Conditioned supernatants derived from gut anaerobes induce proinflammatory responses from specific antigen-presenting cell subsets

The importance of innate cells as the first-line of defence against microorganisms and their microbial products is undisputable. However, while the impact of gut microbial products on gut innate cells is increasingly understood, the consequences of interaction between these components and distant peripheral innate cells remain unclear.

The use of 16S ribosomal RNA (rRNA) gene sequences has been by far the most common methodology to study gut bacterial phylogeny and taxonomy.17 However, whereas this method can be used for a number of bacterial identification purposes, the procedures that enable the acquisition of secreted bacterial components are mainly confined to the use of bacterial culture systems. Therefore, to address our question, we first cultured luminal contents of the small intestine and colon from C57BL/6 mice under aerobic and anaerobic conditions as described in Methods section,18–22 and conditioned supernatants were used to simulate splenocytes isolated from C57BL/6 mice. Four hours later, antigen-presenting cells (APC) were examined and production of proinflammatory cytokines was assessed in response to microbial products derived from cultivable aerobic and anaerobic supernatants (figure 1). We found that supernatants of anaerobes cultured from the small intestine and colon induced substantially tumour necrosis factor (TNF) α and pro-interleukin (IL) 1β than aerobic supernatants (figure 1A, B, and see online supplementary figure S1). Interestingly, the strong TNF-α response was found to be specific to CD11b+ CD11c+ DCs as neither CD11b− CD11c+ DCs nor CD11b+ CD11c− F4/80+ macrophages expressed more TNF-α in response to anaerobic than aerobic supernatants (figure 1A). Likewise, CD11b+ CD11c+ DCs expressed significantly more pro-IL-1β when exposed to conditioned medium from cultivable anaerobes than aerobes of the colon (see online supplementary figure S1). This suggests that proinflammatory responses generated by cultivable bacterial components are primarily induced by gut anaerobes and selectively produced by CD11b+ CD11c+ myeloid DC subset. Of note, the natural ecosystem of the gut microbiota is very complex and culturing luminal faecal contents does not enable the growth of all gut bacteria in culture systems since most species are not cultivable.17 This is not a pitfall of the experimental strategy but rather a limitation of current methodologies for gut bacterial culture procedures.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Proinflammatory responses are induced by components of gut anaerobes from the Bacteroidetes and produced by myeloid dendritic cells (DCs). (A) Bar graphs showing tumour necrosis factor (TNF) α expression as a % of CD11b+ CD11c+ DCs, CD11b− CD11c+ DCs, and CD11b+ CD11c− F4/80+ (n=4). (B) Dot plots showing expression of TNF-α by CD11b+ CD11c+ DCs and CD11b− CD11c+ DCs. Numbers indicate the % TNF-α cells. (C) Bar graphs showing that Firmicutes dominate the aerobic cultures and a large portion of aerobic Firmicutes were Enterococcus (n=4). (D) Bar graph showing that anaerobic cultures of small intestine and colon bacteria have relatively high proportions of Bacteroidetes (n=4–6). (E–G) Mice were gavaged with PBS (untreated) or a combination of neomycin and metronidazole (Neo/Met) used as prophylaxis for bowel surgery for 24 h, or injected intravenously with a combination of piperacillin and tazobactam (Pip/Taz) used as a chemotherapy for sepsis, every 24 h for three days. (E) Bar graph showing the relative number of cultivable anaerobes in antibiotic-treated mice to antibiotic-free animals (n=5–8). (F) Bar graph showing ratio of Bacteroidetes to Firmicutes in untreated and antibiotic-treated mice (n=4). (G) Bar graph showing the proportion of total gut bacteria represented by Bacteroidetes (n=4). *p<0.05, **p<0.005, ***p<0.0005 using two-tailed Student t test. SSC, side scatter.

Aerobic and anaerobic bacterial cultures from the small intestine and colon are dominated by distinct phyla

The gut of mammals is dominated by two major phylotypes of bacteria: Firmicutes and Bacteroidetes.23 ,24 Maintenance of a prescribed equilibrium between these two major groups is critical, as shifts in the balance between these two groups has been associated with obesity, irritable bowel syndrome, Crohn's disease, ulcerative colitis and colon cancer.24–29 Since products derived from cultivable anaerobes but not aerobes were potent inducer of proinflammatory cytokines by peripheral APCs, we next characterised their composition of Firmicutes and Bacteroidetes. To this end, DNA was isolated from both bacterial sources used to generate our conditioned bacterial supernatants and the relative proportions of Firmicutes versus Bacteroidetes was determined by quantitative PCR (QPCR) using specific primers for 16S rRNA gene.30 We found that Firmicutes dominate the aerobic cultures, with significantly more Firmicutes found in the cultures of aerobes compared with their anaerobic counterparts (figure 1C). To further extend this finding, we next examined the relative proportion of a major genus of Firmicutes in the gut, namely Enterococcus.31 We found that a large portion of aerobic Firmicutes cultured from the small intestine and colon were Enterococcus and that there was significantly more Enterococcus in the aerobic cultures compared with anaerobic counterparts (figure 1C). However, unlike aerobic cultures, anaerobic cultures from small intestine and colon revealed significantly higher proportions of Bacteroidetes spp. (figure 1D). Consistent with different effects in driving proinflammatory responses from peripheral APCs, these results suggest that aerobic and anaerobic cultures from small intestine and colon are dominated by distinct phyla, with aerobic cultures dominated by Firmicutes and anaerobic cultures mostly represented by Bacteroidetes.

Changes of Bacteroidetes:Firmicutes ratios may dictate efficacy of antibiotic treatment for sepsis

Our results suggest that secreted microbial components of cultivable anaerobes derived from the Bacteroidetes are potent inducers of proinflammatory cytokines by peripheral APCs. We next sought to understand whether Bacteroidetes could play a role in the induction of systemic inflammation leading to sepsis and death. To address this question, we first treated mice orally with a combination of neomycin and metronidazole (Neo/Met) used as antibiotic prophylaxis for bowel surgery32–34 and examined by QPCR the efficacy of this prophylactic therapy in clearing Bacteroidetes from the murine intestine (figure 1E, F). As expected, we found that numbers of cultivable anaerobes decreased significantly (∼70%) in antibiotic-treated mice compared with untreated littermates (figure 1E). However, while the total number of cultivable anaerobes decreased in response to antibiotic treatment, the ratio of Bacteroidetes to Firmicutes in the anaerobes that survived this treatment revealed a substantially increased (∼30-fold) proportion of Bacteroidetes (figure 1F). Moreover, when examining the proportion of total gut bacteria represented by Bacteroidetes, we found no decrease as a result of Neo/Met treatment (figure 1G), suggesting that therapy with Neo/Met is efficient at clearing most cultivable anaerobes but not Bacteroidetes spp.

To further extend this observation, we next asked whether a treatment for intra-abdominal infections following bowel surgery could be more effective at clearing gut Bacteroidetes. To this end, mice were treated with piperacillin/tazobactam (Pip/Taz) intravenously for 3 days and efficacy of this treatment in clearing Bacteroidetes from the intestine was examined by QPCR as previously described.30 ,35 Unlike treatment with Neo/Met, treatment with Pip/Taz significantly shifted the ratio of Bacteroidetes:Firmicutes in favour of Firmicutes (figure 1F) and drastically reduced both the number and percentage of Bacteroidetes in the gut to <3% of the total bacterial species (figure 1E, G). Thus, while the preventive treatment with Neo/Met was ineffective in clearing Bacteroidetes from the murine intestine, a more effective treatment with Pip/Taz was highly efficient at Bacteroidetes clearance. Together, these data indicate that efficacy of antibiotic treatment as a control strategy for systemic inflammation leading to sepsis may correlate with clearance of Bacteroidetes from the gut.

Prophylactic treatment increases systemic inflammation and causes more rapid death in response to intestinal barrier damage

Since cultivable anaerobes formed by the Bacteroidetes are potent inducers of proinflammatory responses by peripheral APCs and the prophylactic Neo/Met treatment is ineffective at clearing these anaerobes (figure 1), we hypothesised that the preventive antibiotic treatment (the Nichols-Condon bowel preparation) may be detrimental rather than beneficial at reducing risks of systemic inflammation leading to sepsis and death in conditions of intestinal barrier damage. Caecal ligation and puncture has been proposed as a surgical murine model to study the pathogenesis of sepsis.36 However, because this model utilises the surgical perforation of the ligated caecum which results in immediate and consistent drainage of caecal bacteria into the peritoneal cavity,36 its applicability was unsuitable for this study since our intent is to understand the role of secreted bacterial products rather the bacteria themselves during the initial phase of intestinal barrier injury. Therefore, to test our hypothesis, we used a minimally invasive sepsis mouse model that couples antibiotic treatments with dextran sulfate sodium (DSS) induced intestinal barrier damage.37 DSS is toxic to colonic epithelial cells and typically causes a colitis-like disease in BALB-c mice characterised by weight loss and intestinal barrier injury.38 Briefly, BALB-c mice were treated with drinking water only (control) or drinking water supplemented with DSS continuously for 4–12 days (figure 2A, upper panel). By the end of the fourth day, significant physical and immunological changes were observed between both mouse groups (figure 2B first panel, C). DSS-treated BALB-c mice lost significant body weight compared with untreated controls (figure 2B first panel). In addition, percentages of DC-expressing proinflammatory TNF-α and pro-IL-1β were higher in the spleen of DSS-treated mice compared with untreated animals (figure 2C), suggesting the initiation of systemic inflammatory responses at day 4 post-DSS treatment. Next, we asked whether Nichols-Condon bowel preparation can further accelerate or rather attenuate this immune process. To this end, BALB-c mice were treated orally with Neo/Met; 24 h later, animals were provided with sterile drinking water (control) or drinking water supplemented with DSS continuously for 4 days (figure 2A, middle panel). In parallel experiments, mice were treated with Pip/Taz intravenously every 24 h for 3 days. At the end of the first day, mice were given autoclaved drinking water (control) or drinking water supplemented with DSS continuously for 4 days (figure 2A, lower panel). The goal of these experiments was to compare the outcomes of ineffective clearance of Bacteroidetes (Neo/Met) versus effective clearance of Bacteroidetes (Pip/Taz) in response to acute intestinal barrier damage. Consistent with previous results, the use of the prophylactic therapy induced significant weight loss in response to DSS compared with other animal groups (figure 2B). Plasma concentration of TNF-α is frequently elevated in septic patients and mouse models of sepsis and correlates with severity and outcome.39 ,40 In this context, our data revealed that plasma concentration of TNF-α did not decrease as a result of preventive therapy (figure 2D). Furthermore, mice treated with the prophylactic therapy died more rapidly in response to DSS treatment compared with wild type littermates that received DSS only or DSS in combination with the Pip/Taz therapy (figure 3A). Therefore, Neo/Met treatment used as prophylaxis for sepsis appears to be rather inefficient at reducing systemic inflammation and may even accelerate the initiation of this process resulting in increased mortality in intestinally-injured mice.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Neo/Met therapy is inefficient for reducing systemic inflammation and causes faster death in intestinally-injured mice. (A) Mouse model of intestinal injury was developed using 6–8 week old BALB-c mice provided with drinking water only (control) or supplemented with DSS continuously for the duration of the experiments. Mice were orally injected with the Nichols-Condon bowel preparation which consists of a combination of (Neo/Met, 85 and 145 mg/kg, respectively). Twenty-four hours later, animals were transferred to new cages and provided with autoclaved drinking water (control) or drinking water supplemented with DSS (w/v) continuously for 4–12 days (Middle panel). Other mice were injected intravenously with Pip/Taz (1500 and 187.5 mg/kg, respectively) every 24 h for three days. At the end of the third day, mice were transferred to new cages and were given autoclaved drinking water (control) or drinking water supplemented with DSS continuously for 4–12 days (lower panel). (B) Weight loss of mice treated with PBS or antibiotics only (n=4), or DSS only (n=10), or DSS in combination with antibiotic therapies (n=10) for 4 days. (C) Bar graphs showing TNF-α and PROIL1-β production by splenic DCs in mice untreated (n=4) and DSS-treated animals (n=10). (D) Serum concentrations of TNF-α at day 4 post-DSS treatment. Bar graphs showing relative production of TNF-α in the blood (n=4–10). NS (p>0.01), *p<0.05, **p<0.005, using two-tailed Student t test (DCs, dendritic cells; DSS, dextran sulfate sodium; i.v., intravenously; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin and tazobactam; PROIL1-β, prointerleukin 1β; TNF-α, tumour necrosis factor α).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Spread of cultivable anaerobes occurs only at a later time point of intestinal injury and further accelerates death in Neo/Met-treated mice. (A) Survival of mice that received water or antibiotic only (n=4) or water supplemented with DSS (n=10), or DSS in combination with antibiotic therapies (n=20) for 4 days. (B) Amounts of cultivable anaerobes at day 4 post-DSS treatment in the spleen and liver of antibiotic-free and Neo/Met-treated mice (3 independent experiments of n=3). (C) Survival of mice that were injected i.p. with anaerobic broth (vehicle, n=5) or supernatants of cultivable anaerobes from the colon of untreated (n=10) or Neo/Met-treated (n=10) mice. (D) Amounts of cultivable anaerobes at day 12 post-DSS treatment in the spleen and liver of antibiotic-free and antibiotic-treated mice (3 independent experiments of n=3). (E) Survival of mice that received water or antibiotics only (n=4) or water supplemented with DSS (n=20), or DSS in combination with Neo/Met (n=20) for 12 days. NS (p>0.01), *p<0.05, using two-tailed Student t test (DSS, dextran sulfate sodium; i.p, intraperitoneal; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin/tazobactam). CFU, colony-forming unit; PBS, phosphate buffered saline.

Intraperitoneal spread of gut anaerobe products prepared from colonic contents of Neo/Met-treated mice induces rapid death in host mice

Translocation of bacteria, particularly anaerobes, across the intestinal mucosa into the systemic milieu was proposed as a unifying mechanism to explain early postoperative sepsis.7 ,10 ,41 To determine whether faster moribundus or death induced by the preventive therapy is caused by systemic dissemination of gut-derived microbes, we assessed the number of anaerobes in extraintestinal organs at day 4 post-DSS treatment. Our data showed no significant levels of cultivable anaerobes in the spleen and liver of Neo/Met-treated mice compared with untreated controls (figure 3B). These findings, however, do not exclude the possibility of the development of non-cultivable species in these tissues or the development of bacteraemia in the portal circulation that could have been cleared in the liver prior to the initiation of systemic inflammation. Nevertheless, these outcomes suggest that negative cultures in the spleen and liver are associated with a lack of significant level of cultivable anaerobes in these tissues in the early phase of intestinal injury. Since cultivable anaerobe products were potent inducers of proinflammatory cytokines by peripheral APCs (figure 1); we hypothesised that spread of gut anaerobe products may be an aetiological agent inducing the negative outcomes resulting from the prophylactic therapy. To test this hypothesis, we performed experiments to mimic the spread of gut-derived cultivable anaerobe products in extraintestinal organs (figure 3C). Briefly, healthy wild-type mice were injected intraperitoneally with bacteria-free supernatants prepared from cultivable colonic contents of untreated or Neo/Met-treated animals and outcomes from such treatment were assessed in vivo. Unlike injection of bacterial material from antibiotic-free mice, conditioned supernatants from Neo/Met-pretreated mice induced high mortality after only 48 h of treatment (figure 3C), indicating that early death induced by the prophylactic Neo/Met may be caused by the spread of gut anaerobe products, rather than the microbes themselves, in the initial phase of intestinal barrier damage.

Prophylactic treatment for sepsis induces increased dissemination of gut anaerobes in response to chronic, but not acute, intestinal barrier injury

During early intestinal barrier damage, we observed no detectable levels of cultivable anaerobes in the liver and spleen of untreated and antibiotic-treated mice (figure 3B). However, we could not eliminate the possibility that translocation of gut anaerobes across the intestinal mucosa into the systemic milieu may occur at later time points of intestinal injury. To test this possibility, BALB-c mice were treated with DSS to induce intestinal injury, coupled with Neo/Met therapy, for 12 days. Unlike in the early phase of intestinal injury, analysis at a later time point revealed the detection of gut cultivable anaerobes in the spleen and liver of treated mice (figure 3D), and this phenotype correlated with increased death rate in antibiotic-treated animals (figure 3E).

All together, the prophylactic therapy with Neo/Met followed by acute intestinal barrier damage appears to induce more rapid death in mice that may be caused by the spread of gut anaerobe products, rather than the anaerobes themselves. However, the dissemination of gut anaerobes occurs at later time points of intestinal barrier damage and this may further accelerate death.