Introduction

Non-alcohol fatty liver disease (NAFLD) is characterized by excess liver fat in the absence of significant alcohol consumption. It can progress from simple steatosis to nonalcoholic steatohepatitis (NASH) and fibrosis and eventually to chronic progressive diseases such as cirrhosis, end-stage liver failure, and hepatocellular carcinoma (Loomba, Friedman et al., 2021). In 2020, an international panel of experts led a consensus-driven process to develop a more appropriate term for the disease utilizing a 2-stage Delphi consensus, that is, “metabolic dysfunction-associated fatty liver disease (MASLD)” related to systemic metabolic dysregulation (Gofton, Upendran et al., 2023, Rinella, Lazarus et al., 2023). The pathogenesis of MASLD has not been entirely elucidated. Multifarious factors such as genetic and epigenetic factors, nutritional factors, insulin resistance, lipotoxicity, microbiome, fibrogenesis and hormones secreted from the adipose tissue, are recognized to be involved in the development and progression of MASLD (Buzzetti, Pinzani et al., 2016, Friedman, Neuschwander-Tetri et al., 2018, Lee, Kim et al., 2017, Rada, Gonzalez-Rodriguez et al., 2020, Sakurai, Kubota et al., 2021). Free fatty acids (FFAs), which are central to the pathogenesis of MASLD, originate from the periphery, mainly via lipolysis of triglyceride in the adipose tissue, or from increased hepatic de novo lipogenesis (DNL). Fatty acids in hepatocytes undergo mitochondrial β-oxidation and re-esterification to form triglyceride (TG), which are then exported into the blood as very low-density lipoproteins or stored in lipid droplets. Hepatic lipotoxicity occurs when the disposal of fatty acids through β-oxidation or the formation of TG is overwhelmed, which leads to endoplasmic reticulum (ER) stress, oxidative stress and inflammasome activation (Friedman et al., 2018). Collective literature has indicated that a cluster of differentiation 36/fatty acid translocase (CD36) could increase FFAs uptake in the liver and drive hepatosteatosis onset. Under physiological conditions, CD36 expression in hepatocytes was found to be minimal; however, lipid overload or activation of nuclear receptors including peroxisome proliferator-activated receptor α/γ (PPARα/γ) and liver X receptor (LXR), could significantly increase it (Rada et al., 2020).

Epigenetics, an inheritable phenomenon occurring without altering the DNA sequence, can regulate gene expression through different forms, including DNA methylation, histone modifications, chromatin remodeling, transcriptional control, and non-coding RNAs (Mann, 2014). Histone modifications, including acetylation, methylation, phosphorylation, ubiquitination, ribosylation, and ubiquitin-like protein modification (SUMO), are important epigenetic determinants of chromatin tightness and accessibility (Chen & Pikaard, 1997). Histone methylation is associated with chromatin-specific transcriptional activity states; for example, methylation of lysine 4 of histone H3 (H3K4), H3K36 and H3K79 are linked with a transcriptional activation state, and H3K9, H3K27, and H4K20 with transcriptional repression state (Pirola & Sookoian, 2020). Previous studies have illustrated that epigenetics factors including histone modification play key role in lipid metabolism (Bayoumi, Gronbaek et al., 2020, Byun, Seok et al., 2020, Jun, Kim et al., 2012). Furthermore, long non-coding RNAs (lncRNAs) are non-coding RNAs with more than 200 bases in length, can be transcribed by RNA polymerase II, and are comparable to mRNA but lack the crucial open reading framework required for translation (Ng, Lin et al., 2013, Ulitsky & Bartel, 2013). LncRNAs are involved in epigenetic regulation of gene expression at different levels and through different molecular mechanisms such as chromatin remodeling, transcriptional regulation and post-transcriptional processing. Previous studies have indicated that lncRNAs are involved in the pathological progress of MASLD (Bayoumi et al., 2020, Sommerauer & Kutter, 2022). Although histone modification and lncRNAs influence the susceptibility to MASLD, their roles in MASLD remain largely unknown.

Small nucleolar RNA host genes (Snhgs), a type of lncRNA, serve as host genes for producing intronic snoRNAs and are mainly related to tumor pathophysiology by regulating proliferation, apoptosis, invasion, and migration (Sen, Fallmann et al., 2020, Zimta, Tigu et al., 2020). Here, we found that the expression of hepatic Snhg3 was decreased in high-fat-induced obesity (DIO) mice. Experiments conducted using in vivo and in vitro models indicated that Snhg3 was involved in lipid metabolism and hepatic steatosis. Mechanistically, Snhg3 interacted with staphylococcal nuclease and Tudor domain containing 1 (SND1), a well-understood Tudor protein that participates in lipid metabolism and tumoral behavior by modulating cholesterol and glycerophospholipid metabolism and acylglyceride storage in lipid droplets (Navarro-Imaz, Ochoa et al., 2020). Furthermore, Snhg3 increased the expression of SND1 by promoting the stability of SND1 mediated by K63-linked ubiquitination and induced nuclear localization of SND1 protein, thereby reducing tri-methylation at H3K27 (H3K27me3) enrichment and boosting chromatin loose remodeling at PPARγ promoter, eventually enhancing Pparγ and Cd36 expressions.

Results

LncRNA-Snhg3 is downregulated in high-fat diet-induced obesity mice

Firstly, we analyzed the lncRNAs expression profiles in the livers of DIO mice and normal chow-fed mice (control) by RNA-Seq, and found 18072 hepatic lncRNAs, including 338 differentially expressed lncRNAs (q-value ≤ 0.05, Figure 1A-Supplementary Table S1). Of all Snhgs, Snhg3 had the most prominent expression and exhibited more noticeable downregulation in the liver of the DIO compared to the control mice (Figure 1B); thus, it was selected for further study. The downregulation of Snhg3 was confirmed by RT-qPCR (Figure 1C). Additionally, the Coding Potential Calculator indicated that Snhg3 has a coding probability of 0.020757, classifying it as a noncoding sequence (Kang, Yang et al., 2017). Localization of Snhg3 was primarily observed in nuclei with a probability score of 0.451138, as predicted using software prediction (http://lin-group.cn/server/iLoc-LncRNA/predictor.php). The exact nuclear localization of Snhg3 was further confirmed by nuclear/cytoplasmic fractionation (Figure 1D).

Figure 1.

Hepatocyte-specific Snhg3 knock-out alleviates hepatic steatosis in DIO mice

Since Snhg3 was associated with hepatic nutrition change, the role of Snhg3 was further confirmed by constructing hepatocyte-specific Snhg3 knock-out (Snhg3-HKO) mice (Figure 2A). The result indicated that body weight was mildly decreased in DIO Snhg3-HKO mice compared with the control DIO Snhg3-Flox mice (Figure 2B). The energy consumption is mainly reflected as the sum of energy utilization during internal heat production using comprehensive laboratory animal monitoring system (CLAMS). DIO Snhg3-HKO mice had higher heat production, total oxygen consumption, and carbon dioxide production (Figure 2C and D). Conversely, the respiratory exchange ratio (RER) showed no difference (Figure 2E). Furthermore, insulin sensitivity, not glucose tolerance, was improved in DIO Snhg3-HKO mice (Figure 2F). The DIO Snhg3-HKO mice had a decrease in liver weight and the ratio of liver weight/body weight, and improved hepatic steatosis, including decreasing lipid accumulations and the ballooning degeneration of liver cells (Figure 2G-I). Serum alanine transaminase (ALT) and aspartate transaminase (AST) levels were significantly decreased in Snhg3-HKO mice (Figure 2J). In addition, serum FFAs, TG and TC were also reduced in Snhg3-HKO mice (Figure 2K). The DIO Snhg3-HKO mice exhibited a decrease in inguinal white adipose tissue (iWAT) weight and the ratio of iWAT weight/body weight, but not in brown adipose (BAT) weight and the ratio of BAT weight/body weight (Figure 2L). And there was no difference in serum insulin between DIO Snhg3-HKO mice and control mice (Figure 2M). These results suggested that hepatic knockout of Snhg3 improves hepatic steatosis in DIO mice.

Figure 2.

Hepatocyte-specific Snhg3 knock-in aggravates hepatic steatosis in DIO mice

Furthermore, the hepatocyte-specific Snhg3 knock-in (Snhg3-HKI) mice were also constructed to detect the function of Snhg3 in the liver (Figure 3A). The DIO Snhg3-HKI mice showed greater weight gains than the control wild type (WT) mice (Figure 3B). In addition, DIO Snhg3-HKI mice had lower heat production, total oxygen consumption, and carbon dioxide production (Figure 3C and D). The RER result indicated the DIO Snhg3-HKI mice utilized less fat (Figure 3E). Insulin sensitivity was also impaired in DIO Snhg3-HKI mice (Figure 3F). The liver weight and the ratio of liver weight/body weight of DIO Snhg3-HKI mice were markedly increased (Figure 3G). Also, DIO Snhg3-HKI mice exhibited severe hepatic steatosis (Figure 3H and I) and higher serum ALT and AST levels (Figure 3J). Both serum TC and iWAT weight were increased in DIO Snhg3-HKI mice (Figure 3K and L). Similar to DIO Snhg3-HKO mice, there was also no difference in serum insulin between DIO Snhg3-HKI mice and WT mice (Figure 3M). These findings indicated that upregulation of Snhg3 could promote DIO hepatic steatosis.

Figure 3.

Snhg3 promotes hepatic steatosis by regulating chromatin remodeling

To clarify the molecular mechanism of Snhg3 in hepatic steatosis, we investigated the hepatic differentially expressed genes (DEGs) using RNA-Seq. There were 1393 DEGs between the DIO Snhg3-HKI and control WT mice, with 1028 genes being upregulated and 365 genes downregulated (log2FC ≥ 1, q-value < 0.001) in the liver of Snhg3-HKI mice (Figure 4A-Supplementary Table S2). A gene set enrichment analysis (GSEA) of DEGs revealed that Snhg3 exerts a global effect on the expression of genes involved in fatty acid metabolism and the PPAR signaling pathway (Figure 4B).

Figure 4.

LncRNAs in the nucleus can affect gene expression in multiple ways, such as chromatin remodeling, transcriptional regulation, and post-transcriptional processing (Morey & Avner, 2004, Thomson & Dinger, 2016). Since Snhg3 was mainly localized in the nuclei of hepatocytes, we next checked the genome-wide chromatin accessibility (log2FC > 2, p-value < 0.001) in the liver of DIO Snhg3-HKI and WT mice using ATAC-Seq. We discovered that in all 6810 differentially accessible regions (DARs), 4305 (> 63.2%) were more accessible in Snhg3-HKI mice and only 2505 (> 36.8%) of peaks were more accessible in control mice (Supplementary Table S3), indicating that the chromatin states were ‘hyper-accessible’ in the liver of Snhg3-HKI mice. Moreover, DARs were with relatively few promoter-proximal (Up2k) and exon regions in both the control and Snhg3-HKI groups (Figure 4C), supporting the idea that gene activation depends on multiple regulatory regions, is not limited to its promoter and exon regions (Ackermann, Wang et al., 2016). Furthermore, 3966 genes were associated specifically with the accessible regions in the Snhg3-HKI group and only 2451 genes in the WT group (log2FC > 2, p-value < 0.001, Supplementary Table S4). Additionally, PPARγ was identified as a potential transcription factor associated with hyper-accessible regions in the liver of the Snhg3-HKI group by HOMER and CREMA (Figure 4D and E).

To determine whether open chromatin regions were correlated with gene expression, we integrated ATAC-Seq data (genes associated with DARs, log2FC > 2, p-value < 0.001) with RNA-Seq data (DEGs in DIO Snhg3-HKI and control mice, log2FC > 1, q-value < 0.001). Overall, 233 upregulated genes shown in quadrant 2, including Cd36 and cell death-inducing DNA fragmentation factor 45 like effector a/c (Cidea/c), which are critical for MASLD progression (Koonen, Jacobs et al., 2007, Matsusue, Kusakabe et al., 2008, Sans, Bonnafous et al., 2019), had at least one associated open chromatin region, which accounted for >22.67% (total 1028) of DEGs mapped to ATAC-Seq peaks in the liver of Snhg3-HKI mice (Figure 4F-H-Supplementary Table S5). Meanwhile, at least one open chromatin region was associated with 65 downregulated genes in the quadrant 3, which accounted for > 17.81% (total 365) of DEGs mapped to ATAC-Seq peaks in the liver of WT mice (Figure 4F-Supplementary Table S5). Furthermore, the expression levels of Cd36 and Cidea in the liver of Snhg3-HKO and Snhg3-HKI mice were confirmed by RT-qPCR and western blotting (Figure 4I and J).

Snhg3 induces SND1 expression by interacting with SND1 and enhancing the stability of SND1 protein

To further elucidate the molecular mechanism of Snhg3 in hepatic steatosis, an RNA pull-down followed by mass spectrometry (MS) assay was performed in primary hepatocytes. The result identified 234 specific Snhg3-associated proteins, involved in multiple signaling pathways, including PPAR, NAFLD and fatty acid degradation pathways (Figure 5A and B-Supplementary Table S6). Of these proteins, a well-understood Tudor protein SND1 was also predicted to interact with three fragments of Snhg3 by bioinformatic method (RBPsuite (sjtu.edu.cn)) (Figure 5C and D-Supplementary Table S7). Snhg3 coprecipitation with SND1 was confirmed by RNA pull-down coupled with western blotting (Figure 5E), which was consistent with the RNA immunoprecipitation (RIP) assay results (Figure 5F). Meanwhile, Snhg3 regulated the protein, not mRNA, expression of SND1 in vivo and in vitro by promoting the stability of SND1 protein (Figure 5G-I). Furthermore, we tested the effect of Snhg3 on the ubiquitin-modification of SND1 and found that Snhg3 enhanced SND1 ubiquitination in vivo and in vitro (Figure 5K and L). Previous studies indicated that K48-linked polyubiquitination aids in proteasome-mediated recognition and degradation and that K63-linked polyubiquitination participates in signaling assemblies and protein stability (Sun, Liu et al., 2020). As predicted, Snhg3 overexpression increased K63-linked ubiquitination modification in endogenous and exogenous SND1 protein, not K48- or K33-linked (Figure 5M and N). Additionally, Snhg3 overexpression enhanced the nuclear localization of SND1 in Hepa1-6 cells with PA treatment (Figure 5O). Collectively, these results suggested that Snhg3 promoted the K63-linked ubiquitination and stability of SND1 protein through interacting with SND1, thus resulting in SND1 protein increase and nuclear localization.

Figure 5.

Snhg3 promotes PPARγ expression by decreasing H3K27me3 enrichment at the Pparγ promoter

SND1, initially named as p100, is a highly conserved and ubiquitously expressed multifunctional Tudor domain-containing protein that participates in pivotal biological processes like double-stranded RNA editing, pre-mRNA splicing, microRNA-mediating gene silencing and piRNA biogenesis in germlines (Ying & Chen, 2012). Previous studies indicated that Tudor proteins participate in epigenetic regulation by binding to methyl-arginine⁄lysine residues (Ying & Chen, 2012). However, whether SND1 influences histone modification remains unclear. It is well known that histone modification dynamically regulates specific gene expression by altering the organization and function of chromatin and is involved in the pathophysiology of some diseases, such as histone H3 methylation modification, which may contribute to MASLD pathogenesis (Byun et al., 2020, Jun et al., 2012, Tessarz & Kouzarides, 2014). Considering that H3K27me3, a repressive chromatin mark, plays a role in autophagy-mediated lipid degradation (Byun et al., 2020), we tested the effect of SND1 on H3K27me3. The results revealed that both SND1 and Snhg3 overexpression reduced the H3K27me3 level (Figure 6A). Furthermore, disrupting SND1 expression increased the H3K27me3 level and reversed the Snhg3-induced H3K27me3 decrease (Figure 6B and C). Moreover, the hepatic H3K27me3 level was upregulated in Snhg3-HKO mice but downregulated in Snhg3-HKI mice (Figure 6D). The results indicated that Snhg3 negatively regulated the H3K27me3 level through SND1.

Figure 6.

To further clarify whether Snhg3-induced H3K27me3 decrease is involved in hepatic steatosis, we examined the H3K27me3 enrichment in the liver of DIO Snhg3-HKO mice using the CUT&Tag assay and detected 10915 peaks (Supplementary Table S8). The genomic locations of these peaks were divided into eight categories, and the H3K27me3 signals were predominantly (about 54%) enriched at the 2-kb promoter, 5’-untranslated region (5’-UTR), and exon categories. Meanwhile, very few signals (about 14%) were enriched in the 2-kb downstream and intergenic categories in the liver of DIO Snhg3-HKO mice (Figure 6E). Moreover, the exon, upstream2k, 5’-UTR and intron regions of Pparγ were enriched with the H3K27me3 mark (fold_enrichment = 4.15697) in the liver of DIO Snhg3-HKO mice (Supplementary Table S8). Subsequently, ChIP assay revealed that hepatic H3K27me3 enrichment at the Pparγ promoter was increased in DIO Snhg3-HKO mice but decreased in DIO Snhg3-HKI mice (Figure 6F). Snhg3-overexpression in Hepa1-6 cells yielded similar results (Figure 6G).

SND1 mediates Snhg3-induced PPARγ expression increase

Collective studies have indicated that PPARγ plays a crucial role in MASLD progression by regulating target genes such as Cd36 and Cidec (Lee, Muratalla et al., 2023b, Lee, Pusec et al., 2021, Lee, Park et al., 2018, Matsusue et al., 2008, Skat-Rordam, Hojland Ipsen et al., 2019). In this study, the mRNA and protein expression levels of hepatic PPARγ were decreased in DIO Snhg3-HKO mice and increased in DIO Snhg3-HKI mice (Figure 7A and B). The upregulation of Snhg3 and SND1 also increased the expression of Pparγ and Cd36 in vitro (Figure 7C-E). Meanwhile, disruption of SND1 expression alleviated Snhg3-induced PPARγ and CD36 increase and lipid accumulation (Figure 7F-H). Collectively, these results demonstrated that SND1 mediates Snhg3-induced PPARγ and CD36 expression.

Figure 7.

PPARγ mediates Snhg3-induced hepatic steatosis

Hepatocyte-specific depletion of PPARγ is known to protect mice against NASH and boost the therapeutic efficacy of rosiglitazone, a synthetic PPARγ agonist, in the liver (Lee et al., 2021). Furthermore, PPARγ is an inducer of adipocyte differentiation and a reservoir for excess FFAs, thereby potentially preventing lipotoxicity in other tissues and organs (Medina-Gomez, Gray et al., 2007). To this end, we tested the effect of T0070907, a selective PPARγ inhibitor, on Snhg3-induced hepatic steatosis in DIO mice. The result showed that T0070907 treatment for 8 weeks had no effects on body weight, liver and iWAT weight, and serum FFAs, TG and TC in DIO Snhg3-HKI mice, but improved Snhg3-induced hepatic steatosis in DIO Snhg3-HKI mice (Figure 8A-E). Additionally, T0070907 also mitigated the hepatic Cd36 increase in DIO Snhg3-HKI mice and Snhg3- and SND1-induced Cd36 increase in hepa 1-6 cells (Figure 8F and G). Collectively, these results suggested that PPARγ-mediated Snhg3-induced hepatic steatosis and CD36 increase.

Figure 8.

Discussion

Liver steatosis is common in various metabolic diseases and related disorders, including MASLD. Although lncRNAs are implicated in regulating numerous mechanisms related to liver steatosis and MASLD, their exact function remains to be determined. In this study, hepatocyte-specific lncRNA-Snhg3 deficiency improved hepatic steatosis and insulin resistance, while overexpression aggravated hepatic steatosis and insulin resistance in DIO mice. Excessive hepatic lipid deposition owing to increased FFAs uptake and hepatic DNL impairs autophagy and promotes endoplasmic reticulum stress and oxidative stress, insulin resistance, liver tissue damage, and inflammation, ultimately aggravating MASLD progression (Rada et al., 2020). Further evidence suggests that CD36 plays an important role in facilitating intracellular FFAs uptake. Overexpression of CD36 in hepatocytes increased FFAs uptake and TG storage; conversely, its deletion ameliorated hepatic steatosis and insulin resistance in DIO mice (Rada et al., 2020). Therefore, CD36 could be a potential biomarker for MASLD diagnosis. As a transcription regulator of CD36, PPARγ plays major adipogenic and lipogenic roles in adipose tissue. Although PPARγ is most abundantly expressed in adipocytes and plays major adipogenic and lipogenic roles in the tissue, it has been reported to also be expressed in hepatocytes. There is a strong correlation between the onset of MASLD and hepatocyte-specific PPARγ expression. Clinical trials have indicated that increased insulin resistance and hepatic PPARγ expressions were associated with NASH scores in some obese patients (Lee, Muratalla et al., 2023a, Mukherjee, Wanjari et al., 2022). Activation of PPARγ in the liver induces the adipogenic program to store fatty acids in lipid droplets as observed in adipocytes (Lee et al., 2018). Moreover, the inactivation of liver PPARγ abolished rosiglitazone-induced an increase in hepatic TG and improved hepatic steatosis in lipoatrophic AZIP mice (Gavrilova, Haluzik et al., 2003). In this study, Snhg3 overexpression induced an increase in hepatic PPARγ and CD36. Although the administration of PPARγ antagonist T0070907 aggravated hepatic steatosis owing to an inhibition of adipogenesis in adipose tissue, it improved Snhg3-induced hepatic steatosis. Even though PPARγ’s primary function is in adipose tissue, patients with MASLD have much higher hepatic expression levels of PPARγ, reflecting the fact that PPARγ plays different roles in different tissues and cell types (Mukherjee et al., 2022). Thus, understanding how the hepatic PPARγ expression is regulated will help develop prevention and treatments for MASLD (Lee et al., 2018).

Epigenetics plays a crucial role in many physiological and pathological situations (Peixoto, Cartron et al., 2020). Epigenetic regulation induces phenotypic changes that may respond to environmental cues through DNA methylation and histone modification, chromatin remodeling, and noncoding RNAs (Mann, 2014). Epigenetic changes interact with inherited risk factors to modulate the individual risk of MASLD development and the severity of progression. Epigenetic modifications, including DNA methylation, miRNAs, and histone modifications, have been associated with MASLD (Baffy, 2015, Eslam, Valenti et al., 2018, Jonas & Schurmann, 2021). To date, there is no approved pharmacologic therapy for MASLD, and the mainstay of management remains lifestyle changes with exercise and dietary modifications (Bayoumi et al., 2020). Therefore, understanding the epigenetic modifications in MASLD pathogenesis might prove a rational strategy to prevent the disease and develop novel therapeutic interventions (Sodum, Kumar et al., 2021).

LncRNAs, being abundant in the genome participate in regulating the expression of coding genes through various molecular mechanisms, including: (1) transcriptional regulation at the promoter of target genes; (2) inhibiting RNA polymerase II or mediating chromatin remodeling and histone modification; (3) interfering with the splicing and processing of mRNA or producing endogenous siRNA; (4) regulating the activity or cellular localization of the target protein; (5) acting as competitive endogenous RNAs; and (6) riboregulation by forming nucleic acid-protein complex as structural component (Morey & Avner, 2004, Sommerauer & Kutter, 2022, Thomson & Dinger, 2016). However, compared to the large number of lncRNAs, only few have been functionally well-characterized. Collective literature has shown that lncRNAs play a crucial role in MASLD (Sommerauer & Kutter, 2022). This study demonstrated that lncRNA-Snhg3 participated in the pathology of MASLD by epigenetic modification; that is, Snhg3 inhibited the H3K27me3 level, and promoted chromatin relaxation at the Pparγ promoter and eventually increased PPARγ expression. The results from Ruan et al. demonstrated that more than a third of dynamically expressed lncRNAs were deregulated in a human MASLD cohort and the lncRNA human lncRNA metabolic regulator 1 (hLMR1) positively regulated transcription of genes involved in cholesterol metabolism (Ruan, Li et al., 2021). Previous studies have also demonstrated that several lncRNAs, including FLRL2/3/6/7/8, H19, and MALAT-1, were associated with lipogenesis via proteins in the PPAR signaling pathway (Mukherjee et al., 2022). Recently, a murine long noncoding single-cell transcriptome analysis elucidated liver lncRNA cell-type specificities, spatial zonation patterns, associated regulatory networks, and temporal patterns of dysregulation during hepatic disease progression. Moreover, a subset of the liver disease-associated regulatory lncRNAs identified have human orthologs (Karri & Waxman, 2023). Based on the aforementioned information, lncRNAs emerge as promising candidates for biomarkers and therapeutic targets for MASLD.

Tudor proteins play vital roles in normal cell viability and growth by diverse epigenetic functions, including methylation dependent chromatin-remodeling, histone-binding, pre-RNA-processing, RNA-silencing, and transposon silencing in ligands (Ying & Chen, 2012). Tudor proteins are divided into four groups: Group1 Tudor proteins bind to methyl-lysine⁄arginine of histone tails, including Tdrd3, PHF1, PHF20, the JMJD family and TP53BP1; Group 2 Tudor proteins bind to methyl-arginine of ligands and representative members include SMN and SMNDC1; Group 3 is represented by SND1; and Group 4, contains many Tudor proteins, including Tdrd1-9 and Tdrd11, that have been identified in methylation-dependent association with PIWI proteins Ago3, Aub and Piwi. Some studies suggested that SND1 plays important roles in cancer by interacting with other transcription factors, including PPARγ, signal transducer and activator of transcription 5/6 (STAT6/5) and myeloblastosis oncogene (c-Myb) (Duan, Zhao et al., 2014, Navarro-Imaz et al., 2020). SND1 could induce adipogenesis and promote the formation of lipid droplets in adipocytes through working as a co-activator of PPARγ and regulating H3 acetylation (Duan et al., 2014). Our study showed that both Snhg3 and SND1 decreased the H3K27me3 level and promoted the expression of PPARγ. SND1 could interact with Snhg3 and mediate the Snhg3-induced decrease in H3K27me3 and increase in Pparγ and Cd36 expressions. Furthermore, inhibition of PPARγ with T0070907 alleviated Snhg3- and SND1-induced Cd36 increase and improved Snhg3-aggravated hepatic steatosis. In lncRNA riboregulation, the actions of noncoding RNAs mostly rely on interactions with proteins, including canonical or noncanonical RNA-binding proteins (RBPs). Canonical RBPs, such as heterogeneous nuclear ribonucleoproteins (hnRNPs), polypyrimidine tract binding protein 1 (PTBP1) and human antigen R (HUR), are often involved in posttranscriptional regulation, including pre-mRNA processing, RNA stability, RNA decay, or nuclear export (Briata & Gherzi, 2020). Previous studies have demonstrated that some lncRNAs, e.g., LINC01018, MEG3, APOA4-AS, hLMR1, Blnc1 and LncARSR interact with canonical RBPs to govern the progression of MASLD (Sommerauer & Kutter, 2022). In summary, our study demonstrates that lncRNA-Snhg3 increased the expression, stability, and nuclear localization of SND1 protein by interacting with SND1, thus enhancing the expression of PPARγ via reducing H3K27me3 enrichment and boosting chromatin loose remodeling at the Pparγ promoter, ultimately resulting in hepatic steatosis under DIO status (Figure 8H). This study reveals a new signaling pathway, Snhg3/SND1/H3K27me3/PPARγ, responsible for hepatic steatosis and provides evidence of lncRNA-mediated epigenetics in the pathophysiology of MASLD.

Materials and Methods

Animals and treatments

C57BL/6 Snhg3-Flox mice and hepatocyte-specific Snhg3 knock-in (Snhg3-HKI) mice were created using the CRISPR-Cas9 system at Cyagen Biosciences. To engineer the targeting vector for Snhg3-Flox mice, the exon 3 of Snhg3 was selected as the conditional knockout region, and homology arms and the cKO region were generated by PCR using BAC clone as a template. Cas9 and gRNA were co-injected into fertilized eggs with a targeting vector for mice production. The obtained mice were identified by PCR followed by sequence analysis. Hepatocyte-specific Snhg3 knock-out (Snhg3-HKO) mice were generated by crossing Snhg3-Flox mice with C57BL/6-Alb-Cre mice. For Snhg3-HKI mice, the “Alb promoter-mouse Snhg3 cDNA-polyA” cassette was inserted into an H11 locus (∼0.7 kb 5’ of the Eif4enif1 gene and ∼4.5 kb 3’ of the Drg1 gene), and homology arms were generated by PCR using BAC clone as template to engineer the targeting vector. Cas9 and gRNA were co-injected into fertilized eggs with targeting vector for mice production. The obtained mice were identified by PCR followed by sequence analysis. All mice were housed in the pathogen-free conditions (SPF) facility and maintained on a 12 h light-dark cycle and a regular unrestricted diet. All mice were fed either a normal chow diet (9% fat; Lab Diet) or HFD (60% fat, Research Diets) for inducing obesity and libitum with free access to water. Unless otherwise noted, 6∼8-week-old male mice were used for all experiments. 8-week-old mice fed on HFD were injected intraperitoneally (i.p.) with 1mg/kg of T0070907 dissolved in DMSO for 5 days per week for 2 months. Liver tissue samples were analyzed by the High Fatty Sample Total Cholesterol (TC) Content Assay Kit (APPLYGEN, Cat#E1025-105) and the High Fatty Sample Triglyceride (TG) Content Assay Kit (APPLYGEN, Cat#E1026-105), respectively. Serum concentrations of ALT, AST, FFAs, TG and TC were determined using an automated Monarch device (Peking Union Medical College Hospital, Beijing, China). Serum insulin was detected using a mouse insulin ELISA kit (JM-02862M1, Beijing BioDee Biotechnology Co., Ltd.). All animal experiments were conducted under protocols approved by the Animal Research Committee of the Institute of Laboratory Animals, Institute of Basic Medical Sciences Chinese Academy of Medical Sciences & School of Basic Medicine Peking Union Medical College (ACUC-A01-2022-010).

Cell culture

Primary mouse hepatocytes were isolated from 8-week-old male C57BL/6J mice and cultured in RPMI 1640 medium with 10% FBS as previously described (Matsumoto, Ogawa et al., 2002). Hepa 1-6 cells (ATCC, Cat#CRL-1830) were cultured at 37 °C, 5% CO₂ in DMEM (Gibco, Carlsbad, USA) supplemented with 10% FBS, and 1% penicillin-streptomycin. After attachment, the cells were transfected with indicated plasmids or siSnd1 by Lipofectamine 3000 Transfection Kit (Invitrogen). Cells were treated with and without 0.25 μm PA for 12h - 24h before collection. The sequences of siSnd1were #1, TGGAGAACATGCGCAATGACATC; #2, GTGCATGTCTTCTACATCGACTA; #3, GTATTGCCAGCTCAAGCCACAGAGTAT. UUCUCCGAACGUGUCACGUTT was used as a control.

Plasmid construction

Snhg3 was amplified from liver cDNA and was then constructed into pcDNA3.1 and pAd-Track-CMV using Kpn I/EcoR I and Kpn I/EcoR V, respectively. The primers for pcDNA3.1-Snhg3 were ATATCGGGTACCGACTTCCGGGCGTTAC and ATGATCGAATTCAGACATTCAAATGCT; for pAd-Track-CMV-Snhg3 were ATATCGGGTACCGACTTCCGGGCGTTAC and ATGATTGATATCAGACATTCAAATGCT.

Real-Time Quantitative PCR (RT-qPCR)

Total RNA was extracted from mouse tissues using a Trizol-based method. Approximately 2 μg of total RNA was reverse-transcribed into a first-strand cDNA pool using reverse transcriptase and random primers, according to the manufacturer’s instructions. RT-qPCR was performed using SYBR Green PCR Master Mix (A6002, Promega) with the gene-specific primers (Table S9). All gene expression data were normalized to actin expression levels.

Western blotting

Protein was extracted from frozen tissue samples in cell lysis buffer. In total, protein was loaded onto a 10% SDS-polyacrylamide gel, and separated proteins were transferred to PVDF membranes. Western blot assays were performed using indicated specific antibodies (Table S10). The proteins were quantified by ImageJ software.

Coding potential prediction

The coding potential of Snhg3 was evaluated by the Coding Potential Calculator at CPC2@CBI,PKU(gao-lab.org) (Kang et al., 2017).

Histopathologic analysis

Liver tissue sections were fixed in 4% paraformaldehyde, then embedded in paraffin and stained with H&E to visualize the general morphological and structural characteristics of tissues. Lipid droplet accumulation in the liver was visualized using Oil red O staining of frozen liver sections that were prepared in optimum cutting temperature (O.C.T.) compound.

Subcellular fractionation

A Cytoplasmic & Nuclear fraction Kit (Beyotime Biotechnology, China) was used to detect Snhg3 expression in cytoplasmic and nuclear fractions. RNA was extracted from the cytoplasmic and nuclear fractions using a Trizol-based method and subjected to qPCR. Gapdh was used as a cytoplasmic marker, and Neat1 and Xist were used as a nuclear marker. The percentage of the transcript abundance was calculated using the following formula, Nucleus % = 2^^Ct(Nucleus)/(2^{^Ct(Cytoplasm)} + 2^{^Ct(Nucleus)}), Cytoplasm % = 1-Nucleus %.

Mouse Calorimetry

Male mice were housed individually in metabolic chambers of an Oxymax system (Columbus Instruments). The first readings were taken after a 24-hr acclimation period. Heat production, total carbon dioxide production and oxygen consumption, and RER were determined by Comprehensive laboratory animal monitoring system (CLAMS).

Insulin Tolerance test (ITT) and Glucose Tolerance test (GTT)

For ITT, male mice fasted for 4 h and received an intraperitoneal injection of human insulin (0.75 IU/kg). For GTT, male mice fasted for 16 h received an intraperitoneal injection of glucose (1 g/kg). A series of blood glucose concentrations were measured from tail blood at the indicated times using a One-Touch Ultra® glucometer (LifeScan Inc., Milpitas, CA).

Chromatin Immunoprecipitation (ChIP)

The ChIP assay was performed using Sonication ChIP Kit (Abclonal, Cat#RK20258). Briefly, the liver tissues or primary hepatocytes were collected and cross-linking fixed. Cross-linked chromatin fragments were precipitated with Rabbit control IgG (Abclonal, Cat#AC005) or anti-H3K27me3 antibody (Abclonal, Cat# A16199) for subsequent PCR analysis using the amplification primers for mouse PPARγ promoter (+101∼+420bp), TATTGGGTCGCGCGCAGCC and ACACAGTCCTGTCAGAACG.

RNA immunoprecipitation (RIP)

The RIP assay was performed using RIP Assay Kit (BersinBio, Cat#Bes5101). Briefly, heap 1-6 cells were transfected by indicated plasmids, respectively. The cells were collected, cross-linking fixed and precipitated with Mouse Control IgG (Abclonal, Cat#AC011) or anti-FLAG antibody (Abclonal, Cat#AE005) for subsequent RT-qPCR analysis using the amplification primers for Snhg3.

RNA sequencing (RNA-Seq)

The RNA-Seq was performed according to the manufacturer’s protocol (BGI-Shenzhen, https://www.yuque.com/yangyulan-ayaeq/oupzan/fuoao4). Briefly, total RNA was extracted from liver of 3 male DIO (27 weeks) mice and 3 male control mice for RNA-Seq to screen the differentially expressed lncRNAs using Trizol (Invitrogen, Carlsbad, CA, USA) according to manual instruction. rRNA in total RNA removed using RNase H kit, was subsequently to construct library and perform sequencing analysis. The data were mapped to mouse genome (GRCm39) by using Bowtie2. The data for the differentially expressed lncRNAs had been deposited to National Genomics Data Center, China National Center for Bioinformation (NGDC-CNCB) (https://ngdc.cncb.ac.cn/) with the dataset identifier CRA009822. Additionally, total RNA was extracted from livers of 3 male DIO Snhg3-HKI mice and 3 male DIO WT mice for RNA-Seq to screen the differentially expressed mRNAs using Trizol. Total RNA was enriched by oligo (dT)-attached magnetic beads, followed by library construction and sequencing analysis. The data were mapped to mouse genome (GRCm39) by using Bowtie2. The data for the differentially expressed mRNAs have been deposited to the Sequence Read Archive (SRA) (submit.ncbi.nlm.nih.gov) with the dataset identifier PRJNA904058.

Assay for Transposase-Accessible Chromatin with high throughput sequencing (ATAC-Seq)

The ATAC-seq was performed according to manufacturer’s protocols (BGI_Shenzhen, https://www.yuque.com/yangyulan-ayaeq/oupzan/lllmzg). Briefly, fresh liver tissue samples from 3 male DIO Snhg3-HKI mice and 3 male DIO control mice were flash frozen by liquid nitrogen and then ground completely. The transposition reactions were initiated by adding transposase. The PCR reaction system was configured to initiate PCR amplification of the transposition products. The corresponding library quality control protocol would be selected depending on product requirements. Single-stranded PCR products were produced via denaturation. The reaction system and program for circularization were subsequently configured and set up. Single-stranded cyclized products were produced, while uncyclized linear DNA molecules were digested. Single-stranded circle DNA molecules were replicated via rolling cycle amplification, and a DNA nanoball (DNB) which contain multiple copies of DNA was generated. Sufficient quality DNBs were then loaded into patterned nanoarrays using high-intensity DNA nanochip technique and sequenced through combinatorial Probe-Anchor Synthesis (cPAS). Data were filtered by removing adaptor sequences, contamination and low-quality reads from raw reads. Bowtie2 was used to do genome alignment after evaluating its performance.

Peak Calling was performed by MACS (Model-based Analysis for ChIP-Seq). The candidate Peak region was extended to be long enough for modeling. Dynamic Possion Distribution was used to calculate the p-value of the specific region based on the unique mapped reads. The region would be defined as a Peak when the p-value < 1e-05. MACS works well for the identification of the sharp peaks of most sequence-specific transcription factors.

Peak Distribution On Gene Elements. Peaks were classified based on the location (UCSC annotation data) and showed in the following genome regions: promoter (<=1kb), promoter (1−2kb), promoter (2−3kb), 5’-UTR, 3’-UTR, intergenic, introns, and exons.

Differential Peaks were identified using MAnorm. First, the true intensities of the most common peaks were assumed to be the same between two ATAC-Seq samples. Second, the observed differences in sequence read density in common peaks were presumed to reflect the scaling relationship of ATAC-Seq signals between two samples, which could thus be applied to all peaks. Based on these hypotheses, the log2 ratio of read density between two samples M was plotted against the average log2 read density A for all peaks, and robust linear regression was applied to fit the global dependence between the M-A values of common peaks. Then the derived linear model was used as a reference for normalization and extrapolated to all peaks. Finally, the p-value for each Peak was calculated based on the Bayesian model, the significant regions were picked up if |M=log2FC| > 2 and p-value < 0.001. To identify differentially expressed genes (DARs), the count matrix was input into MAnorm with the cutoff of abs(log2FC)>2 and p-value<0.001. Genomic features of DARs were annotated by R package ChIPseeker (v1.30.3).

Differential motifs analysis. After extracting the corresponding peak sequence, Hypergeometric Optimization of Motif EnRichment (HOMER, Homer Software and Data Download (ucsd.edu)) and Cis-Regulatory Element Motif Activities (CREMA - Cis-Regulatory Element Motif Analysis (unibas.ch)) were used for motif analysis. Genomic regions with differential ATAC peaks were shown using IGV software.

The data of ATAC-seq has been deposited to National Genomics Data Center, China National Center for Bioinformation (NGDC-CNCB) (https://ngdc.cncb.ac.cn/) with the dataset identifier CRA009511.

Integrated analysis ATAC-Seq data with RNA-Seq data

The common number and unique number of genes associated with DARs and DEGs were counted using Wayne comparative analysis. The values of each quadrant satisfying the log2FC condition were selected from DARs-associated genes and DEGs respectively to draw a nine-quadrant plot. Pearson correlation was calculated for both sets of data and p-value was calculated using the Z-test.

Cleavage Under Targets and Tagmentation sequencing (CUT&Tag-Seq)

CUT&Tag was performed according to the Hyperactive Universal CUT&Tag Assay Kit for Illumina (Vazyme, Nanjing, Cat#TD903-01). Briefly, the mixed liver tissues from 3 DIO Snhg3-HKO mice were used for the CUT&Tag experiment. pA-Tn5 transposase was used to cut the genome and add a special adaptor sequence to build a library. The single-stranded PCR products were sequenced on illumina/DNBSEQ-T7 platform PE150 (Annoroad Gene Technology Co.Itd). The data were mapped to mouse genome (GRCm38) by using Bowtie2. Genomic features of DARs were annotated by R package ChIPseeker (v1.30.3). The data of Cut&Tag has been deposited to National Genomics Data Center, China National Center for Bioinformation (NGDC-CNCB) (https://ngdc.cncb.ac.cn/) with the dataset identifier CRA009582.

Biotin-RNA pull-down and mass spectrometry assay

Biotin-RNA pull-down assay was performed as described in a previous study(Guo, Zhao et al., 2018). Briefly, Snhg3 DNA was amplified from mouse liver cDNA using the primers listed in the Table 1 and lacZ DNA fragment were constructed into pGEM-T easy vector. The pGEM-T-Snhg3 and pGEM-T-lacZ vectors were linearized by restriction enzyme digestion, then transcribed to Snhg3 and lacZ fragments. Biotinylated RNAs were transcribed in vitro with Biotin-RNA Labeling Mix (Roche, Indianapolis, IN) and purified with quick spin RNA columns (Roche, Indianapolis, IN). Biotin-labeled RNA or unbiotinylated RNAs was dissolved in RNA structure buffer (10 mM Tris, pH 7.0, 0.1 M KCl, 10 mM MgCl2) to allow formation of the secondary structure. Primary hepatocytes lysates were added to biotin-labeled RNA or unbiotinylated RNA. Streptavidin agarose beads (GE Healthcare, Little Chalfont, UK) were mixed with a pull-down reaction and then rotated constantly. RNA affinity captures were subjected to 12% SDS-PAGE followed by coomassie blue staining or Western blotting. The various bands that were visualized by coomassie blue staining were excised and subjected to mass spectrometry analyses (LC-MS/MS, A TripleTOF®, ABsciex, Concord, ON). The data of RNA pull-down for Snhg3, control or lacZ were deposited to the iProX (https://www.iprox.cn/) with the dataset identifier PXD039526.

Ubiquitination assays

For endogenous ubiquitination assays, Hepa1-6 cells were transfected with the indicated combinations of plasmids, including HA-ubiquitin and Snhg3 plasmids. For exogenous ubiquitination assays, Hepa1-6 cells were transfected with the indicated combinations of plasmids, including HA-ubiquitin, HA-K33-ubiquitin, HA-K63-ubiquitin, HA-K48-ubiquitin, Flag-SND1 and Snhg3 plasmids. Cells were treated with 20 µM MG132 proteasome inhibitor (M1902, AbMole) for 6 hours prior to lyse in lysis buffer (200 mM NaCl, 20 mM Tris-HCl (pH 7.4), 2.5 mM MgCl2, 0.5% Triton X-100, 1 mM PMSF, and protease inhibitor cocktail and then were sonicated. After centrifugation at 14,000 g, the cleared lysates were subjected to immunoprecipitation with anti-SND1 antibody (sc-166676, Santa Cruz) for endogenous ubiquitination assays or with anti-DDDDK-tag magnetic beads (M185-10R, MBL) for exogenous ubiquitination assays. The immunocomplexes were collected and subjected to Western blotting with the indicated antibodies.

Statistical analysis

Data analyses were performed with SPSS (Version 17.0, SPSS, Inc.). The curves of body weight and ITT were analyzed using a repeated measure two-way ANOVA. For the other statistical analysis, the Kolmogorov-Smirnov test was firstly used for normality test. For the data conforming to the normal distribution, experiments were analyzed using Independent-Samples T-test or one-way ANOVA. All data were presented as the mean ± SD or the mean ± SEM. * p < 0.05 was considered statistically significant.

Acknowledgements

This work was supported by Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (CIFMS2021-I2M-1-016) and National Key R&D Program of China (2022YFC2504002, 2022YFC2504003). We thank Dr. Yi Li (Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College) for providing the plasmid of HA-Ub, HA-Ub-K63 and HA-Ub-K48. We thank Cyagen Biosciences for the collaborative efforts in the creation of Snhg3-loxP and liver-specific knock-in Snhg3 mice. We thank BGI-Shenzhen for the collaborative efforts in RNA-Seq and ATAC-Seq technology. We thank Anoroad gene technology (Beijing) for the collaborative efforts in Cut&Tag-Seq and Jingjie PTM BioLab (Hangzhou) Co. Ltd for the collaborative efforts in mass spectrometry technology resources.

Author Contributions

Xianghong Xie performed all experiments, acquired and analyzed the data, generated the figures and wrote the manuscript. Mingyue Gao, Wei Zhao, Chunmei Li, Weihong Zhang, Jiahui Yang, Yinliang Zhang, Enhui Chen, Yanfang Guo, Zeyu Guo, Minglong Zhang, Yinghan Zhu, Yiting Wang and Xiaolu Li performed experiments, acquired and analyzed the data. Ebenezeri Erasto Ngowi revised the manuscript. Heping Wang and Xiaoman Wang analyzed the data and generated the figures. Hong Yao, Li Yan and Fude Fang provided additional reagents and revised the manuscript. Meixia Li, Aijun Qiao and Xiaojun Liu devised and supervised the experimental plan and edited the manuscript.

Conflicts of Interest

The author(s) declare(s) that there is no conflict of interest regarding the publication of this article.

Data Availability

All data generated and analysed that support the findings of this study are available from the corresponding author upon reasonable request. The lncRNA-Seq data had been deposited to National Genomics Data Center, China National Center for Bioinformation (NGDC-CNCB) (https://ngdc.cncb.ac.cn/) with the dataset identifier CRA009822. The data of RNA pull-down for Snhg3, control or lacZ have been deposited to the iProX (https://www.iprox.cn/) with the dataset identifier PXD039526. The data of RNA-seq have been deposited to the Sequence Read Archive (SRA) with the dataset identifier PRJNA904058. The data of ATAC-seq and Cut&Tag have been deposited to National Genomics Data Center, China National Center for Bioinformation (NGDC-CNCB) (https://ngdc.cncb.ac.cn/) with the dataset identifier CRA009511 and CRA009582, respectively.

Supplementary Materials

Table S1 to S10.