age of spent medium influences rpoS92 and cpdA fitness
The competitions discussed in the above sections typically represent the environment prevailing during the first week of our evolution experiment. While the competition assays performed in the stationary- phase reflect the correct environmental context for the emergence of rpoS92, most of the cpdA mutations were observed during the second and third weeks of the original evolution experiment. A reasonable approach to recreate a later stage approximation of environmental context is to generate spent medium by growing the ancestral culture till the desired age and use this medium for competition experiments. We chose ten day and twenty one day old conditioned medium for these experiments. In spent medium competition, the rpoS92 transductant ZK92 minority declined by ~4 fold and the ancestor ZK819 majority declined by ~56 fold (Fig. 7A ; paired t-test comparing slopes of the competing populations, N = 4, P = 0.02). In the reciprocal mix, ZK92 majority declined by ~64 fold whereas ancestor ZK819 minority counts dropped by ~1461 fold (Fig. 7B ; paired t-test comparing slopes of competing population, N = 4, P = 0.004). Thus, in ten day old spent medium, the competitive fitness of ZK92 is still higher than that of the ancestor, although the degree to which ZK92 in minority closes the gap with ZK819 majority is less in the spent medium than in fresh LB.
In Δ3cpdA transductant ZK819Δ3cpdA versus ancestor ZK819 competition in ten day old spent medium, ZK819Δ3cpdA minority counts declined by ~130 fold and the ancestor ZK819 majority declined by ~228 fold (Fig. 7C ; paired t-test comparing slopes of the competing populations, N =4, P = 0.1). In the reciprocal mix, ZK819Δ3cpdA majority declined by ~144 fold whereas ancestor ZK819 minority declined by 49110 fold (Fig. 7D; paired t-test comparing slopes of the competing populations, N = 4, P = 0.0002). In ten day old spent medium, there is a reduction in ZK819Δ3cpdA and ancestor ZK819 fitness as reflected by the decline of the competing populations both as majority or minority fraction. However, ancestor fitness is severely reduced when present in minority with more than four orders of magnitude decline as opposed to nearly two and half orders of decline of ZK819Δ3cpdA minority.
During the double mutant ZK92Δ3cpdA versus ancestor ZK819 competition in ten day old spent medium, ZK92Δ3cpdA minority counts remained stable while ancestor ZK819 majority counts dropped by ~493 fold (Fig. 8A ; paired t-test comparing the slopes between competing populations, N = 4, P = 0.001). In the reciprocal mix, ZK92Δ3cpdA majority declined by ~29 fold and the ancestor minority declined by 653818 fold (Fig. 8B ; paired t-test comparing the slopes between competing populations, N = 4, P = 0.001 ). The ten day old spent medium competition of ZK92Δ3cpd indicates that the competitive fitness of the double mutant ZK92Δ3cpd competitive fitness against ancestor is higher compared to that of the either of the single mutants ZK92 or ZK819Δ3cpdA.
We tested the fitness of the double mutant ZK92Δ3cpdA against its more plausible competitor ZK92 in ten day and twenty one day old spent medium. In ten day old spent medium competition between ZK92Δ3cpdA and ZK92, ZK92Δ3cpdA minority counts declined by ~10 fold and ZK92 minority counts declined by ~270 fold (Fig. 8C ; paired t-test comparing slopes of the competing populations, N =4, P = 0.002). In the reciprocal mix, ZK92Δ3cpdA majority declined by ~77 fold and ZK92 minority declined by ~6567 fold (Fig. 8D; paired t-test comparing slopes of the competing populations, N =4, P = 0.001).
In twenty-one day old spent medium competition, ZK92Δ3cpdA minority counts declined by ~6 fold whereas the ZK92 majority declined by ~77 fold (Fig. 8E ; paired t-test comparing slopes of the  A and B) . One-dayold cultures of the two strains were mixed in ratios of 1:1,000 (A) and 1,000:1 (B) in spent medium derived from a 10 Days old parent culture . Competition between ZK92Δ3cpdA (blue) and ZK92 (red ) (C and D) . One-day-old cultures of the two strains were mixed in ratios of 1:1,000 (C) and 1,000:1 (D) in spent medium derived from a 10 Days old ZK819 culture. Competition between ZK92Δ3cpdA (blue) and ZK92 (red ) (E and F) . One-day-old cultures of the two strains were mixed in ratios of 1:1,000 (E) and 1,000:1 (F) in spent medium derived from a 21 days old ZK819 culture. **P < 0.005, *P < 0.05, not significant (ns)> 0.05. The thick curve is the LOESS line fitted to the data. competing populations, N = 4, P = 0.04). In the reciprocal mix, ZK92Δ3cpdA majority declined by ~60 fold and ZK92 minority declined by ~747 fold (Fig. 8F; paired t-test comparing slopes of the competing populations, N = 4, P = 0.01). The competitive fitness of the double mutant ZK92Δ3cpdA is higher than ZK92 in both ten day as well as twenty-one day old spent medium. However, ZK92 fitness improved in twenty-one day old spent medium with more than ten fold reduction in the population decline. Spent medium competition experiment trends show that Δ3cpdA arrests the decline in fitness of rpoS92 in deep stationary-phase.

Discussion
Laboratory evolution experiments combined with genome sequencing has made it possible to catalog mutations and understand adaptation. The growth advantage in stationary phase (GASP) paradigm, first demonstrated in E. coli, illuminates a common survival strategy under persistent starvation (10,12,13,14). The GASP mutants establish continuous growth and death cycles under growth limiting environment of prolonged stationary-phase (10,48). The dynamics of competitive fitness of GASP mutants is reflected in the different GASP flavours displayed by the competing populations (48). One of the earliest GASP alleles to be identified from the survivors of prolonged stationary-phase was a new allele of stationary-phase sigma factor named rpoS819 (10).
In an earlier study, we had reported the mutation spectrum obtained during a second round of long-term stationary phase evolution of ZK819, an E. coli strain with rpoS819 (39). A key observation of this study, as well as other recent studies of evolution in prolonged stationary phase in complex medium is the prevalence of parallel evolution with mutations in genes that have pleiotropic effects (39,(50)(51).
For example, in a very similar prolonged stationary phase experimental evolution setup like ours, though of longer duration, Hershberg and colleagues had reported multiple mutations in the RNA polymerase core subunits, but not in rpoS, which is usually one of the first gene to be hit by mutations under nutrient limitation (51). In addition to parallel mutations at different RNA polymerase subunits, we had observed parallel evolution at stationary-phase sigma factor rpoS and cAMP-phosphodiestersae cpdA in our one month old long-term evolution (39). In the present study, we characterized new alleles of rpoS and cpdA namely rpoS92 and Δ3cpdA and investigated their fitness potential during prolonged stationary-phase.
We show that RpoS status fluctuates at a genetic level through 30 days of stationary phase, and that additional mutations in the cAMP phosphodiesterase CpdA modulates the fitness of RpoS variants.
Heterogeneity at rpoS has been frequently observed particularly in several laboratory strains of E. coli .
Modulation of rpoS activity has been shown to facilitate a trade-off between general stress response and nutritional competence (6-8, 21, 23). The rpoS819 mediated GASP has been directly correlated with enhanced ability to scavenge amino acids, the key nutrients released by dead cells. This can be modulated by altering environmental factors such as pH of the medium (11,15,52). Strong selection for rpoS92, which partially recovers RpoS activity, indicates further tuning of RpoS activity. As stationaryphase deepens in a closed system such as batch cultures, nutrition may drop below a threshold to support growth. Bringing gene-expression closer to that of the wild type RpoS, RpoS92 might increase the general stress response while compromising the nutritional competence. Indeed in spent medium, though rpoS92 failed to manifest GASP, it showed better survival advantage against the rpoS819 ancestor.
Mutations that abrogated the activity of CpdA appeared in multiple lines during the second half of our GASP experiment. The CpdA mutation (Δ3cpdA) studied here appeared to enhance the survival of rpoS92 in spent medium competitions, whereas it dampened the competitive fitness of rpoS92 against rpoS819 in fresh medium competitions. Therefore, the selective advantage conferred by Δ3cpdA appears to take effect only late in stationary phase. However our observation that the negative effect of Δ3cpdA on rpoS92 is not prominent in the genetic background of the original survivor that carried additional mutations beyond rpoS92 and Δ3cpdA, suggests that these mutations may also play a role in stationary phase growth / survival.
To gain a molecular mechanistic insight underlying adaptaion, we performed RNAseq on cpdA strains, which however failed to identify any significant differentially expressed genes when compared to an otherwise isogenic background. Thus, we did not find evidence for increased nutritional competence via cAMP-CRP mediated regulation (unpublished data). However, these experiments were performed in fresh LB, and whether a different set of results will be obtained in a consistent manner in spent medium RNA-seq experiments remains an open question. There is increasing evidence supporting a role for cAMP in negative regulation of persistence, a phenotype observed under extreme stress (53-55).
Identifying the specific background mutation/s modulating the epistatic interaction between Δ3cpdA and rpoS92 will provide additional insights to the genetic regulation of proliferation versus nonproliferation mechanisms under severe stress.

Accession Numbers
Genome sequence data for Sur_Δ3cpdA is available from the Sequence Read Archive (SRA) database under the accession no. SRP094816 (https://www.ncbi.nlm.nih.gov/search/?term=SRP094816). RNA sequencing data from this study are available from the Gene Expression Omnibus (GEO) database under the accession number GSE119046 (https://www.ncbi.nlm.nih.gov/gds /?te rm=GSE 119046).