Human Clostridium difficile infection: inhibition of NHE3 and microbiota profile

Abstract

Clostridium difficile infection (CDI) is principally responsible for hospital acquired, antibiotic-induced diarrhea and colitis and represents a significant financial burden on our healthcare system. Little is known about C. difficile proliferation requirements, and a better understanding of these parameters is critical for development of new therapeutic targets. In cell lines, C. difficile toxin B has been shown to inhibit Na⁺/H⁺ exchanger 3 (NHE3) and loss of NHE3 in mice results in an altered intestinal environment coupled with a transformed gut microbiota composition. However, this has yet to be established in vivo in humans. We hypothesize that C. difficile toxin inhibits NHE3, resulting in alteration of the intestinal environment and gut microbiota. Our results demonstrate that CDI patient biopsy specimens have decreased NHE3 expression and CDI stool has elevated Na⁺ and is more alkaline compared with stool from healthy individuals. CDI stool microbiota have increased Bacteroidetes and Proteobacteria and decreased Firmicutes phyla compared with healthy subjects. In vitro, C. difficile grows optimally in the presence of elevated Na⁺ and alkaline pH, conditions that correlate to changes observed in CDI patients. To confirm that inhibition of NHE3 was specific to C. difficile, human intestinal organoids (HIOs) were injected with C. difficile or healthy and CDI stool supernatant. Injection of C. difficile and CDI stool decreased NHE3 mRNA and protein expression compared with healthy stool and control HIOs. Together these data demonstrate that C. difficile inhibits NHE3 in vivo, which creates an altered environment favored by C. difficile.

clostridium difficile is a Gram-positive anaerobic bacterium from the phylum Firmicutes that is responsible for the majority of antibiotic-associated diarrhea (17). C. difficile infection (CDI) affects thousands of patients each year and treatment costs of over 1 billion dollars in the United States (11, 18, 49, 59). Furthermore, C. difficile-related deaths have been steadily rising since 1999 (54) and will likely remain a problem, especially in the face of current antibiotic regimens. CDI has been associated with a spectrum of symptoms ranging from mild to watery diarrhea and abdominal pain to life-threatening pseudomembranous colitis and toxic megacolon (8). Although most of the symptoms of CDI have been linked to C. difficile toxin production (31, 37, 41), the mechanism of C. difficile colonization is still unclear. Thus a better understanding of C. difficile pathogenesis is critical for developing new therapeutics.

C. difficile pathogenesis has been hypothesized to be a three-step process: 1) antibiotic disruption of the normal gut microbiota provides a potential niche for growth from its normal gut spore form; 2) the colonization phase, which includes bacterial-host interaction and adhesion; and 3) multiplication that maintains high numbers of vegetative C. difficile and toxin production, both of which exacerbate the infectious process (15, 34). Antibiotic use has been shown to decrease the dominant gut microbiota bacterial phyla Bacteroidetes and Firmicutes (40) and increase Proteobacteria (1, 14, 16, 30, 35, 43, 62), resulting in increased gut susceptibility to C. difficile infection (2, 5, 36, 52, 55, 67, 78). Once C. difficile binds to the gastrointestinal (GI) mucus layer (15, 69), the bacterium can deliver two exotoxins, toxin A (TcdA) and toxin B (TcdB) (17, 32, 74). The Tcd toxins bind to uncharacterized host receptors and are then internalized into the enterocyte cytoplasm, where they become enzymatically active and glycosylate the Rho family of GTPases (19, 29). Inhibition of such GTPases has been shown to have several effects including 1) disorganization of the host actin cytoskeleton, 2) loss of cellular tight junctions, 3) disruption of signaling cascades, and 4) arrest of cell cycle progression (3, 15, 19, 33). In addition, toxin B inhibition of Rho GTPase in cell lines leads to the internalization of the Na⁺/H⁺ exchanger isoform 3 (NHE3) (29), but this has yet to be established in vivo in animals or in humans. Inhibition of NHE3 in mice results in chronic diarrhea (25, 60), elevated Na⁺ and alkaline luminal fluid, and an altered microbiota composition with decreased members of Firmicutes and increased Bacteroidetes (20). It has been suggested that the diarrhea associated with CDI is a result of damage to the host epithelium or a response designed to “flush out” the pathogen. However, we hypothesize that C. difficile toxin production inhibits NHE3, creating an altered intestinal microenvironment and gut microbiota composition, which favor C. difficile proliferation and colonization of the mucosal lining. In this study, we demonstrate that biopsy specimens from patients with CDI have decreased NHE3 with increased Bacteroidetes and decreased Firmicutes phyla in their stool. In vitro, C. difficile growth depends on the high Na⁺ concentration ([Na⁺]) and a more alkaline environment, which can be caused by downregulation of NHE3. This study is the first to demonstrate downregulation of NHE3 and an altered luminal environment in patients with CDI.

METHODS

Patient information.

All patients and healthy volunteers at the University of Cincinnati Medical Center Hospital, Cincinnati, OH, provided informed consent approved by the University of Cincinnati Institutional Review Board. Samples were evaluated from patients with recurrent CDI. Initial CDI cases, defined as only one C. difficile-positive laboratory test with no prior history of CDI, were not included in this study. Recurrent CDI was defined as onset of new diarrhea after a symptom-free period of >3 days, more than one C. difficile-positive laboratory test, and completion of at least one round of antibiotic treatment. C. difficile infection (CDI) was defined as a new onset of diarrhea (>3 loose stools/day for >24 h) and at least one positive C. difficile laboratory test. Diagnosis of CDI was determined by at least one ELISA-positive toxin test or a positive lysosome-associated membrane protein (LAMP) test. Over the course of fecal collections, two types of toxin tests were used. From November 2010 to August 2011, the enzyme immunoassay for toxins A and B was used. After August 2011, the Meridian Illumigene LAMP test was used. This shift in toxin testing represents a switch to in-house testing, lowering the cost, and an upgrade to a more sensitive method.

Fecal samples were collected from 12 recurrent CDI patients with an average age of 56, age range of 32–76. This group included eight females and four males. Selected patients did not have history of inflammatory bowel disease, small bowel obstruction, diverticulosis, colostomy, or cancer. Fecal samples were also collected from 12 healthy volunteers with an average age of 41, age range of 28–61. This group included seven females and five males. To address antibiotic use and stool composition, fecal samples were collected from eight patients with diarrhea, no antibiotics (age range: 34–67; mean age: 49; female: 3; male: 5), five patients with normal stool and antibiotics (vancoymcin/clindamycin) but without CDI infection (age range: 29–56; mean age: 39; female: 4; male: 1), and seven patients with diarrhea and antibiotics (vancoymcin/clindamycin) but without CDI infection (age range: 38–70; mean age: 52; female: 4; male: 3). Healthy volunteers and patients with diarrhea and/or antibiotics were without previous or current GI symptoms and history of chronic disease or cancer. All stool samples were processed for total DNA, ion concentration, and pH and stored at −20°C.

Colon biopsy specimens collected from five healthy volunteers were obtained by consent and fixed in neutral buffered formalin and paraffin embedded. Healthy subjects had an average age of 52 and patient age range of 45–63 and included three females and two males. Healthy volunteers were without previous or current GI symptoms and history of chronic disease or cancer. Paraffin sections of biopsies and colon resections were obtained from five de-identified patients with a current CDI diagnosis (C. difficile-positive toxin test) and no other known morbidity/disorder. The average patient age 44 and patient age range of 28–65 and included two females and three males. Selected patients did not have history of inflammatory bowel disease small bowel obstruction, diverticulosis, colostomy, or cancer. Confirmation of C. difficile infection was performed by tissue staining with C. difficile-specific antibody as described below.

Histology.

Healthy and CDI biopsy and surgical resections were obtained from the transverse colon and fixed overnight at 4°C in neutral-buffered formalin and embedded in paraffin. Serial 6- to 7-μm-thick sections were applied to glass slides, and intestinal architecture was examined by hematoxylin and eosin staining. Expression of NHE3 was examined with rabbit anti-human NHE3 antibody (dilution 1:100, NBP1-82574; Novus Biologicals, Littleton, CO), and C. difficile binding was examined with rabbit anti-C. difficile cell surface protein antibody (dilution 1:100, ab93728; Abcam, Cambridge, MA). Briefly, sections were removed of paraffin and incubated for 40 min at 97°C with Tris-EDTA-SDS buffer as previously described (68). Sections were then blocked with PBS containing 10% serum and stained with primary antibody overnight at 4°C. Sections were then washed three times in PBS, incubated with goat-anti-rabbit IgG Alexa Fluor secondary antibody (dilution 1:100; Life Technologies, Grand Island, NY) for 1 h at room temperature, and counterstained with Hoechst (0.1 μg/ml; Fisher Scientific). Sections were analyzed by confocal laser scanning microscopy (Zeiss LSM Confocal 710; Carl Zeiss). Digital images of slides were evaluated by tabulating mean pixel intensity of the respective color channel on each image using ImageJ software (National Institutes of Health) and reported as relative fluorescence. Five regions of interest per image, four images per slide, and n = 5 healthy and CDI patients were used for semiquantification of stain intensity normalized to healthy subjects and referred to as relative fluorescence.

Human intestinal organoids and microinjection.

Organoids resembling human proximal colon, hereafter referred to as human intestinal organoids (HIOs), were generated by the Cincinnati Children's Hospital Medical Center (CCHMC) Pluripotent Stem Cell Facility through directed differentiation of human pluripotent stem cells (hPSC). Differentiation of hPSCs from a single subject was obtained by culturing hPSCs for 3 days in ActivinA, followed by fibroblast growth factor 4 (FGF4) and Wnt3a. HIOs achieved 3-day dimensional growth in Matrigel with epidermal growth factor (EGF), R-spondin, and Noggin as previously described (77). HIOs were obtained in Matrigel from the CCHMC Puripotent Stem Cell core. These organoids have been previously shown to contain all major intestinal epithelial cell types: enterocytes (villin), goblet cells (mucin), paneth cells (lysozyme), and enteroendocrine cells (chromogranin A) (77). The luminal compartment of HIOs was microinjected with bacteria and stool supernatant to analyze host-microbe interactions as previously described (20). Injection needles were pulled on a horizontal bed puller (Sutter Instruments), and the tip was cut to a tip diameter of ∼10–15 μm. HIOs were injected with C. difficile ATTC 1870 or stool from healthy or CDI patients. C. difficile ATTC 1870 was grown in tryptone yeast TY broth as previously described (22). For stool, 0.5 g of healthy or CDI stool were added to 4.5 ml tryptic soy broth (TSB; Fisher Scientific) in an anaerobic hood. Samples were vortexed and centrifuged at 150 g for 10 min to pellet solid materials. Stool supernatant, C. difficile, or C. butryicum cultures or TSB broth was injected into HIOs using a Nanoject microinjector (Drummon Scientific, Broomall, PA). To minimize stretch effects on epithelial cells, injection volumes of ∼10% or less of the organoid luminal volume were used. Under these conditions, no leakage of cultures from the HIOs was observed. HIOs were processed either for RNA or immunostaining. For RNA, organoids were homogenized in Trizol and extracted with chloroform according the manufacturer's instructions (Invitrogen). For staining, HIOs were incubated overnight after microinjection and fixed with 4% paraformaldehyde for 30 min at room temperature. HIOs were washed in PBS and transferred to sucrose (30% in PBS) and incubated overnight at 4°C. The next day, HIOs were placed in optimal cutting temperature (OCT) embedding medium and frozen at −80°C for 1 day. Seven-micrometer sections were cut on a cryostat. Slides were stained with rabbit anti-human NHE3 antibody (dilution 1:100, NBP1-82574; Novus Biologicals) and analyzed by confocal laser scanning microscopy (Zeiss LSM Confocal 710; Carl Zeiss).

Bacterial strains and culture conditions.

C. difficile ATTC BAA-1870 and Blautia producta ATCC 27340D were purchased from American Type Culture Collection (ATCC, Manassas, VA). Micrococcus luteus, Staphylococcus aureus, Escherichia coli, Burkholderia cepacia, and Faecalibacterium prausnitzii were locally available (Hassett Laboratory). Bacteroidetes thetaiotaomicron ATCC 29741 was purchased from Fisher Scientific (Thermo Fisher Scientific, Waltham, MA). Lactobacillus acidophilus, Rhizobium leguminosarum, and C. butryicum were purchased from Carolina Biological Supply (Burlington, NC). S. aureus, M. luteus, L. acidophilus, B. thetaiotaomicron, E. coli, B. cepacia, and R. legaminosarum were used to generate quantitative (q)PCR standard curves as previously described (20). E. coli, S. aureus, M. luteus, B. cepacia, L. acidophilus, and R. legaminosarum were grown in Luria-Burtani (LB; Thermo Fisher Scientific) broth at 37°C in a shaking incubator. B. thetaiotaomicron was grown in TSB (Fisher Scientific), and C. difficile was grown in tryptone-yeast extract-glucose broth (TYG; Thermo Fisher Scientific) at 37°C in a Coy Systems, dual-port anaerobic chamber (Coy Laboratory Products, Grass Lake, MI).

To determine the optimal [Na⁺] for growth, C. difficile (Cluster X1), C. butryicum (Cluster I), Blautia producta ATCC 27340D (C. coccoides Cluster XIVa), and Faecalibacterium prausnitzii (C. leptum Cluster IV) were grown in media where sodium chloride (NaCl) was either removed or replaced with cesium chloride (CsCl) or potassium chloride (KCl) as previously described (9, 10, 20). Briefly, low Na⁺ media was mixed with normal media at various ratios to obtain varying concentrations of Na⁺ for bacterial growth measurements. Actual Na⁺ and K⁺ concentrations ([Na⁺] and [K⁺]) were confirmed by flame photometry (Single-Channel Digital Flame Photometer Model 02655-10; Cole-Parmer Instrument, Vernon Hills, IL), and Cl⁻ concentration ([Cl⁻]) was measured by chloridometry (Digital Chloridometer Model 4425100; Labconco, Kansas City, MO). Bacteria were grown under anaerobic conditions at 37°C to early stationary phase in normal TYG media [12 h, optical density at 560 nm (OD_560nm): ∼1] and used to inoculate media containing varying [Na⁺]. Growth was measured as the optical density (OD_560nm) with an Amersham Biosciences Ultospec 3100 Spectrophotometer (GE Healthcare Life Sciences, Pittsburgh, PA). Clostridial titers were determined by bacterial cell counts using a Petroff-Hauser chamber (Hausser Scientific; Horsham, PA) and also by colony-forming units (CFU) (9, 10). No differences in growth patterns were observed between 4 (early exponential phase)-, 12 (early stationary phase)-, 24-, or 48 (stationary phase)-h time points (data not shown). As a result, all data are represented as the OD_560nm and CFU at the 24-h time point. To determine the optimal pH for growth, C. difficile, C. butryicum, Blautia producta ATCC 27340D (C. coccoides), and Faecalibacterium prausnitzii (C. leptum) were grown in TYG media containing either normal media or low Na⁺ media adjusted to pH values ranging from 5.5 to 7.0 as determined electrochemically using a pH meter (Orion Model 720A; Thermo Fisher Scientific).

Quantitative real time-PCR amplification of 16S sequences.

QIAamp DNA Stool kit (Qiagen, Valencia, CA) was used to isolate total DNA from stool of healthy subjects or patients with recurrent CDI. To improve bacterial cell lysis, the temperature was increased to 95°C and incubation with lysozyme (10 mg/ml, 37°C for 30 min) was used as previously described (12, 20, 24, 48, 56, 57). qPCR was used to access the abundance of total bacteria and specific intestinal bacterial phyla using a Step One Real Time PCR machine [Applied Biosystems (ABI) Life Technologies] with SYBR Green PCR master mix (ABI) and bacteria-specific primers (Table 1) in a 20-μl final volume. Cycle of threshold values (C_T) were correlated to the calculated bacteria number using standard curves from the pure bacterial cultures as previously described (4, 20, 51, 56). Total bacteria were calculated using a universal bacterial primer that recognizes all bacterial groups and represents the total stool microbiota.

Table 1.

Quantitative PCR primer sequences for of total bacteria, bacterial phyla, and C. difficile

Type	Bacteria	Forward	Reverse	Reference
Total	Universal (total bacteria)	ACTCCTACGGGAGGCAGCAG	ATTACCGCGGCTGCTGG	(4, 23, 26)
Phyla	Bacteriodetes	GGCGACCGGCGCACGGG	GRCCTTCCTCTCAGAACCC	(26)
Phyla	Firmicutes	GGAGYATGTGGTTTAATTCGAAGCA	AGCTGACGACAACCATGCAC	(23)
Phyla	Actinobacteria	CGCGGCCTATCAGCTTGTTG	ATTACCGCGGCTGCTGG	(23)
Phyla	α-Proteobacteria	ACTCCTACGGGAGGCAGCAG	TCTACGRATTTCACCYCTAC	(23)
Phyla	β-Proteobacteria	CCGCACAGTTGGCGAGATGA	CGACAGTTATGACGCCCTCC	(23)
Phyla	y-Proteobacteria	GAGTTTGATCATGGCTCA	GTATTACCGCGGCTGCTG	(39)
Class	Clostridium coccoides cluster XIVa	ACTCCTACGGGAGGCAGC	GCTTCTTAGTCAGGTACCGTCAT	(57)
Class	Clostridium leptum cluster IV	GTTGACAAAACGGAGGAAGG	GACGGGCGGTGTGTACAA	(57)
Species	Clostridium difficile	TTGAGCGATTTACTTCGGTAAAGA	CCATCCTGTACTGGCTCACCT	(65)

Quantitative real time-PCR amplification of mRNA.

To examine NHE3 mRNA level, total RNA was extracted from HIOs with Trizol reagent (Invitrogen, Life Technologies) according to the manufacturer's instructions. Briefly, the Matrigel surrounding the HIOs was removed by the addition of ice-cold PBS and 400 μl of Trizol were added to the HIOs and homogenized. RNA was extracted by the addition of chloroform and reverse transcription was performed using 50 μg/ml oligo(dT) 20 primer and SuperScript reverse transcriptase (Invitrogen) according to the manufacturer's instructions. Amplification reactions of NHE3 mRNA were performed using SYBR Green PCR master mix (ABI) on a Step One Real Time PCR Machine (ABI). The following gene specific qRT-PCR primers derived from previous literature were used: human NHE3 forward 5′-GAGCTGAACCTGAAGGATGC-3′, NHE3 reverse 5′-AGCTTGGTCGACTTGAAGGA-3′, human GAPDH forward 5′-TGCACCACCAACTGCTTAGC-3′, and GAPDH reverse 5′-GGCATGGACTGTGGTCATGAG-3′ (53). Data are reported as the ΔΔC_T using GAPDH as the standard. Differences in mRNA expression were determined by qRT-PCR and expressed as the ΔΔC_T relative fold difference.

Ion and pH measurements.

Stool fluid ion composition was analyzed by flame photometry and chloridometry. Briefly, 0.3 g of human stool/liquid were added to tubes and 300 μl of double deionized-water were added and vortexed thoroughly. The samples were centrifuged at 3,000 rpm for 10 min at 4°C to pellet solids, and the supernatant Na⁺ and K⁺ concentrations were determined using a digital Flame photometer (Single-Channel Digital Flame Photometer Model 02655–10; Cole-Parmer Instrument). Cl⁻ ion concentrations were determined by a digital Chloridometer (Model 4425100; Labconco). All values were normalized to weight. pH measurements were performed electrochemically via a pH meter (Orion Model 720A; Thermo Fisher Scientific).

Statistics.

Data are presented as means ± SE. Comparisons between groups were made with either one- or two-way ANOVA, and the Holm-Sidak post hoc (parametric) test was used to determine significance between pairwise comparisons using SigmaPlot (Systat Software, San Jose, CA). A P < 0.05 value was considered significant while n is the number of experiments performed.

RESULTS

NHE3 has been shown to be essential for intestinal absorption of Na⁺ and, therefore, water (25, 60). Work in cell lines (LLC-PK₁: pig kidney; OK: opossum kidney; and BeWo: human placenta) has demonstrated that C. difficile toxin B inhibits NHE3 by dephosphorylation and redistribution of ezrin, which normally anchors NHE3 to the cytoskeleton, resulting in the loss of NHE3 from the apical membrane (29). To determine if NHE3 was inhibited in CDI patients, intestinal architecture was examined by hematoxylin and eosin staining (Fig. 1A) and NHE3 expression was examined by immunofluorescence (Fig. 1B). Colonic biopsy specimens demonstrate normal healthy crypts in healthy subjects (Fig. 1A). Consistent with CDI pathology, colon segments demonstrated pronounced thickening of the colonic wall (black arrows) and pseudomembranes (grey arrows) as previously described (73). To confirm the presence of C. difficile in patients with CDI, slides were stained with an anti-C. difficile antibody. As expected, healthy colonic tissue and adjacent areas did not contain any C. difficile. In patients with CDI, C. difficile was found primarily in the expelled mucus layer (89 ± 3%) and occasionally in the crypts close to the host epithelium (11 ± 2%; n = 5; Fig. 1A).

Fig. 1.

Clostridium difficile infection (CDI) specimens exhibit altered intestinal structure and decreased apical Na⁺/H⁺ exchanger 3 (NHE3) expression. A: hematoxylin and eosin stains of healthy and CDI patient biopsies or surgical resections demonstrate that these CDI patients have regions of pseudomembranes (composed of inflammatory cells, necrotic epithelium, and mucus) (grey arrows) and areas of thickened intestine wall (black arrows). Scale bar = 500 μM. The presence of C. difficile was confirmed with an anti-C. difficile antibody (purple). Healthy tissue contained no C. difficile stain, while CDI specimens contained C. difficile at the level of the mucus (89 ± 3%) and epithelium (11 ± 2%) (n = 5). B: confocal images from healthy and CDI patient biopsies or surgical resections depicting NHE3 (red) and nuclei (blue) stained with Hoechst. Scale bar = 100 μM. Representative micrographs of observations from n = 5 healthy and CDI patient specimens. NHE3 was found to varying degrees in CDI patients, representing a 48% decrease in NHE3 expression compared with healthy colon (arrows). C: semiquantitative analysis of surface NHE3 expression in healthy colon, CDI biopsy nondiseased adjacent tissue, and CDI biopsy diseased tissue. Data presented as relative fluorescence normalized healthy NHE3 expression. *P < 0.005, two-way ANOVA; n = 5.

In healthy subjects, NHE3 is located along the apical membrane of absorptive enterocytes (Fig. 1B). In contrast, CDI patient biopsy specimens demonstrated varying levels of NHE3. In CDI biopsy and surgical resections, there are areas with intact NHE3 and areas with decreased or no NHE3; together CDI samples demonstrate a 48% decrease in NHE3 compared with healthy colon samples (Fig. 1C). ImageJ analysis of C. difficile and NHE3 immunofluorescence revealed decreased NHE3 expression correlated with C. difficile presence (P = 0.042). This is the first study that demonstrates that NHE3 is inhibited in patients infected with C. difficile.

Mice lacking NHE3 have increased intestinal [Na⁺] and an alkaline pH (20, 25, 60). We predicted that loss of NHE3 activity in patients would likewise result in increased [Na⁺] and an altered intestinal pH. To determine if CDI patients had an altered ion composition, stool from healthy subjects with or without diarrhea and antibiotics and CDI patients were examined by flame photometry and chloridometry (Fig. 2, A–D). Patients with CDI had higher [Na⁺] (P = 0.005) (Fig. 2A), and [Cl⁻] (P < 0.001; Fig. 2C) with no change in [K⁺] (P = 0.106; Fig. 2B) (two-way ANOVA; Fig. 2). To account for differences in stool composition and antibiotic use, patients with normal stool and diarrhea with or without antibiotics were included for analysis. For [Na⁺], an interaction existed between the presence of diarrhea and antibiotics (P = 0.045; Fig. 2A). A significant difference was found between healthy and diarrhea patients without antibiotics (P < 0.001) and with antibiotics (P = 0.039). An interaction was also observed for [Cl⁻] (P = 0.002) between healthy and diarrhea patients without antibiotics (P = 0.025) and with antibiotics (P = 0.015; Fig. 2C). Interestingly a difference was observed in [K⁺] between patients with or without antibiotics (P = 0.027; Fig. 2B).

Fig. 2.

Stool from CDI patients has increased Na⁺ and a more alkaline pH compared with healthy subjects with or without antibiotics. Sodium concentration ([Na⁺]) and potassium concentration ([K⁺]) as determined by flame photometry for patients with normal stool (n = 12) and diarrhea without antibiotics (n = 8), for patients with normal stool on antibiotics (n = 5) and diarrhea on antibiotics (n = 7), and for patients with recurrent CDI (n = 12). A: Na⁺ was significantly increased in CDI compared with healthy patient stool (***P = 0.005). Na⁺ was also increased in patients with diarrhea (*P < 0.001) compared with normal stool, between patients on antibiotics vs. no antibiotics (**P = 0.010). B: no changes were observed in [K⁺] between CDI patients and healthy subjects, but differences were observed between patients with or without antibiotics (*P = 0.027). C: chloride concentration ([Cl⁻]) as determined by chloridometry was significantly increased in CDI compared with healthy patient stool samples (***P < 0.001). Differences were also observed between healthy and diarrhea patients without antibiotics (*P = 0.025) and with antibiotics (**P = 0.015). D: non-Cl anion gap was calculated with the following equation: [Na⁺] + [K⁺] − [Cl⁻]. Differences were between patients with or with antibiotics (**P = 0.012), patients with and without diarrhea (*P = 0.032), and patients with and without CDI (***P < 0.001). E: stool pH was determine electrochemically and CDI patient stool was more alkaline compared with all other groups (**P = 0.003). Differences in pH were also found between patients with and without antibiotics (*P = 0.026). *P < 0.05, two-way ANOVA, Holm-Sidak.

Non-Cl anion gap calculations ([Na⁺] + [K⁺] − [Cl⁻]) demonstrated an increase in bulk non-Cl⁻ anions (Fig. 2D) for between patients with or with antibiotics (P = 0.012), patients with and without diarrhea (P = 0.032), and patients with and without CDI (P < 0.001). This suggests increase in bicarbonate or short chain fatty acids. In addition, CDI patients had a more alkaline stool (pH 6.9 ± 0.3) compared with healthy subjects (pH 6.0 ± 0.1) (P = 0.003; Fig. 2E). Differences in pH were also found between patients with and without antibiotics (P = 0.026). These data indicate that changes in ion composition and pH occur with antibiotic use, but infection with C. difficile alters these parameters further.

The altered intestinal environment observed in recurrent CDI patients (increased [Na⁺] and more alkaline pH) is similar to our observations in NHE3 null mice and may promote the growth of a different subset of gut microbiota. Mice lacking NHE3 exhibit an altered gut microbiota with increased Bacteroidetes and decreased Firmicutes phyla (20). To address whether patients with recurrent CDI exhibit a similar profile, stool microbiota extracted from nine healthy and nine CDI patients were examined by qPCR. Total stool bacteria density remained unchanged between the groups (Fig. 3A). In CDI patients C. difficile represented < 2% of total bacteria (data not shown). In healthy subjects, the gut bacterial phylum Firmicutes constituted the most abundant group, followed by Bacteroidetes (Fig. 3B), consistent with other reports (13). In contrast, patients with CDI had increased Bacteroidetes and decreased Firmicutes (P < 0.001, two-way ANOVA; Fig. 3C). CDI patients also had increased αβγ-Proteobacteria (P = 0.02) that may result from antibiotic use, since antibiotics have been shown to increase Proteobacteria titers (1, 14, 16, 30, 35, 43, 62). To determine if resident Clostridial groups (Firmicutes phylum) were changed in CDI, C. coccoides Cluster XIVa and C. leptum Cluster IV titers were examined (Fig. 3D) and both were decreased compared with healthy subjects. In a healthy patient, C. coccoides and C. leptum account in total for ∼50% of the total stool bacteria (28, 38, 63) and decreases in these groups represent a significant decrease in the Firmicutes phylum. These observations point to the proclivity of C. difficile in generating an altered intestinal environment.

Fig. 3.

CDI patients have an altered gut microbiota compared with healthy subjects. quantitative PCR was performed with total bacteria and bacterial phyla specific primers. A: no differences are observed in total bacteria between patients with healthy subjects (black bar) and CDI patients (white bar). B: healthy and CDI stool microbiota bacterial phyla alterations in Firmicutes, Bacteroidetes, and Proteobacteria. C: CDI patients also have decreased resident Firmicutes members C. coccoides (Δ 21%) and C. leptum (Δ 19%) levels. CFU, colony-forming units. *P < 0.05, one-way ANOVA, Holm-Sidak; n = 9.

To address whether C. difficile prefers the environmental conditions mediated by inhibition of NHE3, C. difficile ATTC BAA-1870 was grown in vitro anaerobically in TYG media containing various [Na⁺] (8–106 mM Na⁺). Growth was examined by OD_560nm and CFU enumeration at 4 (early exponential phase)-, 12 (early stationary phase)-, 24-, or 48 (stationary phase)-h time points as previously described (75). Growth patterns were found to be the same for all time points, and the data are represented as CFU at the 24-h time point. C. difficile was found to grow optimally at >16 mM [Na⁺] (media pH 6.0), conditions observed in stool of CDI patients (Fig. 4A). This experiment was repeated using KCl or CsCl replacement, and C. difficile again was noted to grow more efficiently at higher [Na⁺] (data not shown). C. difficile was also grown in vitro at [Na⁺] and pH values designed to mimic human stool (healthy: Na⁺ 8 mM, pH 6.0; CDI: Na⁺ 75 mM pH 7.0, refer to Fig. 2). As shown in Fig. 4B, C. difficile also grew better at pH 7.0 vs. pH 6.0 (P = 0.003) at both [Na⁺], indicating that C. difficile is also influenced by pH. The resident Clostridial members C. butryicum (Cluster I), Blautia producta (C. coccoides Cluster XIVa), and Faecalibacterium prausnitzii (C. leptum Cluster IV) were also grown in TYG media in high and low Na⁺ at pH 6.0 and 7.0 to determine if all Clostridial groups preferred the similar environment conditions as C. difficile (Fig. 5). C. butyricum is a resident bacteria that has been used as a probiotic (58, 61) and has been shown to prevent experimental colitis via an IL-10-dependent mechanism (27). C. butyricum grew well at lower [Na⁺], but proliferation significantly dropped at high [Na⁺] (Fig. 5A). B. productus and F. prausnitzii had similar growth preferences to C. butryicum with decreased growth at high [Na⁺] (Fig. 5, B and C). These data demonstrate that C. difficile is distinct in its preference for a high [Na⁺], alkaline pH environment, adding credence to the hypothesis that C. difficile prefers an altered intestinal environment that may be caused by the loss of NHE3 function. Of note, at low [Na⁺], C. butyricum proliferated to much higher levels compared with C. difficile, even at the high [Na⁺]. This suggests that under healthy conditions, C. butyricum would be able to outcompete C. difficile for a given niche.

Fig. 4.

In vitro growth of C. difficile ATTC BAA-1870 in varying Na⁺ and pH conditions. [Na⁺] ranges for healthy and CDI stool are displayed as bars along the x-axis. A: growth (CFU) of C. difficile in tryptone-yeast extract-glucose (TYG) broth at varying [Na⁺], which mimics those seen in vivo for healthy and CDI stool at 24 h. C. difficile grew optimally >16 mM Na⁺ (pH 6.0), which is above the in vivo concentration of 3 mM Na⁺ for healthy patient stool. B: growth of C. difficile in TYG broth at varying pH, which mimics that seen in vivo for healthy and CDI stool. Growth was determined at 8 mM Na⁺ (●) mimicking healthy stool and 75 mM Na⁺ (○) mimicking CDI stool. Significant difference between 8 mM Na⁺ and 75 mM Na⁺ at pH 5.5 (P = 0.001), 6.0 (P < 0.001), and 6.3 (P = 0.001). In addition, there is a significant difference between growth at pH 5.5 vs. 6.0–6.5 (P < 0.001) and pH 6.0–6.5 and pH 7.0 (P < 0.001) for both 8 mM Na⁺ and 75 mM Na⁺. *P < 0.05, two-way ANOVA, Holm-Sidak.

Fig. 5.

In vitro growth of Clostridial members in a range of Na⁺ and pH conditions. Growth (CFU) of Blautia producta (C. coccoides Cluster XIVa; A), Faecalibacterium prausnitzii (C. leptum Cluster IV; B), and C. butryicum (Cluster I; C) in TYG broth at varying [Na⁺], which mimics that seen in vivo for healthy (pH 6, ●) and CDI stool (pH 7, ○) at 24 h. All Clostridial groups grew optimally from 7–40 mM Na⁺ (pH 6.0 and 7.0) *P < 0.05, two-way ANOVA, Holm-Sidak.

Although CDI patients demonstrate decreased NHE3 expression, it could be argued that a number of different bacterial groups could be responsible for changes in NHE3 levels. We have previously used intestinal organoids to address microbial-host interaction (20). To determine if C. difficile alone was sufficient to decrease NHE3, HIOs were used. HIOs have been shown to mimic native tissue: the cellular diversity and architecture are similar to tissue; HIOs contain all the intestine cell lineages; secretory and absorptive functions are present; and HIOs also contain a significant degree of epithelial and mesenchymal complexity and secrete mucus (77). To confirm that decreases in NHE3 were C. difficile specific, HIOs were injected with C. difficile, C butryicum, and stool supernatant from healthy and CDI patients. mRNA levels of NHE3 (Fig. 6A) demonstrate that C. butryicum and healthy stool does not inhibit NHE3 expression. However, injection of C. difficile and CDI stool supernatant resulted in a substantial decrease in NHE3 mRNA compared with broth-injected (control) organoids. This inhibition was also observed at the protein level (Fig. 6, B and C), demonstrating that C. difficile is sufficient for NHE3 inhibition in patients with CDI. Taken together, these data indicate that C. difficile is capable of decreasing NHE3 expression in vivo and in vitro. Stool from patients with CDI exhibit increased Na⁺ and Cl⁻ and are more alkaline in pH compared with healthy subjects and patients on antibiotics. Stool gut microbiota from patients with CDI are higher in Bacteroidetes and lower in Firmicutes compared with healthy subjects, with decreased resident Clostridial members C. coccoides and C. leptum. In vitro high Na⁺ and pH conditions results in C. difficile proliferation and the suppression of resident Clostridial members. Injection of C. difficile into HIOs confirms that C. difficile alone is capable of eliciting changes in NHE3 at the level of protein and mRNA.

Fig. 6.

Human intestinal organoids (HIOs) grown in 3-dimensional culture microinjected with bacterial or stool supernatant. A: NHE3 mRNA levels indicate decreased expression in HIOs injected with CDI stool and C. difficile. No changes were observed between healthy stool or C. butryicum culture. Left: organoid culture in Matrigel with injection needle. B: NHE3 protein expression as determined by immunofluorescence is decreased in HIOs injected with CDI stool and C. difficile compared with control, healthy, and C. butryicum infected HIOs (white asterisk designates lumen). Left: depicts widefield view of HIO injection. C: semiquantitative analysis of NHE3 florescence. *P < 0.05, two-Way ANOVA, Holm-Sidak; n = 6–9 organoids.

DISCUSSION

C. difficile represents an ever increasing public concern as the major cause of antibiotic-induced diarrhea and colitis. The incidence of CDI has increased in the last decade, and, with the emergence of more virulent strains, C. difficile will likely persist as a major health concern. Herein, we have demonstrated several novel aspects of CDI including 1) CDI patients exhibit decreased NHE3 expression in the apical membrane of intestinal enterocytes and higher [Na⁺] and alkaline stool pH compared with healthy subjects. The altered gut intestinal environment correlates with changes in the dominant bacterial phyla, Firmicutes and Bacteroidetes, with decreased C. coccoides and C. leptum; 2) C. difficile has increased proliferation at Na⁺ concentrations >16 mM and at more alkaline pH level in vitro, a pattern of proliferation that is not observed in resident C. butryicum, B. producta, or F. prausnitzii; and 3) C. difficile alone and in combination with a complex microbiota (CDI stool) are capable of decreasing NHE3 expression in HIOs. These new findings shed light on several novel aspects of the C. difficile colonization phase. This study represents the first in vivo analysis of NHE3 inhibition in response to C. difficile infection. Targeted disruption of the normal intestinal environment via regulation of ion transport may help explain both the diarrhea phenotype and how C. difficile maintains a competitive advantage.

We propose that in healthy individuals the luminal and mucosa-associated gut microbiota compete for a C. difficile niche (see model in Fig. 7). In an antibiotic microbiota-depleted environment, C. difficile spores germinate and vegetative C. difficile likely use intestinal nutrients (such as cleaved mucus oligosaccharides) to enter a colonization phase. After a colonization phase, C. difficile enters a virulence phase and produces toxins that inhibit NHE3. C. difficile toxin inhibition of NHE3 alters the intestinal environment producing a high [Na⁺], and a more alkaline fluid, which enhances C. difficile proliferation and inhibits competitive Clostridial groups proliferation. This altered intestinal environment further shapes the reemerging gut microbiota (which is restructuring after antibiotic use), perhaps so that noninhibitory Bacteroidetes members proliferate. Altered gut microbiota may also play a role in further shaping the intestinal environment, making it more favorable for C. difficile colonization. C. difficile toxin B inhibition of NHE3 was demonstrated in cells lines (29), but this is the first study that has demonstrated that inhibition of NHE3 occurs in infected human patients. Loss of NHE3 in mice appears to mimic the effects of C. difficile toxin production in humans as NHE3^−/− mice have higher [Na⁺], alkaline intestinal fluid, and a distal colon microbiota that are higher in Bacteroidetes and lower in Firmicutes (20). It should be noted that NHE3^−/− mice have higher Bacteroidetes and lower Firmicutes in both the luminal and mucosa-associated bacterial populations, with the mucosa-associated bacterial population being more dramatic than the luminal population. If the mucosa-associated bacteria of CDI patients likewise reflects the luminal bacterial composition, this indicates that this altered composition represents a noncompetitive population to C. difficile.

Fig. 7.

Working model of C. difficile inhibition of NHE3. Under normal conditions, Firmicutes is the dominant member of the human colon at the level of the luminal (our data) and mucosa-associated bacterial populations (17). Antibiotic use alters the gut microbiota and allows for the germination of C. difficile spores. Vegetative C. difficile use fuel sources to proliferate and once in virulence phase produces toxins that inhibit NHE3. Loss of functional NHE3 results in a high in Na⁺ and more alkaline pH luminal environment, which shapes the gut microbiota to be increased in Bacteroidetes and decreased in Firmicutes. It remains unknown what changes are occurring in the mucosa-associated bacteria under these conditions. High Na⁺ and more alkaline pH favors C. difficile proliferation and prevents resident Clostridial groups from outcompeting C. difficile.

NHE3^−/− mice also exhibited increased [Na⁺], alkaline fluid, and Bacteroidetes in the small intestine (terminal ileum) in addition to changes in the colon. This suggests that the altered intestinal environment, due in part to the loss of NHE3, may occur upstream as well as in the colon. Although C. difficile studies have focused on the colon, C. difficile infection has also been reported in the small intestine (47, 70, 76) and can cause small bowel disease (66, 71, 72). These studies suggest that C. difficile infection is not localized solely to the colon and may provide keys areas for the initial pathogenesis of C. difficile. Knowledge of C. difficile colonizing in the intestine (either small or large intestine) is critical for developing better therapies against CDI.

It has been speculated by some that the diarrhea observed in CDI patients is the direct result of epithelial integrity or secretory diarrhea by the host to remove the pathogen. Our work has demonstrated that patients with CDI have a dramatic increase in stool [Na⁺] and a moderate increase in [Cl⁻]. This suggests that loss of Na⁺ absorption, and concomitant water absorption, in combination with increased [Cl⁻], may contribute to in the diarrhea observed in CDI patients. A caveat to such as study is the fact that it is unfeasible to collect samples before patients become infected with C. difficile, so it remains to be determined the extent of alterations in ion composition and pH in each patient before the acquisition of CDI. However, these data demonstrate that [Na⁺] and [Cl⁻] in CDI patients are significantly different from those of patients with or without diarrhea on antibiotics, indicating that the large alterations in Na⁺ and Cl⁻ content in CDI patients are likely due to the presence of C. difficile.

Hayashi et al. (29) demonstrated in placental and renal cell lines that exposure to C. difficile toxin B resulted in decreased expression of NHE3 at the apical membrane and translocation to a subapical endosome. This redistribution in NHE3 location was suggested to contribute to loss of NHE3 activity. However, our work with the HIOs has shown that injection with toxin-producing C. difficile decreased NHE3 expression at the level of the protein and mRNA. This indicates that a dual mechanism of NHE3 inhibition may be occurring in patients with CDI. In renal cell lines, NHE3 activity has also been shown to be diminished at the level of apical protein and mRNA by parathyroid hormone (PTH) (6). PTH addition correlated with a significant decrease in promoter −152/+55 activity. This promoter segment contains putative binding sites for Sp1, AP2, and NF-Y, which seem to be essential for NHE3 gene transcription (42). C. difficile toxin, similar to PTH, may also lead to decreased NHE3 promoter activity and thus transcription, but this remains to be determined. C. difficile toxin production has been demonstrated to disrupt intestinal actin cytoskeleton, which is thought to lead to cell death (44). Loss of cell integrity could also contribute to decreased NHE3 in diseased segments of the intestine of CDI patients. However, decreased NHE3 mRNA may point to a selective inhibition of NHE3 to ensure a more favorable environment and reemerging bacterial composition for C. difficile growth.

In vitro C. difficile has optimal growth at higher sodium and more alkaline pH compared with healthy subjects. This might represent a biphasic event where C. difficile is capable of using [Na⁺] until a more alkaline pH is obtained. Patients on antibiotics (either with or without diarrhea) have elevated [Na⁺] compared with patients not on antibiotics (Fig. 2). Since patients who are on antibiotics and have diarrhea have a lower pH compared with CDI patients, C. difficile may use [Na⁺] until the pH becomes more alkaline. Once the pH is alkaline, in vitro data suggest that C. difficile is insensitive to [Na⁺]. Additional experiments are likely needed to identify the exact conditions that stimulate C. difficile growth in vivo.

Ion transport now represents a new route for combating CDI. For example, were NHE3 to be upregulated, this may provide a means to reestablish the normal intestinal environment and thus shift the microbiota toward one that is considered normal. The effects of normal commensal microbiota on NHE3 expression and function may prove valuable in this regard. Lactobacillus has been used as a probiotic treatment for CDI (7, 45, 50) and Lactobacillus acidophilus has been shown to upregulate NHE3 (64). In addition, Lactobacillus has been shown to produce lactic acid that inhibits C. difficile growth (46). Since normal gut microbiota have been shown to out compete and inhibit C. difficile growth, control of ion transport can provide a novel therapeutic for CDI.

GRANTS

This work was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-079979 (to R. T. Worrell).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: M.A.E. and R.T.W. conception and design of research; M.A.E., K.A.E., M.B.Y., and B.R.Y. performed experiments; M.A.E., K.A.E., M.B.Y., J.W., and R.T.W. analyzed data; M.A.E., M.B.Y., J.W., B.D., B.R.Y., and R.T.W. interpreted results of experiments; M.A.E. and R.T.W. prepared figures; M.A.E. and R.T.W. drafted manuscript; M.A.E., M.B.Y., J.W., D.J.H., B.D., B.R.Y., and R.T.W. edited and revised manuscript.