Endotoxin Acts Synergistically With Clostridioides difficile Toxin B to Increase Interleukin 1β Production: A Potential Role for the Intestinal Biome in Modifying the Severity of C. difficile Colitis

Abstract

Background

Inflammation is a crucial driver of host damage in patients with Clostridioides difficile colitis. We examined the potential for the intestinal microbiome to modify inflammation in patients with C. difficile colitis via the effects of gut-derived endotoxin on cytokine production.

Methods

Endotoxin from Escherichia coli and Pseudomonas aeruginosa as well as stool-derived endotoxin were tested for their ability to enhance interleukin 1β (IL-1β) and tumor necrosis factor alpha (TNF-α) production by toxin B–stimulated peripheral blood mononuclear cells. Inflammasome and Toll-like receptor 4 (TLR4) blocking studies were done to discern the importance of these pathways, while metagenomic studies were done to characterize predominant organisms from stool samples.

Results

Endotoxin significantly enhanced the ability of C. difficile toxin B to promote IL-1β production but not TNF-α. The magnitude of this effect varied by endotoxin type and was dependent on combined inflammasome and TLR4 activation. Stool-derived endotoxin exhibited a similar synergistic effect on IL-1β production with less synergy observed for stools that contained a high proportion of γ-proteobacteria.

Conclusions

The ability of endotoxin to enhance IL-1β production highlights a manner by which the microbiome can modify inflammation and severity of C. difficile disease. This information may be useful in devising new therapies for severe C. difficile colitis.

Clostridioides difficile colitis is the leading cause of hospital-acquired diarrhea. Affected patients exhibit a range of disease states from asymptomatic colonization to a rapidly progressive condition known as fulminant colitis, which occurs in 1%–10% of patients [1–3]. Pathogen-driven inflammation plays a central role in the clinical manifestations of C. difficile colitis, especially for fulminant disease. The importance of inflammation in this process is highlighted by the elevated levels of inflammatory cytokines and chemokines seen in affected patients, including high levels of interleukin 1β (IL-1β) in the stool and serum [4–7].

The 2 toxins produced by C. difficile (toxins A and B) are essential to the pathogenesis of C. difficile disease (reviewed in [8, 9]). Besides their direct effects on the intestinal epithelium, these toxins stimulate inflammatory cytokine production by host cells [10–13]. Increased toxin production has been hypothesized to result in more severe disease. For example, the NAP1/B1/027 strain, which has a deletion in the negative regulator TcdC gene resulting in increased toxin production, has been linked to outbreaks of severe disease [14, 15].

Changes in the microbiome, including a reduction in biodiversity, play a role in the development of C. difficile disease [16–18]. Less well understood is the contribution of the intestinal microbiome to C. difficile disease severity. Endotoxin (also known as lipopolysaccharide [LPS]) is an important cell membrane constituent of enteric gram negatives. It is present in large quantities in the colon and spontaneously released by organisms. This release can be further augmented by antibiotic therapy [19, 20]. LPS activates Toll-like receptor 4 (TLR4) signaling and promotes inflammation via enhanced cytokine and chemokine production. In this study, we explored the potential for endotoxin to modify the inflammatory response to toxin B by modifying IL-1β and tumor necrosis factor alpha (TNF-α) secretion.

MATERIALS AND METHODS

Cell Culture Studies

Using Ficoll-Paque (GE Healthcare, Uppsala, Sweden), peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation of leukocyte-enriched blood samples obtained from the New York Blood Bank. Adherent cells were exposed to C. difficile toxin B (List Labs, Campbell, California) with or without commercially obtained LPS (Sigma, St Louis, Missouri). In some experiments, rabbit polyclonal serum against C. difficile toxin B (Biorbyt, Cambridge) or MCC950 (InvivoGen, San Diego, California), a NOD-like receptor pyrin domain-containing protein 3 inflammasome inhibitor, was added to the cell cultures. In other experiments, antibiotics that bind LPS, colistin (Par Pharmaceutical, Chestnut Ridge, New York), or reagents that block the TLR4 pathway, TAK-242 (MilliporeSigma, Burlington, Massachusetts), were used. Supernatants were removed at 16 hours and tested for IL-1β and TNF levels by enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, Minnesota). All in vitro studies were conducted on 2 separate occasions.

Stool Samples

Stool samples from children were collected in sterile containers upon admission to the hospital, sent to the microbiology laboratory, and refrigerated before further processing. Within 24 hours, samples were aliquoted and stored at –80 °C. A convenience sample of 11 stools was chosen for study.

Crude LPS Stool Extracts

Frozen samples were allowed to thaw, weighed, and placed in pyrogen-free test tubes. LPS-free normal saline was added to the tubes to yield a ratio of 1:4 based on weight. Glass beads that were heat baked (250°C for 30 minutes) were added to the mixture, which was then vortexed for 10 minutes and centrifuged at 9500g for 10 minutes. The supernatant was removed, heated to 100°C for 30 minutes, and diluted again before being passed through a nonpyrogenic 0.2-µm syringe filter (Corning Inc, Corning, New York). Before incubation with cells, all samples were diluted 1:400 in LPS-free normal saline.

To ensure our findings were not related to contaminating proteins or nucleic acids, LPS was extracted from 8 stools using the hot phenol-water method with some modifications [21]. For these studies, stool samples weighing 50–100 mg were treated serially with RNase (40 μg/mL) and DNase I (12.5 U/mL) (both Thermo Fisher Scientific, Waltham, Massachusetts) and 1 µL/mL of 20% magnesium sulfate for 1 hour at 37°C. This was followed by incubation with proteinase K (100 μg/mL) (Abcam, Branford, Connecticut) overnight at 37°C. Samples were then treated with tri-saturated phenol (65°C–70°C) (Fisher Scientific, Pittsburgh, Pennsylvania) with intermittent shaking over 15 minutes. Supernatants were harvested and precipitated with 10 volumes of 100% ethanol and 0.5 M sodium acetate. Precipitates were dissolved in endotoxin-free water and dialyzed against endotoxin-free water followed by lyophilization. Specimens were tested for protein contamination using the bicinchoninic acid (BCA) (Thermo Fisher Scientific) and nucleic acid contamination by running samples on 1% agarose gel followed by ethidium bromide staining.

LPS Determinations

LPS levels of crude stool extracts were determined using the Toxin Sensor Chromogenic LAL LPS Assay (GenScript, Piscataway, New Jersey) according to the manufacturer’s instructions. Prior to testing, all samples were diluted in LPS-free normal saline (1:125) and the pH of sample was determined to ensure it ranged between 6 and 8.

Biome Studies

Stool samples stored at –80°C were used for sequencing. Metagenomic DNA was extracted from stool samples using a PowerSoil DNA Isolation Kit [22]. The Weinstock Laboratory at Jackson Laboratories (Farmington, Connecticut) produced sequencing data from the metagenomic DNA. The V1 through V3 hypervariable regions (V1–V3) of 16S ribosomal RNA genes were amplified from the metagenomic DNA using primers 27F and 534R. The uniquely barcoded amplicons from multiple samples were pooled and sequenced on the Illumina MiSeq sequencing platform using a V3 2 × 300 sequencing protocol. Illumina’s software was used for initial processing of all the raw sequencing data. One mismatch in primer and zero mismatch in barcodes were applied to assign read pairs to the appropriate sample within a pool of samples. Barcode and primers were removed from the reads. Reads were further processed by removing primers using Trimmomatic version 0.32 with the commands HEADCROP:20, TRAILING:10, MINLEN:100 [23] joining paired-end sequences using the FLASH algorithm [24]. Chimeric sequences were removed using UChime [25]. The cleaned, assembled amplicons were the input data for analysis. Institutional review board approval from the Albert Einstein College of Medicine was obtained for these studies. Parts of the data presented here have previously been included in a manuscript detailing the intestinal biome of hospitalized children [26].

Metagenomic Analysis

We used the Qiime2 (version 2019.4) [27] software package to calculate the relative abundances of microbes in each patient sample. We used the Qiime2 implementation of DADA22 for sequence quality control and sequence feature table construction. With the resulting sequence feature tables, we performed taxonomic classification using the Qiime [28] maintained 85% clustered version of the Greengenes [29] database. We classified each full-length sequence feature using the “feature-classifier” tool. The resulting classification was plotted using the “barplot” tool.

Statistical Analysis

All statistics were calculated using Graph Pad Prism (GraphPad, San Diego, California). For numerical data, Shapiro–Wilk test was done to test for normality. For normally distributed data, a t test was used for individual comparisons and analysis of variance when >2 groups were being compared. A Tukey multiple comparisons test was done for subsequent comparisons between individual groups. Simple linear regression was used to determine correlations between endotoxin levels in stool samples and cytokine production. For biome and cytokine comparisons, stool samples were classified as low cytokine (IL-1β or TNF-α) producers based on levels that were ≤25th percentile. Comparisons of the predominant bacteria among these groups were performed using a Fisher exact test.

RESULTS

LPS Acts Synergistically With Toxin B to Increase IL-β Production, but Not TNF-α

Incubation of PBMCs with either C. difficile toxin B or Escherichia coli LPS (ELPS) alone resulted in relatively low IL-1β levels. In contrast, the coincubation of ELPS with toxin B resulted in elevated IL-1β levels over a wide range of LPS concentrations (Figure 1A). The magnitude of this synergistic effect increased with increasing toxin B concentrations (Figure 1C). When these experiments were conducted with Pseudomonas aeruginosa LPS, a similar synergistic effect was observed, though overall, IL-1β levels were lower when compared to ELPS (Figures 1B and 1D). In contrast to IL-1β production, incubation of LPS alone (either from E. coli and P. aeruginosa) resulted in significant TNF-α production (Figure 2A–D), and the addition of toxin B did not increase TNF-α production.

Inhibition of IL-1β Production

To better understand the synergy of LPS and toxin B on IL-1β production, several studies were done in which inhibitors were added to the combination of LPS and toxin B (Figure 3). The addition of the antibiotic colistin, which is known to bind LPS, decreased IL-1β production in a concentration-dependent manner (Figure 3A). Similarly, the TLR4 inhibitor, TAK-242, also inhibited IL-β production, but only at the highest concentration of inhibitor (Figure 3B). Both a polyclonal rabbit serum against toxin B and an NLP3-inflammasome inhibitor also reduced IL-1β production produced by the combination of LPS and toxin B. This effect was lost at the highest toxin B concentrations (Figures 3C and 3D).

Stool Studies

The average age of patients from whom these stools came was 2.9 ± 4.6 years; 4 patients were <1 year of age. Stools came from 8 females and 3 males. One tested positive for C. difficile toxin by polymerase chain reaction, but this likely represented colonization as he was <1 year of age. Of the patients, 1 of 11 (9.1%) had diarrhea while 9 of 11 (82%) were receiving antibiotics. Several patients had underlying diseases including sickle cell disease, cholangitis, cholestasis, malignancy, and ventriculoperitoneal shunt infection.

Stool Experiments

Eleven crude stool LPS extracts were available for IL-1β experiments; enough sample was left for TNF-α studies in 9 samples. Incubation of PBMCs with these extracts resulted in IL-1β expression for all specimens, though there was significant variation between the specimens (Figure 4A). Co-incubation of stool extracts with toxin B consistently elicited higher IL-1β levels when compared to stool extracts alone. Importantly, the magnitude of this effect varied greatly between stool extracts, with an average increase of 23.6-fold (range, 2.5- to 46.8-fold). This synergistic effect was completely abrogated by the addition of colistin to cell cultures (Figure 4B).

Incubation of PBMCs with crude stool LPS extracts resulted in TNF-α production by PBMCs for all specimens. Co-incubation of PBMCs with a combination of stool extracts with toxin B resulted in a significant increase in TNF-α levels relative to stool LPS extracts alone, for most samples. However, the magnitude of this increase was lower when compared with changes in IL-1β production, with a mean increase of 2.4-fold (range, 1.1- to 4.2-fold) (Figure 4C).

Endotoxin concentrations of the 11 stool samples varied from 98.4 to 196.2 EU/mL. There was no correlation between the endotoxin concentrations of individual stool sample and the amount of IL-1β produced by the combination of the sample with toxin (r² = 0.11, P = .32) (not shown).

LPS was obtained by phenol extraction in 8 stool samples. No protein was detected in these samples by BCA assay (limit of detection 50 μg/mL) and no bands were visualized by ethidium bromide staining. Similar to crude extracts, significant synergy for IL-1β production was observed following toxin B coincubation, with an average 41.3-fold increase compared with stool-derived endotoxin alone (range, 8.8- to 132.0-fold increase) (Figure 4D). Levels of IL-1β were similar to those observed following incubation with commercially obtained LPS, but lower than levels observed with crude extract studies (for both stool alone and stool combined with toxin B).

Biome Correlations

The biome and corresponding IL-1β levels produced by the co-incubation of crude stool LPS extracts and toxin B are shown in Figure 5. The bacterial population of specimens, which produced relatively low IL-1β levels, were comprised of a disproportionately high percentage of γ-Proteobacteria and more specifically bacteria from the Aeromonadales order. In this low IL-1β group, 75% (3/4) of samples had Aeromonadales bacteria as the predominant (81%, 72%, 62%) component. A similar pattern was present with for LPS preparations prepared by enzymatic digestion and phenol extraction.

For the 7 remaining samples, the predominant bacterial order was either Bacteroidales (n = 4) or Clostridiales (n = 3) (P = .024). Two of 3 stool specimens, which produced relatively low TNF-α levels when combined with toxin B, contained a majority of bacteria from the Clostridiales order (Figure 6). None of the 6 stool samples from the non–low TNF-α group contained a majority of bacteria from the Clostridiales order; however, this difference was not significant when compared by Fisher exact test (P = .83).

DISCUSSION

Inflammation plays a central role in severe C. difficile colitis. In support of this hypothesis, several studies have shown increased levels of a variety of chemokines and cytokines in association with severe C. difficile disease [5, 7, 30]. Both serum and stool IL-1β levels are elevated in patients with C. difficile colitis [6, 7, 31]. In mice, decreased IL-1β activity as a result of administration of an IL-1 blocking agent or the use of genetically deficient mice was associated with decreased intestinal pathology [32]. Besides its direct proinflammatory activity, increased IL-1β levels in C. difficile disease may exacerbate inflammation through its ability to augment Th17 inflammation [31, 33]. Nonetheless, a protective mechanism for IL-1β in C. difficile colitis has also been proposed, specifically as it relates to preventing translocation of commensal gut bacteria [34].

Findings from our study demonstrate a previously unrecognized mechanism by which the intestinal microbiome can modify the inflammatory response to C. difficile toxin and thereby affect disease severity. We show that endotoxin acts synergistically with toxin B to augment IL-1β secretion, but not TNF-α. This phenomenon resulted in more than a log₁₀ increase in IL-1β levels, with the effect being dependent on the type of endotoxin, and less so on its concentration. Our findings are consistent with a previously reported synergy between C. difficile toxins (A and B) with Salmonella endotoxin on IL-1β production [13].

We hypothesize that this synergy results from the ability of LPS to increase pro–IL-1β production via TLR4 signaling, and of toxin B to activate the inflammasome, thereby promoting pro–IL-1β processing to mature IL-1β. This hypothesis is consistent with previous studies that demonstrate the ability of toxin B to activate the inflammasome [32, 35, 36]. Our hypothesis is further supported by our inhibition studies, which indicate that both TLR and inflammasome activation are necessary for maximum IL-1β production.

The intestinal biome appears to modulate susceptibility to C. difficile infection by several mechanisms including modifications in bile acid metabolism and the production of bacteriocins [17, 37–39]. We found that endotoxin from P. aeruginosa was significantly less effective than endotoxin from E. coli in enhancing IL-1β production when incubated with toxin B. This observation is consistent with the known structural variations in lipid A among different gram-negative organisms and the effects of these differences on TLR4 binding and subsequent inflammatory cytokine secretion (reviewed in [40]). Similarly, we found that stool-derived endotoxin from different individuals demonstrated significant variation in the ability to promote IL-1β production. This effect was present for both crude stool LPS extracts and those prepared by phenol extraction. An overrepresentation of bacterium from the γ-Proteobacteria class, and more specifically from the Aeromonadales order, was found in the samples that produced low IL-1β levels. The high prevalence of Aeromonadales in these samples is unusual but may be accounted for by the high rates of antibiotic usage, underlying diseases, and the relatively young age of this cohort. Additional study is clearly needed to further define the effects of LPS from different bacteria on toxin B–mediated inflammatory responses and its contribution to disease. The extent of synergy is likely to be a dynamic and competitive process, depending on both the predominant bacterial species and the extent of LPS shedding within the colon. We further note that within a given genus, significant variation in the TLR signaling can occur, with some species unable to activate TLR4 signaling (reviewed in [41]) so that distinction at a species level will be important for future studies.

Our findings have important additional potential clinical implications, including the potential utility of endotoxin-binding antibiotics or endotoxin binding resins to help limit IL-1β production and decrease inflammation during C. difficile colitis. This approach might be complicated by the fact that antibiotics like colistin and polymyxin are themselves risk factors for C. difficile disease. Interestingly, successful treatment of severe difficile colitis using hemoperfusion with polymyxin B immobilized fiber has been described [42, 43]. Current recommendations for the treatment of fulminant C. difficile colitis (where surgical intervention is necessary) include the placement of a diverting loop ileostomy with colonic lavage, using vancomycin [44, 45]. For these patients, the use of an endotoxin-binding agent in addition to medical therapy may prove particularly useful, though additional study is clearly indicated.

Antibiotic usage is a known risk factor for the development of C. difficile colitis. Findings from our study imply that antibiotics could also exacerbate colitis by increasing endotoxin release from enteric organisms, thereby increasing inflammation. In this regard, both in vitro and in vivo experiments have demonstrated that antibiotics promote endotoxin release by gram-negative organisms (reviewed in [46]). This effect varies with the type of antibiotic and its mechanism of action.

In summary, our studies highlight a previously unrecognized mechanism by which the intestinal microbiome can alter the inflammatory response to C. difficile disease. The contribution of this mechanism to disease severity remains to be fully demonstrated. We note that our findings are limited by the relatively small number of stool samples we studied as well as the relatively young age of patients who were included. Additional studies examining the biome of patients with differing severity of C. difficile are underway. However, the observed effect is consistent with the known signaling pathways for C. difficile toxin and endotoxin. Furthermore, the high levels of endotoxin in stool, at the site of toxin B activity, highlight the potential biologic relevance of our findings. If confirmed, these findings could have important implications for the treatment of severe C. difficile disease.

Notes

Acknowledgments. We would like to thank Dr George M. Weinstock and his laboratory for the help in DNA sequencing studies.

Potential conflicts of interest. All authors: No reported conflicts of interest.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References