Global discovery of colonization determinants in the squid symbiont Vibrio fischeri

Mutagenesis of V. fischeri and INSeq.

To identify colonization determinants, it was first necessary to develop tools to globally analyze mutant pool composition in V. fischeri. The Bacteroides thetaiotaomicron mutagenesis vector pSAM was modified so that its resistance cassette and transposase promoter functioned in V. fischeri (10, 12). The resulting vector, V. fischeri mariner transposon delivery vector (pMarVF1), can be conjugated from Escherichia coli into V. fischeri, generating stable erythromycin-resistant insertions (Fig. S1). With this approach, we generated a library of over 41,000 independent erythromycin-resistant insertions (Tnerm) in MJM1100. This is the “input” library for our analysis (Fig. 1A). Using the INSeq high-throughput method, the transposon–chromosome junctions were subsequently captured, subjected to limited amplification, and sequenced (10, 12) (SI Materials and Methods and Table S2), facilitating the identification of over 41,000 independent genomic insertion sites with high concordance across 10 technical replicates (Fig. 1B and Fig. S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Use of insertion sequencing to identify putative essential genes in Vibrio fischeri. (A) Overall approach. Dense transposon mutagenesis was performed in V. fischeri. This “input library” was then passaged, either through colonization in the hatchling squid during a 3-h inoculation and 48-h total colonization time or for an equivalent number of generations (n = 15) in LBS medium in vitro. Identification of genes that were missing in the input library and comparison of the output libraries to the input library enabled categorization of V. fischeri genes as putative colonization factors, as putative growth factors (i.e., required for normal medium growth), or as essential genes. (B) The mariner transposon developed for this purpose inserted randomly across the V. fischeri chromosomes and plasmid. Normalized transposon counts per 10,000-bp bin are plotted from the input library. (C) Data points represent individual V. fischeri ES114 genes, with the normalized transposon counts per million (cpm) and transposon sites (TA dinucleotides) in the 5′-most 90% of each gene. Orange circles are predicted nonessential genes and genes for which there is not sufficient information. Blue circles are predicted essential genes that lack an E. coli ortholog (n = 244). Plus signs are genes for which the E. coli ortholog is essential and the V. fischeri ortholog is predicted to be essential (n = 211). Solid black circles represent genes for which the E. coli ortholog is essential and the V. fischeri ortholog is predicted to be nonessential (n = 19).

Essential and Nonessential Genes in Culture.

The mutagenesis approach affords the opportunity to identify potentially essential genes as those with reduced transposon insertion in the input mutant population. We quantified the transposon abundance (i.e., Illumina read counts) from the 5′-most 90% of each gene, and normalized those counts to the respective sample in a counts-per-million (cpm) measure. Genes that exhibited a low cpm (normalized to the size of the gene) represented likely essential genes and operons (including nonessential genes that exhibit polar effects on the downstream essential genes). To identify these loci, the transposon counts were plotted against the size of each respective gene. Mariner transposons insert at TA dinucleotides, so for gene size we used the count of such dinucleotides in the gene. We observed three clusters of genes (Fig. 1C): two clusters of likely essential genes included those with no transposon hits and those with a low number of normalized hits (0.03–4.5, depending on gene size). The third cluster exhibited significantly higher levels of transposon counts, and the genes in this cluster were therefore labeled as putative nonessential genes. Genes with fewer than nine TA dinucleotides were too small to reliably determine essentiality.

Onto the plot in Fig. 1C we mapped the V. fischeri genes that were orthologs of essential genes in E. coli MG1655, the organism for which the best curated list of essential genes exists (13). Most E. coli essential genes mapped to the clusters with no or relatively few transposon counts in V. fischeri (175 of the 230 orthologs encoded in V. fischeri), supporting their assignment as putative essential genes in V. fischeri. For the V. fischeri genes that are orthologs of essential E. coli genes, 55 fell in the higher (putative nonessential) gene cluster. For 36 of these genes, 85–100% of hits fell in the 3′ half of the gene, suggesting that the mutant allele still encodes a (partially) functional truncated protein. Therefore, we classified these genes as “putative essential” in V. fischeri. The remaining 19 genes had a more even distribution of hits across each gene, arguing that the gene is not essential in V. fischeri. The 19 essential E. coli genes that we predict to have nonessential V. fischeri orthologs include multiple genes whose products function as part of the same complex or biochemical pathway, supporting this assignment. Independent analysis using published software further confirmed their assignment as nonessential, as described in SI Materials and Methods. The genes that are predicted to be nonessential in V. fischeri include genes that have been demonstrated to be nonessential in the related species Vibrio cholerae: genes for riboflavin biosynthesis (ribABDE), phospholipid biosynthesis (plsB), other metabolic functions (asd, can, cca), cell division (mukBEF), and secretion (secDF) (Table S3). Mutants in multiple V. fischeri muk genes were identified in a previous study, and in the VFS (Vibrio fischeri sublibrary) collection 9 of the 19 predicted nonessential genes harbor Tn5 transposon insertions (Table S3) (8). Included in the list of essential E. coli genes that are not essential in V. fischeri—and that are essential in V. cholerae or for which there exist conflicting data in different strains or studies—are multiple genes involved in cell division (ftsK, mrdB/rodA, zipA), LPS biosynthesis (kdsA, kdsB), and ubiquinone biosynthesis (ubiA) (Table S3). Comparative analysis of the essential genes therefore suggests conserved and unique features of Vibrio and V. fischeri biology for these factors. Importantly for this study, these results highlight that the INSeq analysis identifies relevant signal within a mutant library, and we therefore proceeded to examine genes that were depleted following squid colonization.

Identification of Squid Colonization Factors.

For experimental inoculation of E. scolopes bacterial concentrations of 10³–10⁶ cfu/mL are typically used (14, 15). It was demonstrated that higher inoculum concentrations led to a greater diversity of strains within the light organ, so we used an inoculum of 2 × 10⁵ cfu/mL (16, 17). To ensure that sufficient strain diversity could be achieved in the light organ, we colonized squid with a defined 96-mutant library and then analyzed the individual animal output samples by INSeq. Each animal contained over 80 of the initial 96 mutants (Fig. S2A), revealing that even with some initial bottleneck we can obtain a large number of independent mutant strains in an individual animal. We estimated that with 250 animals, we could conservatively sample 20,000 mutants (250 animals × 80 mutants minimum per animal). We therefore sought to analyze the distribution of the input mutant library following colonization in six replicates of 250 animals each, which would give average coverage of at least 35-fold per nonessential gene. In many systems, the natural route of inoculation is bypassed to investigate colonization in vivo, in contrast to inoculation of E. scolopes where bacteria are introduced to the seawater and then acquired by the host’s natural mechanisms. We observed some noise in the squid output libraries (Figs. S2B and S3), which likely reflects some variation in the extent to which the host bottleneck influenced symbiont diversity across the large number of animals in our analysis.

To avoid the study of mutants that were simply defective for growth, the input library was passaged in LB salt (LBS)-rich medium and the 115 genes that exhibited >10-fold mutant depletion after 15 generations were excluded from further analysis (Fig. 2A). We proceeded to use the dynamics of known colonization mutants to train our analysis to identify true, novel colonization factors. First, we observed that in most cases, mutants in known colonization factors were observed in every squid output library, and they were diminished in each library with respect to their relative abundance in the input library (Fig. 2C and Fig. S3). In contrast, 967 genes were represented by mutants that were not sampled in at least one squid output library, and because known squid colonization factors were not highly represented in this dataset, we suspect that some of these factors are artifacts (i.e., seawater growth defective; or low representation in the input library and therefore dramatic sampling bias in the squid). For subsequent analysis, we therefore took a conservative approach and focused on the 2,291 genes for which mutants were sampled in all squid libraries, and for which representation in each output library diminished by at least 1.8-fold (Fig. 2B). This set of 380 squid-depleted mutants captured over 75% of known mutants during the first two of the three defined stages of squid colonization (i.e., initiation and accommodation; Fig. 2C and Fig. S3), validating the analytical criteria applied. The INSeq squid-depleted mutants identified only 20% of known persistence-stage colonization factors (Fig. 2C), and finer-scale analysis of these factors revealed significant variation across the different replicates (Fig. S3), which in most cases excluded them from our selection criteria.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

INSeq analysis identifies squid colonization factor candidates. To identify putative colonization factors, the dynamics of the 41,897 mariner transposon library was followed during squid colonization. (A) Flowchart of the manner by which V. fischeri genes were categorized during the analysis. The 155 rRNA and tRNA genes were removed before the analysis shown. Genes that were classified as essential or that grew poorly in LBS medium (depleted by >1 log over 15 generations) were excluded from analysis in the squid host. The remaining 3,258 genes included 2,291 genes with mutants that were sampled in all squid output libraries. This conservative filter ensured that the genes assayed were abundant enough to colonize squid in the experimental conditions provided. (B) These genes were plotted by their best colonizing replicate (i.e., highest count for the mutant), and 380 genes had transposon counts that were depleted in all squid output libraries by at least 1.8-fold. This set of 380 genes represents the colonization factor candidates presented in this report. (C) For each category of colonization mutant, the percentage of the genes that were identified in the screen is shown by the black bars. The gray bars represent the percentage of genes that did not display depletion, and the remaining percentage indicates genes that were not assayed fully in squid (e.g., in vitro growth defect). Additional detail is provide in Fig. S3. (D) COG categorization of depleted genes. COG categories that are present in V. fischeri are shown, and plotted for each, the number of essential genes (blue) and the number of significantly squid-depleted genes (red).

To understand global patterns in the dataset, we categorized genes by the COG (clusters of orthologous groups) classification. First, we examined the V. fischeri essential genes. The categories that were most represented included translation, cell envelope, coenzyme transport and metabolism, lipid transport and metabolism, and replication (Fig. 2D). These assignments track with essential cellular processes, validating the use of COG categories as a tool to understand mutant dynamics in V. fischeri. We proceeded to categorize the squid-depleted genes in a comparable manner. In contrast to the overall essential genes, the colonization factors exhibited a markedly different pattern (Fig. 2D). Over 20% were not included in known groups (i.e., they were referred to as R or S). Categories that included a large number of the colonization factors included those for signal transduction, motility, amino acid transport and metabolism, inorganic ion transport and metabolism, and cell envelope integrity (Fig. 2D). The COG analysis highlighted broad-scale classes of genes that are required for squid colonization. We then proceeded to examine individual factors both in vivo and for specific behaviors in vitro.

Validation of Individual Candidates in 1:1 Competition Assays.

The INSeq analysis involved competition of over 41,000 mutants at high inoculum concentrations. To identify colonization factors that are required across a broad range of inoculation concentrations and conditions, we examined 20 factors predicted by INSeq to have a fitness deficit in the animal. We introduced each mutant in competition with a marked wild-type strain at a 1:1 ratio and then examined the relative ratio of the mutant following colonization. Nine of the strains exhibited fitness deficits in the 1:1 competition assay relative to wild type, ranging from 2- to 65-fold depletion in the host (Fig. 3A and Fig. S4). These nine strains include mutants in genes encoding two copper efflux systems (CusC and CopA); two protein quality control factors (cytoplasmic chaperone DnaJ and periplasmic endoprotease DegS); an inner membrane protein predicted to assist in the secretion of autotransporters (TamB/YtfN); a predicted lysine 2,3-aminomutase (YjeK/EpmB); and three poorly characterized proteins (YdhC, YafD, and YhcB) (Fig. 3A). We conducted the same competitions under squid-free culture conditions to control for the host environment (Fig. 3B); two of the nine mutants (ydhC and yjeK/epmB) exhibited competitive defects in culture that fully explained their respective squid colonization deficiencies, arguing that these genes encode factors that are not specifically required for colonization of the host environment.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Validation of host colonization factors identified by INSeq analysis. (A) Competition (1:1) of the indicated mutant strain versus a LacZ-marked wild-type strain in squid. Hatchling squid were inoculated with 2 × 10³ cfu/mL bacteria, washed at 3 and 24 h, and assayed at 48 h. The competitive index (CI) is the Log₁₀(mutant/wild type), where the mutant/wild-type output ratio is normalized to the input ratio as detailed in SI Materials and Methods. Dots represent the CI for individual squid, and bars represent the median. The value below the strain represents the fold depletion of the strain (e.g., a median CI of −1 would equal a 10-fold depletion). (B) Competition (1:1) of the indicated mutant strain versus a LacZ-marked wild-type strain in culture. Log phase cells were mixed, diluted, and grown for 15 generations in LBS medium. Dots represent the CI for three independent replicates, and bars represent the median. The competitive fold depletion was normalized to the in vitro competition as described in the SI Materials and Methods, and the normalized in vivo depletion is shown in the gray shaded box. (C) Wrinkled colony biofilm analysis at 24 h and 48 h after plating the indicated strains containing the biofilm-inducing pKG11 plasmid on LBS–tetracycline. (D) Single-strain growth curves in LBS medium. Strains were grown in a microplate for 30 h. The gray filled curve is shown similarly in all plots and represents the wild-type growth. Mutant growth curves are shown with the black lines.

V. fischeri must synthesize a symbiotic biofilm to passage through the squid ciliated epithelial field to the light organ pore. Biofilm assembly in vivo correlates strongly with the ability to form wrinkled colonies upon overexpression of the histidine kinase RscS (18). We therefore queried the colonization-defective mutants for their ability to express the wrinkled colony phenotype, and the three mutants that exhibited the most severe colonization defects (normalized to the in vitro results) also exhibited defects in the biofilm assay. The degS and yhcB mutants exhibited modest defects in the wrinkled colony biofilm assay, and dnaJ, which encodes the cytoplasmic cochaperone for DnaK, exhibited no wrinkling in the assay (Fig. 3C). None of the three mutants exhibited a growth defect in culture (Fig. 3D), arguing that the wrinkled colony assay reflected a defect in biofilm development and not the lack of growth in the assay. The other mutants did not exhibit any defect in growth or biofilm formation (except the yjeK/epmB mutant; consistent with its competitive growth defect, it also exhibited a growth defect without competition, and the wrinkled colony pattern was less complex). INSeq analysis has thus suggested new pathways to explore and new regulators of biofilm formation, a behavior that is known to be critical for symbiotic development and host colonization specificity (15, 18).

DnaJ/DnaK Regulate Symbiotic Biofilm Development.

Elaboration of the Syp symbiotic biofilm is critical for squid colonization (15, 18). We therefore sought to further examine the phenotype of dnaJ, which exhibited no wrinkled colony formation in the Syp biofilm assay (Fig. 3B). We constructed a dnaJ deletion and it exhibited the same biofilm defect; complementation with the wild-type allele restored full wrinkled colony expression (Fig. 4A). DnaJ (Hsp40) is conserved from bacteria through metazoans as a cochaperone for the ATP-dependent DnaK (Hsp70) chaperone (19). We therefore asked whether DnaK was similarly involved in the biofilm phenotype. We observed that a dnaK::Tnerm mutant exhibited a significant defect in wrinkled colony formation, and that the phenotype was complemented by dnaK⁺ in trans (Fig. 4A). The dnaK::Tnerm strain exhibited a growth defect, as did dnaK mutants in the INSeq dataset (Fig. S5 and Dataset S1). However, DnaJ and DnaK were similarly required for heat tolerance, arguing that it is reasonable to model the DnaJ/DnaK chaperone machine in V. fischeri on mechanisms proposed from other organisms (19). We therefore proceeded to probe the role of the DnaJ/DnaK system using the ΔdnaJ allele.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The DnaJ/DnaK chaperone system regulates Syp biofilm formation. (A) Biofilm formation, as assayed by wrinkled colony formation, is not observed in dnaJ or dnaK mutant strains but is complemented when the interrupted gene is provided in trans. (B) Single-strain colonization of E. scolopes squid by each of the indicated strains; Apo, aposymbiotic. Hatchling squid were inoculated with 2 × 10³ cfu/mL bacteria, washed at 3 h and 24 h, and assayed at 48 h. (C) Aggregate formation in the squid ciliated epithelial field by strains of the indicated genotype. Hatchling squid were inoculated with 2 × 10⁶ cfu/mL bacteria that constitutively express GFP from the pVSV102 plasmid and assayed at 3 h by epifluorescence microscopy. Large aggregates are readily apparent in the wild-type colonized animals (arrow) but only isolated cells are observed in the isogenic ΔdnaJ mutant (arrow). (Scale bars, 50 μm.) (D) Syp polysaccharide immunoblot with biofilm-specific antibodies. (E) QRT-PCR demonstrating the relative abundance for rscS and sypG transcripts in the indicated strains.

Based on the severe biofilm defect observed in culture, we predicted that the ΔdnaJ mutant would exhibit a colonization defect even in the absence of competition. We inoculated squid hatchlings with the mutant strain for 3 h and observed at 24 h that the ΔdnaJ mutant failed to colonize the light organ (Fig. 4B). The complemented strain was indistinguishable from wild type. This result is similar to that observed for other biofilm mutants, supporting a role for DnaJ during biofilm formation in vivo. We proceeded to test directly whether DnaJ influences biofilm development in the host mucociliary field. We colonized squid hatchlings with GFP-labeled bacteria for 3 h and then imaged the dissected light organ by epifluorescence microscopy. In contrast to the wild-type strain, which exhibited large aggregates of ∼20 μm diameter, the ΔdnaJ mutant was present only as single cells (Fig. 4C). These data support a role for DnaJ in regulating biofilm formation in vivo during the initiation stage of squid colonization.

We next asked whether DnaJ regulates polysaccharide production or polysaccharide secretion. The Syp exopolysaccharide (EPS) is readily detected with an anti-EPS antibody, and upon Western blot analysis a strong signal was observed for EPS produced, regardless of whether it is secreted or retained in the cell (20). The Syp polysaccharide that is synthesized in an RscS-dependent manner is not observed in a strain harboring the ΔdnaJ allele (Fig. 4D). The phenotype resembled that of a strain with an interruption in the SypQ glycosyltransferase, which is required for polysaccharide production (20). The canonical pathway that activates the Syp biofilm relies on a transcriptional regulatory cascade to activate four σ⁵⁴-dependent promoters at the syp locus (18). In a biofilm-induced background, we observed high levels of transcription at sypG, even in the absence of DnaJ (Fig. 4E). Together, these results suggest that DnaJ acts at a step downstream of syp transcription to allow polysaccharide production.