Cholera toxin promotes pathogen acquisition of host-derived nutrients

Abstract

Vibrio cholerae is the causative agent of cholera, a potentially lethal enteric bacterial infection¹. Cholera toxin (CT) is required for V. cholerae to cause severe disease and is also thought to promote transmission of the organism in that victims can shed many liters of diarrheal fluid that typically contains in excess of 10¹¹ organisms per liter. How the pathogen is able to reach such high concentrations in the intestine during infection remains poorly understood. Here we show that CT-mediated disease enhances pathogen growth and induces a distinct V. cholerae transcriptome signature that is indicative an iron-depleted gut niche. During infection, bacterial pathogens need to acquire iron, a nutrient essential for growth². The majority of iron in the mammalian host resides in a chelated form within the porphyrin structure of heme, and V. cholerae genetically encodes the ability to utilize heme as a source of iron³. We show that V. cholerae heme and vibriobactin utilization genes confer a growth advantage to the pathogen only when CT is produced. Furthermore, CT-induced capillary congestion pathology in the terminal ileum correlated with an increased bioavailability of luminal heme. CT-induced disease in the ileum also led to increased luminal concentrations of long-chain fatty acids (LCFAs) and L-lactate metabolites, as well as upregulation of V. cholerae iron-sulfur cluster-containing TCA cycle enzyme genes. Genetic analysis of V. cholerae suggested that heme and LCFA uptake-dependent growth of V. cholerae occurs during infection but only in a strain capable of producing CT in vivo. We conclude that CT-induced disease creates an iron-depleted metabolic niche in the gut that selectively promotes the explosive growth of this pathogen through acquisition of host-derived heme and fatty acids as nutrients.

The enterotoxin activity of CT from supernatants of V. cholerae cultures was first reported almost 60 years ago⁴. The catalytic activity of CT activates adenylyl cyclase, which increases cellular levels of 3’,5’-cyclic AMP (cAMP) leading to changes in numerous metabolic functions regulated by the cAMP-dependent protein kinase PKA⁵. The massive secretory diarrhea caused by CT is thought to occur when PKA phosphorylates and activates ion channels that ultimate lead to electrolyte imbalances in the intestinal lumen⁶. Elevation of cAMP levels is also able to include lipolysis and indeed CT can cause massive breakdown of lipids in adipose cells⁷. Recent studies have indicated that some virulence factors can promote pathogen acquisition of nutrients in the host gut by generation of inflammation-associated metabolites⁸. Although CT is thought to confer a fitness advantage to V. cholerae by increasing the dissemination of the organism during outbreaks, we wondered whether CT stimulates intestinal growth of V. cholerae by affecting host metabolism and nutrient acquisition by this pathogen.

To test this hypothesis, we first asked whether CT can enhance the growth of V. cholerae during infection of 3-day old neonatal (infant) rabbits because these experimental animals show CT-dependent disease symptoms similar to what is observed during human infection⁹. Rabbit pups were infected with either wild type (WT) V. cholerae C6706 or an isogenic in-frame ΔctxAB mutant (FRC38) which is unable to produce either the A or B subunits of CT (from here forward referred to as Δctx) (Extended Data Fig. 1a). As expected, pups infected with the WT strain but not the Δctx mutant developed severe diarrheal disease one day after infection (Fig. 1a). Consistent with previous observations, no difference in bacterial concentrations (CFU/g) in the tissue of the ileum (small intestine) or downstream in the fluid of the cecum were observed between infant rabbits infected with either the WT or Δctx mutant (Fig. 1b and Fig. 1c). However, standard CFU normalization by weight is misleading because the cholera-like disease caused by the WT strain led to massive intestinal fluid accumulation and loss (Fig. 1a). When we corrected for the volume of fluid present in the lumen of the intestinal cecum, we found that the WT V. cholerae reached ~22-fold higher levels than the Δctx mutant in the gut lumen (Fig. 1d). This growth yield difference was not observed when WT was competed with the Δctx mutant in mixed infections (Extended Data Fig. 1b), suggesting that CT-enhanced colonization depends on secreted CT and its effect on the host intestinal environment.

Figure 1. CT induces a unique Vibrio cholerae transcriptome signature during infection.

(a-d) Groups of 3-day old New Zealand white rabbits (N = 4) were mock-infected or infected intragastrically with 1 × 10⁹ CFU per animal with V. cholerae C6706 wild type (wt) or the cholera toxin mutant (Δctx) and samples were collected one day after infection. (a) The fluid accumulation (FA) ratio in the cecum is expressed as percent of the response observed in animals infected by the wild type strain. The FA ratio of the wild type was compared to the Δctx mutant or mock using one-way ANOVA (F (2, 9) = 52.534, p < 0.0001) and post-hoc Sidak's multiple comparisons test. Bars represent percent fluid accumulation relative to wild type ± SE. The CFU/g was determined in the ileum tissue (b), and cecal fluid (c). The total CFU of V. cholerae wild type (black circles) or Δctx mutant (red circles) in the cecal fluid was determined by normalizing the CFU/g by the final weight of the fluid collected from the cecum (d). Error bars represent geometric means ± SD. Bacterial numbers were compared by unpaired two-sided student t-test. (e-h) RNA-seq analysis of Vibrio cholerae during colonization of the ileum and cecum of infant rabbits (e) PCA plots for RNA-seq data from two biological replicates of V. cholerae wild type (black circles) or Δctx mutant (red circles) collected from the ileum or cecum of infant rabbits (N = 2). (f) Venn diagram of genes significantly upregulated in the wild type relative to the Δctx mutant in the ileum (purple) and cecum (yellow). (g) Percent abundance of raw RNA-seq reads aligned to chromosome I (Chr. I; purple) or chromosome II (Chr. II; yellow) from wild type or the Δctx mutant in the ileum. (h) Percent abundance of heme utilization genes (VCA0907-VCA0915, VCA0576; red) from raw RNA-seq reads aligned to chromosome II (Chr. II; yellow) from wild type or the Δctx mutant in the ileum. ns, not statistically significantly different.

Previously, RNA-seq transcriptome analysis has been used to correlate changes in V. cholerae gene expression with the presence of metabolites in the gut during infection of neonatal rabbits¹⁰. To understand how CT enhances the replication of this organism within the intestine, strand-specific Illumina-based RNA-seq analysis¹¹ was performed on two samples each of WT and the Δctx mutant bacteria harvested from either the rabbit ileum or the cecum one-day post-intragastric infection (Fig. 1e-h). We observed a strong correlation between RNA-seq biological replicates in both the WT and Δctx mutant samples obtained (Extended Data Fig. 2a-d). Principle coordinate analysis (PCA) revealed that there was a marked difference in gene expression in WT bacteria harvested from the ileum compared to the adjacent cecum (Extended Data Fig. 3a). Although most open reading frames (ORFs) were not differentially expressed, we found 128 genes that were significantly up-regulated in WT V. cholerae colonizing the ileum relative to the cecum (Supplementary Table 1) and that these included many virulence genes in the ToxT regulon (Extended Data Fig. 3b). Notably, the ToxT-regulated genes encoding CT (ctxA and ctxB) were the most highly expressed ORFs in WT bacteria recovered from the ileum.

We next compared gene expression in the WT and Δctx mutant and found a V. cholerae transcriptome signature that correlated with the capacity to produce CT regardless of anatomical location (Fig. 1e). We found 243 genes that were significantly up-regulated in the WT compared to the Δctx mutant and these included 101 genes in ileum samples, 118 in the cecum samples, and 24 in both samples (Fig. 1f, Supplementary Table 2). Interestingly, although most ORFs were not differentially expressed in the WT compared to the Δctx mutant, virulence factors including many encoded by ToxT-regulated genes were up-regulated in the WT cells compared to the Δctx mutant (Extended Data Fig. 4a and 4b), suggesting that CT induced changes in the gut environment may provide a positive signal for ToxT-regulated virulence gene expression.

The V. cholerae genome consists two circular chromosomes: a large (2,961,146 bp) chromosome (Chr. I), and a smaller (1,072,314 bp) chromosome (Chr. II)¹². Surprisingly, we observed a 16-fold induction in transcription originating from Chr. II in the WT cells relative to the Δctx mutant during colonization of the ileum (Fig. 1g), and 4-fold induction in the cecum (Extended Data Fig. 4c and 4d). Remarkably, we found that expression of just 10 genes involved in heme utilization accounted for almost 14% of the entire Chr. II transcriptome during colonization of the ileum and that these genes were up-regulated nearly 20-fold in WT compared to the Δctx mutant (Fig. 1h). Further analysis revealed that virtually all genes involved in heme uptake and vibriobactin-mediated iron acquisition that were located on either Chr. I and Chr. II were highly up-regulated in WT relative to the Δctx mutant and this induction occurred most prominently in the ileum relative to the cecum (Fig. 2a, Extended Data Fig. 4e, and, Supplementary Table 3).

Figure 2. Heme utilization confers a CT-dependent fitness advantage to Vibrio cholerae.

(a) RNA-seq analysis of Vibrio cholerae heme utilization genes (VCA0907-VCA0915, VCA0576) and vibriobactin genes (VC0200-VC0201, VC2209-VC2211, VC0771-VC0780, VC0475, VCA0227-VCA0230) in the wild type relative to the cholera toxin mutant (Δctx) in the ileum or cecum. (b) Luminal iron concentrations from the ileum of infant rabbits (N = 4) that were mock-infected or infected with the wild type (wt) or the Δctx mutant. The iron concentrations between type (wt) and the Δctx mutant were compared by unpaired two-sided student t-test. Red dot represents sample from animal containing statistical outlier for V. cholerae cholera toxin (ctxA) expression. Error bars represent means ± SD. (c) Luminal hemin measurements from the ileum and cecum of infant rabbits (N = 4) that were mock-infected or infected with the wild type (wt) or the Δctx mutant. The wt heme levels in the ileum or cecum were compared to the Δctx mutant or mock using one-way ANOVA (Ileum: F (2, 9) = 55.88, p < 0.0001; Cecum: F (2, 9) = 16.61, p = 0.0010) followed by post-hoc Sidak's multiple comparisons test. The heme levels in the ileum and cecum between groups were compared using two-way ANOVA followed by post-hoc Sidak's multiple comparisons test (see Supplementary Table 7). Black lines show median for individual animals (black circles). (d) Groups of CD-1 mice were infected with a 1:1 mixture of the indicated V. cholerae strain mixtures (N = 8 for wild type vs hutA viuA; N = 4 for Δctx vs Δctx hutA viuA; N = 5 for Δctx vs Δctx hutA viuA + CT). Some mice received an oral dose of purified CT mixed into the inoculum. The competitive index (CI; the ratio of strains recovered) was determined one day after infection. Dotted lines connect strains recovered from the same animal. The CI was compared using one-way ANOVA (F (2, 14) = 7.634, p = 0.0057) followed by post-hoc Sidak's multiple comparisons test. Different colored lines distinguish animals from different litters. ns, not statistically significantly different.

When iron is a limiting nutrient, V. cholerae is capable of expressing multiple iron acquisition systems that include receptors and TonB-dependent transporters of heme as well as the siderophore vibriobactin, its receptor, and transport systems¹³, a process that is controlled by the Ferric Uptake Regulator (Fur)¹⁴. We found that almost all Fur-controlled genes known to be induced under low-iron growth conditions were up-regulated in WT relative to the Δctx mutant in both the ileum and in the cecum, with a marked up-regulation in the ileum relative to the cecum (Extended Data Fig. 5, Supplementary Table 3). Furthermore, 18 of the 24 genes up-regulated in both the ileum and cecum in WT cells are known to be induced under low iron conditions (Fig. 1f, Supplementary Table 2). We next memeasured the concentration of iron in the ileum of infant rabbits by triple quadrupole ICP-MS (ICP-QQQ-MS). Remarkably, iron levels were significantly reduced in the ileum samples collected from WT infected animals compared to those infected with the Δctx mutant (Fig. 2b). We wondered if CT was inducing local intestinal inflammation which could deplete available iron via host iron-binding proteins such as neutrophil lactorferrin¹⁵. However, we found that a stable marker of inflammation, the iron-sequestration protein Lipocalin-2¹⁶ was not elevated in the gut of infant rabbits infected with the WT relative to the Δctx mutant (Extended Data Fig. 5d).

The presence of free heme has previously been reported in the intestinal lumen of neonatal rabbits infected with V. cholerae⁹. Remarkably, rabbits infected with WT V. cholerae exhibited elevated concentrations of heme (measured as hemin) in luminal samples from the ileum and the cecum relative to animals infected with the Δctx mutant (Fig. 2c). Higher concentrations of hemin were detected in the ileum relative to the cecum for WT infected rabbits and these data are consistent with our RNA-seq data that showed higher expression of heme utilization genes in ileum relative to the cecum (Fig. 2a, Supplementary Table 3). Collectively, these data indicate that infection with V. cholerae leads to a CT-mediated increase of heme in the gut lumen.

Infection with V. cholerae leads to capillary congestion of red blood cells in the ileum of the small intestine^4,17. V. cholerae studies in neonatal rabbits have demonstrated that capillary congestion requires CT⁹. Neonatal (suckling) mice exhibit CT-dependent diarrheal disease, and are commonly used as a model for evaluating V. cholerae mutants for defects in intestinal colonization¹⁸. Interestingly, heme utilization genes have also been found to be upregulated in the infant mouse gut¹⁰. To investigate whether CT induces capillary congestion in mice, 4-day old suckling mice were infected with either WT or with the Δctx mutant, mock infected, or treated with purified CT. While no mice exhibited signs of inflammation (Extended Data Fig. 6a and 6b), infection with the WT or oral treatment with CT resulted in capillary congestion of red blood cells in the small intestine (Extended Data Fig. 6a), which correlated with CT-dependent fluid accumulation and elevated concentrations of heme (Extended Data Fig. 6c and 6d). Interestingly, elevated levels of heme were not detected in the ileum of wild type infected animals, suggesting that heme may be a limiting nutrient in the mouse gut (Extended Data Fig. 6d).

We next determined whether CT could also enhance growth of V. cholerae in the mouse gut. We observed significantly higher total bacterial numbers in the GI tracts of WT infected mice compared to Δctx mutant infected mice (Extended Data Fig. 6e). As we observed with neonatal rabbits (Fig. 1d), the WT strain was recovered in significantly higher numbers in the luminal contents of the small intestine and cecum relative the Δctx mutant, while no difference was observed for tissue-associated bacteria or at an early time-point (6 hr post-infection) (Extended Data Fig. 6f and 6g). We next investigated whether heme uptake was important for CT-enhanced colonization through mutant analysis. V. cholerae encodes three receptors for heme uptake and the receptor, HutA, has previously shown to be required for utilization of hemoglobin as an iron source¹⁹. Interestingly, hutA (VCA0576) was the most highly expressed Fur-regulated gene in WT V. cholerae relative to the Δctx mutant in both the ileum and cecum (Supplementary Table 3). Furthermore, genetic ablation of either hutA or viuA (the later encoding the vibriobactin receptor) leads to a modest growth defect in the gut, which could be complemented by expression of these genes on a low-copy plasmid (Extended Data Fig. 7a). A V. cholerae double mutant of hutA and viuA has a reported pronounced growth defect in both neonatal mice and rabbits^19,20. To investigate whether CT-induced disease is required for heme and vibriobactin uptake to confer a fitness advantage to V. cholerae, mice were infected with either a 1:1 mixture of the wild type and a hutA viuA mutant or with a 1:1 mixture of the Δctx mutant and a Δctx hutA viuA mutant. While the CT-producing V. cholerae exhibited a fitness advantage over the hutA viuA mutant in the gut, no fitness advantage was observed in mice inoculated with a 1:1 mixture of the Δctx mutant and a Δctx hutA viuA mutant (Fig. 2d). Additionally, oral administration of purified CT rescued the growth and the fitness advantage of the Δctx mutant over the Δctx hutA viuA mutant (Fig. 2d). Collectively, these data suggest that CT induces iron-depletion and luminal heme release in vivo (perhaps involving capillary congestion) and these host responses drive heme utilization and luminal growth of V. cholerae.

Iron is essential for bacterial metabolism. We found that metabolic pathways were the highest up-regulated genes in V. cholerae during colonization of the ileum relative to the cecum (Fig. 3a). Notably, genes involved in TCA cycle metabolism, L-Lactate utilization, and LCFA utilization were significantly up-regulated in the WT in the ileum relative to the Δctx mutant during infection (Fig. 3b-c and Extended Data Fig. 7b). Importantly, iron-sulfur cluster-containing TCA cycle enzymes encoding the succinate dehydrogenase (sdh operon), the iron-sulfur assembly protein cysteine desulfurase (VC0748), as well as other iron-binding proteins involved in metabolism, were highly expressed and up-regulated in the WT relative to the Δctx mutant during infection (Extended Data Fig. 7c, Supplementary Table 2). These data suggest that CT-induced disease enhances metabolism in V. cholerae, which in turn, might exacerbate iron depletion in the gut lumen through explosive bacterial growth and iron uptake.

Figure 3. Cholera toxin promotes Vibrio cholerae LCFA utilization during infection.

(a) DAVID bioinformatics pathway analysis of RNA-seq data of significantly upregulated pathways in the ileum relative to the cecum. (b) Differential expression of V. cholerae genes involved in TCA cycle (VC0432, VC1573, VC2084-VC2092, VC2738), LCFA utilization (VC1043, VC2758, VC2231), and L-Lactate utilization (VCA0983, VCA0984), and all open reading frames (ORFs) of wild type C6706 during colonization of the ileum relative to the cecum. (c) Differential expression of V. cholerae genes involved in TCA cycle, LCFA utilization, and L-Lactate utilization, and all ORFs in the wild type relative to the cholera toxin mutant (Δctx) during colonization of the ileum or cecum. (d) Luminal LCFA measurements (8 carbon chain or greater) from the ileum and cecum of infant rabbits (N = 4) that were mock-infected or infected with the wild type (wt) or the cholera toxin mutant (Δctx). Black lines show median for individual animals (black circles). The wt LCFA levels in the ileum or cecum were compared to the Δctx mutant or mock using one-way ANOVA (Ileum: F (2, 9) = 12.69, p = 0.0024; Cecum: F (2, 9) = 2.951, p = 0.1034) followed by post-hoc Sidak's multiple comparisons test. (e) Luminal L-lactate measurements from the ileum and cecum of infant rabbits (N = 4) that were mock-infected or infected with the wild type (wt) or the cholera toxin mutant (Δctx). Black lines show median for individual animals (black circles). The wt L-Lactate levels in the ileum or cecum were compared to the Δctx mutant or mock using one-way ANOVA (Ileum: F (2, 9) = 50.98, p < 0.0001; Cecum: F (2, 9) = 1.916, p = 0.2027) followed by post-hoc Sidak's multiple comparisons test. The L-Lactate levels in the ileum and cecum between groups were compared using two-way ANOVA (see Supplementary Table 7) followed by post-hoc Sidak's multiple comparisons test. (f) Groups of CD-1 mice were infected with a 1:1 mixture of the indicated V. cholerae strain mixtures (N = 8 for wild type vs fadL; N = 4 for Δctx vs Δctx fadL; N = 5 for Δctx vs Δctx fadL + CT). Some mice received an oral dose of purified CT mixed into the inoculum. The competitive index (CI; the ratio of strains recovered) was determined one day after infection. Dotted lines connect strains recovered from the same animal. The competitive indices between groups were compared using one-way ANOVA (F (2, 14) = 57.98, p < 0.0001) followed by post-hoc Sidak's multiple comparisons test. Different colored lines distinguish animals from different litters. ns, not statistically significantly different.

The presence of LCFAs has previously been reported in the intestinal lumen of neonatal rabbits infected with V. cholerae¹⁰. Elevation of cAMP by CT activates eukaryotic Protein Kinase A, which in turn activates hormone-sensitive lipase, leading to lipolysis in target cells²¹. Hydrolysis of triacylglycerol stores results in liberation of glycerol and LCFAs²². Elevation of cAMP by CT can also induce a switch in host cell metabolism, leading to reduced oxygen consumption and increased lactic acid production²³. Remarkably, WT infected rabbits had elevated concentrations of both LCFAs and L-Lactate in the lumen of the ileum relative to Δctx mutant infected animals (Fig. 3d and 3e) and elevated concentrations of LCFAs in the ileum in CT-treated mice (Extended Data Fig. 6h). Collectively, these data indicate that V. cholerae infection leads to a CT-dependent increase in LCFAs and L-Lactate in the small intestine lumen.

The uptake of exogenous LCFAs in bacteria requires the outer membrane transport protein, FadL²⁴. The V. cholerae genome encodes three fadL homologs¹². Our transcriptome analysis indicated that the fadL homolog, VC1043, was highly expressed in the ileum of infant rabbits and up-regulated in the WT relative to the Δctx mutant (Extended Data Fig. 7b and Supplementary Table 2). To investigate whether LCFA-uptake confers a fitness advantage to V. cholerae in the gut, we infected mice with a 1:1 mixture of WT and a fadL mutant (VC1043). The WT was recovered in significantly higher numbers from the small intestine (s.i) than the fadL mutant, suggesting that LCFA-uptake confers a fitness advantage to V. cholerae during infection (Fig. 3f). To determine whether CT-induced disease was required for LCFA-uptake mediated growth of V. cholerae, we infected mice with a 1:1 mixture of the Δctx mutant and Δctx fadL double mutant. In the absence of CT production, LCFA-uptake no longer conferred a fitness advantage to V. cholerae (Fig. 3f). Furthermore, oral administration of purified CT rescued the growth and the fitness advantage of the Δctx mutant over the Δctx fadL mutant (Fig. 3f). Interestingly, one WT V. cholerae sample collected from the rabbit ileum was found to have low expression of ctxA during infection and elevated concentrations of iron in the ileum relative to infected rabbits that had relatively high expression of ctxA (Extended Data Fig. 7d and 7e). The low expression of ctxA in this animal correlated with low expression of hutA, viuB, fadL, and lldD in the gut, similar to what was observed in a Δctx mutant infected animals.

We conclude that modulation of the host cell metabolism by CT produced by V. cholerae in vivo creates an iron-depleted intestinal niche that allows for enhanced growth of this organism through pathogen acquisition of host-derived nutrients including heme and LCFAs, and possibly L-lactate²⁵. This conclusion may be relevant to the emergence and fitness of V. cholerae as a human pathogen in that genes induced by iron depletion in vitro are highly expressed in bacteria present in the stools of cholera patients²⁶. Recent, highly successful variant strains of V. cholerae have also been reported to produce much more CT than earlier 7th pandemic strains²⁷. Our results provide a new paradigm for how a microbial toxin enhances the growth of a pathogen and promotes its own transmission by remodeling both host and microbial metabolism.

METHODS

Bacterial strains

V. cholerae and E. coli strains used in this study are listed in Supplementary Table 4. Unless indicated otherwise, bacteria were routinely grown aerobically at 37°C in LB broth or on LB plates. Antibiotics were added to the media at the following concentrations: 0.1 mg/ml streptomycin, 0.02 mg/ml chloramphenicol (E. coli), 0.005 mg/ml chloramphenicol (V. cholerae), and 0.1 mg/ml carbenicillin.

In vitro iron depletion

For in vitro iron depletion and heme supplementation, overnight cultures of V. cholerae were grown in LB broth with DMSO (control), LB broth supplemented with 300 μM 2,2′-dipyridyl, or with LB broth supplemented 300 μM 2,2′-dipyridyl and 5uM hemin and cells were harvested when the bacterial growth reached approximately an OD600 of 0.9.

Construction of plasmids

The plasmids and primers used in this study are listed in Supplementary Table 5 and 6, respectively. PCR products were cloned into vectors using Gibson Assembly (New England Biolabs, Ipswich, MA) and sequenced.

A suicide plasmid for generating a deletion of ctxA and ctxB was constructed by using primers del_A and del_B to PCR amplify a 557 bp 5′ flanking sequence that included 9 nucleotides of the ctxA coding region. Primers del_C and del_D were used to PCR amplify a 515 bp fragment that included 3 nucleotides of the 3′ coding region of ctxB. The two PCR fragments resulting in-frame ctxAB operon deletion were cloned into pDS132 using Gibson Assembly to generate plasmid pFRC1.

An internal fragment of the fadL gene was amplified by PCR using the primers listed in Supplementary Table 6. The fadL PCR product was then cloned into pGP704 using Gibson Assembly to create pFRC5. A fragment of the hutA gene was amplified by PCR and cloned into pGP704 using Gibson Assembly to create pFRC9. A fragment of the viuA gene was amplified by PCR and cloned into pEP185.2 using Gibson Assembly to create pFRC10.

For complementation of the hutA, viuA, and fadL mutants, the open reading frame including the promoter region was amplified using the primers listed in Supplementary Table 6. The resulting PCR fragment was cloned into pWSK29 using Gibson Assembly to create pFR11, pFR12, and pFR13, respectively.

Construction of V. cholerae mutants

Suicide plasmids were propagated in E. coli DH5α λpir and introduced into the V. cholerae strain El Tor strain C6706 by conjugation using E. coli S17-1 λpir as a donor strain.

The ctxAB operon was deleted by introduction of pFRC1 into V. cholerae C6706. Transconjugants were selected for on LB plates containing Strep and Cm. Sucrose counter-selection was performed as described previously²⁸. A strain that was sucrose-tolerant and Cm^s was verified by PCR to carry a deletion in the ctxAB operon and was denoted FRC38.

Exconjugants were selected on LB agar plates containing streptomycin and antibiotics selecting for integration of the suicide plasmid. Integration of the plasmids in the correct location on the chromosome was verified by PCR. Using this methodology, pFRC5, pFRC9, and pFRC10 were integrated into the fadL, hutA, and viuA genes of C6706 to yield strains FRC122, FRC127 and FRC128, respectively.

pFRC5 and pFRC10 were integrated into the fadL and viuA genes of FRC38 to yield strains FRC123 and FRC130, respectively.

pFRC9 was integrated into the hutA gene of strain FRC130 to yield FRC131.

Animal Experiments

All animal experiments in this study were approved by the Institutional Animal Care and Use Committee at Harvard Medical School.

Rabbit Experiments.

Infant rabbits were infected with V. cholerae as previously described⁹ with the following modifications. Three‐day‐old infant rabbits (N = 4) were treated with Zantac (2 mg/kg body weight) intraperitoneally and 3 hours later orally inoculated with 0.5 ml of 2.5% sodium bicarbonate (mock infection) or with 1 × 10⁹ colony forming units (CFU) of V. cholerae wild type or the ΔctxAB mutant in 0.5 ml of 2.5% sodium bicarbonate. One-day post-infection, rabbits were euthanized and samples from the gastrointestinal tract were collected for RNA purification and bacteriological analysis. Independent infant rabbit litters were used for infection with each strain. Rabbits from different litters are indicated under the “Litter” number in the source data files. Fig. 1, Fig. 2a-c, Fig. 3a-e, Extended Data Fig. 2, Extended Data Fig. 3, Extended Data Fig. 4, and Extended Data Fig. 5, Extended Data Fig. 7b-e, represent data from four independent rabbit litters (litter 1: wild type infected, litter 2: ΔctxAB mutant infected #1, litter 3: mock infected, litter 4: ΔctxAB mutant infected #2). Animals were randomly allocated in treatment groups. Group sizes (at least four animals per treatment group) represents the minimum number animals needed to reach statistical significance (p < 0.05) between experimental groups. The ΔctxAB mutant infection group was repeated in an independent litter of infant rabbits. Animals that were euthanized before an experimental endpoint due to health concerns were excluded from the analysis.

Mouse Experiments.

Suckling mice were infected with V. cholerae as previously described²⁹ with the following modifications. CD-1 mice were obtained from Charles River Laboratories (Wilmington, MA, USA). 4-day old CD-1 suckling mice separated from their mothers were randomly allocated in treatment groups and 2 hours later orally inoculated with 0.05 ml sterile LB broth (mock infection) or with 1 × 10⁷ CFU of V. cholerae in 0.05 ml LB broth. For competitive infections, mice were infected with an equal mixture of 5 × 10⁶ CFU of each V. cholerae strain in 0.05 ml LB broth. To determine the competitive index (CI), the ratio of recovered wild-type bacteria to mutant bacteria (output ratio) was divided by the ratio present in the inoculum (input ratio). Mice from different litters are indicated under the “Litter” number in the source data files. In Fig. 2d, a litter of mice (litter 1) were randomly allocated in treatment groups and infected with either the wild type vs. viuA hutA or ΔctxAB vs. ΔctxAB viuA hutA. The wild type vs. viuA hutA experiment was repeated in an independent litter (litter 3). In Fig. 3f, a litter of mice (litter 1) were randomly allocated in treatment groups and infected with either the wild type vs. fadL or ΔctxAB vs. ΔctxAB fadL. The wild type vs. fadL experiment was repeated in an independent litter (litter 3). In Extended Data Fig. 1 and Extended Data Fig. 6f (ΔctxAB + CT CFU data), mice from one litter (litter 6) were randomly allocated in treatment groups and infected with either the wild type vs. ΔctxAB or with ΔctxAB + 10 μl CT. In Extended Data Fig. 6b, mice from one litter (litter 1) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant and a second litter of mice (litter 2) was mock treated for qRT-PCR analysis. In Extended Data Fig. 6c, two independent litters (litter 3 and litter 4) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant. Mice from an independent litter (litter 5) were randomly allocated in treatment groups and mock treated or treated with with 10 μl CT. For the 6 hrs p.i data, mice from one litter (litter 8) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant. In Extended Data Fig. 6d, mice from one litter (litter 9) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant. Hemin data for mock or 10 μl CT treated mice was generated from same experiment (litter 5) as shown in Extended Data Fig. 6c. For Extended Data Fig. 6e, mice from one litter (litter 10) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant. In Extended Data Fig. 6f, CFU data shown was generated from litters shown in Extended Data Fig. 6c (matching litter numbers). Mice from an independent litter (litter 11) were randomly allocated in treatment groups and infected with either the wild type or ΔctxAB mutant to reproduce data in Data Fig. 6f. In Extended Data Fig. 6g, 6 hrs p.i CFU data was generated from same experiment (litter 8) as shown in Extended Data Fig. 6c. In Extended Data Fig. 6h, LCFA data was generated from same experiments (litter 5 and litter 9) as shown in Extended Data Fig. 6d. In Extended Data Fig. 7a, mice from one litter (litter 1) were randomly allocated in treatment groups and infected with either wild type vs. hutA or wild type vs. viuA. Mice from a second litter (litter 2) were randomly allocated in treatment groups and infected with either pWSK29::hutA vs. hutA(pWSK129) or pWSK29::viuA vs. viuA(pWSK129). Mice from a third litter (litter 3) were infected with pWSK29::fadL vs. fadL(pWSK129). In Extended Data Fig. 1b, Extended Data Fig. 6d, Extended Data Fig. 6e, Extended Data Fig. 6f (CFU/g tissue), and Extended Data Fig. 6g, neonates from one litter were randomly assigned for infection with the wild type or the ΔctxAB mutant and the individual neonate is the experimental unit “N” for statistical analysis (within-litter design). For treatment with purified cholera toxin (CT), mice were treated by oral gavage with 10 μg of purified CT (List Biological Laboratories, Campbell, CA) in 0.05 ml LB broth or by mixing 10 μg of purified CT with 0.05 ml containing the indicated V. cholerae inoculum. After euthanasia, gastrointestinal tracts were collected approximately 22 hours post-infection for histopathological and bacteriological analysis. The ileum was defined as the last 4 cm of the distal small intestine. The fluid accumulation ratio (FA) ratio in mice was determined by weight of entire gastrointestinal tract (gut) / (mouse body weight - gut weight), as previously described³⁰. Group sizes (at least four animals per treatment group) represents the minimum number animals needed to reach statistical significance (p < 0.05) between experimental groups. Animals that were euthanized before an experimental endpoint due to health concerns were excluded from the analysis. Mouse samples were blinded for histopathology analysis.

Histopathology

Formalin fixed small intestine tissue sections were stained with hematoxylin and eosin at the Harvard Medical School Rodent Histopathology Core. Capillary congestion was confirmed by a veterinary pathologist. Representative images were obtained using a Nikon TE2000 inverted microscope and the brightness adjusted (Adobe Photoshop CS2).

RNA isolation of in vivo samples

Luminal samples from the ileum and cecum of 3-day old infant rabbits infected with V. cholerae were snap-frozen in liquid nitrogen for subsequent phenol:chloroform RNA isolation. Contaminating DNA was removed using the DNA-free kit (Applied Biosystems) and RNA was stored at −80°C.

In vivo transcriptome analysis using Illumina RNA sequencing

The Ovation Complete Prokaryotic RNA-Seq Library Systems (NuGEN) was used for cDNA synthesis and library preparation of 200 bp (mean size) fragments. cDNA library quantification was performed using the Agilent 2200 TapeStation. RNA-sequencing library quantification was performed on HiSeq 2500 Rapid PE V2 by 100 bp paired-end high-throughput sequencing. The raw reads in FASTQ format were aligned to chromosomes I and II of the V. cholerae El Tor N16961 genome (RefSeq accession numbers NC_002505 and NC_002506) using the bowtie2³¹ algorithm on Geneious 10.0 (http://www.geneious.com)³². Differential expression and PCA analysis was performed on Geneious using DEseq 2³³. Venn diagram was made using Venny 2.1 (Oliveros, J.C. (2007-2015) Venny. An interactive tool for comparing lists with Venn’s diagrams. http://bioinfogp.cnb.csic.es/tools/venny/index.html). Bioinformatics for metabolic pathway analysis was performed using DAVID³⁴. Upregulated genes were defined as having a positive log2 fold change and adjusted p-value of less than 0.0001.

Quantitative real-time PCR

Reverse transcription (RT) and Quantitative real-time PCR (qRT-PCR) was performed using 50ng of DNase treated RNA with the KAPA SYBR FAST One-Step (Kapa biosystems) kit. The fold-change of gene expression was calculated by using the comparative CT method³⁵ with the housekeeping genes gyrB for V. cholerae and GAPDH for mice.

RNA-seq heatmap

RNA-seq heat map of V. cholerae fur-regulated genes was created using Cluster 3.0 (http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm) and Java TreeView (http://jtreeview.sourceforge.net) software packages from mean-centered log2 transformed expression values (RPKM).

Iron measurements by ICP-MS

Luminal samples from the ileum of infant rabbits were snap-frozen in liquid nitrogen. Samples were first treated with concentrated nitric acid (65%-70%) for 24 hours to kill bacteria. Sample analysis was performed by Brooks Applied Labs (Bothell, WA). Briefly, samples were digested in their entirety with a mix of concentrated nitric acid and hydrochloric acid by the EPA Method 3050B and iron was measured by triple quadrupole ICP-MS using the Agilent 8800.

Heme measurements

Samples from fluid of the ileum and cecum of infant rabbits were centrifuged and the supernatant was sterilized using a 0.2 μm filter. Heme levels from filtered samples were measured using the hemin assay kit (ab65332, Abcam, Cambridge, UK) according to the manufacturer’s instructions.

LCFA measurements

Samples from fluid of the ileum and cecum of infant rabbits were centrifuged and the supernatant was sterilized using a 0.2 μm filter. LCFAs (C-8 octanoate and longer) from filtered samples were measured using the Free Fatty Acid Assay Kit (ab65341) according to the manufacturer’s instructions.

L-Lactate measurements

Samples from fluid of the ileum and cecum of infant rabbits were centrifuged and the supernatant was sterilized using a 0.2 μm filter. L-Lactate levels from filtered samples were measured using the L-Lactate assay kit (ab65331, Abcam, Cambridge, UK) according to the manufacturer’s instructions.

Lipocalin-2 ELISA

Samples from fluid of the ileum and cecum of infant rabbits were centrifuged and the supernatant was sterilized using a 0.2 μm filter. Lipocalin-2 levels from filtered samples were measured using the Rabbit Lipocalin-2 (LCN2) ELISA Kit (MBS017878, MyBioSource, San Diego, CA, USA) according to the manufacturer’s instructions.

Statistical analysis

RNA-seq differential expression analysis p-values were determined and adjusted for multiple testing by the Benjamini-Hochberg method using the R package DESeq2. Statistical analyses were performed with Prism software (GraphPad). Fold changes of ratios were transformed logarithmically prior to statistical analysis. An unpaired two-tailed Student’s t-test was used to determine whether differences in fold-changes between two groups were statistically significant (P < 0.05). For comparisons between three or more groups, one-way ANOVA (between-subjects factors) or two-way ANOVA (between-subjects factors) was used followed by post hoc multiple comparisons test. F-statistics (F(Dfn, DFd)) are shown in the figure legends for one-way ANOVA and in Supplementary Table 7 for two-way ANOVA. Outliers were determined with Prism using the ROUT method (Q = 0.5) from log transformed fold-change expression values. Outlier for ctxA expression was excluded from statistical analysis.

Data availability

The data that support the findings of this study are available from the corresponding author upon request. The RNA-Seq data have been deposited in the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo) under the accession number: GSE132653.

Extended Data

Extended Data Figure 1. V. cholerae CT production rescues Δctx mutant growth (related to Figure 1).

(a) GM-1 cholera toxin ELISA assay from filtered supernatant of V. cholerae C6706 wild type (wt) (N = 5) or cholera toxin mutant (Δctx) (N = 3) grown in vitro under cholera toxin-inducing (AKI) conditions. Error bars represent means ± SD. (b) Groups of CD-1 mice (N = 4) were infected with a 1:1 mixture of the V. cholerae WT and the Δctx mutant. The competitive index (CI; the ratio of strains recovered) was determined one day after infection. Dotted red lines connect strains recovered from the same animal. Bacterial numbers were compared by unpaired two-sided student t-test. Each dot represents data from an individual animal or experiment. ns, not statistically significantly different.

Extended Data Figure 2. RNA-seq analysis of V. cholerae in vivo gene expression (related to Figures 1-3).

(a-d) RNA-seq normalized expression (RPKM = reads mapping to gene)/(length of gene/1000)/(total reads mapped to V. cholerae genome/1000000) from wild type or Δctx mutant V. cholerae during infection of 3-day old infant rabbits. Total expression from chromosome I (Chr. I) or chromosome II (Chr. II) from wild type infecting the ileum (a) or cecum (b) or the Δctx mutant infecting the ileum (c) or cecum (d). Correlations (r) are shown between biological replicates (rabbit 1 and rabbit 2).

Extended Data Figure 3. RNA-seq analysis of the wild type C6706 during colonization of the ileum and cecum (related to Figure 1).

(a-b) RNA-seq analysis of Vibrio cholerae in the ileum and cecum during infection of infant rabbits. (a) PCA plots for RNA-seq data from two biological replicates of V. cholerae wild type from the ileum (black circles) or cecum (red circles) of infant rabbits (N = 2). (b) Differential expression of all V. cholerae ORFs and genes involved in virulence. ToxT (VC0838), Cholera Toxin (VC1456-VC1457), Toxin-Coregulated Pilus (TCP) (VC0825-VC0837), TagA (VC0820), HlyA (VCA0219), and Accessory Colonization Factors (Acf) (VC0840-0844) in the wild type during colonization of the ileum relative to the cecum.

Extended Data Figure 4. RNA-seq analysis of CT-dependent V. cholerae gene expression during colonization of the gut (related to Figures 1-2).

(a-e) RNA-seq analysis of Vibrio cholerae in the ileum and cecum during infection of infant rabbits. (a) Differential expression of all V. cholerae ORFs and genes involved in virulence. ToxT (VC0838), Cholera Toxin (VC1456-VC1457), Toxin-Coregulated Pilus (TCP) (VC0825-VC0837), TagA (VC0820), HlyA (VCA0219), and Accessory Colonization Factors (VC0840-0844) in the wild type relative to the cholera toxin mutant (Δctx) during infection of the ileum or cecum. (b) Transcript levels of tcpA were determined by quantitative real-time PCR in the wild type (N = 3) relative to the cholera toxin mutant (Δctx) (N = 4) colonizing the ileum (red circles), in LB (white circles) (N = 3) and in the ileum relative to LB in the wild type (grey circles) and cholera toxin mutant (Δctx) (light blue circles). Error bars represent means ± SD. (c) Percent abundance of raw RNA-seq reads aligned to chromosome I (Chr. I; purple) or chromosome II (Chr. II; yellow) from wild type or the cholera toxin mutant (Δctx) in the cecum. (d) Percent abundance of heme utilization genes (VCA0907-VCA0915, VCA0576; red) from raw RNA-seq reads aligned to chromosome II (Chr. II; yellow) from wild type or the cholera toxin mutant (Δctx) in the cecum. (e) DAVID bioinformatics pathway analysis from RNA-seq data of significantly upregulated pathways (positive log2 fold change and adjusted p-value of less than 0.0001) in the wild type relative to the cholera toxin mutant (Δctx) during colonization of the ileum of infant rabbits (N = 2). RNA-seq differential expression analysis p-values were determined and adjusted for multiple testing by the Benjamini-Hochberg method using the R package DESeq2.

Extended Data Figure 5. RNA-seq analysis of the Vibrio cholerae fur regulon during infection of infant rabbits (related to Figures 1-2).

(a) Heat-map of RNA-seq data from normalized expression (RPKM) of Vibrio cholerae fur regulated genes (listed in Supplementary Table 3) in the wild type relative and the cholera toxin mutant (Δctx) during colonization of the ileum or cecum of infant rabbits. (b) Differential expression of hutA, viuA, and viuB was determined by quantitative real-time PCR analysis from V. cholerae wild type or the cholera toxin mutant (Δctx) in the ileum (N = 3 wild type hutA, viuA, and viuB; N = 6 for Δctx hutA and viuA, N = 4 for Δctx viuB) or in V. cholerae wild type or the cholera toxin mutant (Δctx) grown in LB (N = 3), in LB with 2,2 dipyridyl (N = 3), or in LB with 2,2 dipyridyl and hemin (N = 3). Error bars represent means of log2 transformed mRNA levels determined by quantitative real-time PCR from ileum relative to LB (red circles), LB with 2,2 dipyridyl relative to LB (white circles), and LB with 2,2 dipyridyl and hemin relative to LB (grey circles) ± SD. A two-way ANOVA (Supplementary Table 7) was used to compare normalized expression (ddCT) values between the strains (wt and Δctx mutant) for each treatment (Ileum, 2,2 dipyridyl, 2,2 dipyridyl + 5 μM hemin, and LB) followed by a post-hoc Sidak’s multiple comparisons test. Blue lines on red circles distinguish animals from different litters of Δctx infected infant rabbits. (c) Schematic representation of the genetic region for viuA, viuB, and viuF. ns, not statistically significantly different. (d) Luminal Lipocalin-2 levels determined by ELISA from the ileum and cecum of infant rabbits (N = 4) that were mock-infected or infected with the wild type (wt) or the cholera toxin mutant (Δctx). Error bars represent means ± SD.

Extended Data Figure 6. Cholera toxin induces capillary congestion and enhances luminal growth of V. cholerae in a mouse model of infection (related to Figures 1-3).

(a-h) Groups of CD-1 mice were mock-infected or infected intragastrically with the wild type (wt), the cholera toxin mutant (Δctx), the Δctx mutant mixed with purified CT, or treated orally with purified cholera toxin (CT) and samples were collected one day after infection. (a) Representative images of H/E-stained ileum tissue sections showing capillary congestion (arrows) from mice infected with the indicated strains or treated orally with purified CT. All images were taken at 60X magnification. (b) Expression of CXCl1 and CXCl2 in the ileal mucosa of mice (N = 4) was determined by quantitative real-time PCR analysis. Error bars represent means of CXCl1 and CXCl2 mRNA levels as fold-change over mRNA levels in mock-infected mice ± SD. An unpaired two-sided Student’s t-test was used to compare the differences in fold-changes between the wt and Δctx mutant. (c) The fluid accumulation ratio (FA) ratio for mice (N = 8 for wild type, Δctx, and mock; N = 4 for CT, and Δctx + CT, N = 5 for wild type, and Δctx 6 h.p.i) treated with the indicated strains or treated orally with purified CT. The FA ratios between the groups were compared using one-way ANOVA (F (6, 35) = 66.03, p < 0.0001) followed by post-hoc Tukey’s multiple comparisons test. Black lines show median for individual animals (black circles). Different shapes indicate animals from different litters. (d) Luminal hemin measurements from the ileum of mice (N = 4) treated with the indicated strains or treated orally with purified CT. Hemin levels in the CT-treated group were compared to the Δctx mutant and mock treated using one-way ANOVA (F (3, 12) = 25.59, p < 0.0001) followed by post-hoc Sidak’s multiple comparisons test. Black lines show median for individual animals (black shapes). (e) The CFU/g and total CFU from the whole gastrointestinal tract (gut) from mice (N = 4) infected with wt or the Δctx mutant. An unpaired two-sided Student’s t-test was used to compare the bacterial concentrations from wt and Δctx mutant infected animals. Error bars represent means ± SD. (f) The CFU/g tissue (ileum) (N = 4) and total CFU from the lumen of the ileum (N = 11 for wild type, N = 9 for Δctx, and N = 4 for Δctx + CT) or lumen of the cecum (N = 8 for wild type, N = 7 for Δctx, and N = 4 for Δctx + CT) of mice infected with the indicated V. cholerae strains. Different shapes indicate animals from different litters. An unpaired two-sided Student’s t-test was used to compare the CFU/g tissue in wt and Δctx mutant infected animals. The wt CFU in the ileum or cecum were compared to the Δctx mutant or mock groups using one-way ANOVA (Ileum: F (2, 21) = 50.24, p < 0.0001; Cecum: F (2, 16) = 51.6, p < 0.0001) followed by post-hoc Sidak’s multiple comparisons test. Error bars represent means ± SD. (g) The CFU in the lumen of the ileum 6 hours post-infection mice infected with V. cholerae. An unpaired two-sided Student’s t-test was used to compare the CFU from wt and Δctx mutant infected animals. Error bars represent means ± SD (h) Luminal LCFA measurements (8 carbon chain or greater) from the ileum of mice (N = 3 for wild type, Δctx, and mock; N = 4 for CT) treated with the indicated strains or treated orally with purified CT. LCFA concentrations in the CT-treated group were compared to the Δctx mutant and mock groups treated using one-way ANOVA (F (3, 9) = 7.814, P = 0.0071) followed by post-hoc Sidak’s multiple comparisons test. Black lines show median for individual animals (black circles). ns, not statistically significantly different.

Extended Data Figure 7. Differential expression of CT-dependent metabolism-related genes in V. cholerae (related to Figures 1-3).

(a) Groups of CD-1 mice (N = 4) were infected with a 1:1 mixture of the wild type and the indicated mutant (white bars) or with a 1:1 mixture of the mutant harboring a control vector (pWSK129) and the mutant harboring the complemented gene in pWSK29 (grey bars). The competitive index for the wild type versus the fadL mutant from the first experiment (N = 4) in Fig. 3f is shown for comparison. The competitive index (CI; the ratio of strains recovered) was determined one day after infection. Error bars represent means ± SD. (b) Transcript levels of fadL and lldD were determined by quantitative real-time PCR in the wild type (N = 3) and cholera toxin mutant (Δctx) (N = 4) during colonization of the ileum relative to LB (white bars) and in the wild type relative to the cholera toxin mutant (Δctx) in LB (grey bars) (N = 3). An unpaired two-sided Student’s t-test was used to compare the log2 fold-change in expression of fadL and lldD (ileum relative to LB) between wt and Δctx mutant. Error bars represent means ± SD. (c) Differential expression of V. cholerae Superoxide Dismutase (VC2694), and Nitric Oxide Dioxygenase (VCA0183) in the wild type relative to the cholera toxin mutant (Δctx) during colonization of the ileum or cecum. (d) Iron concentrations from the ileum of individual rabbits infected with the wild type V. cholerae for experiments shown in Figure 2b. (e) Fold-change of transcript levels of wild type V. cholerae genes: ctxA (VC1456), hutA (VCA0576), viuA (VC2211), viuB (VC2210), tcpA (VC0825), fadL (VC1043), and lldD (VCA0983) in the ileum relative to LB were determined by quantitative real-time PCR from individual rabbits in Extended Data Fig. 4b, 5b, and 7b. Dark red (rabbit 2) circle indicates data for sample collected with low expression of cholera toxin (ctxA) that was determined to be a statistical outlier using the ROUT method (Q = 0.5) from log transformed fold-change expression values.