The LetA/S two-component system regulates transcriptomic changes that are essential for the culturability of Legionella pneumophila in water

Surviving the nutrient-poor aquatic environment for extended periods of time is important for the transmission of various water-borne pathogens, including Legionella pneumophila (Lp). Previous work concluded that the stringent response and the sigma factor RpoS are essential for the survival of Lp in water. In the present study, we investigated the role of the LetA/S two-component signal transduction system in the successful survival of Lp in water. In addition to cell size reduction in the post-exponential phase, LetS also contributes to cell size reduction when Lp is exposed to water. Importantly, absence of the sensor kinase results in a significantly lower survival as measured by CFUs in water at various temperatures and an increased sensitivity to heat shock. According to the transcriptomic analysis, LetA/S orchestrates a general transcriptomic downshift of major metabolic pathways upon exposure to water leading to better culturability, and likely survival, suggesting a potential link with the stringent response. However, the expression of the LetA/S regulated small regulatory RNAs, RsmY and RsmZ, is not changed in a relAspoT mutant, which indicates that the stringent response and the LetA/S response are two distinct regulatory systems contributing to the survival of Lp in water.


Results
LetS is important for the culturability of Lp in water. The culturability of the ∆letS mutant was compared to that of the wild-type (WT) in water at 42 °C using CFU counts (Fig. 1A). An inducible plasmid carrying the letS gene (pletS) was introduced into ∆letS to complement the mutant strain. Expression of the plasmid-borne copy was induced with 0.1 mM IPTG, herein referred to as ON, or was left uninduced, herein referred to as OFF. A significant (p < 0.005) defect was observed in the ability of ∆letS to form colonies compared to the WT starting at early time points ( Fig. 1A and B). Inducing the expression of letS in the ON strain solely in water did not correct this defect (Fig. 1A); however, when induced prior to and during water exposure, ON performed markedly better than the mutant (Fig. 1B). ∆letS and OFF behaved similarly to eachother. These results implicate the LetA/S two-component system as an important tool for the adaptation and potential survival of Lp in water at 42 °C.
Culturability of the strains was also tested at 4 °C and 25 °C to determine whether the defect observed at 42 °C was temperature specific. As a general trend, strains reached the detection limit slower as the incubation temperature decreased (Fig. 1), as previously described 51 . Similar to Fig. 1B, the loss of letS caused a defect when Lp was exposed to water at 4 °C and 25 °C, a phenotype that was complemented in the ON strain ( Fig. 1C and D). At 25 °C, ON mirrored the WT strain better during the early stages of water exposure (Fig. 1D). IPTG is reported to be a stable inducer of gene expression with minimal decay in broth; however, its long-term stability in water over the course of several months has never been investigated to our knowledge. Alternatively, the uptake of IPTG may be compromised over the long-term. Indeed, Lp repressed the majority of its genes 24 hours after water exposure 8 , including channels and transporters that may be involved in IPTG uptake. Taken together, these experiments suggest that the LetA/S TCS is an important regulator that allows Lp to successfully adapt to and, likely, survive the nutrient poor aquatic environment.
The effect of the letS mutation on culturability of Lp was most pronounced at 42 °C (Fig. 1). Similarly, previous reports from our group found that the deleterious effect of some mutants was observed more readily at 42 °C than at lower temperatures 8,11,13 . A higher metabolic rate at elevated temperatures can explain the faster decline in CFU counts in mutant strains that are less fit than the wild-type. As such, when faced with starvation in water, internal energy sources would be depleted more rapidly at warmer temperatures. Accordingly, we expected the transcriptomic and physiological responses initiated by the LetA/S cascade to be apparent at earlier time points at 42 °C. Therefore, subsequent experiments were conducted at 42 °C to facilitate the study of the regulatory effects exerted by LetS.
LetS influences morphological changes in water, pigment production and resistance to heat shock.
WT, ∆letS, ON and OFF strains were exposed to water for 24 hours at 42 °C. Microscopic analysis determined that ∆letS cells were significantly (p < 0.0005) longer than WT cells after water exposure. Interestingly, the ON strain produced Lp cells that were significantly (p < 0.0005) shorter than the WT ( Fig. 2A,C and E), which may be due to a higher level of letS expression. Notably, we observed that filamentous cells were more commonly found within the ∆letS and OFF populations than in the WT and ON populations. In accordance with previously published reports, cell size reduction was also observed in the post-exponential (PE) phase compared to the exponential (E) phase in the WT (Supplementary Figs S1 and S2) 49 . This reduction was absent in the letS and OFF strains. A strain over-expressing CsrA was similarly unable to reduce cell size upon entry into the PE phase 52  highlights the contribution of LetS to cell size reduction in both the post-exponential phase and under the starvation condition of water, but not during the exponential phase of growth.
In agreement with previously published work 46,48 , ∆letS was defective for pigment production during the late PE phase ( Supplementary Fig. S3). It is noteworthy that pigment production was not completely abolished in the ∆letS mutant which produced visible pigmentation. Nonetheless, the amount of pigment produced by the mutant was significantly (p < 0.0005) lower than the WT, a defect that was corrected in the ON strain ( Supplementary Fig. S3).
Lp frequently encounters high temperatures within man-made water distribution systems and is known to persist despite continuous heat treatments [53][54][55][56] . Therefore, we tested the ability of ΔletS to withstand heat shock in 55 °C water. WT CFU counts dropped by 5 orders of magnitude after 30 minutes at 55 °C, but remained above the detection limit after 60 minutes. In contrast, the letS mutant was more sensitive, as CFU counts dropped below the detection limit after a 30-minute heat shock treatment (Fig. 3). Expression of letS in ON afforded Lp an increased resistance to heat compared to the WT. The OFF strain resulted in lower heat tolerance compared to the WT and ON; however, its survival was markedly better than that of the letS mutant.
Transcriptomic analysis of the LetS regulon in water. In order to identify genes affected by LetA/S in response to water, DNA microarrays were used to probe the transcriptomic differences between the WT, ∆letS, ON and OFF strains 2 hours after water exposure at 42 °C. Justification for using an early time point (2 h) were two-fold: 1) Lp is known to reduce transcription dramatically after 6 h in Fraquil 8 , and 2) we expect that the regulatory changes initiated by the LetA/S TCS are important for early adaptation in Lp as evidenced by the need for letS induction on agar prior to water exposure to achieve proper complementation (Fig. 1). To extract sufficient RNA for transcriptomic analysis, a high cell density was required. Therefore, the survival defect observed at low cell density ( Fig. 1 -OD 600 0.02) was confirmed at high cell density ( Supplementary Fig. S4 -OD 600 1).
Gene expression profile of ∆letS vs. the WT was compared to that of OFF vs. ON in Fig. 4. The absence of letS led to the up-regulation of a large portion of genes (Fig. 4). Among the significantly (p < 0.05) up-regulated open reading frames (ORFs), 569 genes and 49 annotated sRNAs (a total of 618 ORFs) were shared in both ∆letS and OFF compared to the WT and ON respectively (Fig. 4B, Supplementary Table S1). On the other hand, 16 genes, including the two sRNAs RsmY and RsmZ, were significantly (p < 0.05) down-regulated in both the letS mutant and OFF (Fig. 4C, Supplementary Table S1). qPCR was performed on six randomly chosen genes; three were upregulated and three were downregulated in ΔletS compared to the WT (Fig. 5). The fold change pattern observed in the microarray analysis was mirrored in the qPCR results (Fig. 5).
Genes that are differentially and significantly (p < 0.05) expressed in both ∆letS and OFF, relative to the WT and ON respectively, were categorized into clusters of orthologous groups (COGs) (Fig. 6). The cellular functions The ON strain in panels C and D was induced on agar prior to water exposure and during water exposure. Strains were suspended in water at an OD 600 of 0.02. ON was induced with 0.1 mM IPTG. DL, detection limit. An unpaired, one-tailed Student's t-test was used to assess statistical significance versus the WT. *p < 0.05; **p < 0.005; ***p < 0.0005.
SCieNtiFiC REPORTS | (2018) 8:6764 | DOI:10.1038/s41598-018-24263-9 most affected by LetS were those associated with exponential growth or nutrient rich conditions and were mostly negatively affected by LetS in water (Fig. 6). These include "Translation", "Amino acid metabolism", "Lipid metabolism" and "Energy metabolism". Notably, 30 of the 38 "Translation" genes that are normally down-regulated by , ON (C) and OFF (D) exposed to water for 24 hours are shown. In Panel E, Image J software was used to quantify the average length of 100 cells after exposure to water for 24 hours. The scale bar is equivalent to 5 μm. An unpaired, one-tailed Student's t-test was used to assess statistical significance versus the WT, unless identified otherwise. *p < 0.05; **p < 0.0005. Figure 3. Deletion of letS affects sensitivity to heat shock. The WT, ∆letS, ON and OFF strains were suspended in water for 2 hours and subsequently exposed to a 55 °C water bath for 15, 30 or 60 minutes. CFU counts were enumerated on CYE agar before and after the heat shock treatment. DL, detection limit. An unpaired, one-tailed Student's t-test was used to assess statistical significance versus WT. *p < 0.05; **p < 0.005. LetS in water encode 30 S or 50 S ribosomal proteins (Fig. 6, Supplementary Table S1). It is also noteworthy that the rpoD transcript coding the housekeeping sigma factor was also highly up-regulated in water in the absence of letS ( Table 1).
The LetA/S cascade is linked to the RpoS regulatory cascade, but its activation is ppGppindependent. Lp responds to water by shutting down major metabolic gene groups, such as replication, transcription, translation and amino acid metabolism 8 . In the absence of letS, Lp is unable to mount such a response, which likely results in overconsumption of internal resources. The transcriptomic response of ∆letS shows similarity to the response of an rpoS insertion mutant exposed to water; both mutants demonstrate abhorrent up-regulation of gene transcription in response to water 14 . Moreover, there is considerable phenotypic overlap between the activation of the stringent response and the LetA/S cascade both in vitro 19,57,58 and in vivo 19,59 . As a result, LetS has been proposed to respond to the stringent response alarmone, ppGpp, during starvation 60 . We decided to investigate whether this hypothesis holds true when Lp is exposed to water. The NCBI Conserved Domain Database (CDD) was used to identify protein domains within LetS that could be responsible for environmental signal recognition.     Figure 7A is a graphical representation of LetS topology as predicted by the bioinformatic tools described herein. The transmitter (T), receiver (R) and histidine phosphotransfer (HPT) domains, characteristic of tripartite sensor kinases, are depicted by blue boxes, preceded by two transmembrane domains (TMDs) in green. In addition to these domains that comply with a previous study 23 , we report the presence of signal sensing domains and their topology, thereby fine-tuning our knowledge of LetS structure. The N-terminus of LetS contains two conserved signal sensing domains; DUF2222 and HAMP. The former is located between residues 37-180, while the latter spans residues 185-241. The two TMDs are located between residues 15 to 34 and 182 to 204 (TMHMM), or 11 to 31 and 181 to 201 (TOPCONS). While the DUF2222 domain is found between the two helices, the N-terminus of the HAMP domain overlaps the second transmembrane (TM) helix by 19 (TMHMM) or 16 residues (TOPCONS) (Fig. 7A). Therefore, approximately 30% the HAMP domain lies within the membrane, and 70% in the cytoplasm. LetS may be activated by an extracellular (via DUF2222), intracellular (via HAMP) or by an intermembrane (via the TM region of HAMP) signal. Here, we investigated the possibility that ppGpp may act as an intracellular signal that activates the LetA/S cascade in response to water exposure as previously postulated 48,61 .
The small RNAs (sRNAs) RsmY and RsmZ were used as reporters for the activation of LetS using Northern blotting. A ∆relAspoT strain (ppGpp°), incapable of producing ppGpp 14 , was used to determine the effect of ppGpp on RsmY/Z production. The probe for RsmY detected two bands (Fig. 7B). Similar banding patterns are reported for RsmZ in Pseudomonas fluorescens and RsmY in P. aeruginosa [62][63][64] . The top band (~110 nucleotides) was considered the active sRNA as per previous reports 46,65 . Corroborating previous studies 65, 66 , the levels of RsmZ increase when cells enter the PE phase (Fig. 7B -bottom panel). This increase was, however, absent in RsmY (Fig. 6 -top panel). We postulate that the increase in RsmY may occur later during the PE phase. Nevertheless, both sRNAs were strongly expressed in water. ∆letS and OFF expressed both RsmY/Z at basal levels, while ON recovered WT level expression of RsmY/Z in water. The basal levels of RsmY/Z observed in the letS mutant is likely due to strong promoters preceding the sRNA coding regions, as the E phase WT cells exhibited similar bands (Fig. 7B). The expression of RsmY/Z was abolished in ∆rsmYZ; RsmY expression was recovered to WT levels when rsmY is expressed in trans. The absence of the alternative sigma factor, RpoS also resulted in a significant decrease in RsmY/Z expression (Fig. 7C), corroborating previous findings that the LetA/S cascade is RpoS-dependent 65 . Importantly, the ppGpp null mutant had WT levels of the RsmY transcript and a slightly lower level of RsmZ compared to the WT, suggesting that the alarmone is dispensable for LetS activation in water and expression of the effector sRNAs.

Discussion
Legionella pneumophila (Lp) is a resident of natural and man-made water systems, and uses aerosols generated by the latter to infect human hosts 5 . Lp leads a biphasic lifestyle, alternating between the replicative phase and transmissive phase during infection, and between the exponential (E) and post-exponential (PE) phase in broth culture 58,67 . The former stages are characterized by nutrient abundance. In contrast, the hallmark of the transmissive/PE phase is nutrient deprivation causing morphological changes, a transcriptomic shift, stress resistance and virulence phenotypes 58,61,68 . The LetA/S two-component system is a key tool for differentiation of Lp in response to starvation, both in artificial medium and inside the host 46,50,66,69,70 . The established role of LetA/S in adaptation to low nutrient conditions led us to test its contribution to surviving starvation experienced in water. The absence of letS did not affect the growth rate of the mutant on solid medium. Pyomelanin production, which is linked to iron acquisition, was impaired in the mutant (Supplementary Fig. S3) 19,46,48 . However, this characteristic may be irrelevant when Lp is exposed to water, as genes related to pigmentation and iron acquisition (lpg2278/hpd/lly, feoA, lpg0232, lpg0124, lpg0746, lpg0467, np20) were repressed by the TCS (Table 1) [71][72][73] . The absence of letS also caused a slightly elongated cellular morphology in water compared to the WT (Fig. 2).
Importantly, an intact LetA/S system is required for culturability of Lp in water (Fig. 1). We corroborate previous studies showing that the LetA/S cascade exerts its regulatory activity via the two small RNAs, RsmY/Z during PE phase in broth (Fig. 7B) 46 . Defects previously associated with the LetA/S system occur under the starvation conditions found during the PE and transmissive phases 19,50,69 . Our data show that LetA/S is also activated in water, leading to RsmY and RsmZ expression (Fig. 7B). Interestingly, the transcriptomic changes initiated by the LetA/S cascade occur and are needed rapidly upon water exposure (Fig. 1).
LetS is a tripartite sensor kinase whose ortholog, BvgS, in Bordetella is known to respond to multiple environmental stimuli, activating specific regulons 22,74 . Accordingly, differences between the LetS regulon of Lp in water and that in the PE phase were expected 46,50,75 . While both are characterized by the absence of nutrients, the latter accumulates waste products likely causing additional stress. Indeed, Lp can survive for several weeks in Fraquil, but dies after only a few days in broth 11 . Over time, WT Lp exposed to water progressively shuts down transcription relative to exponential growth 8 . In stark contrast, an overwhelming majority of differentially expressed genes are up-regulated (97%) in the absence of letS (Fig. 4). In PE phase, approximately equal numbers of genes are positively and negatively influenced by the LetA/S cascade 46 . In water, LetS almost exclusively represses genes encoding functions that are unnecessary for survival, such as translation and metabolism of amino acids, lipids and carbohydrates (Fig. 6).
Over-expression of replicative phase genes in the absence of LetS is likely due to its downstream effect on CsrA. CsrA affects mRNA expression by: 1) blocking the ribosome binding site, 2) increasing degradation by recruiting RNase E, or 3) enhancing the stability and, thereby, expression of the target mRNA 76 . In ∆letS, binding of CsrA to target mRNAs is not relieved by RsmY/Z, which will favor the continued expression of replicative genes, while repressing transmissive genes 48 . A recent study revealed that CsrA directly regulates 479 transcripts involved in amino acid metabolism, carbon metabolism, virulence, flagella expression and iron acquisition 77 . For example, the binding of CsrA to the gap mRNA increases transcription by preventing rho-dependent termination 77 . As a result, gap expression increased in the absence of LetS in water (Supplementary Table S1). Moreover, two out of three Fis transcriptional regulators (Lpg0542/Fis1 and Lpg1743/Fis2), that are stabilized by CsrA 77 , are repressed by LetS in water. Strikingly, 30 ribosomal genes were highly up-regulated in the absence of letS (Supplementary Table S1). In contrast, only five ribosomal genes are reported to bind CsrA directly 77 , suggesting indirect regulation by CsrA. Another notable regulator repressed by LetS in water is the housekeeping sigma factor, RpoD. Increasing the amount of RpoD in the letS mutant strain could interfere with the binding of other sigma factors to the core RNA polymerase, a mechanism termed sigma factor competition 78 . Regulation of RpoD by LetA/S or by CsrA has not been reported in Lp 46,48,77 . As such, it is unclear whether the effect of CsrA on rpoD in water is direct or indirect. The general down-regulation of gene expression upon LetS activation in water is likely mediated by relieving both the direct and indirect effects of CsrA binding to target mRNAs.
The most noticeable difference between the transcriptome of the letS mutant in PE phase and in water is the marked absence of gene up-regulation. Only 16 genes were significantly (p < 0.05) induced by LetS in water, including RsmY/Z that are directly regulated by LetA (Table 1). In contrast, over 300 genes were significantly up-regulated in the PE phase when letA or letS were absent 46 . This is presumably because water is a less stressful condition than PE phase, during which metabolic by-products accumulate. As such, virulence regulators (RpoS, LetE and LqsR) were not affected by LetS in water (Supplementary Table S1). While LetA/S controls flagella genes in broth cultures 19,23,46,69,79 , we report that only one flagella-associated gene was differentially expressed in ∆letS in water (Supplementary Table S1). At 25 °C, flagella genes are maximally induced at 6 hours after exposure to water 8 . At 42 °C, it is possible that their induction was not captured at the 2-hour time point. Furthermore, flagellar gene expression is tightly regulated in Lp, both cooperatively and independently by several high-profile regulatory entities including RpoS, RpoN, FleQ, LetS and LqsR 46,[79][80][81] .
PE phase heat shock (HS) resistance is presumed to be conferred by genes under LetS control 48,50 . Accordingly, ∆letS was also heat sensitive relative to the WT after water exposure (Fig. 3); however, similar to flagella gene expression, HS genes were unaffected by LetS at 42 °C. The HS response is initiated within the first few minutes of exposure to this stress, whose transcriptomic effects subside quickly thereafter 82,83 . It is, therefore, possible that the HS-related transcriptomic changes were not detectable by the microarray analysis, if, indeed, they were activated at 42 °C. Alternatively, the general sensitivity of the letA/S mutants to various stresses, including heat shock, (Fig. 3) 48,50 may be a result of cell structure and therefore, a by-product of the cell's inability to adapt to the respective environments.
Upon sensing starvation, bacteria deploy the stringent response (SR) network governed by the cellular alarmone ppGpp, a key signal in growth phase differentiation 60,84 . SR is characterized by a rapid downshift in the synthesis of stable RNAs like rRNAs and tRNAs 85 . For the most part, ɤ-proteobacteria synthesize ppGpp using RelA, a synthase, and SpoT, a dual-acting hydrolase with weak synthase activity 60,85 . ppGpp positively affects cellular levels of RpoS, an alternative sigma factor 60,86 . Mutation of RpoS leads to the over-expression of replicative genes in water 14 , similar to the letS mutant (Fig. 5). Given these transcriptomic similarities, the LetA/S system was proposed to be integrated into the SR during water exposure, which suggests that the same activating signal initiates the LetS cascade.
Northern blotting was used to probe for the direct downstream targets of LetA/S, RsmY and RsmZ (Fig. 7B). Both sRNAs were LetS-dependent and highly expressed upon water exposure (Fig. 7B). The ppGpp° strain (∆relA∆spoT) that is unable to survive in water 14 was used to determine the effect of the alarmone on RsmY/Z production. We show that the ppGpp° strain did not change RsmY/Z expression considerably (Fig. 7B). Therefore, the transcriptomic changes initiated by the SR seem largely independent of those mediated by LetS, and the survival defect of ppGpp° is not due to an impaired LetS response and vice-versa. We also confirm that RsmY/Z expression is RpoS-dependent in water (Fig. 7C) 65 . ppGpp may contribute to RpoS-mediated RNAP binding to letS or rsmY/Z, but its effect is likely minimal (Fig. 7B). Notwithstanding its effect on transcription, the LetA/S-RsmY/Z-CsrA cascade and the SR seem to be parallel responses contributing to survival under nutrient-deprived conditions. The data presented here suggests the following model. ppGpp and the LetA/S cascade regulate similar regulons in parallel and possibly in different time frames, each promoting adaptation to and survival in water independently. A ppGpp° mutant has a survival defect that is apparent earlier than that of the letS mutant in water 14 , suggesting that the SR is more important than LetS and that its effect is required immediately upon exposure to water. RpoS which is positively affected by ppGpp, positively influences the letS transcript and thereby, levels of RsmY/Z (Fig. 7C). However, whether RpoS increases the level of letS in response to ppGpp in water, or whether basal RpoS levels maintain constant letS expression within the cell is not yet know.
It is likely that LetS responds to a variety of stresses fine-tuning the transcriptomic response to the challenge at hand 23 . In broth, LetA/S was required to increase flagella expression in response to nicotinic acid, and the expression of several virulence traits in response to free fatty acids 75,87 . It is unclear whether these are direct signals and whether they also represent activation signals in water. Recently, the Lp quorum sensing molecule, Legionella autoinducer-1 (LAI-1) was shown to increase RsmY/Z levels in broth 88 ; however, it is unlikely to be a viable signal for the activation of LetS in water, because of the bacterium's low metabolic activity, lowering LAI-1 production under this condition. As such, the environmental signal initiating the LetS/A cascade in water is yet to be determined. This study does not exclude the possibility that ppGpp is one of multiple stimuli sensed by LetS in water; its weak contribution to LetS activation may be masked by other, stronger signals.
In conclusion, we report that the LetA/S-RsmY/Z-CsrA regulatory cascade is essential for the culturability of Lp in water. In contrast to the transcriptome in broth cultures, LetS almost exclusively acts to repress genes related to growth; with RsmY/Z being the prominent exceptions. While there is overlap between the regulons and crosstalk between members of the SR and the LetA/S TCS, activation of the two systems seems to be independent of each other in water.

Experimental Procedures
Bacterial strains and media. KS79 is a ∆comR mutant of the JR32 strain rendering it constitutively competent. JR32 is a salt-sensitive, streptomycin-resistant, restriction negative mutant of Lp strain Philadelphia 1 89 . The increased competence of KS79 renders allelic exchange mutations through natural transformation possible. It was, therefore, used as the wild-type (WT) strain in this study. A complete list of strains used in this study can be found in Table 2. Bacterial strains stored at −80 °C in 10% glycerol were grown on CYE (ACES-buffered charcoal yeast extract) agar supplemented with 0.1 mg ml −1 α-ketoglutarate, 0.25 mg ml −1 L-cysteine and 0.4 mg ml −1 ferric pyrophosphate 90 . AYE broth (CYE without agar and charcoal) was used as the liquid medium 90 . When necessary, media were supplemented with 5 μg ml −1 chloramphenicol, 2.5 μg ml −1 kanamycin and/or 0.1 mM IPTG.
Deletion of rsmY and rsmZ. Construction of the ∆rsmYZ strain was performed by allelic exchange as described before 65 , by replacing, first, rsmY with a kanamycin resistance cassette and, then, rsmZ with a gentamicin resistance cassette. Two 1-kb fragments corresponding to the upstream and downstream of rsmY were amplified using primers rsmY-BF/rsmY-BR and rsmY-EF/rsmY-ER, respectively. A kanamycin cassette was amplified from pSF6 using primers rsmY-BRKN/rsmY-EFKN. PCR fragments were purified on gel using a gel extraction kit (Qiagen). The three fragments were ligated together by PCR using primers rsmY-BF/RsmY-ER. The resulting 3 kb fragment was purified on gel, as described above, and introduced into KS79 by natural transformation 91 . The recombinants were selected for kanamycin resistance and the allelic exchange confirmed by PCR. Deletion of rsmZ was performed similarly by using primers rsmZ-BF/rsmZ-BR and rsmZ-EF/rsmZ-ER to amplify the upstream and downstream fragments, and rsmZ-BRGT and rsmZ-EFGT to amplify a gentamycin cassette from pBBR1MCS-5. The resulting 3 kb fragment was introduced into the ∆rsmY strain, recombinants were selected for kanamycin and gentamicin resistance and the deletion of rsmZ was confirmed by PCR. The resulting strain was named SPF41. Northern blot was used to confirm absence of expression of both sRNA. All PCR amplifications were carried with Phusion polymerase (NEB) according to the manufacturer's protocol. Primer sequences used in this study are found in Supplementary Table S2.
Cloning of rsmY. The rsmY gene was amplified with its own promoter from the KS79 wild-type strain using primers rsmY-F and rsmY-R (Supplementary Table S2). The pXDC39 vector and the amplicon were then digested with BamHI and HindIII (New England Biolabs) according to the manufacturer's protocol, and purified using a MinElute purification kit (Qiagen). The digested vector and insert were ligated overnight using T4 DNA Ligase (New England Biolabs) at 16 °C. The recombinant plasmid was then transformed into E. coli DH5α (pSF86). The transformed population was incubated at 37 °C shaking for 90 minutes before plating on 25ug ml −1 chloramphenicol plates. Colonies that grew on antibiotic plates were patched and tested by PCR for insertion of the rsmY gene into the vector. The recombinant plasmid was extracted using a plasmid extraction kit (Qiagen) and introduced into the ∆rsmYZ mutant to produce the ∆rsmYZ + prsmY strain (SPF291).

Survival in water.
Survival in water was tested in the artificial freshwater medium, Fraquil, as described previously 8,51 . Briefly, Lp strains cultured on CYE agar at 37 °C for 3 days were suspended in Fraquil at an OD 600nm of 0.1 and washed three times with Fraquil. One millilitre of this bacterial suspension was mixed with 4 ml of fresh Fraquil in a 25 cm 2 cell culture flasks (Sarstedt) and incubated at 4 °C, 25 °C or 42 °C. To test survival in water at high cell densities, strains were suspended in Fraquil at an OD 600nm of 1 and washed three times with Fraquil. Five millilitres of this bacterial suspension was placed in a 25 cm 2 cell culture flasks (Sarstedt) and incubated at 42 °C. Survival of the strains in water was monitored using CFU counts.

Microscopic analysis.
Lp strains were grown on CYE agar for 3 days at 37 °C. AYE broth was inoculated with the respective strains and grown to exponential phase (OD 600 0.4-0.7) or post-exponential phase (OD 600 > 3) at 37 °C shaking (200 rpm). To test the effect of water, strains were suspended in Fraquil at an OD 600 of 1 and incubated at 42 °C for 24 hours. 20 μl of each sample was placed on a clean microscope slide, covered with a cover slip and observed under 1000X magnification under oil immersion using digital phase contrast microscopy (Nikon Eclipse 80i). For each strain, 10 images of random microscopic fields were captured using the NIS Element Software (Nikon Instruments, Inc.). ImageJ software 92 was used for quantitative analysis of cell length. 10 cells from 10 different fields of view were randomly chosen and analyzed per strain per treatment (n = 100). Multiplying cells (presence of a septum) were excluded and only individual, non-filamentous cells were used for analysis. Pigment production. Lp strains grown on CYE for 3 days at 37 °C were inoculated into AYE broth. Strains were grown to late post-exponential phase at 37 °C shaking at 200 rpm. 10 ml of each strain was pelleted at 4500 rpm for 10 minutes. The supernatant was then removed and filtered using 0.2μm pore sized syringe filter. The optical density of the supernatant was measured at 550 nm.
Heat shock. Lp strains cultured on CYE agar at 37 °C for 3 days were suspended in Fraquil at an OD 600nm of 0.1, as described above for the survival assay. One milliliter aliquots of each strain were transferred to 13 ml tubes (Sarstedt) and were allowed to acclimate to the water environment for 2 hours at room temperature. At the end of the incubation period, tubes were submerged in a 55 °C water bath. At each time point tested, three biological replicates from each strain were removed from the water bath and the CFU counts enumerated on CYE agar. DNA microarray. The WT, ∆letS, and the complemented strain induced (ON) or not (OFF) with IPTG were grown on CYE agar for 3 days at 37 °C. Each culture was suspended in Fraquil in triplicate at an OD 600 of 1. 20 ml of each strain was placed in 75 cm 2 cell culture flasks and incubated at 42 °C for 2 hours. After incubation, 10 ml aliquots were pelleted for 5 minutes at 5000 g. After centrifugation, the supernatant was removed, and the cell pellets were re-suspended in 1 ml of TRIzol reagent (Invitrogen). Three biological replicates of each strain were used for the transcriptomic analysis. The three remaining replicates of each strain were preserved at −20 °C for further experimentation. RNA extractions were done according to the manufacturer's protocol. To remove DNA contamination, extracted RNA was subsequently treated with Turbo DNase (Ambion) as per the manufacturer's protocol. The purity  Table 2. Strains used in this study. a Sm r , streptomycin resistance; Cm r , chloramphenical resistance; Gm r , gentamicin resistance; Km r , kanamycin resistance. and concentration of RNA were determined by UV spectrophotometry. Fifteen micrograms of RNA was labeled with aminoallyl-dUTP (Sigma) during reverse transcription (ProtoScript II, New England Biolabs) using random hexamers (Life Technologies) as previously described 8,93 . Genomic DNA was used as a reference channel and was labeled by random priming using Klenow fragments (New England Biolabs), aminoallyl-dUTP and random primers as described previously 93 . The cDNA and gDNA were subsequently coupled to the succinimidyl ester fluorescent dye (Life Technologies) AlexaFluor 647 or AlexaFluor 546, respectively, following the manufacturer's protocols. The microarray design (GPL19458) and the protocol for hybridization, data acquisition and data analysis have been published previously 8 . Statistical analyses were performed using an unpaired one-tailed Student's t-test. Genes were considered differentially expressed if they demonstrated a ratio-to-control value of ± 2-fold with a P < 0.05. qPCR validation. One replicate of the WT, ∆letS and the complemented strain induced (ON) or not (OFF) with IPTG exposed to water as described above was used to validate the microarray results using qPCR. RNA was extracted as described above. 1 μg of total RNA was transcribed to cDNA (Supercript II, Life Technologies) using random hexamers. qPCR was performed as described previously using the primers described in Supplementary  Table S2. Ct values were normalized to the 16 S rRNA.
Northern Blotting. Lp strains grown on CYE agar for 3 days at 37 °C were used to inoculate AYE broth. All strains were grown to exponential phase (OD 600 0.4-0.7) at 37 °C shaking (200 rpm). For each strain, 10 ml of exponential phase bacterial culture was centrifuged for 10 minutes at 4500 rpm, the supernatant removed and the pellet re-suspended in 10 ml of Fraquil. Water-exposed bacteria were incubated at 42 °C for 2 hours, after which cells were pelleted and suspended in 1 ml of TRIzol reagent (Invitrogen). RNA was extracted according to the manufacturer's protocol. RNA from the WT was also extracted from 10 ml of exponential phase culture and 5 ml of post-exponential phase culture (OD 600 > 3). 1 ug of RNA was loaded and migrated on a 6% Tris-borate-EDTA-urea polyacrylamide gel (Ambion) at 180 mV. The RNA was transferred onto a positively charged nylon membrane (Thermo Scientific) using a semidry gel blotting system (BioRad) for 20 minutes at 200 mA. The membranes were pre-hybridized in ULTRAhyb-Oligo Hybridization Buffer (Ambion) for 1 hour at 37 °C before hybridization with 5′ biotinylated RsmY and RsmZ probes (Integrated DNA Technologies Availability of materials and data. Microarray data generated during the current study are available from Gene Expression Omnibus (GSE98743).