Impaired Autophagy in Intestinal Epithelial Cells Alters Gut Microbiota and Host Immune Responses

The homeostasis of host-microbiota interactions is of great importance to host health. Previous studies demonstrated that disruption of autophagy was linked to inflammatory bowel disease. However, the interaction mechanism of gut microbiota regulated by autophagy was obscure. In an intestinal epithelium-specific autophagy-related 5 (Atg5) knockout mouse model, we observed a significant alteration and decreased diversity in the gut microbiota of Atg5-deficient mice compared with that of wild-type mice. Although the numbers of some organisms (e.g., Akkermansia muciniphila and members of the Lachnospiraceae family) associated with the control of inflammation decreased, those of proinflammationory bacteria (e.g., “Candidatus Arthromitus”) and potential pathogens (the Pasteurellaceae family) increased in Atg5^−/− mice. Differential gene expression analysis revealed that two key genes, RORC and TBX21, involved in inflammatory bowel disease were upregulated in Atg5^−/− mice. Our study suggests that Atg5 deficiency results in an imbalance of the host-microbe interaction and deterioration of the gut microenvironment.

ABSTRACT

Establishing and maintaining beneficial interactions between the host and associated gut microbiota are pivotal requirements for host health. Autophagy is an important catabolic recycling pathway that degrades long-lived proteins and some organelles by lysosome to maintain cellular homeostasis. Although impaired autophagy is thought to be closely correlated with Crohn's disease (CD), the functional role of autophagy in the maintenance of gut microbiota is poorly understood. As autophagy-related 5 (Atg5) is a key gene associated with the extension of the phagophoric membrane in autophagic vesicles, we established a gut-specific Atg5 knockout mouse model, and we found that the disruption of autophagic flux in the intenstinal epithelium cells dramatically altered the composition of the gut microbiota and reduced alpha diversity. Microbial function prediction indicated that the pathway allocated for infectious diseases was enriched in Atg5^−/− mice. “Candidatus Arthromitus” and the Pasteurellaceae family were increased in Atg5^−/− mice, whereas Akkermansia muciniphila and the Lachnospiraceae family were reduced. Transcriptome analysis revealed that two key inflammatory bowel disease (IBD)-related transcription factors, RORC and TBX21, of host cells were upregulated in Atg5^−/− mice, thus elevating the Muc2-related immunological response. The findings suggest that intestinal autophagy plays a vital role in modulating the diversity and composition of gut microbiota.

IMPORTANCE The homeostasis of host-microbiota interactions is of great importance to host health. Previous studies demonstrated that disruption of autophagy was linked to inflammatory bowel disease. However, the interaction mechanism of gut microbiota regulated by autophagy was obscure. In an intestinal epithelium-specific autophagy-related 5 (Atg5) knockout mouse model, we observed a significant alteration and decreased diversity in the gut microbiota of Atg5-deficient mice compared with that of wild-type mice. Although the numbers of some organisms (e.g., Akkermansia muciniphila and members of the Lachnospiraceae family) associated with the control of inflammation decreased, those of proinflammationory bacteria (e.g., “Candidatus Arthromitus”) and potential pathogens (the Pasteurellaceae family) increased in Atg5^−/− mice. Differential gene expression analysis revealed that two key genes, RORC and TBX21, involved in inflammatory bowel disease were upregulated in Atg5^−/− mice. Our study suggests that Atg5 deficiency results in an imbalance of the host-microbe interaction and deterioration of the gut microenvironment.

INTRODUCTION

The mammalian gastrointestinal tract is colonized by trillions of bacteria, which have interacted with the host over millions of years (1). Intestinal homeostasis involves reciprocal communication between the host and gut microbiota and plays vital roles in the host's physiological metabolism and immune and nervous systems (2–4). A previous review revealed that an imbalance in the composition of the gut microbiota was associated with intestinal diseases (5). Host-related genetic factors, especially those associated with the immune system, shape the gut microbiota, as shown by many studies which demonstrated that deficiencies in specific genes involved in immune responses played important roles in regulating the distribution of gut microbiota. For example, a mutation of nucleotide-binding oligomerization domain protein 1 influenced the structure and composition of gut microbiota due to the failure of peptidoglycan recognition and enhanced systemic innate immunity (6). A deficiency in recombination-activating gene 2 or deletion of immunoglobulin A affected gut microbial composition and led to increased abundance of segmented filamentous bacteria (SFB) (7). Research also demonstrated that specific microbes modulated the generation of some subclasses of immune cell populations and reestablished a new steady-state immune system. Colonization of the gut by SFB induced an increase in T helper 1 (T_H1) and T helper 17 (T_H17) (8) cells. Polysaccharide A of Bacteroides fragilis resulted in interleukin 10 (IL-10) feedback by intestinal T cells, which restrained the accumulation of T_H17 cells and potentially damaged the mucosal barrier (9). Although cross talk between the gut microbiota and host immune system is well known, the interaction mechanisms remain obscure.

Autophagy is a membrane-trafficking process that primarily involves lysosomal degradation of cellular components, such as long-lived proteins and organelles. Autophagy is indispensable for cell growth, adaptation to stressful conditions, intracellular quality control, renovation during development and differentiation, and organelle biogenesis (10–13). The process of autophagy includes the initiation, nucleation, expansion, maturation, and degradation; more than 40 autophagy-related (ATG) proteins have been characterized during autophagy (14, 15). Nutrient, carbon, or nucleic acid starvation and some physiological stress stimuli could trigger the induction of autophagy in many organisms (16, 17). Autophagy primarily serves as a modulator to prevent organisms from suffering due to diverse pathologies (18). For example, research reported that various autophagy-related genes, such as Atg16L1, Atg5, and Atg7, played important roles as intestinal immunological barriers in small intestinal Paneth cells and that the disruption of these genes was correlated with Crohn's disease (CD) (19). Research also reported that Atg7 deficiency influenced the composition and structure of gut microbiota and that it was involved in inhibiting tumorigenesis (20). However, when the mutation was accompanied by Apc deletion, it contributed to the occurrence of colorectal cancer (20). The structural composition of gut microbiota of colorectal cancer patients was imbalanced compared with that of a healthy population (21). The functional role of autophagy in interacting with the gut microbiota is unknown. Given the role of the interaction between the host and gut microbiota in regulating the physiological activities of the host, it is imperative that the potential mechanism of interaction between the gut microbiota and autophagy is explored.

Previous research demonstrated that Atg5 was required for autophagosome formation (22) and that it was necessary for maintaining the function of Paneth cells (19, 23). Antimicrobials secreted by Paneth cells were disseminated in the mucous layer and mucins, especially mucins produced by goblet cells, which, together with antimicrobials, prevented microbes from contacting the epithelial surface and being exposed to immune cells directly by forming a physical and biochemical barrier (24). Other studies reported that Atg5 contributed to antibiosis, especially by increasing susceptibility to infection and controlling dissemination of Listeria monocytogenes, Mycobacterium tuberculosis, and Salmonella (25–29). Thus, Atg5 was selected for further studies to explore the mechanism of interaction between autophagy and the gut microbiota.

A previous study reported that the influence of Toll-like receptor deficiency on gut microbiota was weaker under homeostatic conditions than under long-term breeding (30). In our previous studies, we demonstrated that the host genotype markedly influenced the composition of the gut microbiota after controlling for other factors, including the effects of diet and the environment (31–33). In the present study, all the experimental animals were maintained in essentially identical environments, except for differences in host genotype. To that end, a gut-specific Atg5 knockout mouse model was established. Then we applied next-generation sequencing technology to identify the composition of gut microbiota from the duodenum, jejunum-ileum, cecum, colon, and feces of C57BL/6J mice and C57BL/6J Atg5^−/− mice. In addition, according to the analysis of the transcriptome, we found that some components of the immune system were clearly increased in the gut of the C57BL/6J Atg5^−/− mice. By integrating these results with findings regarding the gut microbiota, we aimed to explore the possible mechanism of interaction between host autophagy and gut microbiota.

RESULTS

Establishment of an intestinal epithelium-specific Atg5 knockout mouse model.

To investigate the functional role of autophagy in intestinal disease, we specifically knocked out Atg5 in the intestinal epithelium by crossing C57BL/6J mice with a floxed allele of Atg5 (34) with C57BL/6J PVillin-Cre mice. The latter, which specifically express Cre recombinase in the intestinal epithelium from the duodenum to the cecum, were generated by the Nanjing Biomedical Research Institute of Nanjing University (T000142), which could specifically conditionally knock out the autophagy gene Atg5 in the intestinal epithelium of the small intestine and colon. These intestinal epithelium-specific Atg5 knockout C57BL/6J mice were called Atg5^−/− mice. To examine whether the knockout was indeed successful, we first detected the protein level of ATG5 in intestinal epithelium homogenates from Atg5^+/+ and Atg5^−/− mice by immunoblotting and found that ATG5 was dramatically reduced in the intestinal epithelium of Atg5^−/− mice compared with that of control groups (Fig. 1a). Consistent with the function of Atg5 in autophagy (35), LC3-I accumulated in Atg5^−/− mice, whereas the membrane-associated form, LC3-II, and the autophagic substrate, SQSTM1/p62, did not (Fig. 1a). An immunofluorescence analysis of LC3 showed the disappearance of punctate structures (representing autophagosomes) in intestinal epithelial cells of Atg5^−/− mice (Fig. 1b). The results suggested that autophagic flux was destroyed in the intestinal epithelium of Atg5 knockout mice. Previous research reported that disruption of autophagy in the intestinal epithelium caused abnormalities in Paneth cells associated with CD (19, 20). Remarkably, Paneth cells of Atg5-deficient mice showed morphological abnormalities (Fig. 1c). These results indicated that we had generated an intestinal epithelium autophagy-deficient mouse model and that disruption of autophagy in the intestinal epithelium was responsible for the observed Paneth cell pathology.

FIG 1.

Intestinal epithelium-specific knockout of Atg5 in mice. (a) Autophagic flux was disrupted in Atg5-deficient intestinal epithelia. The intestinal epithelia of Atg5^+/+ and Atg5^−/− mice were extracted and analyzed by immunoblotting using anti-ATG5, anti-p62, and anti-LC3 antibodies. Actin served as a loading control. (b) LC3 punctate structures disappeared in Atg5-deficient intestinal epithelia. Immunofluorescence detection of LC3 (green) in Atg5^+/+ and Atg5^−/− mouse intestinal epithelia is shown. The nucleus (blue) was counterstained with DAPI. (c) Morphological abnormalities in Atg5-deficient Paneth cells, as shown by a histological examination of the intestines in Atg5^+/+ and Atg5^−/− mice.

Atg5 deficiency altered gut microbiota composition and structure.

To study the effect of knockout of Atg5 on the gut microbiota, we collected the intestinal contents of the duodenum, jejunum-ileum, cecum, and colon, as well as feces, and analyzed the microbiome by deep sequencing. A total of 9,342,844 filtered reads resulted from high-throughput sequencing of 60 samples, and they were classified into different taxonomies and diversity analyses. Taxonomies that appeared in at least one-fourth of 60 samples were regarded as common, and their relative abundance in samples was employed for further analyses. Phylogenetic dissimilarities between communities of microorganisms among the microbiota samples were analyzed using UniFrac metrics (36). First, on the basis of the UniFrac metrics, the canonical analysis of principal coordinates (CAP) showed separation among the samples from different intestinal locations (Fig. 2a). Samples derived from the large intestine (cecum, colon, and feces) clustered with each other, and samples from the small intestine (duodenum and jejunum-ileum) clustered together. The results of a correlation analysis indicated that the correlation coefficient was very high, over 0.85 (P < 0.01), among the large intestinal segments but was less than 0.78 (P < 0.01) for the correlation between large and small intestinal segments (see Table S3 in the supplemental material), suggesting that the distribution of the microbiome in different intestinal segments was diverse. In addition, samples that originated from different genotypes clustered together (Fig. 2a). Atg5 status did influence the composition of gut microbiota, as confirmed by a nonparametric permutational multivariate analysis, Adonis (P = 0.001; duodenum, jejunum-ileum, cecum, colon, and feces from Atg5^+/+ and Atg5^−/− mice). We further confirmed that samples from Atg5^+/+ and Atg5^−/− mice were significantly different using analysis of similarities (ANOSIM) (P = 0.001, P = 0.001, P = 0.05, P = 0.001, and P = 0.05). Alpha diversity was evaluated by the Simpson and Shannon indices. The Simpson and Shannon indices for colon and feces were significantly different (P < 0.05) for Atg5^+/+ and Atg5^−/− mice (Fig. 3a), pointing to substantially decreased alpha diversity in Atg5^−/− mice. A Venn diagram of operational taxonomic unit (OTU) numbers further confirmed the conclusion that reduced for diversity was caused by Atg5 deficiency (Fig. 3b). However, alpha diversity and OTU numbers in other intestinal segments were similar in the two groups (P > 0.05). Thus, these results revealed that host genotypes played a vital role in shaping the gut microbiome.

FIG 2.

(a) Canonical analysis of principal coordinates based on unweighted UniFrac metrics; (b) relative abundance distribution of Verrucomicrobia in Atg5^+/+ and Atg5^−/− mice.

FIG 3.

Comparison of alpha diversity. (a) Shannon and Simpson index comparisons of gut microbiota in the colon and feces of Atg5^−/− and Atg5^+/+ mice; (b) OTUs detected in the colon and feces of Atg5^−/− and Atg5^+/+ mice.

Comparisons of bacterial taxonomy in Atg5^+/+ and Atg5^−/− mice.

To further investigate the gut microbiota in autophagy-deficient mice, we compared the variation in bacterial taxa between Atg5^+/+ and Atg5^−/− mice. Firmicutes, Proteobacteria, and Bacteroidetes, which generally constituted more than 90% of the entire sequences in each group, dominated the gut microbiome of mice. However, Cyanobacteria, Actinobacteria, Tenericutes, and Verrucomicrobia were present as minor constituents (Fig. S1). Verrucomicrobia were significantly decreased in Atg5^−/− mice (P < 0.05) (Fig. 2b).

Using linear discriminant analysis (LDA) effect size (LEfSe), an algorithm that can robustly identify features that are significantly different among biological groups, we identified significant biomarkers at various taxonomic levels. At the family level, Pasteurellaceae were enriched in whole intestinal segments of Atg5^−/− mice, especially in the jejunum-ileum (LDA = 4.02) (Fig. 4a). Clostridiaceae (LDA = 4.08) also accumulated in the jejunum-ileum in Atg5^−/− mice (Fig. 4b). In the duodenum, Pseudomonadaceae (LDA = 4.74) and Ruminococcaceae (LDA = 4.12) were enriched in Atg5^−/− and the Atg5^+/+ groups, respectively (Fig. 4c and d). In large intestinal segments (cecum, colon, and feces) of Atg5^−/− mice, Lachnospiraceae (LDA > 4.40) and Verrucomicrobiaceae (LDA > 3.15) were present at significantly higher levels in Atg5^+/+ mice (Fig. 4e and f). In contrast, Lactobacillaceae (LDA = 3.86) levels were lower in the cecum of Atg5^+/+ mice (Fig. 4g).

FIG 4.

Comparisons of the relative abundances of abundant bacterial families in Atg5^+/+ and Atg5^−/− mice in different sites. *, P < 0.05; **, P < 0.001.

At the genus level, the results showed that 23, 25, 33, 35, and 23 genera were significantly altered in the duodenum, jejunum-ileum, cecum, colon, and feces, respectively, of Atg5^−/− mice compared with the respective anatomical sites in control mice (Table S4). Among these altered genera, 20 were identified to be the most altered biomarkers (LDA score > 3.5 and P < 0.05) (Table 1). Of these, we observed that with the exception of the Sutterella and Veillonella genera, seven genera (Pseudomonas, Aggregatibacter, Klebsiella, “Candidatus Arthromitus,” Mannheimia, Gemella, and Streptococcus) were all enriched in the jejunum-ileum of Atg5^−/− mice. In particular, the abundance of “Candidatus Arthromitus” was dramatically increased (LDA score = 4.09 and P < 0.05) (Table 1). In the colons of Atg5^−/− mice, the relative abundance of seven genera (Acinetobacter, Akkermansia, Ochrobactrum, Brevibacterium, Ruminococcus, Sphingomonas and Meiothermus) was all significantly decreased. The abundance of Akkermansia was decreased in large intestine segments of Atg5^−/− mice, including the cecum, colon, and feces (LDA score > 3.5 and P < 0.05), whereas Lactobacillus was enriched in the cecum of Atg5^−/− mice (LDA score = 4.36 and P < 0.05) (Table 1).

TABLE 1.

The most different genera (LDA score > 3.5) between Atg5^+/+ and Atg5^−/− mice in various intestinal segments

Gut position	Genus	High groupa	LDA score	P valueb
Duodenum	Klebsiella	Atg5−/−	4.558	0.013*
	Labrys	Atg5−/−	3.948	0.000**
	Mannheimia	Atg5−/−	3.885	0.009**
	Pseudomonas	Atg5−/−	4.953	0.009**
	Serratia	Atg5+/+	3.613	0.004**
	Meiothermus	Atg5+/+	4.546	0.012*
Jejunum-ileum	Aggregatibacter	Atg5−/−	4.241	0.009**
	Klebsiella	Atg5−/−	4.108	0.016*
	Mannheimia	Atg5−/−	4.081	0.041*
	Pseudomonas	Atg5−/−	4.661	0.021*
	Sutterella	Atg5+/+	4.95	0.033*
	Veillonella	Atg5+/+	3.728	0.000**
	“Candidatus Arthromitus”	Atg5−/−	4.094	0.032*
	Gemella	Atg5−/−	3.585	0.001**
	Streptococcus	Atg5−/−	3.578	0.041*
Cecum	Akkermansia	Atg5+/+	3.541	0.001**
	Lactobacillus	Atg5−/−	4.367	0.008**
Colon	Acinetobacter	Atg5+/+	4.091	0.013*
	Ochrobactrum	Atg5+/+	3.764	0.013*
	Sphingomonas	Atg5+/+	3.602	0.004**
	Akkermansia	Atg5+/+	3.939	0.000**
	Brevibacterium	Atg5+/+	3.667	0.001**
	Meiothermus	Atg5+/+	3.577	0.000**
	Ruminococcus	Atg5+/+	3.606	0.016*
Feces	Akkermansia	Atg5+/+	40136	0.001**
	Bacteroides	Atg5−/−	3.625	0.043*

^aThe line with the significantly greater relative abundance.

^b*, P < 0.05; **, P < 0.01.

Next, we investigated the influence of autophagy deficiency at the species level. In Atg5^−/− mice, 7, 10, 12, 14, and 13 species were significantly different from the controls in the duodenum, jejunum-ileum, cecum, colon, and feces (Fig. 5; see also Fig. S2), respectively. Interestingly, Akkermansia muciniphila was the most significant biomarker (LDA score > 4.4 and P < 0.05) and was remarkably enriched in the large intestine segments of Atg5^+/+ mice (cecum, colon, and fecal samples). However, the levels of “Candidatus Arthromitus” spp. (LDA score > 4.2 and P < 0.05) were distinctly increased in all the intestinal samples from Atg5^−/− mice, except for the duodenum (Fig. 5; see also Fig. S2). In addition, Aggregatibacter pneumotropica, the most significantly altered species, was expanded throughout the whole intestine of Atg5^−/− mice, particularly in the jejunum-ileum (LDA score = 4.66 and P < 0.05) (Fig. 5; see also Fig. S2). These results hinted that disruption of autophagy had a severe impact on the composition and population structure of the gut microbiota.

FIG 5.

LEfSe identified species with significantly different abundances in different intestinal segments in Atg5^+/+ and Atg5^−/− mice. Taxa enriched in Atg5^+/+ mice are shown with a positive linear discriminant analysis (LDA) score (green), and taxa enriched in Atg5^−/− mice are shown with a negative score (red). Only taxa meeting an LDA significant threshold of >2 and P value of <0.05 are shown.

Gene functional enrichment of gut microbiota.

Functional prediction of microbiota was performed using PICRUSt to predict differences in gene functional enrichment in different intestinal segments of Atg5^+/+ and Atg5^−/− mice. After P value correction, the results revealed that microbiota functions shifted remarkably in small intestinal segments after the disruption of Atg5. In the duodenum, environmental adaptation and nucleotide metabolism pathways were the most significant in Atg5^+/+ mice (Fig. S3). Endocrine system and energy metabolism pathways were enriched in the jejunum-ileum of Atg5^+/+ mice. However, pathways of infectious diseases were enriched in small intestinal segments (duodenum and jejunum-ileum) of Atg5^−/− mice (Fig. 6a; see also Table S5), indicating that deletion of Atg5 may potentially expose the host to attack by pathogenic bacteria. As shown in Fig. 4a, the Pasteurellaceae family, which contains a number of potential pathogens (37), was enriched in Atg5^−/− mice, further supporting this hypothesis. Thus, autophagy-deficient mice may be susceptible to attack by pathogenic bacteria.

FIG 6.

(a) Comparisons of functional microbiota pathways in Atg5^+/+ and Atg5^−/− mice. (b) The lysozyme distribution in Atg5-deficient intestinal epithelia was detected using immunofluorescence detection of LAMP1 (red) in Atg5^+/+ and Atg5^−/− mice. The nucleus (blue) was counterstained with DAPI. (c) Immunohistochemistry detection of MUC2 in Atg5^+/+ and Atg5^−/− mice.

Altered host immune response.

In intestinal epithelium, autophagy exerts selective influences on the cell biology and specialized regulatory properties of Paneth cells, which are critically important intestinal innate immune cells. Thus, we speculated that disruption of Atg5 might influence the immune response. As indicated by levels of immunofluorescence of LAMP1, cytoplasmic lysosomes were clearly decreased in Atg5-deficient Paneth cells (Fig. 6b). This finding was consistent with those of previous reports (19, 23). In addition to the decline in cytoplasmic lysozymes levels, we observed that the host cell may be vulnerable to pathogenic infections. The latter was also indicated by microbial functionality predictions (Fig. 6a). In support of this idea, we found that key protective molecules, including MUC2, originated from goblet cells. In addition, MUC2 was increased in Atg5-deficient intestinal epithelia based on immunohistochemistry analysis (Fig. 6c). These findings suggested that the inner layer of the intestine was altered and that the immune response was overactivated in Atg5-deficient tissues.

To further analyze global shifts in the host immune response, we performed transcriptome sequencing to detect differentially expressed genes (DEGs). We used DESeq to identify the DEGs. A total of 349 DEGs were identified (P ≤ 0.05; fold change ≥ 2) in either Atg5^+/+ or Atg5^−/− mice (Table S6). A hierarchical-clustering heat map was used to cluster samples into groups based on Atg5 status (Fig. S4). Of the DEGs, 222 genes were upregulated and 127 were downregulated in Atg5^−/− mice (Table S6). Characterization of Gene Ontology Consortium (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations of the 349 DEGs by DAVID identified 40 DEGs associated with the immune response. Furthermore, the expression of Defa5, a gene that encodes a protein that synthesizes an important antimicrobial peptide in Paneth cells of the ileum, was markedly reduced in Atg5^−/− mice (Fig. 7a). Interestingly, two key transcription factors, RORC and TBX21, which are related to T_H17 and T_H1 cell differentiation, respectively, and are known to be involved in the inflammatory bowel disease (IBD) pathway, were significantly elevated in Atg5^−/− mice (Fig. 7a). In addition, the expression levels of CD5, CD6, CD7, and CD96 were all increased in Atg5-deficient mice (Fig. 7b). Some inflammation-associated genes, such as CCL5, TLE, and THEMIS, were also differentially expressed due to the loss of Atg5 (Fig. 7b). These results suggested that the immune response in the Atg5^−/− mice was dramatically altered.

FIG 7.

(a) Comparisons of transduction-related Defa5, RORC, and TBX21 mRNAs, which differed in expression in Atg5^+/+ and Atg5^−/− mice. (b) Heat map of 40 DEGs associated with the immune response in Atg5^+/+ and Atg5^−/− mice.

DISCUSSION

Evidence indicates that the imbalance between host genes and microbes is an important contributor to disease occurrence (5, 38–40). Autophagy plays a protective role in combatting intestinal inflammation by maintaining the shape and function of Paneth cells (19, 23). In this study, we used an intestinal epithelium-specific Atg5 knockout mouse model to characterize an additional vital role of autophagy: the regulation of gut microbial communities. We observed strong alteration of the composition of the gut microbiota and a reduction of alpha diversity in the colon of Atg5^−/− mice (Fig. 2), similar to previous findings for patients with CD (41). Many factors regulate the gut microbiota, of which the host immune system is the major factor among them. We found that dysfunction of Paneth cells led to decreased expression of defensins (Fig. 7a). Previous research reported that defensins played a vital role in the composition of the microbiota (42) and that dysfunction or reduced expression of defensins was associated with greater susceptibility to enteric infection and the initiation of ileal CD (43). Research also showed that elevated MUC2 was associated with the induction of high levels of reactive oxygen species production, which further stimulated endoplasmic reticulum stress and apoptosis in goblet cells and subsequently damaged the intestinal mucus barrier function in some pathologies (44). In this study, autophagy played an important role in maintaining Paneth cell morphology and function, as well as host intestinal health.

Many protective intestinal bacterial groups are involved in the maintenance of gut homeostasis (45, 46). Numerous IBD-related genes regulate antimicrobial defenses and intestinal immune homeostasis (47). Previous research demonstrated that the Nlrp12 gene was associated with innate immunity and that basal colonic inflammation increased in Nlrp12-deficient mice, which resulted in a less diverse microbiome and a reduction in the relative abundance of Lachnospiraceae (48). This finding was consistent with that in the present study, in which the relative abundances of Lachnospiraceaeae and Ruminococcaceae were reduced in the large intestine and duodenum, respectively, in Atg5^−/− mice. A previous study showed that both these bacterial families were negatively associated with the occurrence of CD disease (49). Although the relative abundance of Lachnospiraceae is decreased in patients with IBD, limited information is available on how these protective microbes regulate the host immune system. Previous research reported that Lachnospiraceae and Ruminococcaceae were preferentially enriched in mucosal folds (50) and that they engaged in cross talk with intestinal immune cells directly, suggesting that these microbes may serve as immunological regulators to prevent enteric pathogen adhesion and colonization. Bacteria belonging to the Lachnospiraceae and Ruminococcaceae are mainly responsible for the production of butyrate (51), which enables dendritic cells to accelerate regulatory T (Treg) cell differentiation (52). Thus, the loss of bacteria belonging to these two families in Atg5^−/− mice might lead to a reduction in short-chain fatty acids, which could be associated with elevated inflammation in Atg5^−/− mice. In this study, the relative abundance of Pasteurellaceae was significantly enriched in the jejunum-ileum of Atg5^−/− mice (Fig. 4a). The Pasteurellaceae family comprises mainly Gram-negative bacteria and includes multiple pathogens. These bacteria possess an outer membrane consisting mainly of lipopolysaccharides (37), which induce recruitment and activation of inflammatory cells and proinflammatory cytokine production (53). In previous research, an increase in the relative abundance of Pasteurellaceae was positively correlated with IBD (49). In our study, according to the prediction analysis of microbial function, the pathway allocated for infectious diseases was enriched in Atg5^−/− mice (Fig. 6a). Taken together, these results indicate that impaired autophagy may alter the colonization patterns of gut microbiota, thereby disrupting microbiota as a physical barrier to maintain intestinal immune homeostasis.

A. muciniphila is a mucin-degrading bacterium which preferably inhabits the colon of humans and rodents (54). In this study, A. muciniphila was reduced in the large intestine of Atg5^−/− mice and was the most significant biomarker (Fig. 5). Previous studies indicated that mice fed a high-fat diet and treated with A. muciniphila via the oral route had elevated levels of intestinal endocannabinoids and that these controlled inflammation, enhanced mucus layer thickness and the gut barrier, increased the ability of the mucus layer to harvest energy from the diet, and increased gut peptide secretion (55, 56). A reduction in the relative abundance of A. muciniphila was linked to IBD (57) and obesity (56). Thus, the absence of A. muciniphila caused by Atg5 deficiency may contribute to aggravating colonic inflammation and increase the risk of obesity.

In this study, “Candidatus Arthromitus” (also known as SFB) was another important biomarker regulated by autophagy. “Candidatus Arthromitus” was predominantly and persistently expanded in Atg5^−/− mice (Fig. 5). This phenomenon was consistent with findings in a previous study, which demonstrated distinct expansion in the abundance of SFB in the small intestines of mice lacking activation-induced cytidine deaminase and immunoglobulin A (7, 58). In addition, treatment with antibiotics inhibited colonization by SFB (59). These results hint that loss of selective stressors, including immune pressure or antibiotics, may enable overcolonization of SFB in the intestine. SFB are tightly associated with the epithelium (60), and they play an important role in stimulating the mucosal immune system. Thus, autophagy might participate in the immune response against SFB expansion. Colonization of the small intestine by SFB is sufficient to induce the appearance of T_H17 cells (8, 61), which have a potential function in the pathogenesis of IBD (62) and autoimmune disease (61). Although there is no strong evidence demonstrating that SFB can induce IBD, SFB have been discovered at some inflammatory sites in IBD patients (63), suggesting that overexpansion of SFB may result in excessive immune responses and further promote inflammation of the intestine. However, a previous study reported that SFB were not “Candidatus Arthromitus” but “Candidatus Savagella” (64). Thus, much attention should be focused on the classification of SFB and distinguishing the functions of “Candidatus Arthromitus” and “Candidatus Savagella.” There are still some scientists who consider “Candidatus Arthromitus” to be SFB (65–67).

Atg5 deficiency alters the composition and diversity of the gut microbiota. The microbiome can also induce adaptive changes in host immunity. In this study, some immune-associated genes were highly expressed in Atg5^−/− mice, including two key IBD-related transcription factors, RORC and TBX21. RORC is a key regulator of immune homeostasis and promotes the development of T_H17 cells (68). T_H17 cells and their cytokines are linked to several autoimmune and inflammatory diseases, such as allergy and asthma, psoriasis, rheumatoid arthritis, multiple sclerosis, systemic lupus erythematosus, and IBD (69). TBX21 has been characterized as the “master regulator” of T_H1 cell development (70). Excess activation of T_H1 cells was also associated with autoimmunity and IBD (71). In addition, the chemokines CCL5, TLE1, and Themis were also elevated in Atg5^−/− mice in our study. Previous research reported that the composition of gut microbiota of NLRP6-deficient mice was altered and that the altered composition stimulated epithelial cells to secrete the chemokine CCL5, which recruited immune cells, such as neutrophils, to initiate a chronic inflammatory response that can present as IBD (72). Research also reported that TLE1 inhibited the activation of nuclear factor-κB via NOD2 and that the dysfunction of this gene was associated with IBD pathogenesis (73). Themis was linked to the suppressive function of Treg cells and was reported to be involved in intestinal inflammation. Based on the above-described results, we propose that the host intestine may be vulnerable to pathogen invasion and an excessive immune burden in Atg5-deficient mice.

Considering the cross talk that occurs between the host and the gut microbiota during autophagy, we propose that overcolonization by SFB may promote excessive differentiation of T_H17 cells. In addition, decreases in butyrate-producing bacteria, such as those belonging to the Lachnospiraceae and Ruminococcaceae, are related to limited Treg cell differentiation. All the results obtained in this study indicated that autophagy deficiency in the intestinal epidermal cell may induce an imbalance in the ratio of T_H17/Treg cells. T_H17 and Treg cells have adverse function in the development of autoimmune and inflammatory diseases. T_H17 cells can execute strong proinflammatory impacts and are highly pathogenic, leading to the development of inflammation and severe autoimmunity (69). Treg cells serve as commanders in the control of self-tolerance and regulate overall immune responses against infections. After elimination of invading pathogens, Treg cells inhibit an excessive immune response, thereby protecting the host from the chronic immunopathology related to chronic infections (74). In a previous study, the balance between T_H17 and Treg cells was disrupted, and a high T_H17/Treg ratio was identified in patients with rheumatoid arthritis, which further confirms the importance of immune homeostasis of T_H17 and Treg cells (75). Taken together, these results indicate that imbalanced microbial communities induced by autophagy deficiency may activate a persistent immune response and expose the host to excessive immune activation, which may further enhance basal intestinal inflammation.

MATERIALS AND METHODS

Animals and sample collection.

Three C57BL/6J Atg5^flox/flox female mice (RBRC02975) which were purchased from the RIKEN BioResource Center (34) were mated with two C57BL/6J PVillin-Cre male mice which were obtained from the Nanjing Biomedical Research Institute of Nanjing University (T000142) in order to generate heterozygous (Atg5^+/−) mice. A total of 12 littermates consisting of 7 C57BL/6J Atg5^+/+ mice and 5 C57BL/6J Atg5^−/− were generated via intercrossing of heterozygous (Atg5^+/−) mice. Because the PVillin-Cre mice could specifically express Cre recombinase in the entire intestinal epithelium from the duodenum to the cecum (76), the Atg5 gene can be conditionally knocked out in the intestinal epithelium of the small intestine and colon of Atg5^−/− mice. All mice were cohoused in the same room and on the same shelf of a maximum-barrier, specific-pathogen-free facility at State Key Laboratory of Stem Cell and Reproductive Biology, Institute of Zoology, Chinese Academy of Sciences, Beijing, China; a 12-h light-dark cycle was maintained. After 4 weeks, the 12 mice were randomly assigned to individual cages. The mice were fed with a mixed diet containing the standard diet, soybeans, peanuts, and sunflower seeds. After 1 month, fresh fecal samples were obtained directly from individual cages. Then mice were euthanized immediately, and the intestinal contents of the duodenum, jejunum-ileum, cecum, and colon were collected within 20 min after euthanasia. All animal experiments were approved by the Animal Research Panel of the Committee on Research Practice of the University of Chinese Academy of Sciences. In total, there were 60 gut content samples from 7 C57BL/6J Atg5^+/+ mice and 5 C57BL/6J Atg5^−/− mice. Microbial genomic DNA was extracted from these contents (detailed information on samples is provided in Table S1). Distal ileum tissues were collected from each individual mouse for histological examination and RNA transcriptome sequencing.

Immunoblotting.

Intestinal tissue was resuspended (homogenized) in cold RIPA buffer (25 mM Tris-HCl [pH 7.6], 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with 1 mM phenylmethylsulfonyl fluoride and a protein inhibitor cocktail (04693132001; Roche Diagnostics, Basel, Switzerland) for 30 min on ice. The homogenates were centrifuged at 12,000 rpm for 10 min, and then the supernatant was transferred to a new tube. Next, 4× SDS loading buffer was added to samples, which were then boiled and subjected to immunoblotting. Immunoblots were visualized using the ODYSSEY Sa infrared imaging system (LI-COR Biosciences, USA). (Antibody information is given in Table S2.)

Immunohistochemistry and immunofluorescence.

For tissue sections, the ileum was dissected and then fixed and prepared as described previously (77). Serial 5-μm-thick sections were cut perpendicular to the crypt-surface epithelial cuff axis and parallel to the cephalocaudal axis. Sections were stained with hematoxylin and eosin (H&E). For a subset of immunohistochemical studies, sections were boiled for 15 min in sodium citrate buffer for antigen retrieval. Next, 3% H₂O₂ was added to the sections to eliminate internal peroxidase activity. After blocking with 5% bovine serum albumin (BSA; Sigma) for 30 min, each section was incubated with the primary antibody in 1% BSA at 4°C overnight, followed by staining with horseradish peroxidase (HRP)-conjugated secondary antibody. Negative controls were prepared without the primary antibody. Finally, the sections were stained with 3,3′-diaminobenzidine (DAB), and the nuclei were stained with hematoxylin. Images were taken using a Nikon 80i inverted microscope with a charge-coupled device (CCD).

Paraffin sections were boiled for 15 min in the sodium citrate buffer for antigen retrieval. After blocking with 5% BSA for 30 min, each section was incubated with the primary antibody in 1% BSA at 4°C overnight. After a washing with phosphate-buffered saline (PBS), the sample was incubated with fluorescein isothiocyanate (FITC)- or tetramethyl rhodamine isocyanate-conjugated secondary antibody diluted in PBS for 1 h at 37°C. Next, the slides were washed in PBS, and nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). Images were taken immediately using an LSM 780 microscope (Zeiss).

Gut bacterial 16S rRNA gene sequencing.

Microbial genomic DNA was extracted from feces and intestinal content samples using a TIANGEN DNA stool minikit (catalog number DP328) following the manufacturer's recommendations. The V4 hypervariable regions of the 16S rRNA were PCR amplified from microbial genomic DNA, which was harvested from intestinal contents and fecal samples using barcoded fusion primers (forward primer, 5′-AYTGGGYDTAAAGNG-3′; reverse primer, 5′-TACNVGGGTATCTAATCC-3′). The PCR program was as follows: initial denaturation at 94°C for 5 min, 27 cycles of 94°C denaturation for 30 s, 50°C annealing for 30 s, and 72°C extension for 30 s, and then a final extension at 72°C for 7 min. PCR products were purified using a Qiagen quick gel extraction kit (catalog number 28706). Barcoded V4 PCR amplicons were sequenced by an Illumina Miseq platform. Amplification of the V4 regions of 16S rRNA genes and sequencing services were provided by Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China).

Analysis of classification and abundance.

Sequences were filtered as in our previous study (31). Barcodes and sequencing primers were trimmed from sequencing reads. The trimmed and assembled sequences were uploaded to the QIIME and MG-RAST pipelines for further analysis. The trimmed sequence of each sample was compared to the RDP and M5RNA databases using the best hit classification option to classify the abundance in QIIME (78) (http://qiime.org/index.html) and MG-RAST (79) (http://metagenomics.anl.gov/), respectively. For QIIME, data on the OTU level were generated using the uclust script (http://qiime.org/scripts/pick_otus.html), and then according to these OTUs, QIIME automatically generated phylum to genus level data for different intestinal segments and genotypes. Simpson and Shannon indices were calculated by the mothur program. For statistical analysis, an unpaired Student's t test was performed using GraphPad. LEfSe was applied to identify microbes from different taxa among lines using the default parameters (LDA score > 2; P < 0.05) (80). Correlation analyses among different gut positions were carried out in Excel. For MG-RAST, data on the species level were generated, and parameter classification was done using cutoffs of 5 for the maximum E value, 97 for the minimum percent identity, and 60 bp for the minimum alignment length. The 16S rRNA gene sequences used in this study were deposited in the MG-RAST server under the project name “Autophagy_gut microbiota.”

Functional annotation and prediction of microbiota.

Microbial function was predicted using PICRUSt (81). The OTUs were mapped to the gg13.5 database with 97% similarity by the QIIME command “pick_closed_otus.” The OTU abundance was normalized automatically using 16S rRNA gene copy numbers from known bacterial genomes in Integrated Microbial Genomes and Microbiomes (IMG/M [http://img.jgi.doe.gov/]). The predicted genes and their functions were aligned to the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.kegg.jp/) database, and the differences between groups were compared with STAMP (http://kiwi.cs.dal.ca/Software/STAMP) (82). A two-sided Welch t test and the Benjamini-Hochberg false-discovery rate (FDR) (P < 0.05) correction were used in the two-group analysis.

Transcriptome sequencing of intestinal tissues.

Six samples from 3 Atg5^+/+ and 3 Atg5^−/− mice were used for transcriptome analysis. Ileal total RNAs were extracted using TRIzol (Invitrogen). Oligo(dT) beads that bind poly(A) were used to purify mRNA. After high-temperature fragmentation, 275-bp mRNAs were obtained (TruSeq RNA sample prep kit; Illumina). Mouse mRNA was reverse transcribed to cDNA. Following processes to repair ends, adenylate 3′ ends, ligate adapters, purify ligation products, enrich DNA fragments, and validate the library and pool cDNAs, the DNA library was constructed for sequencing by the Illumina Miseq platform. Reads were mapped to the mouse genome and were annotated. Reads per kilobase per million mapped reads (RPKM) were calculated as the normalized abundance of each gene as follows: total exon reads divided by the product of mapped reads (millions) and exon length (in kilobases). Differentially expressed genes (fold change > 2, P < 0.05, and after Benjamini-Hochberg adjustment) were calculated by DESeq (http://www.bioconductor.org/packages/release/bioc/html/DESeq.html). Enrichment analysis was performed using DAVID (https://david.ncifcrf.gov/).

Accession number(s).

The transcriptome sequencing data were deposited in the NCBI SRA database under project number PRJNA396813 and accession number SRP115402.