The microbiota regulates susceptibility to Fas-mediated acute hepatic injury

Animal breeding and cohousing

Female BALB/cJ mice obtained from the Jackson Laboratory (JAX) were established as a breeder colony at the Geisel (GSL) School of Medicine for >4 generations according to Association for Assessment and Accreditation of Laboratory Animal Care guidelines, allowing them to habituate to GSL environmental conditions. In some experiments, 3-wk-old female BALB/cJ mice purchased from JAX were co-housed with 3-wk-old GSL-habituated BALB/cJ mice: per cage, two JAX BALB/cJ mice were co-housed with two GSL-habituated BALB/cJ mice for an additional 5 wk, and then challenged with ConA. To assess the influence of vendor environment, age-matched BALB/c and C57Bl/6 mice were purchased from JAX, Taconic Farms (TAC), and National Cancer Institute (NCI), and housed at GSL for < 1 wk (to minimize shifts in microbiota composition) before challenge. For Myd88 KO and Fas KO experiments, B6.129P2(SJL)-Myd88^tm1Defr/J mice and B6.MRL-Fas^lpr/J mice were purchased from JAX, and either used within 1 wk, or bred at GSL with JAX C57BL/6J mice, in order to allow for the generation of all genotypes as littermates that would be exposed to similar microbiota. Heterozygote offspring were set up as brother-sister breeders, and offspring genotyped (wild-type, heterozygote, or homozygote), and challenged to induce liver damage, as described below.

Liver injury models

6-8 wk old mice were challenged IV with 15 μg/g ConA (Sigma-Aldrich) diluted in sterile saline. Blood was collected retro-orbitally just before, and 6 h after ConA administration. Mice were sacrificed at 24 h post challenge for collection of blood and liver tissue. For Fas-induced liver injury, mice received an IP injection of the agonistic anti-Fas mAb Jo-2 (BD Pharmingen) at either 0.05, 0.1, or 0.2 μg/g as indicated; at various time points, liver and blood were collected. For carbon tetrachloride (CCl₄) liver injury, mice received a single IP injection (2 μL/g in corn oil), and blood and tissue were collected at 24 h. Alanine aminotransferase (ALT), aspartate aminotransferase (AST), and lactate dehydrogenase (LDH) activities were measured in plasma as described (15). For every mouse, the pre-injection plasma ALT level was measured, and was consistently in the range of 20-80 U/L; these baseline data are not depicted.

Antibiotic treatment and microbiota reconstitution

JAX BALB/c mice were treated for 2 wk with 1.0 g/L ampicillin, 1.0 g/L neomycin, 1.0 g/L metronidazole and 0.5 g/L vancomycin administered with Splenda in the drinking water. Control mice received Splenda water. Fecal matter was collected before and after treatment. Bacterial DNA was isolated and quantitated as below. For reconstitution of microbiota, fecal pellets from 5-wk-old JAX BALB/c donor mice were homogenized in phosphate-buffered saline (PBS). 100 μL slurries were gavaged into 5-wk-old GF Swiss-Webster mice from TAC.

Flow cytometry

Following liver perfusion via the portal vein, tissue was dissociated in PBS with 10% fetal calf serum. Following a low speed spin to pellet hepatocytes, the supernatant was subject to a 40% Percoll gradient to isolate liver mononuclear cells (MNC). MNC were passed twice through 70 μm filters. Single-cell suspensions in PBS were incubated with viability dye, Fc-blocked (anti-CD16/CD32; eBioscience), and stained with mAbs to CD45, CD4, CD8α, TCRβ, CD11b, CD49b (clone DX5), GR-1, CD11c, and CD19 (Biolegend). iNKT cells were identified using PBS-57 tetramer, with empty CD1d tetramer as control (NIH tetramer facility). Cells were washed and fixed in 1% paraformaldehyde and acquired on a MACSQuant (Miltenyi). Data analysis used FlowJo.

Cytokine quantification

Livers were homogenized in 0.25M Sucrose, 10mM Tris, pH 7.4, Roche Protease Inhibitors at 120 mg (wet weight) per mL. Lysates were spun (10,000 RCF), and supernatants collected and stored at −80°C. Millipore mouse 32-plex cytokine bead array, or enzyme-linked immunosorbent assay (for IL-22; Biolegend), were used for cytokine measurement per manufacturer's protocol.

Analysis of genetic background

Mouse genetic backgrounds were assessed at the DartMouse™ Speed Congenic Core Facility at GSL. DartMouse uses the GoldenGate Genotyping Assay (Illumina, San Diego, CA) to interrogate 1449 single nucleotide polymorphisms (SNPs) distributed throughout the genome. Raw SNP data were analyzed using DartMouse's SNaP-Map™ and Map-Synth™ software.

Fecal DNA extraction and 16S quantification

As described (16), fecal samples were freshly collected and frozen at −80°C until processing. DNA extraction used the QIAamp DNA Stool Mini Kit (Qiagen). 16S gene copy quantification used 10 ng fecal DNA and universal bacteria primers. PCR reactions were as described (17). SFB was quantified by specific QPCR as described (18).

16S rRNA deep sequencing

Deep sequencing, bioinformatic quality filtering and operational taxonomic unit assignments were performed as described (19). Statistical analyses, abundance, Simpson diversity, and principal coordinates were calculated as described (20). The microbiome data sets have been deposited in the SRA database (SRP039453).

Histology and histopathological scoring

Livers were dissected and fixed in formalin at the time of euthanasia. Paraffin embedding, sectioning, and staining with hemotoxylin and eosin (H&E) were per routine. An arbitrary scoring system was used to evaluate severity of liver damage as follows: 1, normal liver, no inflammation or hepatocyte necrosis; 2, spotty necrosis/apoptosis or rare confluent necrosis; 3, confluent necrosis/apoptosis; 4, large area of confluent necrosis with/without bridging necrosis. Scoring of livers was done in a blinded fashion, such that the scorer was unaware of the treatment or group identity of any sample.

Statistical Analyses

Error bars represent standard deviation (SD). Statistical significance between two groups was determined using the Mann-Whitney U test, as appropriate. Differences among >2 mouse groups were evaluated using the Kruskal-Wallis one-way analysis of variance. For cytokine profile comparisons, the statistical test used was the two-way repeated measures ANOVA, followed by a post-hoc Tukey test. Significance is defined as P<0.05.

Mice from different vendors exhibit different susceptibility to ConA-induced liver injury

We began with the broad hypothesis that environment has a critical effect on susceptibility to acute T cell mediated liver injury. We administered ConA to groups of BALB/c mice born and raised in distinct environments: Taconic Farms (TAC), National Cancer Institute (NCI), and the Jackson Laboratory (JAX). Although all groups exhibited liver injury, there was a clear vendor-specific effect. Specifically, JAX mice were more susceptible, exhibiting a ∼ten-fold-greater increase in plasma ALT and AST levels (Fig. 1A,B). In addition, JAX BALB/c mice showed significantly more liver damage than did TAC BALB/c mice, as measured both by LDH (Fig. 1C) and by histopathological scoring (Fig. 1D). A similar vendor effect was observed for C57Bl/6 mice (Supplemental Fig. 1). For subsequent experiments, we primarily compared JAX BALB/c mice with TAC BALB/c mice.

Genetic drift after decades of separate breeding may account for the phenotypic differences; however, genetic polymorphism analysis using a dense SNP panel showed minimal differences between JAX BALB/c mice and TAC BALB/c mice (Supplemental Fig. 2). To definitively rule out genetics, we ConA-challenged genetically identical BALB/cJ mice raised in different environments. BALB/cJ mice were imported from JAX and bred at Geisel (GSL) for several generations, to allow them to habituate to the new environment. ConA caused markedly less liver injury in GSL-bred BALB/cJ mice than in freshly imported JAX BALB/cJ mice (Fig. 2A).

Susceptibility to ConA-induced liver injury is transmissible between co-housed mice and modulated by commensal intestinal flora

Environmental factors to be considered include not only the microbiota, but also food, bedding, handling, etc. Among these, microbiota's effects should be transmissible between co-housed mice, whereas effects due to the other factors should not. Therefore, we co-housed JAX mice with GSL-habituated mice for several weeks, then challenged the mice with ConA. Control JAX mice and GSL mice responded with high and low susceptibility, respectively, as before. Co-housed mice exhibited intermediate susceptibility (Fig. 2B; Supplemental Fig. 3A). Importantly, co-housed GSL mice experienced the identical environment as did control GSL mice except for exposure to JAX mice, yet became more susceptible. Reconstitution of GF Swiss-Webster mice with fecal microbiota also enhanced susceptibility to ConA (Fig. 2C; Supplemental Fig. 3B). Together, these data demonstrate that environment modifies susceptibility to ConA-mediated liver injury in adult mice, and that such susceptibility is altered through manipulation of intestinal flora, as evidenced by the co-housing and GF-mouse reconstitution results.

Composition of the microbiota in JAX mice and in TAC mice

The above results suggest that JAX BALB/c mice and TAC BALB/c mice harbor distinct intestinal microbiota profiles, similar to results previously shown for C57Bl/6 mice from these vendors (17).To test this hypothesis directly, we evaluated fecal microbiota composition using 16S rRNA gene sequencing. Overall, JAX mice and TAC mice had similar diversity in fecal bacteria (Fig. 3A). Principal Coordinate Analysis confirmed that the microbial community structures were strongly influenced by vendor (Fig. 3B). Among 441 operational taxonomic units (OTU) identified at the family/genus level, some OTU were identifiably different in frequency between JAX samples and TAC samples (Fig. 3C).

We sought to identify bacterial taxa that may be associated with liver injury. We first examined segmented filamentous bacteria (SFB), a Clostridial species that has previously been associated with intestinal inflammation in mice and is known to be a component of the intestinal flora in C57Bl/6 mice from TAC but not from JAX (17). By SFB-specific qPCR (18), SFB was abundant in BALB/c TAC samples, but at or below detection in BALB/c JAX samples (Fig. 4A), similar to results for C57Bl/6 mice (17). Only TAC samples yielded signal high enough for assessment of relationship to liver injury; among these, however, there was no correlation between SFB abundance and ALT levels (Fig. 4B). Next, we applied a rigorous filter of the microbiome data to identify taxa of reasonable abundance (>0.05%), whose frequency correlated with liver injury as quantified using ALT (R²>0.35; two-tailed Pearson correlation: p<0.05). This analysis yielded a single family (Ruminococcaceae; frequency: 3% to 17%). The relative amount of Ruminococcaceae in the fecal microbiota among these samples was significantly and positively correlated with susceptibility to ConA induced liver injury (Fig. 4C). Ruminococcaceae is a family of anaerobic Gram (+) bacteria within the class Clostridia, and is distinct from SFB.

Hepatic immune cell composition and cytokine profiles in JAX mice and in TAC mice

We sought to identify possible mechanism(s) through which microbiota may regulate susceptibility to ConA. First, we considered whether liver resident immune cell number or composition differed significantly between JAX BALB/c mice and TAC BALB/c mice. Livers from JAX BALB/c mice harbored twice as many CD45⁺ leukocytes (Fig. 5A). Among lymphoid cells, there were no statistically significant differences in frequencies of total T cells, CD4⁺ T cells, INKT cells, and B cells. TAC livers harbored a higher frequency of CD8⁺ T cells, whereas JAX livers harbored a higher frequency of NK cells (Fig. 5B). Among myeloid cells, there were similar frequencies of CD11b⁺Gr1⁺ cell and CD11b⁻Gr1⁺ cell sub-populations, with JAX livers exhibiting slightly higher frequencies of CD11b⁺Gr1⁻ cells, and of CD11c⁺ (dendritic) cells (Fig. 5C). As the frequencies of lymphoid cells and myeloid cells were overall similar between JAX and TAC, the higher numbers of CD45⁺ cells in JAX livers reflects a largely non-specific increase in all CD45⁺ cells. Since differences in environment result in differences in liver injury that range from 5- to 20-fold (Figs. 1, 2), a difference in liver immune cell numbers of two-fold, with overall similarity in cell composition, seems unlikely to significantly account for differences in ConA susceptibility, and we therefore looked at additional factors.

ConA-induced liver damage is heavily dependent on specific cytokines and chemokines; we therefore examined cytokine production profiles. We prepared lysates of mouse livers 0 to 24 h following ConA, and measured numerous cytokines and chemokines by cytokine bead array. Shortly after ConA administration, activated iNKT cells release large amounts of interferon (IFN)-γ, interleukin (IL)-4, and IL-17A respectively, each of which is required for ConA-induced liver damage (21-23). Here, IFN-γ and IL-4 were rapidly produced in both groups, but were not different between JAX samples and TAC samples, whereas IL-17 was actually higher in TAC samples (Fig. 5D). Thus, these data argue against differential Th1/2/17 cytokine release as a relevant mechanism to explain differential injury. For some other cytokines, there were statistically significant but modest differences between JAX samples and TAC samples (Supplemental Fig. 4 A,B), but most showed no difference (Supplemental Fig. 4C), Again, since environment induces dramatic differences in liver injury, the cytokine/chemokine profiles observed do not satisfactorily explain the effect of environment.

Fas triggering results in greater liver injury in JAX mice than in TAC mice

We next considered the hypothesis that JAX mice and TAC mice differ in response to triggering of the Fas pathway. Hepatocytes constitutively express the Fas death receptor, and its cross-linking results in acute liver injury. Several reports show that ConA liver injury is dependent on an intact Fas/FasL system (14,24), and we independently verified these observations (Supplemental Fig. 5). Importantly, Fas-deficient mice and Fas-intact mice raised as littermates, and therefore exposed to similar microbiota, showed significant differences in response to ConA (Fig. 6A).

For subsequent experiments we induced liver injury using a reagent that would directly trigger the Fas response pathway. Jo-2 is an agonistic Fas mAb that rapidly cross-links Fas and causes acute hepatocyte apoptosis. We injected Jo-2 at varying doses to JAX BALB/c mice and to TAC BALB/c mice, and measured ALT at 10 and 24 hours. Liver damage was much more severe in JAX mice than in TAC mice, particularly at higher doses (Fig. 6B,C), a finding confirmed in multiple repeat experiments using a single dose of Jo-2 (Supplemental Fig. 6). In additional studies, we observed no difference in liver damage between JAX mice and TAC mice following acute challenge with CCl₄ (Supplemental Fig. 7), a hepatotoxic agent that injures liver through oxidative stress. Thus, a difference in susceptibility to liver injury is observed between JAX mice and TAC mice in response to some types of injury (ConA and Fas), but not others (CCl₄).

Microbiota regulates sensitivity to Fas-induced liver injury: requirement for MyD88 signaling

To test whether the sensitivity of mice to liver damage induced by Fas triggering is modulated by microbiota, we treated JAX mice with oral antibiotics for two weeks. Quantitative 16S PCR of fecal contents confirmed effective reduction in fecal bacteria load by ∼500-fold (Fig. 7A); in addition, enlargement of the cecum was observed, as commonly found after severe reduction in commensal organisms (Fig. 7B; compare top versus bottom). Furthermore, treatment with antibiotics did not impact liver baseline ALT levels (Supplemental Fig. 8A). Treatment with antibiotics greatly reduced Fas-induced liver damage (Fig. 7C; Supplemental Fig. 8B). Conversely, GF Swiss-Webster mice that had been reconstituted with JAX commensal bacteria were significantly more sensitive to Fas-induced liver damage than were non-reconstituted GF mice (Fig. 7D; Supplemental Fig. 8C). Finally, mice deficient in the TLR adaptor molecule MyD88 mice were more resistant to Fas-induced liver damage than were littermate MyD88-intact mice (Fig. 7E; Supplemental Fig. 8D).