The gut microbiome alters immunophenotype and survival from sepsis

Abstract

The microbiome is increasingly implicated in immune regulation and mortality from sepsis. Mice with identical genetic backgrounds but distinct microbiomes were obtained from different vendors and analyzed following cecal ligation and puncture (CLP). β diversity of the microbiome measured from feces demonstrated significant differences between The Jackson Laboratory (Jax; Bar Harbor, ME, USA) and Charles River Laboratories (CR; Wilmington, MA, USA) C57/B6 mice. Jax mice had 7-d mortality of 90% following CLP, whereas CR mice had a mortality of 53%. Differences in vendor were associated with altered immunophenotype with increased splenic IFN-γ⁺CD4⁺ T cells, effector memory CD4⁺ T cells, and central memory CD4⁺ T cells and increased Peyer’s patch effector memory CD4⁺ T cells in septic CR mice. To determine whether differences in the microbiome were responsible for these differences, Jax and CR mice were cohoused for 3 wk, after which they assumed a similar microbiota composition. Cohoused mice had improved survival following CLP compared to Jax mice and had similar survival regardless of their vendor of origin. All differences in immunophenotype between septic Jax and CR mice disappeared following cohousing. These findings suggest that the microbiome plays a crucial role in survival and the host immune response from sepsis and represents a potential target for therapeutic intervention.—Fay, K. T., Klingensmith, N. J., Chen, C.-W., Zhang, W., Sun, Y., Morrow, K. N., Liang, Z., Burd, E. M., Ford, M. L., Coopersmith, C. M. The gut microbiome alters immunophenotype and survival from sepsis.

Sepsis is a life-threatening organ dysfunction caused by dysregulated host response to infection (1). The syndrome can be induced by a wide variety of microbes and, by definition, involves a maladaptive response to a pathogen. Sepsis kills at least 270,000 patients in the United States alone (2) and over 5 million people worldwide (3), leading to the World Health Organization recognizing sepsis as a global health priority (4). Despite its prevalence, there are no specific therapies for sepsis beyond targeted antimicrobial therapy (5), leading to mortality rates as high as 40% (6).

The gut microbiome plays a crucial role not only in gastrointestinal health but also in overall immune development and host health (7–9). The gut microbiome contains ∼40 trillion microorganisms, a similar number of cells as occurs in the host (10). This includes up to 1000 different microbial species and 100 times more bacterial genes than human genes (11, 12). The role of the microbiome in development and health is complex, and there is communication between the host immune system and the microbiome, where bilateral signals play a crucial role in shaping each (13).

The microbiome is markedly altered in critical illness (14–17). Microbial diversity is diminished within 6 h of admission to the intensive care unit, and lack of diversity has been associated with poor outcomes in critically ill patients (18). In addition, remaining microbes alter their own virulence factors in response to stress signals from the host (19). Further, intentionally altering the microbiome—by probiotics, fecal microbial transplant, or selective decontamination of the digestive tract—positively effects numerous outcomes in the intensive care unit (20–22). Together, the transition of a microbiome into a pathobiome has been hypothesized to be a driver of mortality from sepsis (23–26), at least in part by the ability of invading bacteria to act as antigens and modulate the host immune response (7).

Despite this, the role of the microbiome as a variable is both rarely examined and incompletely understood in preclinical studies of sepsis in contrast to rigorous control of variables such as genetic background, age, gender, and comorbidities. To understand the role of the microbiome in outcomes and immunophenotype from sepsis, we examined sepsis in genetically identical, age-matched, gender-matched mice, subjected to cecal ligation and puncture (CLP), the most commonly used preclinical model of sepsis. This experimental design takes advantage of the fact that mice from different vendors have different microbiomes based on their origin yet are otherwise identical, and these differences can directly impact the host response to infection and inflammation (27–29). To determine whether the microbiome is directly responsible for the results generated, unmanipulated mice from different vendors were then cohoused for 3 wk, after which they developed a common microbiome and were then subjected to CLP to determine the impact on survival and immunophenotype.

MATERIALS AND METHODS

Animals

Six-week-old male and female C57BL/6 mice were obtained from either The Jackson Laboratory (Jax) or Charles River Laboratories (CR). Unless otherwise stated, all mice were used for experiments within 2 d of arrival to the animal facility at Emory University. Mice were randomized to undergo sham laparotomy or CLP (see below) and were either euthanized at 24 h for phenotypic analysis or followed 7 d for survival. Experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, Bethesda, MD, USA) and were approved by the Institutional Animal Care and Use Committee at Emory University School of Medicine (Protocol DAR 2002717). All animals had free access to chow and water throughout.

Sepsis model

CLP was used as a model of sepsis as previously published by Baker et al. (30). Mice were anesthetized with isoflurane and a small midline incision was made. The cecum was externalized and ligated 1 cm from its base. The cecum was then punctured twice with a 25-gauge needle, a small amount of stool was extruded into the abdominal cavity, and the intestine was placed back within the abdomen. The midline incision was sutured closed and the skin incision with approximated with animal surgical adhesive. Sham-treated control animals underwent the same procedure without ligation and puncture of the cecum. All animals received buprenorphine (0.1 mg/kg; McKesson Medical, San Francisco, CA, USA) preoperatively for pain relief. To mimic clinical care of patients with sepsis (5), all animals (sham-treated and septic) also received fluid resuscitation with 1 ml of normal saline as well as antibiotic therapy (ceftriaxone 25 mg/kg; MilliporeSigma, Burlington, MA, USA and metronidazole 12.5 mg/kg; Apotex, Weston, FL, USA) administered subcuaneously postoperatively. Antibiotics were continued on a q12 h dosing schedule for 48 h postoperatively. Animals were monitored continuously following anesthesia and then every 15 min until they were able to walk. A heating pad was placed under the recovery cage to prevent hypothermia, and animals were checked twice daily to determine if they were moribund. Animals that were identified as moribund were euthanized using humane end points. Moribund animals were identified as follows: 1) loss of 25% of body weight from baseline weight; 2) surgical complications unresponsive to immediate intervention (wound dehiscence, bleeding, infection); 3) medical conditions unresponsive to treatment such as self-mutilation, severe respiratory distress, icterus, or intractable diarrhea; 4) clinical or behavioral signs unresponsive to appropriate intervention persisting for 1 d including 1 or more unresolving skin ulcers and abnormal vocalization when handled; or 5) inability to ambulate.

Cohousing

Immediately upon arrival to the animal facility at Emory University, 6-wk-old female Jax and CR mice were placed in the same cage with access to the same chow and water. The mice were housed together for a total of 3 wk, after which stool samples were collected to assess the microbiome. Cohoused mice were then subjected to CLP and analyzed for immunophenotyping and survival as detailed elsewhere. Of note, the rationale for using only female mice for this subset of experiments relates to the fact that male mice raised under different conditions fight when placed in a cage together, rendering this experimental design impossible without causing animal suffering in male mice.

Microbiome evaluation

Fecal samples were collected at time of euthanizing and frozen until DNA extraction samples were sent to the Emory Genomics Core Laboratory where bacterial DNA was extracted. The hypervariable V3-V4 region of the 16s rRNA gene was amplified with the Illumina MiSeq sequencing platform (Illumina, San Diego, CA, USA; https://www.illumina.com/systems/sequencing-platforms/miseq.html). The resulting raw sequences were quality controlled with default parameters, including eliminating inadequate sequences and chimeras, using the QIIME v.1.9.1 software package (http://qiime.org/index.html). The resulting sequences were clustered using a closed-reference frame at 97% similarity against the GreenGenes database (v.13_8; https://greengenes.secondgenome.com/) to assign bacterial operational taxonomic units. Samples were rarified to a depth of 1700 copies. α diversity was determined using Shannon indices and β diversity using unweighted UniFrac distances which were then plotted on a Principal Coordinates Analysis (PCoA) plot, again using the QIIME software (31).

Phenotypic flow cytometric analysis

Spleen, Peyer’s patches, and mesenteric lymph nodes (MLNs) were harvested at time of euthanizing. The tissues were processed into single-cell suspensions and stained for CD4-Pacific Blue (BD Biosciences, San Diego, CA, USA), CD8-Pacific Orange (Thermo Fisher Scientific, Waltham, MA, USA), CD44-PerCP (BioLegend, San Diego, CA, USA), CD62L-APC (Thermo Fisher Scientific), CD3-A700 (BD Biosciences), CD25-FITC (BioLegend), FOXP3-APC (Thermo Fisher Scientific), CD62L-PE (BD Biosciences), B220-A700 (BioLegend), anti-Ly6C-Pacific blue (BioLegend), anti-F4/80-FITC (BioLegend), anti-CD11b-PerCp (BioLegend), anti-CD11c-PE-Cy7 (BioLegend), anti-NK1.1-APC (BioLegend), anti-Ly6G-AF700 (BioLegend), and anti-CD3-APC-Cy7 (BioLegend). TruCount Beads (BD Pharmingen, Franklin Lakes, NJ, USA) were prepared according to the manufacturer’s instructions and used to determine absolute cell counts.

Intracellular cytokine staining

A total of 2 × 10⁶ splenic cells were plated into a 96-well plate, suspended in Roswell Park Memorial Institute (RPMI) 1640 culture medium (Mediatech, Herndon, VA, USA), and incubated for 4 h using phorbol 12-myristate 13-acetate (30 ng/ml) and ionomycin (400 ng/ml) with 10 μg/ml of Brefeldin A at 37°C. After stimulation, cells were then stained for CD4-Pacific Blue (BD Biosciences), CD8-Pacific Orange, IL-2-FITC (BD Pharmingen), TNF-PECy7 (BioLegend), IFN-γ-A700 (BD Pharmingen), and CD3-APCCy7 (BioLegend). For experiments examining IL-10 in the innate immune system, cells were stimulated with LPS (10 μg/ml) in the presence of Golgiplug (BD Pharmingen) for 5 h at 37°C. Cells were surface-stained and then were permeabilized and stained with anti-IL-10-PE (BioLegend). All samples were run on an LSR II Flow Cytometer (BD Biosciences), and subsequent data were analyzed with FlowJo 10.0.8rl software (Treestar, Ashland, OR, USA).

IL-10 levels

Serum IL-10 levels were determined by using a commercially available bead array (Bio-Plex Pro Mouse Cytokine Bead Array; Bio-Rad, Hercules, CA, USA) according to the manufacturer’s protocol. All samples were run in duplicate.

Statistical analysis

Statistical analyses were performed using Prism 6.0 software (GraphPad Software, La Jolla, CA, USA) and are presented as means ± sem. Data were tested for Gaussian distribution using the D’Agostino-Pearson omnibus normality test. For 2 group comparisons found to have a gaussian distribution, the Student’s t test was used. If not, the nonparametric Mann-Whitney U test was used. For 16s β diversity, analysis of similarity testing was performed, followed by a post hoc Kruskall-Wallis pair-wise comparison test accounting for multiple comparisons. For multiple group comparisons, 1-way ANOVA followed by Tukey posttest or nonparametric Kruskall-Wallis test, followed by Dunn’s test were used. A value of P < 0.05 was considered statistically significant.

RESULTS

Mortality in genetically identical septic mice from different vendors

Age- and gender-matched, genetically identical C57BL/6 mice from Jax and CR were subjected to CLP. Jax mice had a 7-d mortality of 90% following the onset of sepsis, whereas CR mice had a 7-d mortality of 53% (Fig. 1A).

Figure 1.

Mortality following sepsis, α and β microbial diversity and bacterial phyla and genera vary in genetically identical mice from different vendors. A) Mortality was significantly higher in Jax mice than CR mice (P = 0.0006). B) α diversity was higher in fecal samples from sham-treated CR mice than sham-treated Jax mice, with significantly higher Shannon indices indicating a greater diversity of microbiome composition. Total number of reads: 11,913,820; median number of reads: 138,532; number for rarefaction: 1700 (P = 0.001). C) β diversity was significantly different in fecal samples between CR sham-treated mice and Jax sham-treated mice (P = 0.03). β diversity was different between CR septic mice and Jax septic mice (P = 0.001). There was a trend toward different β diversity between Jax sham-treated and Jax septic mice (P = 0.06), whereas CR sham-treated mice differed significantly from CR septic mice (P = 0.03). PC1 17.3, PC3 7.2%, PC2 13.8%. D) Firmicutes abundance was similar in sham-treated Jax and CR mice. Firmicutes abundance increased following CLP in mice from both vendors (P = 0.04 for Jax, P < 0.0001 for CR) but was higher in CR mice than Jax mice (P < 0.0001). E) Bacteroidetes abundance was similar in sham-treated JAX and CR mice. Stool from CR mice had significantly reduced Bacteroidetes after CLP (P < 0.0001), and septic CR mice had significantly lower abundance of Bacteroidetes compared to septic Jax mice (P < 0.0001). F) Firmicutes: Bacteroidetes was increased in CR mice following CLP (P = 0.03). G) Oscillospira abundance was higher in septic Jax mice than sham-treated Jax mice (P < 0.0001) and was also higher than septic CR mice (P = 0.002). H) A. muciniphila abundance was lower in CR mice compared to Jax mice under both sham (P < 0.0001) and septic conditions (P = 0.01). I) Rikenellaceae abundance was lower in CR mice compared to Jax mice under both sham (P < 0.0001) and septic conditions (P < 0.0001). Sepsis decreased Rikenellaceae abundance in both Jax and CR mice (P < 0.0001 for both). J) Bacteroides of unknown species abundance was higher in CR under both sham (P = 0.0007) and septic conditions (P = 0.005); n = 5–9 mice/group for all comparisons.

Microbiome diversity and Firmicutes:Bacteroidetes ratio in genetically identical sham-treated and septic mice from different vendors

α diversity, an analysis of within-group bacterial diversity, showed sham-treated CR mice had significantly higher diversity than sham-treated Jax mice (Fig. 1B). β diversity, an analysis of between-sample bacterial variance, showed sham-treated Jax and CR mice had distinct microbiota compositions (Fig. 1C). Following CLP, the microbiome of Jax mice and CR were distinct from each other and were also distinct from sham-treated mice from the same vendor (Fig. 1C).

At the phyla level, sham-treated Jax and CR mice had similar abundance of Firmicutes (Fig. 1D). Sepsis induced a small increase in Firmicutes in Jax mice; and, a further disproportionate increase in Firmicutes occurred following CLP in CR mice (Fig. 1D). Sham-treated Jax and CR mice also had a similar abundance of Bacteroidetes (Fig. 1E). Sepsis did not change abundance of Bacteroidetes in Jax mice. In contrast, a significant decrease in Bacteroidetes was observed following CLP in CR mice (Fig. 1E). These differences resulted in an elevated Firmicutes:Bacteroidetes in CR septic mice (Fig. 1F).

Alterations in bacterial genera in stool samples from genetically identical sham-treated and septic mice from different vendors

Individual bacterial genera that differed between Jax and CR mice were then identified. Three bacterial genera—Oscillospira, Akkermansia muciniphilia, and Rikenellaceae—had a reduced mean relative abundance in septic CR mice compared to septic Jax mice (Fig. 1G–I). Of these, the decrease was sepsis specific in Oscillospira because bacterial abundance was similar between sham-treated CR mice and sham-treated Jax mice. In contrast, bacterial abundance was already lower in sham-treated CR mice than sham-treated Jax mice for A. muciniphilia and Rikenellaceae. Both of these were also lower in septic CR mice compared to septic Jax mice, although absolute levels were lowest in septic CR mice, with Rikenellaceae being nearly undetectable.

A single bacterial genus had a higher mean abundance in septic CR mice than septic Jax mice—Bacteroides of unknown species (Fig. 1J). However, Bacteroides of unknown species type was elevated in sham-treated CR mice compared to sham-treated Jax mice, and abundance was not further altered by the presence of sepsis.

Immunophenotype of splenocytes in genetically identical sham-treated and septic mice from different vendors

The frequency of CD4⁺ T cells was similar between sham-treated Jax and CR mice, and frequency for both was decreased following CLP (Fig. 2A, B). The frequency of effector memory CD4⁺ T cells was also similar between sham-treated Jax and CR mice (Fig. 2C, D); however, the frequency of effector memory CD4⁺ T cells was higher in CR mice following CLP than Jax mice following CLP. Although the frequency of central memory CD4⁺ T cells was similar in sham-treated Jax and CR mice, the frequency was higher in septic CR mice than septic Jax mice (Fig. 2C, E). Naive CD4⁺ T cells were unaffected by both vendor and sepsis (Fig. 2C, F). The frequency of CD8⁺ T cells was also unaffected by both vendor and sepsis (Supplemental Fig. S1A, B), as were the frequency of effector and central memory CD8⁺ T cells and naïve CD8⁺ T cells (Supplemental Fig. S1C–F). The frequency of regulatory T (T_reg) cells was similar between Jax and CR sham-treated mice, whereas T_reg cell frequency increased after CLP in CR mice but not Jax mice (Fig. 2G, H). The number of dendritic cells, macrophages, and neutrophils in the spleen did not differ between septic Jax and CR mice (unpublished results).

Figure 2.

Frequency of splenic CD4⁺ T cells and CD4⁺ memory compartment cells in genetically identical mice from different vendors. A) Representative flow cytometry plots of CD4⁺ T cells from 2 independent experiments with 5 animals in each. B) CD4⁺ T cell frequency was decreased following sepsis in both Jax mice (P = 0.02) and CR mice (P < 0.0001), although there was no difference between septic Jax and CR mice (P = 0.6). C) Representative flow cytometry plots of CD4⁺ memory cell compartments from 2 independent experiments with 5 animals in each. D) Septic CR mice had increased frequency of effector memory (CD44^HICD62l^LO) CD4⁺ cells) compared to both septic Jax mice (P = 0.004) and sham-treated CR sham-treated mice (P = 0.002). E) Septic CR mice had increased central memory (CD44^HICD62l^HI) CD4⁺ cells compared to septic Jax mice (P = 0.04). F) No differences were detected in any groups in the naïve CD4⁺ cell (CD44^LOCD62l^LO) compartment. G) Representative flow cytometry plots of T_reg cells from 2 independent experiments with 5 animals in each. H) Septic CR mice had a significant increase in T_reg (FOXP3⁺CD25⁺CD4⁺) cells compared to sham-treated CR mice (P = 0.03).

Functional status of CD4⁺ T cells was then assessed by production of intracellular cytokines. Sham-treated Jax and CR mice had similar percentage of IFN-γ⁺ cells within the CD4⁺ compartment. However, septic CR mice had an increase in percentage of IFN-γ⁺ of CD4⁺ T cells following CLP compared to Jax mice (Fig. 3A, B). No differences were detected between sham-treated Jax and CR mice in either IL-2⁺ or TNF⁺ CD4⁺ T cells (Fig. 3C–F). Sepsis induced a decrease in the frequency of IL-2⁺cells within the CD4⁺ T cell population and an increase in TNF⁺ cells as a percentage of CD4⁺ T cells in CR mice. The percent of IL-10–positive macrophages were similar between sham-treated Jax and CR mice as well as between septic JAX and CR mice (Supplemental Fig. S2A). The percent of IL-10–positive neutrophils were lower in sham-treated CR mice than sham-treated Jax mice. A similar decrease was seen in septic CR mice compared to septic Jax mice, suggesting this difference was dependent on vendor but not changed following CLP (Supplemental Fig. S2B). No differences were detected on percent of IL-10–positive dendritic cells between sham-treated or septic Jax or CR mice (Supplemental Fig. S2C). In addition, systemic IL-10 levels were similar between sham-treated Jax and CR mice and between septic Jax and CR mice (Supplemental Fig. S2D). No differences in bacteremia were detected following CLP between Jax and CR mice (Supplemental Fig. S3).

Figure 3.

Cytokine production of splenic CD4⁺ cells in genetically identical mice from different vendors. A) Representative flow cytometry plots of percentage of IFN-γ⁺ of CD4⁺ T cells from 2 independent experiments with 5 animals in each. B) Septic CR mice had significantly increased frequency of IFN-γ⁺ of CD4⁺ cells compared to septic Jax mice (P = 0.003). C) Representative flow cytometry plots of percentage of IL-2⁺ of CD4⁺ T cells from 2 independent experiments with 5 animals in each. D) No differences were detected in frequency of IL-2⁺ of CD4⁺ T cells between septic CR and septic JAX mice (P = 0.6), although septic CR mice had a lower frequency than sham-treated CR mice (P = 0.01). E) Representative flow cytometry plots of percentage of TNF⁺ of CD4⁺ T cells from 2 independent experiments with 5 animals in each. F) No differences were detected in frequency of TNF⁺ of CD4⁺ T cells between septic CR and septic Jax mice (P = 0.08) although septic CR mice had a higher frequency compared to sham-treated CR mice (P < 0.0001). G) CR mice had increased effector memory CD4⁺ cells compared to Jax mice in Peyer’s patches under both sham (P = 0.006) and septic (P = 0.02) conditions. H) No significant differences were detected in MLN effector memory CD4⁺ cells in both Jax and CR mice in sham-treated and septic mice.

Immunophenotype of Peyer’s patches and MLNs in genetically identical sham-treated and septic mice from different vendors

To determine whether the immunophenotype seen in splenocytes was compartment specific, immunophenotype was evaluated in MLNs and Peyer’s patches. The frequency of Peyer’s patch effector memory CD4⁺ T cells was higher in sham-treated CR mice than sham-treated Jax mice (Fig. 3G). A similar increase was seen between CR and Jax mice following sepsis; however, there was no difference in the frequency of effector memory CD4⁺ T cells between CR sham-treated and CR septic mice. Septic Jax and CR mice had similar Peyer’s patch IFN-γ⁺ (Supplemental Fig. S4A, B), IL-2⁺ (Supplemental Fig. S4C, D), and TNF⁺ (Supplemental Fig. S4E, F) as a percentage of CD4⁺ T cells. In contrast, the frequency of MLN effector memory CD4⁺ T cells in Jax mice and CR mice were unchanged in sham-treated mice and following CLP (Supplemental Fig. S4H). Similarly, septic Jax and CR mice had similar MLN IFN-γ⁺ (Supplemental Fig. S4G, H), IL-2⁺ (Supplemental Fig. S4I, J), and TNF⁺ (Supplemental Fig. S4K, L) as a percentage of CD4⁺ T cells.

Microbiome composition in genetically identical mice from different vendors before and after cohousing

To determine if the alterations in the microbiome were mechanistically responsible for the survival advantage seen in septic CR mice compared to Jax mice, unmanipulated mice were cohoused in order to develop a common microbiome. Mice were given 3 wk of exposure to the same bedding, food, and water, and fecal samples were collected and compared to previous CR and Jax samples. Unlike sham-treated Jax and CR mice (Fig. 1B, C), sham-treated Jax cohoused (JaxCH) and CR cohoused (CRCH) mice had similar α diversity and β diversity (Fig. 4A, B). After cohousing, an evaluation of genera diversity demonstrated that all mice—regardless of whether they were CR or Jax in origin—developed a microbiome similar to CR mice and distinct from JAX at baseline (Fig. 4C).

Figure 4.

Differences in α and β microbial diversity and mortality in genetically identical mice from different vendors disappear after cohousing. A) α diversity in fecal samples from sham-treated CRCH mice and sham-treated JaxCH mice demonstrated no significant differences in observed operational taxonomic units (P = 0.8) after 3 wk of cohousing, indicating a similar level of diversity within their microbiome composition. B) β diversity was similar between sham-treated CRCH mice and sham-treated JaxCH mice (P = 0.6), indicating a similar level of diversity between their microbiome composition. Unweighted UniFrac distances were plotted on a PCoA plot to visually represent β diversity. Statistical significance between those same distances were analyzed using the analysis of similarity test within the QIIME software. C) An overview of bacteria taxa represented in each cohort’s microbiome demonstrated that after cohousing, sham-treated CRCH mice maintained a similar composition to sham-treated CR mice. However, sham-treated JaxCH mice diverged from sham-treated Jax mice and assumed a microbiota phenotype similar to both sham-treated CR and sham-treated CRCH mice. Legend shows the 5 most common taxa for each group as well as the small number of taxa that could not be identified in each group. All taxa presented are genus unless otherwise specified when genus was unable to be identified. D) CRCH and JaxCH mice had similar survival following CLP despite original vendor (P = 0.83). Additionally, septic cohoused mice had a significantly improved survival compared to septic Jax mice regardless of their vendor of origin (P = 0.02 for both CRCH vs. Jax and JaxCH vs. Jax).

Mortality following cohousing in septic genetically identical mice from different vendors

Both JaxCH and CRCH mice had improved survival following CLP compared to Jax mice (Fig. 4D). Notably, cohoused mice had similar survival regardless of the vendor of origin.

Immunophenotype following cohousing in septic genetically identical mice from different vendors

To determine whether differences in immunophenotype between septic Jax and CR mice persist after cohousing, JaxCH and CRCH mice were compared. Despite differences in septic CR and Jax mice (Figs. 2C, D and 3A, B, G), JaxCH and CRCH mice had similar frequencies of splenic IFN-γ⁺ as a percentage of CD4⁺ T cells (Fig. 5A, B), splenic effector and central memory CD4⁺ T cells (Fig. 5C, D), and Peyer’s patch effector memory CD4⁺ T cells (Fig. 5E, F) following CLP. In addition, similar to septic Jax and CR mice, no differences were detected in JaxCH and CRCH mice in frequency of splenic CD4⁺ T cells (Fig. 5G, H), splenic CD8⁺ T cells (Fig. 5G, I) and MLN effector memory CD4⁺ T cells (Fig. 5J, K). To determine whether septic cohoused mice had similar immunophenotype to septic mice that were not cohoused, CR, JaxCH, and CRCH mice were compared in a separate set of experiments. This experimental design was chosen because of similarity in mortality between these groups at 47, 42, and 48% respectively (Figs. 1 and 4). CR mice had slightly lower frequency of splenic IFN-γ⁺ as a percentage of CD4⁺ T cells than JaxCH mice, although no difference was noted between CR and CRCH mice or CRCH and JaxCH mice (Supplemental Fig. S5A). CR mice had a similar frequency of splenic effector memory CD4⁺ T cells as JaxCH mice. CRCH mice had a higher frequency than CR mice, although no difference was noted between JaxCH and CRCH (Supplemental Fig. S5B). No difference was detected between frequency of central memory CD4⁺ T cells between CR, JaxCH, and CRCH mice (Supplemental Fig. S5C). CR mice also had a lower frequency of Peyer’s patch effector memory CD4⁺ T cells than JaxCH mice, although no difference was noted between CR and CRCH mice or CRCH and JaxCH mice (Supplemental Fig. S5D). Together, these data show JaxCH and CRCH were generally similar to CR mice.

Figure 5.

Immunophenotype of sham-treated and septic mice in genetically identical mice from different vendors disappear after cohousing. A) Representative flow cytometry plots of percentage of splenic IFN-γ⁺ of CD4 + T cells from 2 independent experiments. B) Frequency of splenic IFN-γ⁺ of CD4 + T cells was similar between sham-treated and septic JaxCH and CRCH mice. C) Representative flow cytometry plots of splenic effector memory CD4⁺ T cells from 2 independent experiments. D) Frequency of splenic effector (top panel) and central (lower panel) memory CD4⁺ T cells were similar between sham-treated and septic JaxCH and CRCH mice. E) Representative flow cytometry plots of Peyer’s patch effector memory CD4⁺ T cells from 2 independent experiments. F) Frequency of Peyer’s patch effector memory CD4⁺ T cells was similar between sham-treated and septic JaxCH and CRCH mice. G) Representative flow cytometry plots of splenic CD4⁺ and CD8⁺ T cells from 2 independent experiments. H) Frequency of splenic CD4⁺ T cells was similar between sham-treated and septic JaxCH and CRCH mice. I) Frequency of splenic effector memory CD8⁺ T cells was similar between sham-treated and septic JaxCH and CRCH mice. J) Representative flow cytometry plots of MLN effector memory CD4⁺ T cells from 2 independent experiments. K) Frequency of MLN effector memory CD4⁺ T cells was similar between sham-treated and septic JaxCH and CRCH mice.

DISCUSSION

This study demonstrates that the microbiome plays a crucial role in survival and immunophenotype following sepsis as evidenced by the response of genetically identical age- and gender-matched mice from different vendors to CLP. Although there are no obvious differences between mice except vendor of origin, it was possible that unknown factors independent of the microbiome were responsible for these findings. As such, cohousing experiments were performed taking advantage of the fact that unmanipulated mice that share an identical local environment will end up developing a common microbiome. All immunophenotypic differences between septic Jax and CR mice disappeared following cohousing, with similar results identified in all elements measured between JaxCH and CRCH mice. Notably, the significant discrepancy in mortality between septic Jax and CR mice disappeared after cohousing, and cohoused mice had an improved survival compared to Jax mice, regardless of their vendor of origin. The significance of these results is demonstrating that the microbiome independently mediates mortality following sepsis, and alterations in the microbiome are associated with significant compartmentalized changes in the host immune response.

An extensive analysis of the immunophenotype of Jax and CR mice across 3 different compartments (spleen, Peyer’s patch, MLN) demonstrated the following differences following CLP: 1) the frequency of splenic effector memory CD4⁺ T cells was higher in CR mice, 2) the frequency of splenic central memory CD4⁺ T cells was higher in CR mice, 3) CR mice had an increase in percentage of splenic IFN-γ⁺ of CD4⁺ T cells, and 4) the frequency of Peyer’s patch effector memory CD4⁺ T cells was higher in CR mice. These results demonstrate an increase in both percentage of memory cells (in more than 1 compartment) and functionality of cells in septic CR mice compared to septic Jax mice. Memory cells function to augment the immune response to previously encountered antigen, so these findings suggest that CR mice have the capacity to respond more robustly to infection than Jax mice. Further, an increase in IFN-γ production suggests memory cells in CR mice are more functional, with the capacity to increase cytokine production that serves to further up-regulate the immune response. Because the only obvious difference between genetically identical CR and Jax mice is their microbiome, this suggests that the flora in CR leads to a more robust immune response when bacteria translocate following abdominal perforation in CLP. In turn, this is associated with improved mortality in septic CR mice, and we hypothesize that the augmented memory response may be responsible for their increased survival.

The conceptual benefit of examining genetically identical mice from different vendors allows for mice to develop naturally (within the confines a controlled animal facility), with control of multiple variables and without need for intervention prior to the onset of sepsis.

This approach has previously been used by other laboratories to demonstrate the microbiome plays a crucial role in the host response. Villarino et al. (32) examined genetically identical C57BL/6 mice from 4 vendors to malarial infection. Significant differences were noted in parasite burden and mortality following infection with multiple Plasmodium species. In addition, germ-free mice that received cecal content transplants from resistant or susceptible mice had differential parasitic burden. A separate study by Hilbert et al. (33) examined the impact of the gut microbiota in genetically identical C57BL/6 mice from 3 vendors to intraperitoneal stool injection. Disease severity at 18 h varied between vendors as assessed by “clinical score,” body weight, and temperature with differences in cytokine levels and vascular leakage. Neither of these studies examined any elements of the immunophenotype. An elegant study by the Griffith group compared Jax mice to mice from a different vendor following CLP, taking advantage of the fact that these mice differed in whether they had segmented filamentous bacteria (SFB) (34). SFB-specific CD4⁺ T cells underwent antigen-driven proliferation following sepsis in mice that contained this normal gut microbe but not in genetically identical mice that lacked the microbe. Cohousing mice prior to CLP induced a similar phenotype in mice that initially lacked SFB but subsequently acquired it via horizontal transfer during cohousing. Further, mice containing SFB were resistant to a secondary lethal infection with an SFB antigen-expressing virulent Listeria, suggesting sepsis primed for a protective response via the SFB-specific CD4⁺ T cells. Similar to the results contained herein, this demonstrates that the microbiome alters survival from sepsis and changes host immunophenotype. However, the studies differ in multiple key ways, leading to complementary insights. Notably, our study examines a broader view of the adaptive immune system in multiple compartments as opposed to a comprehensive examination of the antigen-specific response to a single bacterium in the spleen. In addition, our study examines survival directly after CLP as opposed to a secondary virulent Listeria infection. A recent mechanistic study by Wilmore et al. (35) compared Jax C57Bl/6 mice to genetically identical mice bred in a university animal facility and found a differential survival following CLP. Notably, the university-bred mice had significantly higher levels of IgA at baseline. Cohousing mice converted low IgA mice into higher IgA mice, an effect proven to be due to the microbiome by replicating the effect by fecal transplant and abrogating it with antibiotics. Several members of the gut microbiome induced IgA, including Burkholderia, Sphingomonas, and Bacillus genera in a T-cell–dependent manner. Notably, these are entirely distinct from genera identified as being distinct between sham-treated and septic Jax and CR mice in this study. The “low IgA mice” (from Jax) were more susceptible to mortality from CLP than high IgA mice but became more resistant after cohousing mice for 10 wk and somewhat more resistant after cohousing for 4 wk. Interestingly, despite a transfer of microbiome, no survival difference was seen when cohousing was performed for 1.5 wk. Similar to our results, Wilmore et al. (35) demonstrate that the microbiome alters survival following CLP and also identifies Jax mice as being particularly susceptible to sepsis. However, the take home points from the 2 studies are almost entirely distinct. Whereas the Wilmore et al. (35) study specifically focused on the role of IgA (not examined herein), T cells were examined only insofar as IgA production was determined to be dependent on T cells. Thus, the detailed immunophenotyping of CD4⁺ and CD8⁺ cells in multiple body compartments are unique to this study. Further, neither the microbiome nor immunophenotyping was examined after sepsis in the Wilmore et al. (35) study as the sole post-CLP endpoint was mortality. Because sepsis alters both the microbiome and immunophenotyping both in isolation and in a vendor-specific manner, the insights gained by examining Jax and CR mice both before and after CLP here are also unique to this study.

Four genera of bacteria were identified that were different between Jax and CR mice following sham surgery or sepsis. Of these, 3 (Oscillospira, A. muciniphilia, and Rikenellaceae) had a reduced mean relative abundance in septic CR mice compared to septic JAX mice, whereas 1 (Bacteroides of unknown specie) was increased in septic CR mice. Increased levels of Rikenellaceae have been associated with exacerbation of inflammation, especially in the setting of high-fat diets (36). As such, it is possible that the higher levels seen in Jax mice potentially induce a detrimental proinflammatory state, although this would need to be tested in future experiments. In contrast, increases in A. muciniphilia and Oscillospira have been associated with reduced intestinal inflammation and improved metabolic disorders such as obesity (37, 38), and it is less clear why a decrease would be beneficial in septic CR mice. Although not a prominent species within the normal commensal flora, Bacteroides fragiilis has been shown to have multiple effects on the local immune system. The bacterium’s immunogenic effects are derived from its polysaccharide capsule A (PSA), which is strongly antigenic. Germ-free mice colonized with PSA⁺ B. fragiilis have an expansion of T_reg cells, which have been shown to mitigate colonic inflammation in the setting of excessive T helper (T_h)17 production and have exacerbated immunosuppressive characteristics through the expression of IL-10 (39). Additionally, PSA influences dendritic cells to stimulate immunosuppressive T helper cells to control inflammation (40). Although we were unable to fully speciate the Bacteroides present in the gastrointestinal tract of CR and Jax mice, our findings of a higher mean relative abundance of Bacteroides, expansion of CD4⁺ effector memory cells, and IFN-γ+ CD4⁺ T cells suggest these bacteria may potentially play a role in the altered immunophenotype and improved survival in septic CR mice. Additionally, given the elevation of Bacteroides in CR mice at baseline, it could be a chronic antigenic source for development of memory cells that improve bacterial clearance after a septic insult.

No differences were noted in bacteremia between septic Jax and CR mice despite differing survival and immunophenotype. This is a common finding in preclinical studies of sepsis (41). This potentially could be explained by induction of disease tolerance, in which the host protects against microbes by reducing the negative impact they have on host fitness (42). The mechanisms underlying disease tolerance are complex, but damage control allows for the preservation of functional output of parenchymal tissue, thereby maintaining homeostatic parameters within an acceptable range (43). Additionally, multiple studies have demonstrated that neither bacteremia nor fungemia discriminates mice that live or die within a treatment group, and bacterial translocation to the lymph nodes, spleen, or liver does not discriminate progression to recovery vs. death (24).

This manuscript has a number of limitations. First, it is possible that there is an unrecognized upstream factor present at the original vendor that is responsible for both different microbiomes and immunophenotypes that is eliminated after cohousing. Next, although changes in the immunophenotype were associated with changes in the microbiome, we cannot conclude that changes in the host immune response were directly responsible for differences in mortality seen in genetically identical mice from different vendors or following cohousing especially considering that some of the changes were relatively small and are of unclear biologic significance. Further, the mechanisms through which the microbiome alter the host immunophenotype remain to be discovered as do which bacteria are responsible for the changes. In addition, immunophenotype was measured at only a single time point (24 h), so mechanistic insights at other time points would be missed by our experimental design. Finally, although multiple bacteria were identified as distinct between sham-treated and septic Jax and CR mice, we did not identify whether the differences identified were due to a single strain of bacteria or a select subset. Although testing the impact of each individual bacteria for immunophenotype in sham-treated and septic mice was outside the scope of the experimental design, an overview at the phylum, class, order, family, and genus level between the 4 groups (Supplemental Fig. S6) demonstrates differences that could be examined in future experiments.

Despite these limitations, this work demonstrates that the microbiome independently alters mortality and immune response following sepsis. Controlling for the microbiome [similar to controlling for age, gender, genetics, and surgical details such as needle gauge and length of cecum ligated (44, 45)] should be considered when performing preclinical studies on sepsis in order to maximize the translation of findings. Further work needs to be done to understand the mechanisms underlying the role of the microbiome in mortality that can potentially be manipulated for therapeutic gain in patients with sepsis.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

CLP

cecal ligation and puncture

CR

Charles River Laboratories

CRCH

CR cohoused

Jax

The Jackson Laboratory

JaxCH

Jax cohoused

MLN

mesenteric lymph node

PSA

polysaccharide capsule A

SFB

segmented filamentous bacteria

T_reg

regulatory T

AUTHOR CONTRIBUTIONS

K. T. Fay, M. L. Ford, and C. M. Coopersmith designed research; K. T. Fay, N. J. Klingensmith, C.-W. Chen, W. Zhang, Y. Sun, K. N. Morrow, Z. Liang, and E. M. Burd performed research; K. T. Fay, N. J. Klingensmith, C.-W. Chen, W. Zhang, Y. Sun, K. N. Morrow, M. L. Ford, and C. M. Coopersmith analyzed data; and K. T. Fay and C. M. Coopersmith wrote the manuscript.