Research Article

Virulence factors enhance Citrobacter rodentium expansion through aerobic respiration

See allHide authors and affiliations

Science  16 Sep 2016:
Vol. 353, Issue 6305, pp. 1249-1253
DOI: 10.1126/science.aag3042

Abstract

Citrobacter rodentium uses a type III secretion system (T3SS) to induce colonic crypt hyperplasia in mice, thereby gaining an edge during its competition with the gut microbiota through an unknown mechanism. Here, we show that by triggering colonic crypt hyperplasia, the C. rodentium T3SS induced an excessive expansion of undifferentiated Ki67-positive epithelial cells, which increased oxygenation of the mucosal surface and drove an aerobic C. rodentium expansion in the colon. Treatment of mice with the γ-secretase inhibitor dibenzazepine to diminish Notch-driven colonic crypt hyperplasia curtailed the fitness advantage conferred by aerobic respiration during C. rodentium infection. We conclude that C. rodentium uses its T3SS to induce histopathological lesions that generate an intestinal microenvironment in which growth of the pathogen is fueled by aerobic respiration.

Attaching and effacing (AE) pathogens are defined by virulence characteristics encoded by a pathogenicity island called the locus of enterocyte effacement (LEE) (1, 2), which contains genes encoding a type III secretion system (T3SS) (3) and an adhesin termed intimin (4). The T3SS injects the intimin receptor Tir into the host cell cytosol (5), resulting in intimate attachment of bacteria to the effaced epithelial surface (6). The LEE-encoded T3SS of the AE pathogen C. rodentium injects additional effector proteins that are required for causing transmissible colonic crypt hyperplasia in mice (7, 8) (fig. S1). After the development of colonic crypt hyperplasia, C. rodentium blooms in the lumen of the murine large bowel (9). The LEE-encoded T3SS is required for this rapid luminal expansion possibly by allowing C. rodentium to compete with the microbiota for carbon sources, because the T3SS provides no benefit in germ-free mice (10). These data suggest that the T3SS places C. rodentium in a microenvironment that somehow provides a growth advantage during its competition with the resident microbiota, but it remains obscure which resources might become available in this niche to fuel pathogen expansion.

Nitrate respiration is dispensable for pathogen growth during inflammation

Electron acceptors, such as nitrate, are produced as a by-product of the inflammatory host response and boost luminal growth of pathogenic Salmonella enterica or commensal Escherichia coli by anaerobic respiration in mouse models of colitis (1113). Because C. rodentium infection triggers colonic crypt hyperplasia, we wanted to determine whether the inflammatory host response would enable the pathogen to grow by anaerobic respiration. The respiratory reductases for nitrate, dimethyl sulfoxide (DMSO), and trimethylamine N-oxide (TMAO), as well as the formate dehydrogenases FdnGHI and FdoGHI, contain a molybdopterin cofactor. Thus, to explore a possible role of anaerobic respiration during C. rodentium growth in the mouse gut, we constructed a C. rodentium mutant lacking a gene required for molybdopterin cofactor biosynthesis (moaA mutant) (fig. S2A) (14). Mice (C57BL/6) were inoculated with an equal mixture of wild-type C. rodentium and an isogenic moaA mutant to compare the fitness of both strains. Mice developed intestinal inflammation as indicated by increased transcript levels of proinflammatory markers in the colonic mucosa (fig. S3, A and B). The C. rodentium wild type was recovered in significantly (P < 0.05) higher numbers than the moaA mutant (Fig. 1A). Similar results were observed with genetically susceptible C3H/HeJ mice that experience more severe intestinal inflammation during C. rodentium infection (fig. S3, C and D). In contrast, when germ-free mice were inoculated with an equal mixture of the C. rodentium wild type and a moaA mutant, both strains were recovered in similar numbers (Fig. 1B and fig. S3E), suggesting that either anaerobic respiration or the utilization of formate provided an edge during competition of the pathogen with the resident microbiota.

Fig. 1 Oxygen respiration supports C. rodentium expansion in the mouse colon.

(A) C57BL/6 (C57) mice were infected with C. rodentium wild type (wt, DBS100) and either a moaA mutant (CAL142) or a napA narG narZ mutant (CAL93). (B) Conventional C57 or germ-free Swiss Webster (SW) mice were infected with wt C. rodentium and either a moaA mutant, a fdnG mutant [CAL210 (pWSK129)], or a fdoG mutant (CAL261). N is indicated in fig. S3E. (C) Competitive in vitro growth (N = 8 mice) of C. rodentium wild type (wt) and a cydAB mutant (CAL247) for 16 hours in minimal medium in the presence of the indicated oxygen levels (% O2). (D) Conventional or germ-free mice were infected with an equal mixture of the C. rodentium wild type (wt) and a cydAB mutant. (A and D) N = 4 mice. (E) C. rodentium was grown in minimal medium supplemented with mannose as a carbon source under either microaerobic or anaerobic conditions. (F) Bacterial RNA was isolated from either mucus scrapings or colon contents of C. rodentium–infected mice. (E and F) The transcript levels of sucA were quantified by real-time polymerase chain reaction, normalized to 16S ribosomal RNA (rRNA) levels, and shown as fold-changes. N is shown in fig. S4, B and C. (G and I) Mice [N indicated in (I)] were either mock-treated or infected with the C. rodentium wild type (wt) or a cydAB mutant (CAL247). (G) Enumeration of C. rodentium by plating on selective media. (A to G) Bars represent geometric mean ± SE. (H) Boxes in whisker plots represent the second and third quartiles of combined histopathology scores, while lines indicate the first and fourth quartiles. (I) Microbial representation at the class level based on 16S rRNA gene sequencing of colon contents 7 days after infection. Color coding for classes is shown on the right. *P < 0.05; **P < 0.01; ns, not statistically significantly different.

C. rodentium infection resulted in a markedly increased colonic expression of Nos2 (fig. S3, B and D), the gene encoding inducible nitric oxide synthase (iNOS), an enzyme necessary for generating nitric oxide, which is converted to nitrate in the intestinal lumen (11). To determine whether nitrate respiration enhances growth, we constructed a mutant deficient for the three nitrate reductases encoded by C. rodentium (narG napA narZ mutant) (fig. S2, B and F). Notably, in mice inoculated with an equal mixture of the C. rodentium wild type and a narG napA narZ mutant, both strains were recovered in equal numbers from colon contents and feces (Fig. 1A and fig. S3, C and E), suggesting that nitrate respiration did not provide a fitness advantage.

We next examined the possibility that the phenotype of the moaA mutant was due to an inability to utilize formate as an electron donor rather than nitrate as an electron acceptor. Because formate dehydrogenases FdnGHI and FdoGHI can couple electron transfer from formate through the quinone pool to nitrate (15) or oxygen (16), respectively, we envisioned that analysis of this pathway would provide indirect information about the electron acceptor used by C. rodentium in the gut. To test whether the nitrate-dependent and/or the oxygen-dependent formate dehydrogenase contributed to growth of C. rodentium during colitis, we inoculated mice with an equal mixture of the C. rodentium wild type and either an fdnG mutant or fdoG mutant (fig. S2, C and D). Consistent with our observation that nitrate respiration did not contribute to fitness (Fig. 1A), inactivation of the nitrate-dependent fdnG did not reduce growth of C. rodentium (Fig. 1B). However, inactivation of the oxygen-dependent fdoG reduced fitness of C. rodentium in conventional (C57BL/6) mice (P < 0.05), whereas little benefit was observed in germ-free (Swiss Webster) mice (Fig. 1B and fig. S3E). The finding that fdoG was required for pathogen expansion was intriguing, because it suggested that C. rodentium might utilize oxygen as a terminal electron acceptor during luminal growth.

Aerobic respiration drives a luminal pathogen expansion

To test the hypothesis that aerobic respiration fuels C. rodentium growth during infection, we deleted the C. rodentium cydAB genes (fig. S2E), which encode a high-affinity cytochrome bd oxidase required for aerobic respiration in microaerobic environments. In vitro, the C. rodentium wild type was recovered in higher numbers than the cydAB mutant when both strains were cocultured under conditions mimicking hypoxia (1% O2) or normal tissue oxygenation (4 or 8% O2), but not under anaerobic conditions (0% O2) or at atmospheric oxygen levels (21% O2) (Fig. 1C). Next, we compared the fitness of the C. rodentium wild type and a cydAB mutant by infecting C57BL/6 mice with a 1:1 mixture of both strains. Notably, the C. rodentium wild type was recovered in 20,000-fold higher numbers (P < 0.01) than the cydAB mutant from colon contents 7 days after infection (Fig. 1D), suggesting that aerobic respiration contributed to growth of the pathogen. Expression of ler, encoding the master regulator of T3SS synthesis, was reduced in the cydAB mutant, but this difference did not reach statistical significance (fig. S4A). When infection with a 1:1 mixture of the C. rodentium wild type and a cydAB mutant was repeated in germ-free (Swiss Webster) mice, the fitness advantage provided by cydAB was greatly reduced (Fig. 1D). Collectively, these data suggested that aerobic respiration drove an uncontrolled luminal expansion of C. rodentium in an environment occupied by the gut microbiota.

To test whether tissue- and mucus-associated bacteria have a respiratory metabolism compared to luminal bacteria, we investigated transcription of the sucA gene, which encodes a subunit of 2-ketoglutarate dehydrogenase, an enzyme required for the aerobic tricarboxylic acid (TCA) cycle. Under anaerobic conditions, SucA is no longer required as the normally cyclic TCA pathway switches to a noncyclic series of reactions. Higher expression of sucA under aerobic than anaerobic conditions can serve as an indicator for a respiratory central metabolism (17). C. rodentium sucA was transcribed at significantly (P < 0.05) higher levels when bacteria were cultured in vitro under microaerobic conditions compared to anaerobic conditions (Fig. 1E and fig. S4B). To investigate sucA expression in vivo, we isolated RNA from mucus scrapings or colon contents of mice infected with C. rodentium. C. rodentium residing in close proximity to tissue (i.e., mucus scrapings) transcribed sucA at significantly (P < 0.05) higher levels than bacteria located in colon contents (Fig. 1F and fig. S4C). These results provided further support for the idea that bacteria in close proximity to the mucosal surface had an oxidative metabolism in vivo.

Intestinal inflammation caused by Salmonella enterica drives a depletion of Clostridia, which in turn increases epithelial oxygenation (18). We thus wanted to determine whether C. rodentium–induced colonic crypt hyperplasia would increase oxygen availability in the gut by driving a depletion of Clostridia. The numbers of C. rodentium recovered from colon contents of mice infected with the C. rodentium wild type were 1000-fold higher (P < 0.05) than those recovered from colon contents of mice infected with a cydAB mutant (Fig. 1G), which correlated with reduced colonic inflammation (P < 0.05) (Fig. 1H and fig. S5). Consistent with a previous report (9), infection of mice with the C. rodentium wild type resulted in a shift in the microbial community structure (Fig. 1I and fig. S6). The absolute abundances of members of the Enterobacteriaceae (fig. S7A) and Erythrobacteraceae (fig. S7B) were significantly (P < 0.05) increased in mice infected with the C. rodentium wild type, a shift in the community structure previously associated with increased oxygen availability in the gut (19). However, infection with the C. rodentium wild type resulted in an increased (P < 0.05) abundance of Clostridia (Fig. 1I), which was most pronounced for members of the Lachnospiraceae belonging to the genus Dorea (fig. S7C) and the genus Coprococcus (fig. S7D). These changes in the microbial population profile were reduced or absent in mice infected with a cydAB mutant (Fig. 1I and figs. S6 and S7). In conclusion, our data were not consistent with the hypothesis that colonic crypt hyperplasia increased oxygen availability by depleting Clostridia.

T3SS-mediated niche construction

C. rodentium attaches to murine colonic epithelial cells (colonocytes) using a type IV pilus encoded by the cfc operon (20). To test whether this adhesin contributed to establishing a niche in which the pathogen could grow by CydAB-mediated respiration, we constructed a cfcH mutant (fig. S1F). Genetic ablation of type IV pilus biosynthesis did not change the CydAB-dependent growth advantage 7 days after infection of mice (Fig. 2A and fig. S8A), possibly because a cfcH mutant still occupied a niche near the epithelium (fig. S8B).

Fig. 2 A functional T3SS increases epithelial oxygenation and drives a CydAB-dependent C. rodentium expansion.

(A) Mice (N is indicated in fig. S8A) were infected with the indicated C. rodentium strain mixtures. Bars represent geometric means of the competitive index (CI) ± SE. (B) Representative images of colonic sections stained to detect pimonidazole hypoxia stain (red fluorescence) and counterstained with DAPI (4′,6-diamidino-2-phenylindole) nuclear stain (blue fluorescence). Arrowheads point to the mucosal surface. E, epithelial cell; L, lumen. (C) Pimonidazole staining (PMDZ+) was quantified by flow cytometry in colonic epithelial (CD45 CD326+) cells. White bars indicate the level of background staining observed in mice that were not injected with PMDZ. Bars represent geometric mean ± SE. **P < 0.01; *P < 0.05; ns, not statistically significantly different; wt, DBS100; cydAB, CAL247; cfcH, BMM12; cfcH cydAB, BMM11; eae, CAL290; eae cydAB, CAL291; escN, CAL286; escN cydAB, CAL287; espH cesF map, BMM30; espH cesF map cydAB, BMM40.

C. rodentium intimately attaches to colonocytes in a T3SS-dependent fashion using intimin. To test whether the T3SS and intimin contributed to CydAB-dependent growth, we constructed mutations in escN, encoding a component of the T3SS, and eae, the gene encoding intimin (fig. S1G and S1H). The fitness advantage conferred by CydAB-dependent respiration 7 days after infection was significantly reduced in C. rodentium strains lacking eae or escN (Fig. 2A and fig. S8A), suggesting that in the absence of intimate attachment or a functional T3SS, oxygen respiration no longer conferred a marked fitness advantage. However, at 3 days after infection, when intestinal inflammation was just beginning to develop (fig. S8, C and D), inactivation of escN did not significantly reduce the fitness advantage conferred by the cydAB genes (fig. S8E). Because the escN mutation did not reduce the fitness advantage conferred by the cydAB genes in the absence of overt crypt hyperplasia (i.e., at 3 days after infection), we considered the possibility that enhanced access to oxygen might not depend on intimate attachment, but might require T3SS-mediated host responses.

Colonocytes mature as they migrate from the bottom of the crypt to the luminal surface. The mature surface colonocyte population functions in water absorption, which is driven by electrolyte transport (21). To energize absorption of electrolytes, mature colonocytes respire butyrate in their mitochondria (22), a process consuming oxygen and rendering the mucosal surface hypoxic (< 7.6 mmHg pO2 or < 1% O2) (23, 24) (fig. S1). To investigate whether T3SS-induced colonic crypt hyperplasia, which developed by seven days after infection (fig. S9A and S9B), would alter epithelial oxygenation in vivo, we visualized the “physiologic hypoxia” of surface colonocytes using the exogenous hypoxic marker pimonidazole (25) (Fig. 2B). Notably, a lack of hypoxia staining at the mucosal surface suggested that infection with the C. rodentium wild type significantly (P < 0.05) reduced epithelial hypoxia by 7 days after infection (Fig. 2, B and C, and fig. S10), indicative of a marked increase in epithelial oxygenation. This increased epithelial oxygenation was T3SS-dependent, because it was not observed in mice infected with the escN mutant (Fig. 2B and 2C), which did not develop inflammation in the colon (fig. S8).

T3SS-induced colonic crypt hyperplasia is characterized by an excessive intestinal epithelial repair response [reviewed in (26)]. To further investigate the mechanism by which the T3SS increases epithelial oxygenation, we deleted the map, cesF, and espH genes to reduce damage to colonocytes (fig. S2H). Map and EspH are T3SS effector proteins that damage mitochondria (which might impair respiration of butyrate) or activate caspase-3 to induce cytotoxicity, respectively (27, 28), whereas the cesF gene encodes a chaperone for the mitochondria-associated T3SS effector EspF (29). Deletion of the map, cesF, and espH genes significantly (P < 0.05) reduced the fitness advantage conferred by CydAB-dependent respiration 7 days after infection of mice (Fig. 2A), but it was not clear whether this was simply due to an overall lower level of colonization (fig. S8A). When mice were infected with either the C. rodentium wild type or an espH cesF map mutant, both strains were detected in association with the mucosal surface, although the espH cesF map mutant appeared impaired for colonization (fig. S11A) and the C. rodentium wild type was recovered in higher numbers than a espH cesF map mutant from colon contents 7 days after infection (Fig. 3A). Only infection with the C. rodentium wild type significantly reduced epithelial hypoxia, while the colonic surface remained hypoxic in mice infected with a espH cesF map mutant (fig. S11, B and C).

Fig. 3 Colonic crypt hyperplasia drives a CydAB-dependent C. rodentium expansion.

(A to C) C57BL/6 mice were mock infected or infected with the C. rodentium wild type (wt, DBS100) or an espH cesF map mutant (BMM30). (D to F) C57BL/6 mice were mock infected or infected with a 1:1 mixture of the C. rodentium wild type (wt) and a cydAB mutant (CAL247). Mice were treated with either dibenzazepine (DBZ) or vehicle control. (A to F) Organs were collected 7 days after infection. (A) Colony-forming units (CFU) recovered from colon contents. (B and D) Representative images of colonic sections were stained to detect Ki67 (yellow fluorescence) and counterstained with DAPI nuclear stain (blue fluorescence). (C and E) Ki67 staining was quantified by image analysis. (F) Enumeration of wt and cydAB mutant recovered from colon contents was used to calculate the CI. (A, C, E, and F) Bars represent geometric mean ± SE. **P < 0.01; *P < 0.05; ns, not statistically significantly different.

Whereas mature colonocytes at the luminal surface respire butyrate, undifferentiated colonocytes in the crypts exhibit the Warburg metabolism of dividing cells, which is characterized by fermenting glucose to lactate (30). C. rodentium infection induces epithelial regeneration and repair mechanisms, which drives a marked expansion of undifferentiated colonocytes, resulting in crypt elongation and the presence of undifferentiated colonocytes at the luminal surface (31, 32) (fig. S1). Because the Warburg metabolism does not consume oxygen, we wanted to determine whether reduced hypoxia of surface colonocytes observed during C. rodentium infection (Fig. 2, B and C, and fig. S11, B and C) was due to an accumulation of undifferentiated colonocytes at the luminal surface. Infection with a espH cesF map mutant triggered significantly (P < 0.05) less mitotic divisions in the colonic epithelium than infection with wild-type C. rodentium (fig. S8C), suggesting that deletion of the espH, cesF, and map genes might reduce crypt hyperplasia. To test this idea, we stained colonic sections from mock-infected mice or mice infected with the C. rodentium wild type or an espH cesF map mutant with Ki67, a cellular marker for proliferation. Infection with the C. rodentium wild type induced a marked expansion of Ki67-positive colonocytes, resulting in crypt elongation, a thickening of the colonic mucosa, and an accumulation of Ki67-positive colonocytes along the length of the crypt and at the luminal surface (Fig. 3, B and C). This host response was blunted in mice infected with a C. rodentium espH cesF map mutant (Fig. 3, B and C), which correlated with increased epithelial hypoxia (fig. S11, B and C).

Colonic crypt hyperplasia drives aerobic pathogen expansion

We next wanted to determine whether colonic crypt hyperplasia was a driver of CydAB-dependent pathogen expansion. To this end, mice were mock infected or infected with a 1:1 mixture of the C. rodentium wild type and cydAB mutant and then treated with dibenzazepine (DBZ) or with vehicle control. DBZ is a Notch and Wnt (wingless-related integration site) signaling pathway inhibitor that prevents colonic crypt hyperplasia during C. rodentium infection (31). Ki67 staining suggested that mice infected with the C. rodentium strain mixture developed colonic crypt hyperplasia, which was significantly (P < 0.05) blunted by DBZ treatment (Fig. 3, D and E), although DBZ treatment did not reduce the severity of other histopathological changes (fig. S12, A and B). Notably, inhibition of colonic crypt hyperplasia by DBZ treatment abrogated the fitness advantage conferred by the cydAB genes (Fig. 3F), suggesting that crypt hyperplasia was a main driver of aerobic pathogen expansion.

Our finding that the LEE-encoded T3SS provides C. rodentium access to oxygen in vivo helps explain why the mechanism by which this trademark virulence factor confers a benefit to AE pathogens has not been apparent from in vitro studies (33, 34), because such experiments are commonly performed in a 95% air (79% N2/21% O2) atmosphere, supplemented by 5% CO2, thus providing 19.95% O2 (150 mmHg). This oxygenation is considerably higher than normal tissue oxygenation (23 to 70 mmHg or 3 to 10% O2) (35) or oxygenation of colonocytes in vivo (< 7.6 mmHg or < 1% O2) (23, 24). Although oxygen emanating from the mucosal surface is a very limited resource in the lumen of the large intestine (36), epithelial oxygenation was markedly elevated during T3SS-induced colonic crypt hyperplasia. As a result, T3SS-induced colonic crypt hyperplasia drove growth of C. rodentium through CydAB-mediated aerobic respiration, presumably because a respiratory metabolism enables the pathogen to utilize carbon sources more effectively than competing microbes that rely on fermentation for growth. In conclusion, our results revealed how histopathological changes triggered by a trademark virulence factor of AE pathogens create a unique nutrient niche to fuel an uncontrolled luminal expansion of C. rodentium.

Supplementary Materials

www.sciencemag.org/content/353/6305/1249/suppl/DC1

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

References (3752)

References and Notes

  1. Acknowledgments: We acknowledge the Host-Microbe Systems Biology Core (HMSB Core) at the University of California Davis School of Medicine for expert technical assistance with microbiota sequence analysis. The data reported in the manuscript are tabulated in the main paper and in the supplementary materials and are available in the National Center for Biotechnology Information BioSample database under accession numbers SAMN05420840 through SAMN05420856. Work in A.J.B.’s laboratory was supported by Public Health Service Grants AI044170, AI096528, AI107393, and AI112949. Work in R.M.T.’s laboratory was supported by Public Health Service Grant AI098078. Work in S.E.W.’s laboratory was supported by Public Health Service Grant AI103248. C.A.L was supported by Public Health Service Grant AI112241. E.M.V. was supported by Public Health Service Grant OD010931.
View Abstract

Navigate This Article