ABSTRACT
The mutualistic symbiont Vibrio fischeri builds a symbiotic biofilm during colonization of squid hosts. Regulation of the exopolysaccharide component, termed Syp, has been examined in strain ES114, where production is controlled by a phosphorelay that includes the inner membrane hybrid histidine kinase RscS. Most strains that lack RscS or encode divergent RscS proteins cannot colonize the squid host unless RscS from a squid symbiont is heterologously expressed. In this study, we examine V. fischeri isolates worldwide to understand the landscape of biofilm regulation during beneficial colonization. We provide a detailed study of three distinct evolutionary groups of V. fischeri and find that while the RscS-Syp biofilm pathway is required in one of the groups, two other groups of squid symbionts require Syp independent of RscS. Mediterranean squid symbionts, including V. fischeri SR5, colonize without an RscS homolog encoded in their genome. Additionally, Group A V. fischeri strains, which form a tightly-related clade of Hawaii isolates, have a frameshift in rscS and do not require the gene for squid colonization or competitive fitness. These same strains have a frameshift in sypE, and we provide evidence that this Group A sypE allele leads to an upregulation in biofilm activity. This work thus describes the central importance of Syp biofilm in colonization of diverse isolates, and demonstrates that significant evolutionary transitions correspond to regulatory changes in the syp pathway.
IMPORTANCE Biofilms are surface-associated, matrix-encased bacterial aggregates that exhibit enhanced protection to antimicrobial agents. Biofilms are prevalent in environmental, beneficial, and pathogenic interactions, though their role in beneficial colonization has been understudied. Previous work has established the importance of biofilm formation by a strain of luminous Vibrio fischeri bacteria as the bacteria colonize their host, the Hawaiian bobtail squid. In this study, expansion of this work to many natural isolates revealed that biofilm genes are universally required, yet there has been a shuffling of the regulators of those genes. This work provides evidence that even when bacterial behaviors are conserved, dynamic regulation of those behaviors can underlie evolution of the host colonization phenotype. Furthermore, this work emphasizes the importance of investigating natural diversity as we seek to understand molecular mechanisms in bacteria.
INTRODUCTION
A fundamental question in studying host-associated bacterial communities is understanding how specific microbial taxa assemble reproducibly in their host. Key insights into these processes were first obtained by studying plant-associated microbes, and the discovery and characterization of Nod factors in Rhizobia was valuable to understand how partner choice between microbe and host could be mediated at the molecular level (1, 2). There are complex communities in humans and other vertebrate animals, yet metagenomic and imaging analyses of these communities have revealed striking reproducibility in the taxa present and in the spatial arrangement of those taxa (3–5). Invertebrate animal microbiomes provide appealing systems in which to study microbiome assembly in an animal host: the number of taxa are relatively small, and examination and manipulation of these organisms have yielded abundant information about processes underlying host colonization (6). For this work we focused on the binary symbiosis between Vibrio fischeri and bobtail squids, including the Hawaiian bobtail squid, Euprymna scolopes. Bobtail squid have an organ for the symbiont termed the light organ, and passage of specific molecules between the newly-hatched host and the symbiont leads to light organ colonization specifically by planktonic V. fischeri and not by other bacteria (7–9). The colonization process involves initiation, accommodation, and persistence steps, resulting in light organ crypt colonization by V. fischeri. Upon colonization of the squid light organ, bacteria accumulate to high density and produce light. The bacterial light is modulated by the host to camouflage the moonlight shadow produced by the nighttime foraging squid in a cloaking process termed counter-illumination (10, 11). A diel rhythm leads to a daily clearing of 90-95% of the bacteria from the crypts and regrowth of the remaining cells (12). However, the initial colonization process, including biofilm-based aggregation on the host ciliated appendages, occurs only in newly-hatched squid. This work examines regulation of biofilm formation in diverse squid-colonizing V. fischeri strains.
In the well-studied V. fischeri strain ES114, biofilm formation is required to gain entry into the squid host. RscS is a hybrid histidine kinase that regulates V. fischeri biofilm formation through a phosphorelay involving the hybrid histidine kinase SypF and the response regulator and σ54-dependent activator SypG (13–15). This pathway regulates transcription of the symbiosis polysaccharide (Syp) locus, which encodes regulatory proteins (SypA, SypE, SypF, and SypG), glycosyltransferases, factors involved in polysaccharide export, and other biofilm-associated factors (14, 16). The products of the ES114 syp locus direct synthesis and export of a biofilm exopolysaccharide that is critical for colonization. Additional pathways have been identified to influence biofilm regulation in ES114, including the SypE-SypA pathway and inhibition of biofilm formation by BinK and HahK (17–21).
V. fischeri biofilm regulation is connected to host colonization specificity. In the Pacific Ocean, the presence of rscS DNA is strongly correlated to the ability to colonize squid (22). As one example, while the fish symbiont MJ11 encodes a complete syp locus, it lacks RscS and does not robustly colonize squid. Heterologous expression of ES114 RscS in MJ11 activates the biofilm pathway and is sufficient to enable squid colonization (22). Similarly, addition of ES114 RscS to mjapo.8.1--a fish symbiont that encodes a divergent RscS that is not functional for squid colonization--allows the strain to colonize squid (22). RscS has also been shown to be necessary for squid colonization. In addition to ES114, interruption of rscS in V. fischeri strains KB1A97 and MJ12 renders them unable to colonize squid. Previous phylogenetic analysis revealed that ancestral V. fischeri do not encode rscS, and that it was acquired once during the organism’s evolution, likely allowing for an expansion in host range. From this analysis, it was concluded that strains with rscS can colonize squid, with the only exception being the fish symbionts that harbor the divergent RscS, including mjapo.8.1 (22).
There are similar Vibrio-squid associations worldwide, yet only V. fischeri and the closely-related Vibrio logei have been isolated from light organs (23–26). Our 2009 study revealed that although most symbionts have rscS DNA, there are Mediterranean V. fischeri (e.g., SR5) that do not have rscS yet can colonize squid (22, 24, 27). This unexpected finding prompted the current work to examine whether strains such as SR5 colonize with the known biofilm pathway or with a novel pathway. Here, we show that all V. fischeri strains tested require the Syp biofilm to colonize a squid host, and we identify two groups of isolates that colonize with novel regulation. Given the exquisite specificity by which V. fischeri bacteria colonize squid hosts, this work reinforces the importance of biofilm formation and reveals different regulatory modes across the evolutionary tree.
RESULTS
Most V. fischeri strains synthesize biofilm in response to RscS overexpression
Biofilm formation is required for squid colonization, and overexpression of the biofilm regulator RscS in strain ES114 stimulates a colony biofilm on agar plates (15). Our previous work demonstrated that V. fischeri strain MJ11 synthesizes a colony biofilm under similar inducing conditions, which is notable because MJ11 does not encode RscS in its chromosome (22). While the ancestral strain MJ11 did not encode RscS, it had what seemed to be an intact syp locus, and overexpression of the heterologous RscS from ES114 was sufficient to induce syp biofilm production and enable robust squid colonization (22). We examined a phylogenetic tree of V. fischeri isolates (Fig. 1), and in this study we expand our analysis of RscS-Syp biofilm regulation in a wider group of V. fischeri strains.
Initially, we asked whether responsiveness to RscS overexpression would yield a similar colony biofilm in this diverse group of strains. We took the same approach as our previous study and introduced plasmid pKG11, which overexpressed ES114 RscS, into strains across the evolutionary tree (22, 28). We observed that almost all strains tested, including those that lack rscS, were responsive to overexpression of ES114 RscS (Fig. 2). The morphology of the colony biofilms differed across isolates; but in most cases colony biofilm was evident at 24 h and prominent at 48 h. All of the strains exhibited some wrinkled colony morphology at 48 h with the exception of CG101, which was isolated from the pineapplefish Cleidopus gloriamaris (25). These results demonstrated that most V. fischeri strains can produce biofilm in response to RscS overexpression, and this includes strains that have not encountered rscS in their evolutionary history.
One unexpected observation was that there was a subset of rscS-encoding strains that were reproducibly delayed in their colony biofilm, and had only a mild wrinkled colony phenotype at 48 h (strains MB11B1, ES213, KB2B1; Fig. 2). We considered whether this was due to differential growth of the strains, but resuspension of spots and dilution plating to determine CFU/spot demonstrated no significant growth difference between these strains and ES114 under these conditions. The strains are closely-related (Fig. 1) and a previous study had noted that this group shared a number of phenotypic characteristics, e.g. reduced motility in soft agar (29). Those authors termed this tight clade as “Group A” V. fischeri (30). Our results in Figure 2 argue that Group A strains do not respond to RscS in the same manner as other V. fischeri strains, which prompted us to investigate the evolution of the RscS-Syp signaling pathway. We have maintained the Group A nomenclature here, and furthermore we introduce the nomenclature of Group B (a paraphyletic group of strains that contain rscS; this group includes the common ancestor of all rscS-containing strains) and Group C (a paraphyletic group of strains that contains the common ancestor of all V. fischeri – these strains do not contain rscS), as shown in Figure 1.
Ancestral Group C squid isolates colonize E. scolopes independent of RscS
Group C strains generally cannot colonize squid, yet there are Mediterranean squid isolates that appear in this group (Fig. 1; (22)). The best-studied of these strains, SR5, was isolated from Sepiola robusta, is highly luminous, and colonizes the Hawaiian bobtail squid E. scolopes (24). Nonetheless, this strain lacks rscS (27). We first asked whether the strain can colonize in our laboratory conditions, and we confirmed that it colonizes robustly, consistent with the result result previously published by Fidopiastis et al. (24) (Fig. 3). Next, we asked whether it uses the Syp biofilm to colonize. To address this question, we deleted the 18 kb syp locus (i.e., sypA through sypR) in strains SR5 and ES114. Deletion of rscS or the syp locus in ES114 led to a substantial defect in colonization, consistent with a known role for these factors (Fig. 3). Similarly, deletion of the syp locus in SR5, a strain that does not encode rscS, led to a dramatic reduction in colonization (Fig. 3). Therefore, even though strain SR5 does not encode rscS, it can colonize squid, and it requires the syp locus to colonize normally.
RscS is dispensable for colonization in Group A strains
We noted in the wrinkled colony biofilm assays shown in Figure 2 that Group A strains exhibited a more modest response to overexpression of RscS. Sequencing of the native rscS gene in these strains revealed a predicted −1 frameshift (ΔA1141) between the PAS domain and the histidine kinase CA domain.
Whereas ES114 and other Group B strains have nine adenines at this position, the Group A strains have eight, leading to a frameshift and then truncation at an amber stop codon, raising the possibility that Group A strains have a divergent biofilm signaling pathway (Fig. 4A). Given the importance of RscS in the Group B strains including ES114, we considered the possibility that this apparent frameshift encoded a functional protein, either through ribosomal frameshifting or through the production of two polypeptides that together provided RscS function; there is precedent for both of these concepts in the literature (31, 32). We first introduced a comparable frameshift into a plasmid-borne overexpression allele of ES114 rscS, and this allele did not function with the deletion of the single adenine (Fig. 4B). This result suggested to us that the frameshift in the Group A strains may not be functional. Therefore, we proceeded to delete rscS in two Group A strains (MB11B1, ES213) and two Group B strains (ES114, MB15A4). The Group B strains required RscS for squid colonization (Fig. 5A).
However, the Group A strains exhibited no deficit in the absence of rscS DNA (Fig. 5A). We next attempted a more sensitive assay in which a Group A strain was competed against MB15A4. Previous studies have demonstrated that in many cases Group A strains outcompete what we term Group B strains (30, 33). We competed Group A strain MB11B1 against Group B strain MB15A4 and observed a significant competitive advantage for the Group A strain, as was observed previously (30). Deletion of rscS in the Group A strain did not affect competitive fitness, demonstrating that MB11B1 can outcompete a Group B strain even if MB11B1 lacks RscS (Fig. 5B).
Syp biofilm genes are broadly required for squid colonization
Given that Group A strains seemed to represent a tight phylogenetic group in which RscS was not required for colonization or competitive fitness, we next asked whether this group requires the Syp biofilm for colonization. We proceeded to delete the entire syp locus in two Group A and two Group B strains and to conduct single-strain colonization analysis. In each strain assayed, the syp locus was required for full colonization, and we observed a 2-4 log reduction in CFU per animal in the absence of the syp genes, pointing to a critical role for Syp biofilm in these strains (Fig. 6). In Group A strain in particular, no colonization was detected in the absence of the syp locus.
Group A strains encode an alternate allele of SypE
It seemed curious to us that Group A strains do not encode a functional RscS and do not require rscS for colonization, yet in many cases Group A strains can outcompete Group B strains (e.g. MB11B1 in Fig. 5B; and Refs. (30, 33)). We reasoned that if the Syp biofilm had a different regulatory architecture in Group A strains--e.g., constitutively activated or activated by a different regulatory protein--then this could explain the Syp regulation independent of RscS. Genome sequencing of SR5 and MB11B1 did not identify a unique histidine kinase that was likely to directly substitute for RscS (27, 33).
Given that the syp locus encodes biofilm regulatory proteins, we examined syp conservation. We used TBLASTN with the ES114 Syp proteins as queries to determine amino acid conservation in the other V. fischeri Group A strain MB11B1, Group C strain SR5, and the Vibrio vulnificus type strain ATCC 27562 (34, 35). As shown in Figure 7, ES114 SypE, a response regulator and serine kinase/phosphatase that is a negative regulator of the Syp biofilm (17, 36), exhibited the lowest level of conservation among syp locus products. V. vulnificus does not encode a SypE ortholog (37), as the syntenic (but not homologous) RbdE encodes a predicted ABC transporter substrate-binding protein. The closest hit for SypE was AOT11_RS12130 (9% identity), compared to 7% identity for the RbdE. Due to the reduced conservation at both the strain and species levels, we analyzed V. fischeri MB11B1 SypE in greater detail. Examination of the sypE coding sequence revealed an apparent −1 frameshift mutation in which the position 33 (guanine in ES114 and adenine in other Group B and C strains examined) is absent in Group A strains (Fig. 7B). We therefore considered the hypothesis that SypE is nonfunctional in Group A, and that these strains can colonize because they are lacking a functional copy of this negative regulator that is itself regulated by RscS.
To test this hypothesis, we relied on knowledge of the biofilm regulatory pathway from ES114, in which overexpression of SypG produces a wrinkled colony phenotype, but only in strains lacking SypE activity (38, 39). Therefore, we introduced the SypG-overexpressing plasmid pEAH73 into strains as a measure of whether the SypE pathway was intact. In the ES114 strain background, we observed cohesive wrinkled colony formation at 48 h in an ES114 ΔsypE strain, but not in the wild-type parent (Fig. 8A). If the sypE frameshift observed in MB11B1 led to a loss of function, then introduction of that frameshift into ES114 would lead to a strain that is equivalent to the ΔsypE strain. We constructed this strain and upon SypG overexpression we observed wrinkled colony formation. Surprisingly, the biofilm phenotype was observed earlier (i.e., by 24 h) and leads to more defined colony biofilm architecture at 48 h. While the lack of SypE leads to increased and more rapid biofilm formation, in this assay we observed an even greater increase as a result of the frameshift in sypE (Fig. 8A).
We proceeded to conduct a similar assay in the MB11B1 strain background. The colony biofilm phenotypes were muted compared to the ES114 background, but the pattern observed is the same. Strains lacking the additional nucleotide at position 33 (i.e., the native MB11B1 allele) exhibited the strongest cohesion, whereas strains with the nucleotide to mimic ES114 sypE (i.e., added back in MB11B1 sypE(nt::33G)) were not cohesive (Fig. 8B). These results argue that a novel allele of sypE is found in Group A strains and this allele results in more substantial biofilm formation than in a ΔsypE strain.
Our finding that the MB11B1 sypE allele promotes biofilm formation bolstered the model that this allele contributes to the ability of MB11B1 to colonize squid independent of RscS. To test this model, we introduced the frameshift into ES114 or “corrected” the frameshift in MB11B1.
We then conducted single-strain colonization assays, and in each case the sypE allele alone was not sufficient to alter the overall colonization behavior of the strain (Fig. 9). Therefore, these data suggest that the frameshift in the MB11B1 sypE is not sufficient to explain its ability to colonize independent of RscS, and therefore other regions of SypE and/or other loci in the MB11B1 genome contribute to its ability to colonize independent of RscS.
BinK is active in Group A and Group C strains
We recently described the histidine kinase, BinK, which negatively regulates syp transcription and Syp biofilm formation (18). In ES114, overexpression of BinK impairs the ability of V. fischeri to colonize. We therefore reasoned that if BinK could function in Group A strains and acted similarly to repress Syp biofilm, then overexpression of BinK would reduce colonization of these strains. We introduced the pBinK plasmid (i.e., ES114 binK (18)) and asked whether multicopy BinK would affect colonization. In strain MB11B1, BinK overexpression led to a dramatic reduction in colonization (Fig. 10A). Therefore, there is a clear effect for BinK overexpression on the colonization of the Group A strain MB11B1.
We attempted to ask the same question in Group C strain SR5, but the pES213-origin plasmids were not retained during squid colonization. Therefore, we instead asked whether deletion of the BinK, a negative regulator of ES114 colonization, has a comparable effect in SR5 (18). We deleted binK and observed a 2.4-fold competitive advantage during squid competition (Fig. 10B), arguing that BinK in this Group C strain is active and performs an inhibitory function similar to that in ES114.
We next examined the colony biofilm phenotype for strains lacking BinK. MB11B1 ΔbinK exhibited a mild colony biofilm phenotype at 48 h, as evidenced by the cohesiveness of the spot when disrupted with a toothpick (Fig. 10C). The colonies also exhibited an opaque phenotype.
In a minority of experimental replicates, wrinkled colony morphology was evident at 48 h, but in all samples wrinkled colony morphology was visible at 7 d (data not shown). The SR5 ΔbinK strain also exhibited slightly elevated biofilm morphology at 48 h, though the cells were not as cohesive as those of MB11B1 ΔbinK (Fig. 10C). Together, the results in Figure 7 argue that BinK, a factor that has been characterized as a negative regulator of Syp biofilm, plays similar roles in Group A and Group C strains and has a widely-conserved function across the V. fischeri evolutionary tree.
DISCUSSION
This study examines regulation of a beneficial biofilm that is critical to host colonization specificity in V. fischeri. The Syp biofilm was discovered thirteen years ago and has been characterized extensively for its role in facilitating squid colonization by V. fischeri. This work establishes that the syp locus is required broadly across squid symbionts, and it uncovers three groups of V. fischeri that use different regulatory programs upstream of the syp locus. A simplified phylogenetic tree showing key features of squid symbionts in these three groups is shown in Figure 11.
There are three nested evolutionary groups of V. fischeri that have been described separately in the literature and here we formalize the nomenclature of Groups C, B, and A. Group A is a monophyletic group, as are Groups AB and ABC (Fig. 1). This work provides evidence that squid symbionts in each group have a distinct biofilm regulatory architecture. Most V. fischeri isolates that have been examined from the ancestral Group C cannot colonize squid; however, those that can colonize do so without the canonical biofilm regulator RscS. We show that the known targets of RscS regulation—genes in the syp biofilm locus—are nonetheless required for squid colonization by this group. Group B strains include the well-characterized ES114 strain, which requires RscS and the syp locus to colonize squid. Group A strains differ phenotypically and behaviorally from the sister Group B strains (30), and we demonstrate that these strains have altered biofilm regulation. Group A strains have a frameshift in rscS that renders it nonfunctional, and a 1 bp deletion in sypE, and we provide evidence that the sypE allele promotes biofilm development in the absence of RscS. Additionally, we note that the sypE frameshift is not present in SR5, arguing for distinct modes of biofilm regulation in Groups A, B, and C.
At the same time, this study provides evidence that some aspects of biofilm regulation are conserved in diverse squid symbionts, such as the effects of the strong biofilm negative regulator BinK. Published data indicate that evolved BinK alleles can alter colonization of H905 (Group B) and MJ11 (Group C), and that a deletion of MJ11 binK leads to enhanced colonization (20). Our experiments in Figure 10 show a clear effect for BinK in all three phylogenetic groups. We also observed responsiveness to RscS overexpression in all squid symbionts examined (Fig. 2). CG101 was the only V. fischeri strain examined that did not exhibit a colony biofilm in response to RscS overexpression. CG101 was isolated from the Australian fish Cleidopus gloriamaris; based on these findings, we suspect that the strain does not have an intact syp locus or otherwise has divergent biofilm regulation.
It remains a formal possibility that the entire syp locus is not required in Group A or Group C, but instead that only one or a subset of genes in the locus are needed. We have constructed Campbell-type (insertion duplication) alleles to interrupt sypG in MB11B1 and SR5, and additionally have isolated a transposon insertion in SR5 sypJ, and none were able to colonize well. Additionally, aggregation in squid mucus has been observed for the Group A strain MB13B2, and this aggregation is dependent on sypQ (40). In our data we note that Group A strains were completely unable to colonize in the absence of the syp locus, unlike the tested Group B & C strains that exhibited reduced colonization in their respective mutants (Figs. 3, 6).
Therefore, the simplest explanation is that the syp locus is required in divergent strains in a manner similar to how it is used in ES114. We think that the ability to completely delete the syp locus is a clean way to ask whether the locus is required for specific phenotypes, and our strains are likely to be useful tools in probing Syp protein function in diverse V. fischeri isolates.
It is intriguing to speculate as to how the two frameshifts in the Group A strains arose, and why the nonfunctional RscS is tolerated in this group. One possible scenario is that the Group A strains acquired a new regulatory input into the Syp pathway, and that the presence of this new regulator bypassed the requirement for RscS. This is supported by comparative genomic analysis of Hawaiian D (dominant)-type strains--which largely overlap with Group A--that demonstrated an additional 250 kb of genomic DNA compared to other isolates (33). A related possibility is that rscS-independent colonization results from altered regulation of the syp locus, either due to changes in regulators (e.g. SypF) or sites that are conserved with Group B. An additional possibility is that the sypE frameshift arose, enabling Group A strains to colonize independent of rscS. Given that correction of this frameshift in MB11B1 does not significantly affect colonization ability (Fig. 9), this sequence seems less likely, and we expect that another regulator in MB11B1 is required for the RscS-independent colonization phenotype. There is evidence that under some conditions LuxU can regulate the syp biofilm (41), and as this protein is conserved in V. fischeri it may play an important role in Group A or Group C.
Results from two experimental conditions suggest that the Group A strains may have an elevated baseline level of biofilm formation. Our data indicate that in the absence of BinK or upon SypG overexpression, MB11B1 colonies exhibit strong cohesion under conditions in which ES114 does not (Figs. 8, 10). Furthermore, we note that the Group A strain MB11B1, when lacking BinK, also exhibits a darker, or more opaque, colony phenotype (Fig. 10). This phenotype has been observed in some ES114 mutants (16) but not in the corresponding ES114 ΔbinK strain (Fig. 10). The entire colonization lifecycle likely requires a balance between biofilm formation/cohesion and biofilm dispersal, and these data argue that Group A strains may be more strongly tilted toward the biofilm-producing state. There is evidence that strains lacking BinK exhibit a colonization advantage in the laboratory (18, 20), suggesting that this strategy of more readily forming biofilms may provide a fitness advantage in nature. At the same time, the biofilm negative regulator BinK is conserved among V. fischeri strains examined (including MB11B1; Fig. 10), arguing that there is a benefit to reducing biofilm formation under some conditions.
Our study provides hints as to the role of SypE in MB11B1 and other Group A strains. In ES114, the C-terminus is a PP2C serine kinase domain, whereas the N-terminus of SypE is an RsbW serine phosphatase domain. SypE acts to phosphorylate and dephosphorylate SypA Ser-56, with the unphosphorylated SypA being the active form to promote biofilm development (17). The balance between SypE kinase and phosphatase is modulated by a central two-component receiver domain (17). Our data that the MB11B1 sypE allele promotes biofilm formation suggest that the protein is tilted toward the phosphatase activity. In MB11B1, the frameshift early in sypE suggests that there is a different start codon and therefore a later start codon. An alternate GTG start codon in MB11B1 occurs corresponding to codon 18 in ES114 sypE (Fig. 7), and this is likely the earliest start for the MB11B1 polypeptide. We attempted to directly identify the SypE N-terminus by mass spectrometry, yet we could not identify the protein from either strain. Additional study is required to elucidate how MB11B1 SypE acts to promote biofilm formation.
V. fischeri strains are valuable symbionts in which to probe the molecular basis to host colonization specificity in animals (22, 25, 26). A paradigm has emerged in which biofilm formation through the RscS-Syp pathway is required for squid colonization but not for fish colonization. This study affirms a role of the Syp biofilm, but at the same time points out divergent (RscS-independent) regulation in Group C and Group A isolates. In another well-studied example of symbiotic specificity, Rhizobial Nod factors are key to generating specificity with the plant host, yet strains have been identified that do not use this canonical pathway (42, 43). Future work will elaborate on these RscS-independent pathways to determine how non-canonical squid colonization occurs in diverse natural isolates.
MATERIALS & METHODS
Bacterial strains and growth conditions
V. fischeri and E. coli strains used in this study can be found in Table 1. E. coli strains, used for cloning and conjugation, were grown in Luria-Bertani (LB) medium (25 g Difco LB Broth [BD] per liter). V. fischeri strains were grown in Luria-Bertani salt (LBS) medium (25 g Difco LB Broth [BD], 10 g NaCl, and 50 ml 1 M Tris buffer pH 7.0, per liter). Growth media were solidified by adding 15 g Bacto agar (BD) per liter. When necessary, antibiotics (Gold Biotechnology) were added at the following concentrations: tetracycline, 5 μg/ml for V. fischeri; erythromycin, 5 μg/ml for V. fischeri; kanamycin, 50 μg/ml for E. coli and 100 μg/ml for V. fischeri; and chloramphenicol, 25 μg/ml for E. coli, 2.5 −5 μg/ml for Group B V. fischeri, and 1 – 2.5 μg/ml for Group A V. fischeri. The two MB11B1 / pKV69 strains listed reflect two separate constructions of this strain, though we have not identified any differences between them.
Phylogenetic analysis
Phylogenetic reconstructions assuming a tree-like topology were created with three methods: maximum parsimony (MP); maximum likelihood (ML); and Bayesian inference (Bayes) as previously described (22, 30). Briefly, MP reconstructions were performed by treating gaps as missing, searching heuristically using random addition, tree-bisection reconnection with a maximum of 8 for swaps, and swapping on best only with 1000 repetitions. For ML and Bayesian analyses, likelihood scores of 1500+ potential evolutionary models were evaluated using both the corrected and uncorrected Akaike Information Criterion, the Bayesian Information Criterion, and Decision Theory (Performance Based Selection) as implemented by jModelTest2.1 (44). For all information criteria, the most optimal evolutionary model was a symmetric model with a proportion of invariable sites and a gamma distribution of rate heterogeneity (SYM+I+Γ).
ML reconstruction was implemented via PAUP*4.0a163 (45) by treating gaps as missing, searching heuristically using random addition, tree-bisection reconnection for swaps, and swapping on best only with 1000 repetitions. Bayesian inference was done by invoking the ‘nst=6’ and ‘rates=invgamma’ and ‘statefreqpr=fixed(equal)’ settings in the software package MrBayes3.2.6 (46). The Metropolis-coupled Markov chain Monte Carlo (MCMCMC) algorithm used to estimate the posterior probability distribution for the sequences was set up with ‘temp=0.2’ and one incrementally ‘heated’ chain with three ‘cold’ chains; these four chains were replicated two times per analysis to establish convergence of the Markov chains (i.e., ‘stationarity’ as defined by (47) and interpreted previously in (30)). For this work, stationarity was achieved after approximately 50,000 samples (5,000,000 generations) were collected, with 25% discarded. The ~37,500 samples included were used to construct a 50% majority-rule consensus tree from the sample distribution generated by MCMCMC and assess clades’ posterior probabilities. For ML and MP analyses, the statistical confidence in the topology of each reconstruction was assessed using 1000 bootstrap replicates. Phylogenetic trees were visualized with FigTree 1.4.3 (http://tree.bio.ed.ac.uk/software/figtree); the final tree was edited for publication with Inkscape 0.91 (http://inkscape.org/) and GIMP 2.8.22 (http://www.gimp.org/).
DNA synthesis and sequencing
Each of the primers listed in Table 3 was synthesized by Integrated DNA Technologies (Coralville, IA). Full inserts from all cloned constructs were verified by Sanger DNA sequencing through ACGT, Inc via the Northwestern University Feinberg School of Medicine NUSeq Core Facility; or the University of Wisconsin-Madison Biotechnology Center. Sequence data was analyzed with SeqMan Pro (DNAStar software), SnapGene (GSL Biotech), and Benchling.
Construction of gene deletions
Deletions in V. fischeri strains ES114 and MB11B1 were made according to the lab’s gene deletion protocol: doi:10.5281/zenodo.1470836. In brief, 1.6 kb upstream and 1.6 kb downstream of the targeted gene or locus were cloned into linearized plasmid pEVS79 (amplified with primers pEVS79_rev_690/pEVS79_for_691) using Gibson Assembly (NEBuilder HiFi DNA Assembly cloning kit) with the primer combinations listed in Table S1. The Gibson mix, linking together the upstream and downstream flanking regions, was transformed into E. coli on plates containing X-gal, with several white colonies selected for further screening by PCR using primers flanking the upstream/downstream junction (Tables 3 and S1). The resulting plasmid candidate was confirmed by sequencing and conjugated into the V. fischeri recipient by tri-parental mating with helper plasmid pEVS104, selecting for the chloramphenicol resistance of the plasmid backbone. V. fischeri colonies were first screened for single recombination into the chromosome by maintaining antibiotic resistance in the absence of selection and then screened for double recombination by the loss of both the antibiotic resistant cassette and the gene/locus of interest. Constructs were verified by PCR (Table 3) and sequencing.
Deletion of SR5 binK was conducted using Splicing by Overlap Extension PCR (SOE-PCR) and natural transformation (method modified from (48)). Oligos binK-F1 and binK-R1-LUH, and oligos binK-F2-RUH and binK-R2 were used in a PCR with MJM1125 (SR5) genomic DNA as the template to amplify DNA fragments containing ~1 kb of sequence upstream and downstream relative to binK, respectively. Using SOE-PCR, these fragments were fused on either side to a third DNA fragment containing an ErmR cassette, which was amplified using pHB1 as template and oligos HB41 and HB42. We then used natural transformation with pLostfoX (49) to insert this mutagenic DNA into MJM1125, where the flanking sequences guide the ErmR cassette to replace binK, generating the desired gene deletion. Candidate SR5 ΔbinK mutants were selected after growth on LBS-Erm5 plates. Oligos binK-F1 and binK-R2, and HB8 and binK-FO were used to screen candidates for the correct deletion scar by PCR, and oligos KMB_036 and KMB_037 were used to confirm the absence of binK in the genome. The deletion was verified by Sanger sequencing with primers HB8, HB9, HB42, and HB146. The base plasmid pHB1 contains an erythromycin resistance cassette flanked by FRT sites, and was constructed using oligos HB23 and HB39 with gBlock gHB1 (sequence in Supplementary File S1; Integrated DNA Technologies, Inc.) as template to amplify the ErmR cassette flanked by HindIII and BamHI sites, which was then cloned into the corresponding site in pUC19.
For most constructs, the deleted genetic material was between the start codon and last six amino acids (50), with two exceptions: the ΔsypE in MJM1130 included the ATG that is two amino acids upstream of the predicted start codon, but not the canonical start codon; and the ΔbinK alleles in MJM1117, MJM1130, and MJM2114, which were constructed to be equivalent to MJM2251 (ΔbinK in ES114) (18). The ΔbinK alleles in these strains include the start codon, the next six codons, two codons resulting from ATCGAT (ClaI site), and the last three codons for a predicted 12 amino acid peptide.
Construction of sypE alleles
To create sypE(ntG33Δ) in MJM1100 and sypE(nt33::G) in MJM1130, the single point mutation was created by amplifying the gene in two halves, with the N-terminal portion consisting of approximately 300 bp upstream of sypE up through nucleotide 33 and the C-terminal portion consisting of nucleotide 33 and the remaining sypE gene. The overlap between the two halves contained the single nucleotide polymorphism in the primers that connected them. The altered sypE alleles were initially cloned into plasmid pEVS107 (linearized with primers pEVS107_3837/pEVS107_3838) using Gibson Assembly and then the entire altered sypE allele was subcloned into pEVS79 with Gibson Assembly (Table S1). After double recombination of the vector into V. fischeri, candidate colonies for the altered sypE in MJM1100 were screened with primers ES114_indel_for/ES114_indel_rev. The primer set anneals more strongly to the wildtype sypE sequence than to sypE(ntG33::Δ). Candidates in the MJM1100 background with a fainter PCR band were sequenced and confirmed to have the sypE(ntG33::Δ) allele. For MJM1130, the primer set MB11B1_indel_for/MB11B1_indel_rev anneals more strongly to the sypE(nt33::G) allele than to the naturally occurring sypE allele and candidates in MJM1130 that contained a more robust PCR band were selected for sequencing to be confirmed as being sypE(nt33::G).
Construction of pKG11 rscS1(ntA1141::Δ)
Plasmid pKG11 encodes an overexpression allele of RscS, termed rscS1 (15, 28). rscS nucleotide A1141 was deleted on the plasmid using the Stratagene Quikchange II Site-Directed Mutagenesis Kit with primers rscS_del1F and rscS_del1R. The resulting plasmid, pMJM33, was sequenced with primers MJM-154F and MJM-306R to confirm the single base pair deletion.
Squid colonization
Hatchling E. scolopes were colonized by exposure to approximately 3 x 103 CFU/ml (ranging from 5.2 x 102 – 1.4 x 104 CFU/ml; as specified in figure legends) of each strain in a total volume of 40 ml of FSIO (filter-sterilized Instant Ocean) for 3 hours. Squid were then transferred to 100 ml of FSIO to stop the inoculation and then transferred to 40 ml FSIO for an additional 45 hours with a water change at 24 hours post inoculation. For Figure 10A, kanamycin was added to the FSIO to keep selective pressure on the plasmid. After 48 hours of colonization, the squid were euthanized and surface sterilized by storage at −80 °C, according to standard practices (51). For determination of CFU per light organ, hatchlings were thawed, homogenized, and 50 μl of homogenate dilutions was plated onto LBS plates. Bacterial colonies from each plate were counted and recorded. Mock treated, uncolonized hatchlings (“apo-symbiotic”) were used to determine the limit of detection in the assay. The competitive index (CI) was calculated from the relative CFU of each sample in the output (light organ) versus the input (inoculum) as follows:
Log10 ((Test strain[light organ] / Control strain[light organ]) / (Test strain[inoculum] / Control strain[inoculum])). For competitions of natural isolates, the Group A strain (or its ΔrscS derivative) was the test strain and the Group B strain was the control strain. Colony color was used to enumerate colonies from each--white for Group A strains MB11B1 and ES213; yellow for Group B strains ES114 and MB15A4--along with PCR verification of selected colonies. For competition between SR5 and SR5 ΔbinK, 100 colonies per squid were patched onto LBS-Erm5 and LBS.
Colony biofilm assays
Bacterial strains were grown in LBS media (Fig. 10C) or LBS-Cam2.5 media (Figs. 2, 8) for approximately 17 hours, then 10 μl (Fig. 2) or 8 μl (Fig. 8, 10C) was spotted onto LBS plates (Fig. 10C) or LBS-Tet5 plates (Figs. 2, 8). Spots were allowed to dry and the plates incubated at 25 °C for 48 hours. Images of the spots were taken at 24 and 48 h post-spotting using a Leica M60 microscope and Leica DFC295 camera. After 48 h of growth, the spots were disrupted using a flat toothpick and imaged similarly.
Analysis of DNA and protein sequences in silico
Amino acid sequences for V. fischeri ES114 syp genes were obtained from RefSeq accession NC_006841.2. Local TBLASTN queries were performed for each protein against nucleotide databases for the following strains, each of which were derived from the RefSeq cds_from_genomic.fna file: V. fischeri SR5 (GCA_000241785.1), V. fischeri MB11B1 (GCA_001640385.1) and V. vulnificus ATCC27562 (GCA_002224265.1). Percent amino acid identity was calculated as the identity in the BLAST query divided by the length of the amino acid sequence in ES114. Domain information is from the PFAM database (52).
FUNDING INFORMATION
This work was funded by NIH grants R35GM119627 (to M.J.M.) and R21AI117262 (M.J.M.) and NSF grant IOS-1757297 (M.J.M.). Support for trainees was provided on NIGMS grants T32GM008061 (J.F.B.) and T32GM008349 (K.M.B.). This work was funded by the Chicago Biomedical Consortium with support from the Searle Funds at The Chicago Community Trust (supporting E.R.R.).
ACKNOWLEDGMENTS
The authors thank Elizabeth Bacon, Jacklyn Duple, Cheeneng Moua, Lynn Naughton, Olivia Sauls, and Denise Tarnowski for assistance with experiments.