Long Noncoding RNA lncSHGL Recruits hnRNPA1 to Suppress Hepatic Gluconeogenesis and Lipogenesis

The human and other mammalian genomes produce a huge number of transcripts (1,2), 80–90% of which are not traditional protein-coding RNAs and termed long noncoding RNAs (lncRNAs) with the length >200 nucleotides (3–6). To date, a great number of lncRNAs have been identified in the tissues and circulation of humans and other mammals (6–10), and dysregulated lncRNA expression profiles are involved in the pathogenesis of many diseases (6,11–15). So far, the lncRNAs with function annotations are very few in number. Clearly, lncRNA is a huge treasure vault full of unknown but exciting molecules awaiting exploration.

There has been increasing evidence that lncRNAs regulate glucose and lipid metabolism. Dysregulated lncRNA expression profile is associated with islet dysfunction in humans (16). Knockdown of lncRNA TUG1 causes pancreatic β-cell dysfunction (17). lncRNA H19 regulates glucose metabolism by functioning as a sponge for microRNA let-7 (18,19). lncRNAs long noncoding liver-specific triglyceride regulator (lncLSTR) and MEG3 also regulate hepatic glucose and lipid metabolism (20,21). On one hand, the effects and mechanisms of the reported lncRNAs in glucose and lipid metabolism in various tissues need further validation (16,18–21). On the other hand, to further characterize new lncRNAs that regulate glucose and lipid metabolism is also of great importance. Particularly, identifying new lncRNAs that regulate hepatic gluconeogenesis will shed light on the pathogenesis of type 2 diabetes.

A great number of new lncRNAs were identified in mouse liver and plasma in our previous studies (7,8). Eleven lncRNAs dysregulated in liver after ischemia/reperfusion injury (IRI) had been validated and identified with high expression (7,8). Among these 11 lncRNAs, AK143693 is a nonsecretory lncRNA that exhibits high expression in mouse liver with unknown function(s) (7,8). We found in the preliminary experiment that AK143693 silencing increased lipid deposition in liver after IRI (Supplementary Fig. 1A and B), suggesting it may play roles in regulating hepatic glucose/lipid metabolism. The current study revealed that lncRNA AK143693 functions as a novel suppressor of hepatic gluconeogenesis and lipogenesis (SHGL) and is renamed as lncSHGL.

lncSHGL expression was reduced in obese mouse livers. lncRNA B4GALT1-AS1, the human homologous sequence of mouse lncSHGL, was also reduced in human livers with steatosis. Hepatic lncSHGL overexpression suppressed gluconeogenesis and attenuated hyperglycemia and fatty liver in mice fed a high-fat diet (HFD), whereas hepatic lncSHGL repression promoted hyperglycemia and lipid deposition in normal mice. Mechanistically, lncSHGL recruited heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1) to enhance calmodulin (CaM) mRNAs translation. An increase in the CaM protein level finally suppressed gluconeogenic and lipogenic pathways in an insulin- and calcium-independent manner in hepatocytes.

Male C57BL/6 mice (8 to 10 weeks old) were fed a 45% HFD for 3 months to induce diabetic and steatotic phenotypes (22,23). The study also used 10- to 16-week-old male db/db mice on a BKS background. All animal experimental protocols complied with the Animal Management Rules of the Ministry of Health of the People’s Republic of China and the Peking University Guide for the Care and Use of the Laboratory Animals.

Anti-phosphorylated (p)Akt (phosphorylation at Ser473 site) and Akt antibodies were purchased from CST. Other antibodies were obtained from Santa Cruz Biotechnology, CST, or other commercial companies.

An adenovirus (Ad) expressing lncSHGL was constructed according to the AK143693 sequence in the PubMed database (https://www.ncbi.nlm.nih.gov/nuccore/AK143693) (7,8). To overexpress lncSHGL in mouse livers, 1.0 × 10⁹ plaque forming units Ad-lncSHGL or Ad-green fluorescent protein (GFP) were injected into mice via tail vein. Oral glucose tolerance tests (OGTTs), insulin tolerance tests (ITTs), and pyruvate tolerance tests were performed at the 7th day after viral injection using different sets of mice. On the 9th day, the mice were sacrificed on fed state. The serum and tissues were collected for biochemical analyses.

Stealth small interfering (si)-lncSHGL were synthesized by Invitrogen (sequences provided in Supplementary Table 1). The siRNA mixture was injected into C57BL/6 mice via tail vein (2.5 mg/kg body weight in 100 μL sterile saline) (23,24), the same dose of scrambled siRNA (Invitrogen) was used as the control. OGTTs were performed 72 h after the siRNA injection. On the 4th day, the mice were sacrificed for experimental analyses.

Cell lines or primary mouse hepatocytes were infected with 50 multiplicity of infection of Ad-lncSHGL or Ad-GFP for 24 h. For insulin-stimulation experiments, infected cells were serum starved for 12 h, followed by treating with 100 nmol/L insulin for 5 min. For inhibition of phosphatidyl inositol 3-kinase (PI3K) or P2 receptors or calcium signaling, infected cells were treated with 1 μmol/L wortmannin, 50 μmol/L pyridoxalphosphate-6-azophenyl-2',4'-disulfonic acid, 50 μmol/L suramin, 100 μmol/L chlorpromazine (CPZ), an inhibitor of CaM (25), 10 μmol/L nifedipine, or 10 μmol/L 2-aminoethoxydiphenyl borate for 1 h before experimental assays. For depleting extracellular Ca²⁺, infected cells were treated with Ca²⁺-free medium plus 0.5 mmol/L EGTA for 2 h, and chlorpromazine was added 1 h before experimental assays.

Cells plated in six-well plates were transfected with 5 μg plasmid for 24 h. Plasmids expressing human CALM1-2 and hnRNPA1 plasmids were purchased from OriGene (CALM1, Cat No. SC115829) and Vigene Biosciences China (CALM2, Cat No, CH809926; hnRNPA1, Cat No CH877838), respectively.

Target gene mRNA level was normalized to that of β-actin in real-time PCR assays. Each sample was assayed in duplicate in each experiment. All primer sequences are provided in Supplementary Table 2. The protein levels were analyzed by immunoblotting assay. For protein blot quantitation, each protein (nonphosphorylated) was first normalized to the corresponding housekeeping protein β-actin, and then the value was normalized to the control value in each experiment. For quantitation of phosphorylated protein, phosphorylated protein was first normalized to the corresponding total protein and then was normalized to the control value.

Cells seeded on coverslips were loaded with 1 μmol/L Fura-2 acetoxymethyl ester for 30 min, followed by Ca²⁺ level determination under Olympus IX71 fluorescence microscope. The emission intensities at the wavelengths of 340 nm and 380 nm were recorded every 1 s, and the average ratio of the emission densities (F340-to-F380) in 300 s reflected the basal free calcium level (22,23).

Cells seeded on coverslips were infected with Ad-lncSHGL for 24 h. The coverslips were blocked in 1% BSA for 30 min at 37°C after washing with PBS. The coverslips were incubated with anti-FOXO1 antibodies at 4°C overnight, and washed with PBS, followed by detection with goat anti-rabbit Alexa Fluor 594. After nuclear staining with DAPI, coverslips were mounted on glass slides using 50% glycerol in PBS. Imaged were visualized by fluorescence microscopy using a confocal laser scanning microscope.

As detailed previously (26), cells were infected with Ad-lncSHGL for 24 h and then treated with DMEM/high-glucose medium containing 20 mmol/L sodium lactate/2 mmol/L sodium pyruvate without glucose or phenol red (Gibco) for another 16 h, followed by treatment with 10 nmol/L insulin for 3 h. The culture medium was centrifuged to collect the supernatant for analyzing glucose content with the Glucose Assay Kit (Sigma-Aldrich). The glucose data were normalized with the cellular protein content (micrograms per milligram of protein).

PCR products labeled with or without biotin were amplified by PCR (each PCR product position in lncSHGL is provided in Supplementary Table 2) and purified. National Collection of Type Cultures (NCTC)-1469 cells infected with Ad-lncSHGL were cross-linked by ultraviolet exposure (400 mJ/cm²). Cells were lysed in RNA immunoprecipitation (RIP) buffer (20) and centrifuged at 12,000 rpm for 10 min to collect the supernatant. Purified PCR products (1–2 μg) were added into 1 mg supernatant protein after shortly heating at 95°C, and then quickly cooled on ice. The mixture of DNA and protein was incubated at 37°C for 4 h. Washed Dynabeads M-280 streptavidin (80 μL; Invitrogen) were added to each sample and incubated at room temperature for 30 min. Beads were washed five times with binding and washing buffer, and then boiled in 1×SDS loading buffer for 5 min. The samples were separated by PAGE and then subjected to silver staining and Western blotting assays. The target protein bands were analyzed by mass spectrometry (MS).

The method for RIP was described previously (27). Anti-hnRNPA1 antibody or rabbit IgG (4 μg) was added to 40 μL Protein A Resin (TransGen Biotech) with 100 μL coupling buffer (Thermo Scientific), and the mixture was incubated at 4°C for 1 h with gentle rotation. Cells treated with Ad-lncSHGL were crosslinked by ultraviolet exposure. Cells were lysed and centrifuged at 12,000 rpm for 10 min to collect the supernatant. Supernatant protein (1 mg) was added into each binding reaction and incubated at 4°C overnight. The Protein A Resin/protein mixture was washed with RIP buffer and resuspended in 500 μL Trizol for RNA extraction. The isolated RNAs were detected by real-time PCR with the normalization to control value.

Epididymal adipose tissue (50 mg) was dissected from mice and suspended in 500 μL phenol red-free DMEM (fDMEM) containing 5 mmol/L glucose, and horseradish peroxidase–conjugated IgG on ice and cut into 1 mm³ pieces. The samples were incubated at 37°C in 5% CO₂-95% air in 500 μL fDMEM for 30 min, washed three times, and incubated in 500 μL fDMEM. The medium was collected at 30 and 120 min for glycerol release measurement (28). The glycerol level was normalized with protein content.

The protocol for ribosome extraction was detailed previously (29). Treated cells were lysed in 0.5 mL radioimmunoprecipitation assay buffer (BioTeke Corporation). After being incubated 30 min on ice, the samples were centrifuged at 12,000 rpm at 4°C for 10 min to collect the supernatant. The supernatant was added to 9.5 mL precooled 30% sucrose buffer (20 mmol/L HEPES [pH 7.4], 50 mmol/L potassium acetate, and 5 mmol/L magnesium acetate; 1 mmol/L dithiothreitol, protease inhibitor cocktail, and recombinant RNase inhibitor were added before using), and centrifuged at 30,000 rpm at 4°C for 2 h. The pellets were resuspended in 1 mL precooled Buffer A (10 mmol/L HEPES [pH 7.9], 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.3 mol/L NaCl, and recombinant RNase inhibitor were added before using). After being incubated on ice for 1 h, the samples were centrifuged at 10,000 rpm at 4°C for 15 min to collect the supernatant for RNA determination. The same amount of RNAs was analyzed by real-time PCR with the data normalized to control values.

Hydrodynamics-based transfection in animals by tail vein administration of naked plasmid DNA was detailed previously (30–32). Briefly, HFD mice were divided into three groups based on OGTT. Endotoxin-free plasmid (50 μg) was dissolved in sterile saline with the volume of 8% of the body weight at room temperature and injected into tail vein in 7 s. OGTTs were performed 72 h after the plasmid injection. Mice were sacrificed 24 h later for assays, as described above.

Six steatotic and healthy liver samples were selected from 11 pairs of liver samples in the previous study (33). The ethics issue and biochemical characteristics of these six pairs of steatotic patients and healthy subjects were described previously (28,33).

Data are presented as the mean ± SEM. Statistical significance of differences between groups was analyzed by the t test. P values <0.05 were considered statistically significant.

Mouse lncSHGL is located in chromosome 17 between protein-coding sodium/calcium exchanger one isoform ×5 and extracellular tyrosine-protein kinase PKDCC precursor genes (Supplementary Fig. 1A) (7,8). In the preliminary study, we found that lncSHGL inhibition increased lipid deposition in the liver after IRI (Supplementary Fig. 1B). There is no open read frame >200 base pairs (66 aa) in the lncSHGL sequence (data not shown). lncSHGL is highly expressed in mouse metabolic tissues (Supplementary Fig. 1C). lncSHGL and AK139328 were reduced, whereas the other nine lncRNAs remained unchanged in HFD mouse livers (Fig. 1A). lncSHGL, AK139328, AK143294, and ENSMUST00000151138 were reduced, AK054386 was increased, and the other six lncRNAs were not changed in db/db mouse livers (Fig. 1B). In contrast, lncSHGL expression remained unchanged in muscle, heart, pancreas, and adipose tissues of obese diabetic mice (Supplementary Fig. 1D and E). Time-course analyses of HFD feeding on hepatic lncSHGL expression revealed that it was increased at 1 and 2 months but was decreased at 3 months after HFD feeding (Supplementary Fig. 2A). In NCTC-1469 cells (normal mouse liver cell line), free fatty acids (FFAs) increased lncSHGL expression at 6 h of treatment but repressed it after 18 h of treatment. Glucose similarly repressed lncSHGL expression after 18 h of treatment (Supplementary Fig. 2B and C). lncRNAs generally exhibit low sequence conservation across species. We had searched human homologous sequences of lncSHGL using National Center for Biotechnology Information Basic Local Alignment Search Tool (BLAST). As a result, the noncoding RNA B4GALT1-AS1 (34), whose biological function(s) also remain unknown, exhibits the best sequence match with lncSHGL (total score = 57.2, E value = 8e−5, identity = 73%). The B4GALT1-AS1 encodes three noncoding RNA isoforms. Multiple sequence alignment of lncSHGL and B4GALT1-AS1 isoforms was performed using the ClustalX tool (Supplementary Fig. 3A). In human livers with steatosis, total B4GALT1-AS1 levels were reduced (Fig. 1C). Hematoxylin and eosin staining revealed significant lipid droplet in human livers with steatosis (Supplementary Fig. 3B). Patients with fatty liver have higher serum triglyceride (TG) and cholesterol (CHO) levels but have similar BMI, waist circumference, liver function, and other characteristics as healthy subjects (28). Overall, a decrease in hepatic lncSHGL and B4GALT1-AS1 expression is associated with dysregulated glucose/lipid metabolism in mice and humans.

lncSHGL had been overexpressed in livers of HFD mice via tail vein injection of Ad-lncSHGL. Fasting hyperglycemia was improved at the 4th and 7th day after the viral injection (Fig. 1D). Glucose intolerance was improved at the 7th day after the viral injection (Fig. 1E). ITTs and hyperinsulinemic-euglycemic clamp revealed the improvement of global insulin resistance at the 7th day after the Ad-lncSHGL injection (Fig. 1F and Supplementary Fig. 4A–D). Hepatic gluconeogenesis was decreased at the 7th day after the Ad-lncSHGL injection (Supplementary Fig. 5A and B). Serum aspartate aminotransferase and alanine aminotransferase activities were not different between HFD mice treated with Ad-GFP or Ad-lncSHGL (Supplementary Fig. 5C). The Ad-lncSHGL injection resulted in specific lncSHGL overexpression in mouse livers (Fig. 2A). Morphological, Oil Red O staining, and quantitative assays indicated that lncSHGL overexpression reduced hepatic but not serum TG and CHO content (Fig. 2B–D). Hepatic lncSHGL overexpression reduced serum insulin levels (Fig. 2E). Consistent with enhanced insulin sensitivity, hepatic lncSHGL overexpression repressed the lipolysis of white adipose tissue and reduced serum FFA levels in HFD mice (Fig. 2F).

To confirm the metabolic roles of lncSHGL, its expression in normal C57BL/6 mouse livers was knocked down by tail vein injection of siRNAs. The injection of siRNAs specifically repressed lncSHGL expression in the livers (Fig. 3A). Mice treated with si-lncSHGL exhibited fasting hyperglycemia compared with control mice (Fig. 3B). Oil Red O staining and quantitative assays revealed that lncSHGL inhibition increased hepatic but not serum TG and CHO content (Fig. 3C–E).

lncSHGL overexpression increased pAkt with reduced protein levels of PEPCK, G6Pase, and fatty acid synthase (FAS) in HFD mouse livers (Fig. 4A). lncSHGL overexpression reduced the mRNA levels of PEPCK and G6Pase in HFD mouse livers (Fig. 4B). lncSHGL inhibition increased the mRNA level of PEPCK in normal mouse livers (Supplementary Fig. 6A). lncSHGL silencing reduced pAkt level with increased protein levels of gluconeogenic and lipogenic enzymes (Supplementary Fig. 6B). In HFD mouse livers, lncSHGL overexpression increased pFOXO1 but reduced pmTOR, with little effect on pGSK3 (Fig. 4C). Hepatic lncSHGL overexpression reduced SREBP-1C precursor protein expression in mouse livers (Fig. 4C). lncSHGL overexpression increased the pFOXO1 level and reduced pmTORC1 and SREBP-1C protein levels without affecting pGSK3 levels in primary mouse hepatocytes (Fig. 4D) and HepG2 cells (Supplementary Fig. 7A). Furthermore, lncSHGL overexpression reduced the levels of pS6K1 and pS6, two downstream molecules of mTORC1, and the protein level of Rictor, one of the key components of mTORC2, but had little effect on pAMPK in diabetic mouse livers (Fig. 4C) and primary mouse hepatocytes (Fig. 4D). In support of the changes in SREBP-1C and FAS protein levels, lncSHGL overexpression reduced whereas lncSHGL inhibition increased their mRNA levels in mouse livers (Supplementary Fig. 7B and C). lncSHGL overexpression increased whereas lncSHGL inhibition reduced the mRNA level of ACACB, one of the important genes controlling fatty acid oxidation, in mouse livers. In contrast, lncSHGL overexpression or silencing had little effect on the mRNA levels of SCD1, DGAT1, AOX, ApoB, FABP1, CPT1α, and PDK4 in mouse livers (Supplementary Fig. 7B and C). Overall, these findings revealed that lncSHGL suppressed hepatic lipogenesis and likely increased lipid oxidation.

Because Akt activity and its downstream molecules were changed after lncSHGL overexpression or inhibition in mouse livers, whether lncSHGL directly modulated Akt activity was further evaluated in cultured hepatocytes. lncSHGL overexpression in HepG2 cells promoted Akt activation independent of insulin (Fig. 5A). lncSHGL-induced Akt activation was blocked by inhibitors of PI3K (wortmannin) and CaM (CPZ) but was not affected by inhibitors of P2 receptors (Fig. 5B), L-type calcium channel, and inositol trisphosphate receptor (IP3R) (Supplementary Fig. 8A). lncSHGL had little effect on the cellular Ca²⁺ level (Supplementary Fig. 8B), and depleting extracellular calcium failed to affect its stimulatory effect on Akt phosphorylation in HepG2 cells (Fig. 5C). However, lncSHGL-induced Akt activation was completely blocked by the CaM inhibitor in the presence or absence of extracellular calcium in HepG2 cells (Fig. 5C). lncSHGL similarly activated Akt in an insulin-independent but CaM-dependent manner in primary mouse hepatocytes (Supplementary Fig. 9A and Fig. 5D). lncSHGL promoted FOXO1 nuclear exclusion in a CaM-dependent manner in HepG2 cells (Fig. 5E) and mouse hepatocytes (Supplementary Fig. 9B). lncSHGL repressed gluconeogenic gene expression and gluconeogenesis in HepG2 cells (Fig. 5F) and mouse hepatocytes (Supplementary Fig. 9C and D). Overall, lncSHGL activated the PI3K/Akt pathway in an insulin-independent but CaM-dependent manner.

To further determine the CaM-dependent mechanism of Akt activation induced by lncSHG, its effect on CaM expression was analyzed. Three calmodulin genes, designated as CALM1, CALM2, and CALM3, respectively, encode one identical CaM protein (35). lncSHGL overexpression increased whereas lncSHGL silencing decreased CaM protein level with little effect on CALM1-3 mRNA levels in mouse livers (Fig. 6A and B). lncSHGL overexpression also increased the CaM protein level without affecting CALM1-3 mRNA levels in mouse hepatocytes (Fig. 6C) and HepG2 cells (Fig. 6D and E). In HFD mouse livers, lncSHGL overexpression had little effect on the ubiquitination of CaM protein (Supplementary Fig. 10). CALM2 plasmid transfection activated Akt independent of insulin in HepG2 cells (Fig. 6F). lncSHGL clearly increased CaM protein to induce Akt activation in hepatocytes. Moreover, CaM overexpression reduced pmTORC1 and SREBP-1C protein levels in mouse hepatocytes (Supplementary Fig. 11A). CaM interacted with mTOR, and the interaction was increased after lncSHGL and CaM overexpression in mouse hepatocytes (Supplementary Fig. 11B and C).

An RNA pull-down assay was performed to further pinpoint the mechanism(s) of the lncSHGL-induced increase in the CaM protein level in liver cells. MS analysis had identified hnRNPA1 as the target protein of the indicated band in Fig. 7A (MS data are reported in Supplementary Table 3). Immunoblotting assay confirmed that lncSHGL interacted with hnRNPA1 in NCTC cells (Fig. 7B). lncSHGL overexpression increased the lncSHGL-hnRNPA1 interaction without significantly affecting the hnRNPA1 protein level (Fig. 7B). RIP revealed that hnRNPA1 bound with CALM1-3 mRNAs and that the bindings were enhanced by lncSHGL overexpression in NCTC cells (Fig. 7C). lncSHGL overexpression also increased the bindings of ribosomes with CALM1-3 mRNAs (Fig. 7D). hnRNPA1 overexpression increased CaM and pAkt protein levels with little effect on CALM1-3 mRNA levels in primary mouse hepatocytes (Fig. 7E and F) and HepG2 cells (Supplementary Fig. 12A and B), whereas hnRNPA1 silencing reduced CaM and pAkt protein levels without a significant effect on CALM1-3 mRNA levels in HepG2 cells (Supplementary Fig. 13A and B). hnRNPA1 knockdown inhibited the lncSHGL-induced increase in CaM and pAkt levels without a significant effect on CALM1-3 mRNA levels (Supplementary Fig. 13C and D). Consistent with the changes in CaM and pAkt levels, hnRNPA1 silencing increased glucose production and impaired repression of lncSHGL on glucose production in HepG2 cells (Supplementary Fig. 13E).

Hepatic overexpression of hnRNPA1 and CaM ameliorated hyperglycemia and steatosis of HFD mice (Fig. 8A and B). hnRNPA1 overexpression increased the CaM protein level without a significant effect on CALM1-3 mRNA levels in HFD mouse livers (Fig. 8C and D). hnRNPA1 overexpression activated Akt and reduced pmTORC1 and SREBP-1C levels with the repression of gluconeogenic and lipogenic gene expression (Fig. 8C). hnRNPA1 overexpression also reduced pS6K1 and pS6, and Rictor protein levels in HFD mouse livers (Supplementary Fig. 14A).

CaM overexpression also increased pAkt level with repressed gluconeogenic and lipogenic expression in HFD mouse livers (Supplementary Fig. 14B and C). The expression levels of lncSHGL and CaM and pAkt proteins were reduced in fasting mouse livers but were restored after refeeding (Supplementary Fig. 15A and C).

Insulin physiologically activates Akt through the insulin receptor/PI3K pathway to inactivate FOXO1 by phosphorylating and promoting its nuclear exclusion. Excessive FOXO1 activation due to insulin resistance or deficiency promotes hepatic gluconeogenesis and fasting hyperglycemia (36–38). Deletion of hepatic FOXO1 rescues hyperglycemia by repressing gluconeogenic gene expression in mice with liver-specific knockout of the insulin receptor (39). Moreover, FOXO1 also induces the expression in the liver of lipogenic genes such as FAS and PANDER (24,40). Transgenic or Ad overexpression of FOXO1 in the liver promoted lipid deposition (24,40). Under severe insulin resistance, inactivating hepatic FOXO1 via insulin-independent mechanism(s) holds great promise for the treatment of type 2 diabetes and fatty liver.

Gain- and loss-of-function studies revealed that lncSHGL suppresses hepatic gluconeogenesis and lipogenesis. In the previous studies, we reported that the ATP-P2 receptor signaling–mediated increase in cellular calcium level functionally activates CaM to induce Akt activation and suppress hepatic gluconeogenesis and lipogenesis in obese diabetic mice (22,23). These previous findings had established an insulin-independent but calcium-dependent activation of the CaM/Akt metabolic pathway. We also found that FAM3C activates HSF1 to directly induce CALM1 transcription and elevate the CaM protein level to activate the Akt pathway in a Ca²⁺-independent manner (26). The current study further revealed that lncSHGL modulates the CaM protein level at the posttranscriptional level to induce Akt activation. lncSHGL also inhibits mTORC1 and mTORC2 pathways, which promote hepatic lipid deposition (41). That lncSHGL activates Akt but represses mTOR pathways in hepatocytes is clear; thus far, however, the mechanism of lncSHGL’s inhibition effect on mTOR pathways in hepatocytes remains unclear. lncSHGL likely elevates CaM protein to repress pmTOR via a direct interaction. Similarly, FAM3C activates HSF1 to induce CALM1 transcription, elevating CaM protein to activate Akt independent of calcium. However, FAM3C also represses the mTORC1 pathway with unknown mechanism(s) in hepatocytes (26).

Although Akt activates mTOR in hepatocytes (42), our findings revealed a new regulatory network in which an increase in CaM protein activates the Akt pathway and represses mTOR pathways in hepatocytes (Fig. 8E). CaM regulates many physiological and pathophysiological processes by interacting with hundreds of target proteins (43). CaM is generally activated by an increase in cellular Ca²⁺ level (22,23); however, CaM also interacts with its targets via a Ca²⁺-independent manner in some conditions (43). The upstream signals determine the interaction between CaM and its target proteins in various conditions (43). CaM has been reported to interact and activate mTORC1 in a Ca²⁺-dependent manner in HEK293 cells (44). Our previous and current findings suggested that transcriptional and translational upregulation of the CaM protein level represses the mTOR pathway via a Ca²⁺-independent mechanism in hepatocytes (26). That CaM interacts with mTOR to repress or activate it likely depends on upstream and cellular Ca²⁺ signals in various cell types.

Insulin resistance also plays important roles in the development of fatty liver by promoting lipid transfer from adipose tissue to the liver (ectopic fat deposition) (45). Overall, the repression of FOXO1 and mTOR/SREBP-1C pathways and the improvement of global insulin resistance together contribute to the beneficial effects of lncSHGL on fatty liver. Moreover, a reduction in lncSHGL expression may exert a deleterious effect on liver IRI observed in our previous studies (7,8) by impairing Akt activity.

hnRNPA1 is an RNA-binding protein regulating the translation efficiency of mRNAs (46). Although we found that hnRNPA1 promotes Akt activation, Akt also phosphorylates hnRNPA1 (47), revealing a cross-regulation between hnRNPA1 and Akt activities. Several recent lines of evidence suggest that hnRNPA1 plays important roles in regulating glucose and lipid metabolism. FFAs stimulate the expression of hnRNPA1, which binds to SREBP-1a mRNA and increases its translation in HepG2 cells (48). hnRNPA1 also splices glycolytic enzyme pyruvate kinase precursor mRNA to form mature pyruvate kinase isoform 2 (PKM2) mRNA (49,50). In omental adipose tissue of morbidly obese patients, hnRNPA1 expression is decreased (51). Moreover, hnRNPA1 is also associated with insulin receptor gene splicing in adipose tissue of humans with body weight loss (52).

We previously showed that the CaM protein level is reduced in obese mouse livers due to transcription repression of CALM1 (26). The current study revealed that lncSHGL recruited hnRNPA1 to enhance the translation efficiency of CALM mRNAs without affecting their transcription in hepatocytes. Clearly, the inhibition at the transcriptional and translational levels of CALM1-3 mRNAs together contributes to the decreased hepatic CaM protein under the obese state (26). Beyond the transcriptional regulation of CALM genes by transcription factors such as HSF1 (26), modulating the translation efficiency of CALM mRNAs by the lncSHGL/hnRNPA1 axis is also important for maintaining CaM protein level, Akt activity, and glucose/lipid homeostasis in the basal condition when the insulin level is low in hepatocytes. Administration of CPZ, an inhibitor of CaM, had been reported to induce hyperglycemia with unclear mechanism(s) (53). Our previous (22,23,28) plus current findings suggested that the enhancement of hepatic gluconeogenesis induced by inhibiting the CaM/Akt pathway is a novel mechanism for explaining the hyperglycemic effect of CPZ. Our previous (26) and current findings revealed that an increase in CaM protein triggered by transcriptional or posttranscriptional mechanisms activates the PI3K/Akt pathway and represses the mTOR pathway in hepatocytes independent of calcium. These findings shed light on the Ca²⁺-independent action modes of CaM. Overall, the lncSHGL/hnRNPA1 axis plays important roles in suppressing hepatic gluconeogenesis and lipogenesis via the modulation of CaM protein levels at the posttranscriptional level.

Moreover, we identified human lncRNA B4GALT1-AS1 as the homologous sequence of mouse lncSHGL. Although B4GALT1-AS1 levels were reduced in steatotic human livers, we regret that we were not able to analyze the CaM protein level due to the unavailability of sufficient human liver samples. However, SREBP-1 and FAS expression were increased in steatotic human livers in the previous study (33). The lncSHGL/hnRNPA1/CaM pathway is also involved in regulating Akt activity and hepatic glucose production in physiological conditions such as fasting/refeeding. Repression of the lncSHGL/hnRNPA1/CaM axis for a short time is likely beneficial for increasing gluconeogenesis in fasting status; however, long-term inhibition triggers fasting hyperglycemia and steatosis under the obese state. Regarding the roles of lncSHGL in regulating hepatic glucose/lipid metabolism, several issues should be noted. That lncSHGL also recruits hnRNPA1 to regulate the translation efficiency of other mRNA(s) beyond CALM1-3 mRNAs is also possible. To further identify the target mRNAs of the lncSHGL/hnRNPA1 axis is of great significance. lncSHGL may also regulate hepatic glucose and lipid metabolism via other mechanism(s) beyond hnRNPA1/CaM pathway.

In summary, the new lncRNA lncSHGL recruits hnRNPA1 to enhance the translation efficiency of CALM mRNAs without affecting their transcription, elevating CaM protein level to suppress hepatic gluconeogenesis/lipogenesis independent of insulin and calcium (Fig. 8E). Activating lncSHGL/hnRNPA1 axis represents a potential strategy for the treatment of type 2 diabetes and steatosis.

Funding. This study was supported by grants from National Key Research Program of China (2016YFC1304803 and 2017YFC0909600), the Natural Science Foundation of China (81670748, 81471035, 81322011, 81670462, and 81422006), and Beijing Natural Science Foundation (7171006).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.W. and W.Y. wrote the manuscript. J.W., W.Y., and Z.C. researched data and contributed to discussion. J.W., W.Y., Q.C., and J.Y. designed the study and revised and edited the manuscript. J.C., Y.M., and B.F. provided the technical assistance and animal model preparation. L.S., L.D., and J.L. provided human liver samples and contributed to discussion. Q.C. and J.Y. are the guarantors of this work and had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.