A novel long non-coding RNA connects obesity to impaired adipocyte function

Abstract

Background: Long non-coding RNAs (lncRNAs) can perform tasks of key relevance in fat cells, contributing, when defective, to the burden of obesity and its sequelae. Here, scrutiny of adipose tissue transcriptomes before and after bariatric surgery (GSE53378) granted identification of 496 lncRNAs linked to the obese phenotype. Only expression of linc-GALNTL6-4 displayed an average recovery over 2-fold and FDR-adjusted p-value <0.0001 after weight loss. The aim of the present study was to investigate the impact on adipocyte function and potential clinical value of impaired adipose linc-GALNTL6-4 in obese subjects.

Methods: We employed transcriptomic analysis of public dataset GSE199063, and cross validations in two large transversal cohorts to report evidence of a previously unknown association of adipose linc-GALNTL6-4 with obesity. We then performed functional analyses in human adipocyte cultures, genome-wide transcriptomics, and untargeted lipidomics in cell models of loss and gain of function to explore the molecular implications of its associations with obesity and weight loss.

Results: The expression of linc-GALNTL6-4 in human adipose tissue is adipocyte-specific and co-segregates with obesity, being normalized upon weight loss. This co-segregation is demonstrated in two longitudinal weight loss studies and two cross-sectional samples. While compromised expression of linc-GALNTL6-4 in obese subjects is primarily due to the inflammatory component in the context of obesity, adipogenesis requires the transcriptional upregulation of linc-GALNTL6-4, the expression of which reaches an apex in terminally differentiated adipocytes. Functionally, we demonstrated that the knockdown of linc-GALNTL6-4 impairs adipogenesis, induces alterations in the lipidome, and leads to the downregulation of genes related to cell cycle, while propelling in adipocytes inflammation, impaired fatty acid metabolism, and altered gene expression patterns, including that of apolipoprotein C1 (APOC1). Conversely, the genetic gain of linc-GALNTL6-4 ameliorated differentiation and adipocyte phenotype, putatively by constraining APOC1, also contributing to the metabolism of triglycerides in adipose.

Conclusions: Current data unveil the unforeseen connection of adipocyte-specific linc-GALNTL6-4 as a modulator of lipid homeostasis challenged by excessive body weight and meta-inflammation.

Keywords: Adipocytes; Adipose tissue; Obesity; Triglycerides; linc-GALNTL6-4.

Graphical abstract

Figure 1

linc-GALNTL6-4 in adipose tissue is opposite to obesity. (A) Volcano plot of upregulated (red) and downregulated (green) lncRNAs in SC adipose tissue after weight loss (n = 16), as per inclusion criteria noted in Methods, and listed in (B). Dynamic adaptations affecting adipose linc-GALNTL6-4 after weight loss were confirmed by means of (C) real time PCR (n = 23), and in (D) an independent dataset of 50 participants (microarray) [24]. linc-GALNTL6-4 levels are provided as box plots of 75th to 25th percentiles with the median, and whiskers at maximum and minimum values. (E) Mean and S.D. for linc-GALNTL6-4 values in SC (for obese and non-obese (BMI<30 kg/m²) subjects, n = 73 and 136, respectively) and OM (n = 59 and 111) fat samples from an independent cohort of 212 participants, and in (F) association with BMI. ∗∗p < 0.001 for comparisons between groups of subjects segregated according to their BMI, and ^$p < 0.05 and ^$$p < 0.001 for comparisons of OM vs SC. Coloured numbers depict Spearman's rank (r) correlations for men (green and blue) and women (red and purple), and ∗∗p < 0.001. Continuous and staggered black lines show regression and 95% confidence interval. Plots (G) and (H) depict linc-GALNTL6-4 values in the METSIM study, a population-based study including bulk RNA-seq in the SC fat of 335 Finnish men, in association with BMI and fat mass, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Figure 2

linc-GALNTL6-4 is adipocyte-specific. (A) Heatmap of pairwise Spearman correlation coefficients for linc-GALNTL6-4 (ENSG00000250266) and reference genes for different adipose tissue resident cell populations (see also in Fig. S1B). The lack of values for mesothelial cells (MesoC) and neutrophils (NP) is due to the low presence of these cell types in SC fat, as explained in reference [28]. (B) Digestion of adipose tissue and the study of disaggregated adipose-derived stromal vascular cells (SVC) and mature adipocytes (MA) further confirmed (C) the prevalent expression of linc-GALNTL6-4 in the latter. Single-fat cell RNA-seq measures (RPKM) of linc-GALNTL6-4 in the dataset with ascension number GSE135776 [30] also confirmed (D) its most prominent expression in MA and (E) the higher amounts observed in OM adipocytes, when compared to SC adipocytes. (F) Increased linc-GALNTL6-4 expression was observed in cultures of preadipocytes (PA) while differentiating into lipid-containing MA. DM-2 (first) and AM-1 (second week) stand for differentiation and adipocyte media, respectively (see in Methods). In an independent assay (g), expression of linc-GALNTL6-4 was measured in differentiating adipocytes and in MA when challenged with a macrophage LPS-conditioned media (MCM). (H) RNA Fluorescence In Situ Hybridization (FISH) revealed increased cytoplasmic and nuclear linc-GALNTL6-4 in human PA while differentiating towards MA, (I) as further confirmed by real time PCR measures assessed in the nucleus and cytoplasm of MA. The scale bars denote 100 μm length in representative 20x immunofluorescent images. Plots show mean and S.E.M. Dots show results for each biological replicate (wells of the same 12-well plate). Statistical significance was assessed by ANOVA (post-hoc Bonferroni's multiple comparisons test) to assess the significance of dynamic changes in linc-GALNTL6-4 levels during adipogenesis, and two-tailed Student t-tests for comparisons treated versus control adipocytes. ns, not significant; ∗ and ^#p < 0.05, ∗∗ and ^##p < 0.001.

Figure 3

linc-GALNTL6-4 depletion alters adipocyte function. (A) Decreased linc-GALNTL6-4 signal (FISH) in adipocytes challenged with sh-RNA lentiviral particles drove impaired adipogenesis, as depicted by decreased (B) lipid droplets content and (C) expression of adipogenic markers (mean and S.E.M.; ∗p < 0.05 and ∗∗p < 0.001). The scale bars denote 100 μm length in representative 20x immunofluorescent images. (D) Schematic representation of our experimental approach to evaluate the loss of linc-GALNTL6-4 (LoF) in MA. (E) Principal component analysis (PCA), (F) hierarchical clustering, and (G) the heatmap of DE genes highlighted a great consistency within groups. (H) Gene Set Enrichment Analysis (GSEA) of Human Molecular Signatures Database (MSigDB) categories altered in our genetic LoF model. In an independent experiment (I), the knockdown of linc-GALNTL6-4 modified the (J) lipid landscape of MA, with major effects on a number of (K) SM, PS and PC species. Plots show mean and S.E.M. Dots show results for each biological replicate (wells of the same 12-well plate). Statistical significance was assessed by two-tailed Student t-test. ∗p < 0.05, ∗∗p < 0.001.

Figure 4

Increased linc-GALNTL6-4 boosts adipocyte phenotype. (A) Enhanced linc-GALNTL6-4 signal (FISH) resumes improved adipogenesis, as shown by (B) the amount of lipid droplets and (C) expression of adipogenic markers in differentiated human PA (mean and S.E.M.; ∗p < 0.05, ∗∗p < 0.001). The scale bars denote 100 μm length in representative 20x immunofluorescent images. (D) Schematic representation of the protocol used to evaluate the impact of linc-GALNTL6-4 (GoF) in human MA. (E) PCA and (F) hierarchical clustering indicated poor sample aggregation in treatment and control, but a hub of 29 transcripts depicted opposite changes in our models of GoF and LoF, as pointed in (G). Forest plots show overall suppression or activation of signalling pathways in response to linc-GALNTL6-4 knockdown (LoF) and overexpression (GoF), as disclosed by integrative analysis of our transcriptomic data using (H) C2 and (I) Hallmark gene set collections of the Molecular Signature Database (MSigDB), and (J) the Kyoto Encyclopaedia of Genes and Genomes (KEGG). LoF vs control is inked in blue, and red indicates comparisons in GoF vs control. Circles and triangles show non-significant (adj. p-value>0.05) and significant results, respectively. The size of each symbol stands for the -Log₁₀ adj. p-value. Also using quantitative lipidomics (K), an opposite effect in the GoF model was revealed with regard to the LoF (L), including significant impacts on a number of key lipid species, as shown in (M). Plots depict mean and S.E.M. Dots show results for each biological replicate (wells of the same 12-well plate). Statistical significance was assessed by two-tailed Student t-test. ∗p < 0.05, ∗∗p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Figure 5

linc-GALNTL6-4 may restrain triglyceride metabolism in adipocytes. (A) General pipeline for Weighted Gene Correlation Network Analysis (WGCNA). (B) Linear regression analysis between RNA-seq APOC1 and linc-GALNTL6-4 RPKM values in SC and OM adipocytes [30]. (C) Heatmap of correlations between modules and linc-GALNTL6-4 expression levels (0 vs. >1 RPKM). Degrees of associations are indicted by colours, positive correlation, red; negative correlation, blue; no correlation, white. ∗∗∗p < 0.001. Example of this correlation is provided for the “brown” module in (D), which included APOC1 and a hub of genes connected to this effector and linc-GALNTL6-4, as shown in (E). Correlation coefficients and p-values between Module Membership (MM) and Gene Significance (GS) for all genes within each module were calculated using a linear regression model. The x-axis in (D) denotes the correlation between genes and modules, and the y-axis indicates the correlation between genes and linc-GALNTL6-4 expression levels. The size and colour intensity of each gene in (E) represent the number of interactions within the node. Forest plots show the hallmarks and pathways mostly characterized in human adipocytes for highly positive correlated genes with our selected lncRNA within the “brown” module, as interpreted by using bioinformatics resource of (F) GO, (G) KEGG, and (H) REACTOME. (I) Additional evaluation of our genetic models of LoF and GoF further validated the impact on APOC1 gene expression and protein levels, as measured by enzyme-linked immunoassays applied to the media, mirroring to some extent the triglycerides (TG) released by adipocytes when linc-GALNTL6-4 was (J) knocked-down (LoF) or (K) overexpressed (GoF). Data is presented as mean ± S.E.M. (n > 4 biological replicates/group); ∗p < 0.05, ∗∗p < 0.001. Scatter dot plots show (L) SC and OM APOC1 gene expressions opposite to linc-GALNTL6-4, and (M) directly associated with circulating TG, independently of sex, age and weight, as indicated by Spearman's rank (r)-order correlation tests. ∗p < 0.05, ∗∗p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)