New Target to Lower Glucose, Lipid Levels in Obesity

Identification of GULL, a metabolic disease–associated human lncRNA that has a functionally conserved equivalent in mice. Since most human lncRNAs are not conserved, confidently identifying those that play important roles in metabolic physiology and can serve as therapeutic targets remains a significant challenge. However, recent reports, including our own work, suggest that some lncRNAs across multiple species share conserved function despite their lack of sequence similarity, a phenomenon termed FCLs (8, 9). Leveraging this concept, we have developed a 2-step approach to select human lncRNAs genetically associated with metabolic diseases that also have functionally conserved counterparts in mice (Figure 1A).

Identification of h/mGULLs as a pair of functionally conserved lncRNAs closely related to metabolic disorders. (A) Flowchart for screening the functional conserved lncRNA candidates that are closely associated with metabolic disorders with human GWAS data and nanopore sequencing data (GSE224278). hGULL/mGULL was selected for further study. (B) Graphical representation of location and details of hGULL/mGULL on human or mouse chromosomes from Integrative Genomics Viewer and UCSC Genome Browser (human GRCh38/hg38 and mouse GRCm39/mm39). (C) Graphical representation of the rescue experiment. Mice were injected with 3 groups of adenoviruses: Ad-Ctrl (Lac+pAdV5), mGULL knockdown (Ad-shmGULL+pAdV5), or rescue group (Ad-shmGULL+Ad-hGULL). (D) Heatmap showing the top 600 significantly expressed genes in the mGULL knockdown and mGULL knockdown with hGULL overexpression rescue groups. (E) Gene Ontology (GO) enrichment analysis focused on the biological processes that can be rescued, summarized based on |normalized enrichment score| > 1.5, P < 0.01, and adjusted P value (FDR) < 0.1 from gene set enrichment analysis. The differentially expressed genes were defined as |log2(fold change)| > 0.5 and P < 0.05. (F) Quantitative reverse transcriptase PCR (qRT-PCR) analysis of mRNA levels of lipid synthesis genes (Elovl6, Acc1, Acly, Gck, Scd1, and Fasn), gluconeogenesis genes (G6pc and Pck1), and β-oxidation genes (Cpt1a and Mcad) in rescue assay. *Comparison between Ad-Ctrl and Ad-shmGULL groups; #comparison between Ad-shmGULL and Ad-shmGULL+hGULL groups. (G and H) Plasma/liver TG level, GTT, and ITT were determined in the rescue assay. *Comparison between Ad-Ctrl and Ad-shmGULL groups; #comparison between Ad-shmGULL and Ad-shmGULL+hGULL groups. (I) Immunoblot assay determined the protein levels of p-AKT, AKT, p-GSK3β, and GSK3β in the rescue assay with or without insulin treatment. Data are shown as mean ± SD, 1-way ANOVA, in F–H. *P < 0.05, **P < 0.01, ***P < 0.001; #P < 0.05, ##P < 0.01, ###P < 0.001.

First, we performed eQTL mapping to identify human lncRNAs whose expressions are regulated by single nucleotide polymorphisms (SNPs) associated with metabolic disorders. To do this, we extracted all SNPs associated with metabolic disorders from the NIH GWAS Catalog database (Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/JCI186355DS1), then used a novel annotation of human liver lncRNAs to identify those expressed in the liver (10). We then performed eQTL analysis using Genotype-Tissue Expression Project (GTEx) human liver gene expression data, identifying 287 lncRNAs whose expressions are affected by SNPs linked to metabolic diseases (Supplemental Table 2). Second, we cross-referenced these lncRNAs against a database of functionally conserved lncRNAs between humans and mice that we recently established (9). We selected these FCLs based on criteria of being syntenic, similarly regulated, and correlated to the same metabolic function. Ultimately, we identified 7 human lncRNAs that are genetically associated with metabolic diseases and have functionally conserved equivalents in mice (Figure 1A and Supplemental Table 3).

Careful definition of the pathophysiological function of a potential drug target gene is crucial to understand its disease relevance. For human protein-coding genes, this is often achieved by studying their homologs in mice or other animal models. However, as most human lncRNAs are not conserved, it remains challenging to experimentally define their physiological roles and mechanisms of action within a pathophysiological context because of the lack of suitable models. The 7 human lncRNAs we identified all have potential functional equivalents in mice, allowing their function to be studied in conventional mouse models. Among the 7 candidates, we selected the only pair of intergenic lncRNAs to streamline downstream functional analysis. We named them human and mouse GULLs (glucose and lipid lowering) on the basis of the function identified in this study (Figure 1A).

The newly discovered human and mouse GULLs (h/mGULLs) reside in the syntenic regions between the coding genes MLYCD/Mlycd and OSGIN1/Osgin1 in humans and mice, respectively (Figure 1B). Coding potential analysis indicates that both lncRNAs are indeed noncoding transcripts (data not shown). After successfully cloning and sequencing this pair of lncRNAs from both species, we performed in vitro translation assays to verify that they do not encode peptides (Supplemental Figure 1A).

As noted, we identified the human lncRNA, hGULL, through a GWAS-eQTL analysis, in which its locus was associated with HDL-C levels. To further assess its disease relevance, we analyzed HDL-C levels in individuals with high and low hGULL expression using public datasets. The results revealed significantly higher HDL-C levels in individuals with elevated hGULL expression, consistent with our eQTL findings and suggesting a potential linkage between hGULL and HDL-C (Supplemental Figure 1B). To further investigate the role of hGULL in metabolic disease pathogenesis, we examined its levels in patients with type 2 diabetes (T2D) and hypertension. Results showed that the hGULL was significantly decreased in the livers of patients with T2D and hypertension compared with healthy individuals (Supplemental Figure 1, C and D). Furthermore, hGULL expression was markedly reduced in livers of individuals with fatty liver, indicating a negative correlation between hGULL expression and fatty liver disease (Supplemental Figure 1E). We also investigated the expression levels of mGULL in the mouse livers under physiological and pathological conditions. Our results demonstrated that mGULL RNA levels were significantly upregulated during refeeding compared with fasting (Supplemental Figure 1F). Moreover, mGULL expression was markedly reduced in mice fed a high-fat diet compared with wild-type controls (Supplemental Figure 1F). These findings highlight a potential connection between GULL and metabolic diseases.

Another key criterion for selecting functionally conserved lncRNAs in humans and mice was their similar regulation by the same metabolic stimuli (9). We initially identified hGULL/mGULL because RNA-Seq showed that the expression levels of both human and mouse GULL in the chimeric human-mouse livers of humanized mice were increased by treatment with the FXR agonist GW4064; this result was verified by real-time PCR (Supplemental Figure 1G). FXR is a bile acid receptor known to regulate glucose and lipid metabolism, and FXR agonists are currently under clinical trials for metabolic dysfunction–associated steatohepatitis and fibrosis (11). To further understand whether FXR directly regulates the expression level of mGULL, we performed a chromatin immunoprecipitation (ChIP) assay and found that the occupancy of FXR and its coreceptor, RXR, at the promoter of mGULL in mouse liver was significantly increased by GW4064 treatment (Supplemental Figure 1H), indicating that FXR directly activates mGULL transcription in mouse liver.

Given the identification of hGULL and mGULL as potential functionally conserved lncRNAs (FCLs), we conducted a rescue experiment in mice to examine the functional similarities between them. We used shRNAs delivered by adenovirus to knock down mGULL in the mouse liver and subsequently coexpressed hGULL to evaluate its ability to rescue the loss-of-function effect of mGULL (Figure 1C and Supplemental Figure 1I). Since lncRNAs are known to regulate gene expression, we performed RNA-Seq on liver tissues and used global gene expression levels as an indicator to assess the functional similarities between hGULL and mGULL. This experiment revealed that a substantial number of genes whose expression levels were regulated by mGULL knockdown exhibited reversed expression patterns in the rescue group with concurrent expression of hGULL, often returning to control group levels (Figure 1D). Additionally, Gene Ontology enrichment analysis of genes regulated by hGULL/mGULL showed significant enrichment in genes involved in the fatty acid metabolic process (Figure 1E).

To validate the RNA-Seq results and further explore the role of GULL in lipid metabolism, we examined the expression levels of a panel of genes in lipid and glucose metabolic pathways in mGULL-knockdown and rescue mice. We found that the expression levels of lipogenic genes, such as Elovl6, Acc1, Acly, Gck, Scd1, and Fasn, and gluconeogenic genes, such as G6pc and Pck1, were significantly increased in the livers of mGULL-knockdown mice and were reversed by concurrent expression of hGULL (Figure 1F). Meanwhile, the expression of β-oxidation genes showed the opposite pattern (Figure 1F). To understand whether the altered gene expression led to changes in metabolic response, we examined how hGULL/mGULL affected lipid and glucose metabolism. We found that knockdown of mGULL increased the levels of triglycerides (TGs) in the plasma and liver, and the effects were completely reversed by concurrent expression of hGULL (Figure 1G). Similarly, mGULL knockdown resulted in impaired glucose metabolism as shown by a glucose tolerance test (GTT) and insulin tolerance test (ITT), and the effect was rescued by concurrent hGULL expression (Figure 1H). The impaired glucose metabolism in mGULL-knockdown mice suggests that they might have reduced insulin sensitivity. Indeed, an acute insulin stimulation experiment showed that phosphorylation of Akt and Gsk3β — 2 key proteins in the insulin receptor signaling pathway — was dampened in mGULL-knockdown mice and recovered by concurrent hGULL expression (Figure 1I). Taken together, our data support that human and mouse GULLs are legitimate FCLs and have similar lipid- and glucose-lowering effects in mice.

Both human and mouse GULLs bind to CRTC2 to regulate lipid and glucose metabolism. LncRNAs often carry out their biological functions via interaction with specific binding proteins. To identify potential protein-binding partners of hGULL and mGULL, we used biotinylated versions of these RNAs to perform RNA pull-down assays coupled with proteomic analysis, using chimeric human-mouse liver tissues from GW4064-treated humanized mice. We identified 2 proteins, CREB-regulated transcription coactivator 2 (CRTC2) and argininosuccinate synthase 1 (ASS1), that bind to both hGULL and mGULL (Supplemental Table 4). As CRTC2 is a well-known regulator of lipid and glucose metabolism (12), we focused on its role in the function of hGULL/mGULL. Immunoblot assay verified that the sense, but not antisense, of hGULL and mGULL specifically binds to CRTC2 (Figure 2A). In addition, we performed 2 sets of RNA immunoprecipitation (RIP) assays to examine hGULL/mGULL and CRTC2 interaction. First, we immunoprecipitated endogenous CRTC2 in human and mouse primary hepatocytes using CRTC2 antibody. Second, we overexpressed FLAG-tagged CRTC2 or Crtc2 in human and mouse primary hepatocytes, respectively, and then used anti-FLAG antibody to isolate these proteins. Both experiments showed that CRTC2 strongly binds to hGULL and mGULL (Figure 2B), supporting its role in the biological functions of hGULL and mGULL. CRTC2, functioning as a CREB coactivator, plays a crucial role in regulating gluconeogenesis in the liver. Studies show that prolonged activation of CRTC2 during insulin resistance contributes significantly to hyperglycemia (13–15). To determine whether hGULL and mGULL regulate CRTC2 function, we overexpressed Crtc2 to mimic obesity condition (16) and subsequently expressed hGULL or mGULL to examine their functional interaction (Figure 2C). Overexpression of Crtc2 resulted in significantly increased liver size and liver/body weight ratio, indicators of fatty liver. This phenotype was reversed by overexpression of either hGULL or mGULL, suggesting that GULL might inhibit Crtc2 function (Figure 2D). Consistently, overexpression of Crtc2 significantly increased neutral lipid levels in liver tissue and hepatic and plasma TG levels, all of which were reversed by concurrent expression of hGULL or mGULL (Figure 2E). Furthermore, hGULL and mGULL expression rescued impaired glucose metabolism induced by Crtc2 expression, as shown in glucose and insulin tolerance tests (Figure 2F). Crtc2 overexpression also increased the expression levels of lipogenic and gluconeogenic genes, an effect reversed by hGULL or mGULL expression, while β-oxidation gene expression showed the opposite pattern (Figure 2G). Additionally, hGULL and mGULL expression rescued the reduced insulin-stimulated phosphorylation of AKT and GSK3β caused by Crtc2 expression (Figure 3A), further supporting that hGULL and mGULL inhibit CRTC2 function.

Beneficial effects of hGULL/mGULL on fatty liver are CRTC2 dependent. (A) RNA pull-down coupled with immunoblot assay to verify the associations of hGULL/mGULL with CRTC2 in humanized mouse tissues. The antisense and beads-only groups served as negative control. (B) RNA immunoprecipitation (RIP) assays were performed with endogenous CRTC2 antibody and FLAG antibodies in human primary hepatocyte (PH) or mouse PH cells, and the coprecipitated RNA was subjected to qRT-PCR with the primers of h/mGULL. ***P < 0.001; data shown as mean ± SD, 2-tailed unpaired Student’s t test. (C) Graphical representation of mouse model to explore the phenotype and molecular mechanisms between CRTC2 and h/mGULL. Mice were injected with 4 groups of adenoviruses: the Ad-Ctrl group (Lac+pAdV5), Crtc2 overexpression group (Lac+Ad-Crtc2), Crtc2 overexpression with hGULL knockdown group (Ad-Crtc2+hGULL), and Crtc2 overexpression with mGULL knockdown group (Ad-Crtc2+mGULL). (D) Representative images of liver and liver/body weight ratio analysis in the 4 groups, including the Ad-Ctrl, Ad-Crtc2 overexpression, and AdCrtc2+hGULL/mGULL overexpression groups. *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA. (E) Representative images of Oil Red O and H&E staining and plasma/liver TG level analysis of the Ad-Ctrl, Ad-Crtc2 overexpression, and AdCrtc2+hGULL/mGULL overexpression groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. Scale bar: 50 μm. (F) GTT and ITT were determined in the Ad-Ctrl, Ad-Crtc2 overexpression, and Ad-Crtc2+hGULL or mGULL overexpression groups. *Comparison between Ad-Ctrl and Ad-Crtc2 groups; *P < 0.05, **P < 0.01. #Comparison between Ad-Crtc2 and AdCrtc2+hGULL or mGULL groups; #P < 0.05, ##P < 0.01. Data shown as mean ± SD, 1-way ANOVA. (G) The mRNA levels of lipid synthesis genes (Elovl6, Acc1, Acly, Gck, Scd1, and Fasn), gluconeogenesis genes (G6pc and Pck1), and β-oxidation genes (Cpt1a and Mcad) were quantified in Ad-Ctrl, Ad-Crtc2 overexpression, and AdCrtc2+hGULL or mGULL overexpression groups using quantitative PCR (qPCR). *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA.

hGULL/mGULL bind to CRTC2 and inhibit its activity in regulating gluconeogenesis and lipogenesis. (A) Immunoblotting of p-AKT, AKT, p-GSK3β, and GSK3β in the Ad-Ctrl, Ad-Crtc2 overexpression, and Ad-Crtc2+hGULL or mGULL overexpression groups with or without insulin treatment. (B) Graphical representation of mouse model to explore the phenotype and molecular mechanisms of h/mGULL. Mice were injected with 3 groups of adenoviruses: the Ad-Ctrl group (Lac+pAdV5), hGULL overexpression group (Ad-hGULL), or mGULL overexpression group (Ad-mGULL). (C) Coimmunoprecipitation to detect the association between CRTC2 and CREB after hGULL/mGULL overexpression. Mice were injected with Ctrl (pAdV5) and hGULL or mGULL adenoviruses and treated with 6 hours of fasting before sacrifice. (D) ChIP-qPCR detection of occupancy of CREB or CRTC2 and RNA polymerase II level on G6pc and Pck1 promoter in Ad-Ctrl, Ad-Crtc2 overexpression, and Ad-Crtc2+hGULL or mGULL groups. IgG was used as negative control. **P < 0.01; data shown as mean ± SD, 2-way ANOVA. (E) ChIP-qPCR determined SREBP1 occupancy on Fasn, Acly, or Elovl6 promoter in the Ad-Ctrl, Ad-Crtc2 overexpression, and Ad-Crtc2+hGULL or mGULL groups. IgG was used as negative control. **P < 0.01; data shown as mean ± SD, 2-way ANOVA. (F) Graphical representation of working model of hGULL/mGULL-mediated glucose or lipid lowering. A higher RNA level of hGULL/mGULL specifically binds with CRTC2 and holds it in the cytoplasm, decreasing the occupancy of CREB or SREBP1 at the promoter of gluconeogenesis and lipogenesis genes.

To understand the molecular detail of the interaction between hGULL/mGULL and CRTC2, we generated 3 deletion mutants of CRTC2 (Supplemental Figure 2A) to map the key domains required for GULL binding. Results showed that deletion of either the CREB binding domain (CBD) or regulatory domain (REG) of CRTC2 impaired its interaction with hGULL/mGULL. The affinity of hGULL and mGULL toward each domain was similar (Supplemental Figure 2A), suggesting that they might share binding sites on CRTC2. A binding competition assay showed that hGULL and mGULL could effectively compete with each other for interaction with CRTC2 (Supplemental Figure 2B), supporting this claim.

It is well established that CRTC2 acts as a transcription coactivator of CREB to regulate the transcription of gluconeogenic genes during fasting and glycemic control in both physiological and pathological conditions (17, 18). As the CBD of CRTC2 mediates its interaction with CREB (19, 20), we examined whether hGULL/mGULL affects the CREB-CRTC2 interaction. Overexpression of hGULL or mGULL significantly decreased the CREB-CRTC2 interaction in mouse livers (Figure 3, B and C). Consistent with these results, a ChIP assay revealed that overexpression of Crtc2 increased the occupancy of Crtc2 and RNA polymerase II at the promoter regions of gluconeogenesis genes (G6pc and Pck1), and this occupancy was decreased by hGULL or mGULL overexpression (Figure 3D). Crtc2 is also known to coordinate the actions of downstream effectors such as Sec23A and Sec31A to regulate the processing of SREBP1 maturation and nuclear translocation (21). We performed an SREBP1 ChIP assay and found that hGULL or mGULL overexpression significantly decreased SREBP1 occupancy at the promoter regions of lipogenic genes such as Fasn, Acly, and Elovl6 (Figure 3E). Taken together, these data demonstrate that hGULL and mGULL exhibit potent lipid- and glucose-lowering effects by specific binding to CRTC2 and modulation of the function of its downstream effectors (Figure 3F).

Human and mouse GULLs ameliorate obesity-induced metabolic abnormalities in a CRTC2-dependent manner. The robust lipid- and glucose-lowering effects of hGULL/mGULL suggest that they might be able to improve defective lipid and glucose metabolism in obese mice. To investigate this, we overexpressed hGULL or mGULL in the livers of mice with diet-induced obesity (Figure 4A) and found that both resulted in significant decrease in liver size and the liver/body weight ratio (Figure 4B). Overexpression of hGULL or mGULL also reduced neutral lipid levels in liver tissue and decreased liver and plasma TG levels (Figure 4, C and D). Moreover, GTT and ITT showed significantly improved glucose disposal rates in hGULL- or mGULL-overexpressing mice (Figure 4, E and F). Consistent with these results, the expression levels of lipogenic genes and gluconeogenesis genes were all decreased by hGULL/mGULL overexpression, while β-oxidation gene expression showed the opposite pattern (Figure 4, G and H). Remarkably, all these beneficial effects mediated by hGULL/mGULL overexpression were significantly attenuated by Crtc2 overexpression (Figure 4, B–H). These results not only reinforce our earlier findings that hGULL/mGULL are FCLs that modulate the function of CRTC2 but also indicate potential therapeutic benefits against obesity-induced metabolic abnormalities.

hGULL/mGULL ameliorate obesity-induced metabolic abnormalities in a CRTC2-dependent manner. (A) Graphical representation of obesity mouse model to explore the phenotype and molecular mechanisms of hGULL/mGULL and CRTC2. Mice were injected with 5 groups of adenoviruses: the Ad-Ctrl group (pAdV5), hGULL overexpression group, hGULL+Crtc2 overexpression group, mGULL overexpression group, or mGULL+Crtc2 overexpression group. (B–D) Representative images of liver and liver/body weight ratio analysis (B), Oil Red O and H&E staining (C), and plasma/liver TG level analysis (D) in Ad-Ctrl group, hGULL or mGULL overexpression group, and hGULL+Crtc2 or mGULL+Crtc2 overexpression group. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. Scale bar: 50 μm. (E and F) GTT and ITT were determined in the Ad-Ctrl group, hGULL or mGULL overexpression group, and hGULL+Crtc2 or mGULL+Crtc2 overexpression group. **P < 0.01, *P < 0.05, ##P < 0.01, #P < 0.05; data shown as mean ± SD, 1-way ANOVA. (G and H) The mRNA levels of lipid synthesis genes (Elovl6, Acc1, Acly, Gck, Scd1, and Fasn), gluconeogenesis genes (G6pc and Pck1), and β-oxidation genes (Cpt1a and Mcad) were quantified in Ad-Ctrl group, hGULL or mGULL overexpression group, and hGULL+Crtc2 or mGULL+Crtc2 overexpression group using qPCR. *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA.

Long-term expression of human and mouse GULLs by adeno-associated virus robustly ameliorates obesity-induced metabolic abnormalities. Adenovirus-mediated gene expression in the liver is typically short-lived and often triggers strong inflammatory responses (22). To examine the long-term effects of hGULL/mGULL while minimizing potential immune response effects, we used adeno-associated virus (AAV) to express hGULL/mGULL in mice with high-fat diet–induced obesity (Figure 5A). After 3 months, mice receiving hGULL or mGULL AAVs (AAV-hGULL/mGULL) exhibited a significant decrease in body weight compared with the control group (Figure 5B). The liver size and liver/body weight ratio also decreased in the AAV-hGULL/mGULL group (Figure 5C). Consistent with the adenovirus-mediated overexpression results, neutral lipid levels in liver tissue and TG levels in liver and plasma were significantly decreased in the AAV-hGULL/mGULL group (Figure 5D). Furthermore, glucose disposal in GTT and ITT were significantly improved in the AAV-hGULL/mGULL group (Figure 5E). Fasting insulin levels in plasma and the homeostatic model assessment of insulin resistance (HOMA-IR) index were also decreased in the AAV-hGULL/mGULL group (Figure 5F). To assess the long-term effects of hGULL/mGULL on blood glucose levels, we measured hemoglobin A1c levels, which were significantly decreased in the AAV-hGULL/mGULL group (Figure 5G). Additionally, the expression levels of lipogenic genes and gluconeogenesis genes were significantly decreased, whereas those of β-oxidation genes were significantly increased, in AAV-hGULL/mGULL mice (Figure 5H). Collectively, these results demonstrate that long-term hGULL or mGULL expression substantially ameliorates fatty liver and improves insulin sensitivity and glucose metabolism in obese mice.

AAV-mediated hGULL/mGULL overexpression improves metabolic health of obese mice. (A) Graphical representation of AAV-obesity mouse model to explore the phenotype of hGULL/mGULL. Mice were injected with 3 groups of AAVs: AAV-Ctrl, AAV-hGULL, and AAV-mGULL. (B) Statistical analysis of body weight in the AAV-Ctrl, AAV-hGULL, and AAV-mGULL groups. ***P < 0.001. (C and D) Representative images of liver and liver/body weight ratio analysis (C), Oil Red O and H&E staining, and plasma/liver TG level analysis (D) in the AAV-Ctrl, AAV-hGULL, and AAV-mGULL groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. Scale bar: 50 μm. (E) GTT and ITT were determined in the AAV-Ctrl, AAV-hGULL, and AAV-mGULL groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. (F and G) Plasma insulin level, HOMA-IR (F), and hemoglobin A1c level (G) were analyzed in the AAV-Ctrl, AAV-hGULL, and AAV-mGULL groups. *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA. (H) The mRNA levels of lipid synthesis, gluconeogenesis, and β-oxidation genes were quantified in the AAV-Ctrl, AAV-hGULL, and AAV-mGULL groups using qPCR. *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA.

Functional motifs of GULLs, GULFs, strongly improve metabolic health in obese mice. The challenge of developing effective lncRNA drugs is partly rooted in the sequence divergence of lncRNAs among species, which impedes the identification of key functional elements on human lncRNAs that can serve as direct targets of therapy development. As hGULL and mGULL carry out similar functions via their specific binding to CRTC2 in humans and mice, their respective CRTC2 binding motifs might be essential to their function and can serve as potential therapeutic targets. To identify the core motifs of hGULL/mGULL essential for the CRTC2 interactions, we conducted an in vitro biotin-RNA-protein binding coupled with dot blot assay to scan hGULL/mGULL for their specific binding sites to CRTC2. We identified that CRTC2 interacted with hGULL at 136–166 nt (dot 6, A6: 5′-UGUGGCAUGAAGAGGUCAGGCCAUUCCAGC-3′) and with mGULL at 1,349–1,379 nt (dot 46, F6: 5′-UUGCCAAACCUUUCUUGGUGAGCUGGAUGC-3′) (Figure 6A). We named these motifs GULFs (GULLs’ functional motifs). To determine the significance of human and mouse GULFs to the interaction between GULL and CRTC2, we performed competitive binding assays and found that mGULF was able to effectively compete with full-length hGULL in its binding to CRTC2 in a dose-dependent manner (Figure 6B). Similarly, hGULF was also able to compete with full-length mGULL (Figure 6B). These findings indicate that hGULF or mGULF specifically mediates the interaction between hGULL/mGULL and CRTC2. Furthermore, the predicted secondary structures of hGULF and mGULF are similar, providing a possible reason for their interaction with CRTC2 (Supplemental Figure 2C).

hGULF/mGULF improve metabolic health of obese mice. (A) RNA dot blot assays were performed to identify the core binding motif between CRTC2 and hGULL/mGULL. (B) hGULL/mGULF competition assay coupled with RNA pull-down and visualized by immunoblot assay. Non-biotin-labeled hGULF or mGULF with different doses was used to compete with biotin-labeled hGULL or mGULL (mole ratio of hGULL/mGULL to hGULF/mGULF: 1:1, 1:10, 1:20, 1:30, and 1:40) for binding to the CRTC2. (C) Graphical illustration shows the treatment schedules of tri-GalNAc–tagged hGULF/mGULF mimics and NC dot oligos in the high-fat diet (HFD) mouse model. Mice were injected with tri-GalNAc-NC oligos, tri-GalNAc-hGULF oligos, and tri-GalNAc-mGULF oligos every 2 days for 2 weeks. (D and E) Representative images of the liver and liver/body weight ratio analysis (D), Oil Red O and H&E staining, and plasma/liver TG level analysis (E) in tri-GalNAc-NC, tri-GalNAc-hGULF, and tri-GalNAc-mGULF groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. Scale bar: 50 μm. (F) GTT and ITT were determined in the tri-GalNAc-NC, tri-GalNAc-hGULF, and tri-GalNAc-mGULF groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. (G and H) Plasma insulin level (G) and HOMA-IR (H) were determined in tri-GalNAc-NC, tri-GalNAc-hGULF, and tri-GalNAc-mGULF groups. **P < 0.01; data shown as mean ± SD, 1-way ANOVA. (I) The mRNA levels of lipid synthesis, gluconeogenesis, and β-oxidation genes were quantified in the tri-GalNAc-NC, tri-GalNAc-hGULF, and tri-GalNAc-mGULF groups with qPCR. *P < 0.05, **P < 0.01; data shown as mean ± SD, 1-way ANOVA.