The gut microbiota reprograms intestinal lipid metabolism through long noncoding RNA Snhg9

1The intestinal microbiota has a significant impact on mammalian metabolism. In mice, resident intestinal bacteria enhance digestion of dietary polysaccharides, promote dietary lipid absorption by intestinal epithelial cells, and alter the function of adipose tissue. Consistent with the findings in mice, alterations in the composition of the human intestinal microbiota are associated with metabolic disorders including obesity, type 2 diabetes and cardiovascular disease. Given the rapidly increasing worldwide prevalence of metabolic disease, there is a pressing need to understand the mechanisms by which the intestinal microbiota affects host metabolism.

2Long non-coding RNAs (lncRNAs) are RNA transcripts that are not translated but are nevertheless biologically functional. LncRNAs regulate biological processes such as cell proliferation, cell death, tumorigenesis, and immunity, but little is known about their involvement in the regulation of host metabolism by the gut microbiota. We therefore investigated whether the presence of the gut microbiota impacts the transcription of lncRNAs in small intestinal epithelial cells, which are central to microbial modulation of lipid absorption and metabolism.

3Whole transcriptome sequencing (RNA-seq) of small intestinal epithelial cells from conventionally raised (conventional) and germ-free mice identified 60 differentially expressed non-protein coding genes, including 42 genes that encode lncRNAs (Fig. 1A, fig. S1, A and B). In particular, the lncRNA encoded by Snhg9 (small nucleolar RNA host gene 9) showed reduced abundance in epithelial cells from conventional as compared to germ-free mice (Fig. 1B). Although lncRNAs are generally defined as >200 nucleotides and mouse Snhg9 RNA is only 183 nucleotides long, human SNHG9 RNA is 233 nucleotides long and thus both the mouse and human RNAs are designated as lncRNAs.

4Snhg9 transcripts were detected in small intestinal epithelial cells of antibiotic-treated mice by in situ hybridization (fig. S2A). Analysis of published single cell RNA sequencing data from mouse intestinal epithelial cells indicated that Snhg9 transcripts were mostly from stem cells, enterocytes and enteroendocrine cells (a specialized epithelial cell lineage) (fig. S2B). Using quantitative PCR (qPCR), we confirmed that epithelial cells from conventional mice express less Snhg9 RNA when compared with epithelial cells from both germ-free and antibiotic treated mice (Fig. 1C). Thus, Snhg9 expression in small intestinal epithelial cells is reduced in the presence of the microbiota.

5We next sought to illuminate the biological function(s) of Snhg9. Since other lncRNAs bind to and regulate the activity of proteins, we screened for Snhg9-protein interactions in small intestinal epithelial cells. We performed an RNA-protein pull-down assay in epithelial cell lysates using in vitro transcribed Snhg9 or antisense Snhg9. Mass spectrometry identified the most abundant interacting protein as cell cycle and apoptosis protein 2 (CCAR2; also known as deleted in breast cancer 1, or DBC1) (Fig. 2, A and B). This finding was supported by immunoblot detection of CCAR2 among the proteins precipitated by Snhg9 (Fig. 2C).

6To test whether Snhg9 binds directly to CCAR2, we performed pull-down assays with in vitro transcribed Snhg9 and recombinant CCAR2. Consistent with our findings in epithelial cell lysates, recombinant CCAR2 was precipitated by Snhg9 RNA but not by antisense Snhg9 RNA or polyA RNA (Fig. 2D), indicating that CCAR2 binds directly to Snhg9 RNA. In addition, mutations in Snhg9 impacted binding to CCAR2. Deletion of 28 nucleotides in the middle of the Snhg9 sequence did not affect Snhg9 RNA binding to CCAR2. However, binding was reduced by 24-nucleotide deletions in the 3’ and 5’ regions, which target a predicted loop in the Snhg9 secondary structure (Fig. 2E; fig. S3, A and B). These data support a direct binding interaction and indicate that Snhg9 is a protein-binding lncRNA that binds directly to CCAR2.

7We next investigated the cellular and physiological consequences of lncRNA Snhg9 binding to CCAR2. CCAR2 is an endogenous inhibitor of the deacetylase Sirtuin 1 (SIRT1). Among its several functions, SIRT1 regulates lipid metabolism by interacting with the transcription factor peroxisome proliferator-activated receptor gamma (PPARγ), which directs transcription of lipid metabolic genes. SIRT1 represses PPARγ activity through two mechanisms: by deacetylating PPARγ, or by docking with nuclear receptor corepressor 1 (NcoR1), a PPARγ cofactor. Both mechanisms decrease Pparg expression and reduce lipid metabolism. We reasoned that by binding to CCAR2, Snhg9 might inhibit its interaction with SIRT1, thereby rescuing SIRT1 deacetylase and NcoR1 binding activities from CCAR2 inhibition. Indeed, overexpression of Snhg9 in HEK-293T cells largely abolished CCAR2 binding to SIRT1 (Fig. 3, A and B), and increased both SIRT1 deacetylase activity (Fig. 3C) and binding to NcoR1 (Fig. 3, D and E). These data indicate that Snhg9 promotes SIRT1 activity by sequestering the inhibitory protein CCAR2.

8Since Snhg9 rescued SIRT1 activity from CCAR2 inhibition, we predicted that Snhg9 would repress PPARγ expression and activity. We therefore evaluated the impact of Snhg9 on Pparg expression and lipid metabolism in cells. PPARγ controls differentiation of the mouse fibroblast cell line 3T3-L1 into adipocytes following chemical induction. We generated 3T3-L1 cells that stably express Snhg9 (fig. S4A) and confirmed binding of Snhg9 to CCAR2 by RNA immunoprecipitation (RIP) assay (fig. S4B). Stable expression of Snhg9 inhibited expression of Pparg and its protein product PPARγ, whereas co-expression of Ccar2 rescued expression of both Pparg and PPARγ (Fig. 3, F and G). Accordingly, there was reduced transcription of PPARγ-controlled genes, including Cd36 (encoding a fatty acid transporter), Fabp4 (encoding a fatty acid binding protein), and Lpl (encoding lipoprotein lipase) (fig. S4, C to E). Similarly, stable expression of Snhg9 in mouse small intestinal organoids reduced expression of Pparg and PPARγ-controlled genes (fig. S4, F to J). These effects were not due to expression of the putative Snhg9 open-reading frame (ORF) (fig. S5A), since expression of the ORF alone failed to reduce expression of Pparg and its target genes, and Snhg9 with mutated start and stop codons retained the ability to reduce Pparg expression (fig. S5B). Together, these results indicate that Snhg9 represses PPARγ expression by binding to CCAR2.

9Further supporting this idea, inactivation of Snhg9 in 3T3-L1 cells (Snhg9−/− cells) by CRISPR-Cas9 genome editing (fig. S6, A to D) increased expression of Pparg, PPARγ, and downstream target genes (Fig. 3, H and I, fig. S6E). This increase occurred despite the high frequency of cell passage during the selection process, which tends to suppress Pparg expression. In comparison, cells edited with a non-targeting single guide RNA (sgRNA) showed minimal Pparg expression as a result of a high number of cell passages (Fig. 3, H and I), and re-expression of Snhg9 in Snhg9−/− 3T3-L1 cells largely reversed Pparg expression (Fig. 3, H and I). Of note, overexpression of Snora78, which is flanked by the Snhg9 exons, did not alter Pparg expression (fig. S6F), suggesting that Snora78 deletion did not cause the increased expression of Pparg in Snhg9−/− cells, and deleting the Snhg9 locus did not interfere with the expression of a nearby gene, Rps2 (fig. S6G). These data support the idea that Snhg9 released SIRT1 from CCAR2 inhibition.

10Consistent with the reduced Pparg expression, Snhg9 overexpression restrained the differentiation of 3T3-L1 cells to adipocytes (Fig. 3, J and K). Conversely, Snhg9−/− 3T3-L1 cells maintained the ability to differentiate and form lipid droplets (Fig. 3, L and M). However, cells edited with non-targeting sgRNA lost the ability to differentiate as a result of insufficient expression of Pparg arising from multiple cell passages (Fig. 3, L and M). Taken together, our results reveal that lncRNA Snhg9 inhibits Pparg expression and lipid metabolism by dissociating the CCAR2-SIRT1 complex.

11The presence of the intestinal microbiota enhances dietary lipid absorption and promotes obesity in mice fed a high fat diet. Germ-free mice, which are microbiologically sterile and thus lack an intestinal microbiota, tend to absorb less lipid than conventional mice and thus are largely protected from high fat diet-induced obesity. Since Snhg9 suppresses lipid metabolism in cells (Fig. 3) and is upregulated in the intestines of germ-free mice (Fig. 1C), we hypothesized that Snhg9 limits body fat accumulation in mice.

12To test this hypothesis, we used the Villin promoter to force expression of Snhg9 in intestinal epithelial cells of conventional mice (Villin-Snhg9 transgenic mice; fig. S7, A to D). Consistent with our findings in cultured cells, SIRT1 deacetylase activity was higher in intestinal epithelial cells from conventional Villin-Snhg9 transgenic mice as compared to wild-type littermates, with activity levels comparable to those of wild-type germ-free mice (Fig. 4A).

13To investigate whether Snhg9 overexpression impacts lipid metabolism in vivo, we compared the small intestinal transcriptomes of Villin-Snhg9 transgenic mice and their wild-type littermates by RNA-seq. KEGG pathway analysis confirmed that Snhg9 overexpression suppressed expression of genes involved in metabolic pathways including the PPAR signaling pathway (Fig. 4B). Villin-Snhg9 transgenic mice showed reduced expression not only of Pparg but also genes involved in fatty acid absorption (such as Cd36), transport (such as Fabp4), and synthesis (such as Scd1) (Fig. 4C). Expression of these genes and their protein products was also reduced in germ-free mice in comparison to conventional wild-type mice (fig. S7, E and F). Consequently, Villin-Snhg9 transgenic mice were similar to germ-free mice in that they had less lipid in their intestinal epithelial cells and more in their feces when compared to wild-type littermates (Fig. 4, D to F).

14When fed a normal chow diet, Villin-Snhg9 transgenic mice had body weights similar to their wild-type littermates (fig. S7G). However, their body fat percentages and epididymal fat pad weights were reduced (fig. S7, H and I) and they were more glucose tolerant (fig. S7J). When switched to a high fat diet for 10 weeks, Villin-Snhg9 transgenic mice exhibited decreased body weights (fig. S7K), lower overall body fat percentages (Fig. 4G), smaller epididymal fat pads (Fig. 4H, fig. S7L) and milder liver steatosis (Fig. 4I) than their wild-type littermates. They also had lower serum triglycerides and free fatty acids (fig. S7, M and N), increased glucose tolerance, and decreased insulin resistance (Fig. 4, J and K). These phenotypes did not result from altered food intake, physical activity, respiratory exchange ratio or microbiota composition (fig. S8, A to C).

15Because expression of intestinal Snhg9 is suppressed by the microbiota (Fig. 1C), we further assessed the requirement for the microbiota in Snhg9-regulated lipid metabolism. When we depleted the microbiota by antibiotic treatment, wild-type mice fed a high fat diet had lowered body fat percentages similar to those of Villin-Snhg9 transgenic littermates (Fig. 4L). This is consistent with the increased Snhg9 expression in antibiotic treated mice (Fig. 1C). For comparison, we generated Snhg9−/− mice by CRISPR-Cas9-mediated gene targeting (fig. S9, A and B). Although the Snhg9−/− mice had body weights and body fat percentages similar to wild-type littermates when fed a normal chow diet (fig. S9, C and D), they had increased body fat percentages and weight gain when fed a high fat diet even when their microbiota were depleted with antibiotics (Fig. 4M, fig. S9E). These data support the conclusion that the microbiota promotes body fat accumulation in part by repressing intestinal Snhg9 expression.

16Bacteria activate intestinal epithelial cell gene expression through Toll-like receptors (TLRs) and their common signaling adaptor MyD88. Expression of intestinal Snhg9 was increased in Myd88−/− mice as compared to wild-type controls (Fig. 5A), suggesting that MyD88 is required for microbial repression of intestinal Snhg9 expression. Although epithelial cell Myd88 was dispensable for repression of Snhg9 expression (Fig. 5B), mice with Myd88 selectively deleted in CD11c+ cells showed elevated Snhg9 expression (Fig. 5C), suggesting a role for CD11c+ cells in repressing Snhg9 expression. This idea was supported by studies of a mouse model of CD11c+ cell depletion in which Diphtheria toxin receptor (DTR) is expressed from the Cd11c promoter. Selective depletion of CD11c+ cells by Diphtheria toxin administration increased Snhg9 expression relative to controls (Fig. 5D). Since CD11c marks myeloid cells, including dendritic cells and macrophages, these results indicate that myeloid cells are required for microbial repression of intestinal Snhg9 expression.

17Bacteria regulate the expression of several key intestinal epithelial cell genes through an immune cell signaling relay involving primarily myeloid cells and group 3 innate lymphoid cells (ILC3). In this relay, bacteria activate myeloid cells via TLRs and MyD88, which then signal to ILC3 through the cytokine interleukin 23 (IL-23). Activated ILC3 then signal to intestinal epithelial cells through IL-22. Having established a role for myeloid cells in microbial repression of Snhg9 expression, we next tested for the involvement of ILCs. Rag1−/− mice, which lack T and B cells, had decreased expression of intestinal Snhg9 as compared to wild-type mice (Fig. 5E). This indicated that T and B cells are dispensable for the microbial repression of Snhg9, while decreased expression of Snhg9 may be a result of the increased bacterial loads in the intestines of Rag1−/− mice. In contrast, depleting ILCs in Rag1−/− mice with the CD90.2 antibody elevated Snhg9 expression (Fig. 5F), indicating a requirement for ILCs in microbial repression of Snhg9 expression. Similarly, Rag2−/−;Il2rg−/− mice, which lack immune cells (including ILCs) that dependent on the IL-2 receptor γ chain, showed increased expression of intestinal Snhg9 as compared to Rag1−/− mice (Fig. 5G).

18To further assess the requirement for ILC3, we analyzed Snhg9 expression in Rorcgfp/gfp mice, which lack IL-22 producing cells including ILC3. Rorcgfp/gfp mice showed increased Snhg9 expression as compared to wild-type littermates, and antibiotic depletion of gut microbiota abolished this difference (Fig. 5H), supporting the idea that ILC3s relay microbial signals that repress Snhg9 expression. In addition, supplementing Myd88−/− mice with either IL-22 or IL-23 repressed Snhg9 expression to the levels observed in conventional wild-type mice (Fig. 5I), consistent with the known involvement of these cytokines in the myeloid cell-ILC3 signaling circuit. Further, Snhg9 expression was repressed by monocolonization of germ-free mice with bacterial species known to activate myeloid cell-ILC3 signaling, including Salmonella enterica Serovar Typhimurium (S. Typhimurium), a Gram-negative intestinal pathogen, and segmented filamentous bacteria (SFB), Gram-positive members of the intestinal microbiota (fig. S10, A and B). Thus, the microbiota suppresses Snhg9 expression through a myeloid cell-ILC3 signaling relay.

19This study shows that the gut microbiota promotes lipid absorption and metabolism by repressing expression of lncRNA Snhg9. This finding raises several additional questions for future study. First, are there specific components of the gut microbiota that promote lipid absorption through Snhg9 repression? We found that Snhg9 expression was selectively repressed by bacterial species that activate intestinal myeloid cell-ILC3 signaling, providing a potential clue. However, more studies are needed to determine what bacterial components or characteristics enable activation of myeloid cell-ILC3 signaling, and how microbial community composition impacts Snhg9 expression and lipid metabolism. Second, what is the evolutionary rationale for microbial regulation of lipid absorption? It is possible that intestinal lipid metabolism is linked to innate immune sensing of microbes in order to cope with an increased energy demand during colonization or infection, to provide lipid substrates or mediators that regulate intestinal immune cell development, or to enhance epithelial barrier function through reactive oxygen species production via lipid oxidation. Third, do these findings provide insight into the regulation of human lipid metabolism? SNHG9 is conserved and expressed in humans, raising the possibility that human SNHG9 functions in a similar manner.

20In this study, we show that Snhg9 RNA regulates PPARγ activity by dissociating SIRT1 from CCAR2, providing insight into how a lncRNA regulates intestinal lipid metabolism. These findings advance our understanding of the complex epithelial cell networks that regulate lipid metabolism in response to microbial signals(fig. S11). Ultimately, these results could suggest strategies for treating metabolic disease by targeting Snhg9 and the microbiota.