Aberrant overexpression of HOTAIR inhibits abdominal adipogenesis through remodelling of genome-wide DNA methylation and transcription

Abstract

Objective

Abdominal adiposity is strongly associated with diabetic and cardiovascular comorbidities. The long noncoding RNA HOTAIR (HOX Transcript Antisense Intergenic RNA) is an important epigenetic regulator with fat depot-specific expression. Its functional roles and epigenetic regulation in abdominal adipogenesis remain uncertain.

Methods

We collected different fat depots from healthy, severely obese, and uraemic subjects to measure fat-depot specific gene expression and quantify regional adiposity via dual-energy X-ray absorptiometry (DXA). HOTAIR was overexpressed to evaluate its functional roles. Reduced representation bisulfite sequencing (RRBS), RNA-sequencing, real-time qPCR and RNA/chromatin immunoprecipitation were performed to analyse HOTAIR-mediated epigenetic regulation.

Results

A negative correlation between adipose tissue HOTAIR expression (arm or abdominal subcutaneous fat depots) and regional adiposity under the status of severe obesity or uraemia was observed. HOTAIR overexpression using human immortalized abdominal preadipocytes further revealed its anti-adipogenic effects. Integrative analysis of genome-wide DNA methylation by reduced representation bisulfite sequencing (RRBS) and gene expression was performed. Overall, the differentially methylated genes were functionally enriched for nervous system development, suggesting that HOTAIR may be epigenetically associated with cell lineage commitment. We specifically found that HOTAIR-mediated genes showed strong changes in both DNA methylation and gene expression during abdominal adipogenesis. We observed that two HOTAIR-repressed genes, SLITRK4 and PITPNC1, present an obesity-driven fat-depot specific expression pattern that is positively correlated with the central body fat distribution.

Conclusions

Our study indicated that HOTAIR is a key regulator of abdominal adipogenesis via intricate DNA methylation and is likely to be associated with the transcriptional regulation of genes involved in nervous system development and lipid metabolism, such as SLITRK4 and PITPNC1.

Highlights

• •HOTAIR was lowly expressed in abdominal and arm fats compared to the gluteal fat.
• •Fat-depot-specific HOTAIR expression could be altered in the obese or uraemic status.
• •HOTAIR overexpression suppressed abdominal adipogenesis and modulated methylome.
• •HOTAIR-suppressed genes were associated with neural development and lipid metabolism.

1. Introduction

Obesity significantly increases the risks for cardiometabolic disorders, such as myocardial infarction and type 2 diabetes [1,2]. In addition to general adiposity and regional body fat distribution, central fat accumulation is an even stronger risk factor than obesity per se for triggering cardiovascular diseases [3,4]. Studies have revealed that upper-body fat and lower-body fat have different biological functions [5,6]. Lower-body fat, such as gluteofemoral subcutaneous adipose tissue (SAT), functions as a stable metabolic reservoir, which could maintain long-term fatty acid storage and expand healthily via predominant adipocyte hyperplasia. Upper-body fat, including abdominal SAT and visceral adipose tissue (VAT), presents dynamically quick uptake and rapid release of free fatty acids and tends to expand its storage capacity via adipocyte hypertrophy [5]. Hypertrophic adipocytes frequently induce a hypoxic microenvironment, increase the recruitment of macrophages, and subsequently induce hepatic and systemic insulin resistance with frequent lipid overspill to initiate the harmful atherosclerotic cascade [7]. Therefore, investigating the determinants for the regulation of regional adipocyte biology, such as abdominal adipogenesis, will help to identify novel therapeutic targets for ameliorating obesity-associated cardiovascular risks.

Recent research has shown that the functional discrepancy between upper-body and lower-body fat might stem from their distinct developmental origins with differential expression on several homeobox (HOX) genes [6,[8], [9], [10]]. In particular, HOTAIR (HOX Transcript Antisense Intergenic RNA), as a long noncoding RNA, is one of the most differentially expressed genes between abdominal and gluteal fat depots, and its expression level is much higher in gluteal SAT than in abdominal SAT [6]. HOTAIR resides within the HOXC locus (chromosome 12) and is well known as an important epigenetic regulator that causes chromatin silencing in trans at the HOXD locus (chromosome 2) by interacting with polycomb repressive complex 2 (PRC2), which contains histone methyltransferase EZH2 (enhancer of zeste homologue 2) at the 5′ end and lysine-specific histone demethylase 1 (LSD1) at the 3’ end to modulate H3K27 trimethylation and H3K4 demethylation, respectively [11]. HOTAIR may be involved in adipocyte differentiation through various mechanisms mediated by these key epigenetic regulators. For instance, previous research has shown that LSD1 and EZH2 of PRC2 could regulate adipogenesis by demethylating H3K4 and inducing H3K27me3, respectively, on the promoter regions of the Wnt signalling genes [12,13]. HOTAIR is also highly dysregulated among various cancers and could possibly epigenetically regulate hundreds of genes to promote carcinogenesis, cell proliferation, or tumour metastases [14,15]. Moreover, HOTAIR is sited near the HOXC13 locus, which is one of the genome-wide association studies (GWAS) that identified loci for human fat distribution [16]. Taken together, HOTAIR is very likely a functional epigenetic regulator to maintain or regulate the transcriptome profile in regional adipose tissue and plays a critical role in adipocyte biology, such as abdominal adipogenesis.

Histone modifications might alter chromatin accessibility for transcription factors to influence genetic transcription [17]. Recent studies suggest that HOTAIR also modulates DNA methylation on the promoter regions of specific genes, such as PTEN and PCDH10 [18,19]. In addition, PTEN and PCDH10 could inhibit the PI3K/Akt signalling pathway, which is known to play important roles in adipogenesis [20,21]. Considering that HOTAIR-related epigenetic modifications might present cellular-specific properties, we aim to utilize human immortalized abdominal preadipocytes to decipher HOTAIR-regulated temporal genome-wide DNA methylation and transcriptional patterns during abdominal adipocyte differentiation.

Currently, only a few studies have investigated the expression of the HOTAIR gene in human adipose tissue, and they merely studied abdominal or gluteal SAT in healthy individuals [6,22]. We hypothesized that HOTAIR expression in adipose tissue might be dysregulated under metabolic or inflammatory conditions and might subsequently influence regional adiposity, such as in patients with severe obesity or end stage renal disease (ESRD). Therefore, we conducted the 1st Asian study comparing HOTAIR expression in different fat depots from healthy, severely obese, or uraemic patients; examined the correlation between expression levels and regional adiposity; induced HOTAIR overexpression in the human abdominal preadipocyte cellular model to assess its phenotypic effects and epigenetic regulation; and linked the HOTAIR-repressed genes to the obesity-driven human body fat distribution.

2. Results

2.1. HOTAIR was differentially expressed among different fat depots and might be modulated under the status of severe obesity or uraemia to influence human body fat distribution

In line with previous reports [6,22], we observed that HOTAIR is also highly expressed in gluteal SAT among Asian subjects, whereas arm, abdominal and visceral SAT all showed low expression levels (Supplementary Figure 1), confirming that HOTAIR is critical in fat depot distribution. In particular, differential expression was observed in SAT between normal-weight individuals and severely obese or uraemic patients (Figure 1A). To be more specific, we found that abdominal SAT HOTAIR expression was suppressed in the status of severe obesity, whereas arm SAT HOTAIR expression could be aberrantly increased in the uraemic status (Figure 1A). Furthermore, SAT HOTAIR expression is negatively correlated with regional body fat percentage. The lower expression of abdominal SAT HOTAIR was correlated with increased abdominal adiposity in severely obese subjects, and a higher expression of arm SAT HOTAIR was found to be correlated with decreased arm adiposity in uraemic patients (Figure 1B). Using dual-energy X-ray absorptiometry (DXA) to precisely measure regional body fat composition, both severely obese and uraemic patients showed a higher android-to-gynoid fat ratio than normal-weighted individuals (Table 1). It is not surprising that central fat accumulation with metabolic disturbance (increase in systemic blood pressure, triglyceride, low-density lipoprotein, aspartate aminotransferase and uric acid and decrease in high-density lipoprotein) was observed in the obese patients, but it is worth noting that uraemic patients also present central obesity even under relatively low general adiposity (Table 1). These results suggest that HOTAIR levels might be dynamically changed under metabolic or inflammatory conditions with potential functional roles in regulating regional adiposity or human fat distribution.

Figure 1.

HOTAIR expression was dysregulated under the status of severe obesity or ESRD with anti-adipogenic effect during in vitro abdominal adipogenesis. (A) Paired arm subcutaneous adipose tissue (Arm SAT) and abdominal subcutaneous adipose tissue (Abd SAT) were collected from 6 healthy individuals. Abd SAT was collected from 5 severe obese individuals. Arm SAT was also collected from 5 patients with end stage renal disease (ESRD). HOTAIR expression in different fat depots was measured. (B) Body composition and regional adiposity were measured via Dual-energy X-ray Absorptiometry (DXA). (C) The immortalized human abdominal preadipocyte cell line derived from abdominal subcutaneous adipose tissue was used to assess human regional adipocyte biology. Constitutive HOTAIR overexpression (Abd HOTAIR-OE cell) was conducted via lentiviral transduction and empty vector was used to generate the Abd Control cell. In vitro adipogenesis of Abd HOTAIR-OE and Abd Control cells were performed and HOTAIR expression was measured during adipogenic differentiation (n = 3). (D) The status of lipid droplet formation was quantified with Oil red O staining during adipogenic differentiation (n = 3) and cells were photographed on D0 and D14 (x200 magnification). (E) The expression of two master regulator transcription factors (CEBPA and PPARG) for adipocyte differentiation was measured during adipogenic differentiation (n = 3). All data are presented as means ± SEM. Gene expression was measured via real-time qPCR. Statistical significance was assessed using independent Student T-test to compare different fat-depots or two cells at the same time point & two-way ANOVA to evaluate the effects of two factors (HOTAIR overexpression & time points). The correlation between regional adiposity and the HOTAIR expression in regional adipose tissue was assessed with Spearman's correlation test, ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

Table 1.

Anthropometric and biochemical characteristics of the study population.

	Normal weight	Morbid Obesity	Uremia
	(n = 6, F:M = 6:0)	(n = 5, F:M = 2:3)	(n = 5, F:M = 3:2)
Age (years)	36.8 [24.7; 38.5]	31.1 [26.0; 39.3]	66.0 [52.0; 71.2]∗†
BMI (kg/m2)	20.8 [19.3; 23.4]	38.9 [34.9; 45.9]∗	24.6 [21.3; 26.9]†
Waist-to-hip ratio	0.796 [0.757; 0.849]	0.954 [0.878; 1.102]∗	0.897 [0.859; 0.936]
Systolic BP (mmHg)	113.5 [111.3; 125.8]	153 [142; 165.5]∗	149 [125.5; 158]∗
Diastolic BP (mmHg)	71.5 [63.8; 77.8]	100 [78.5; 115.5]	73 [59.5; 96]
Fasting glucose (mg/dL)	87 [82.5; 89]	101 [85; 105.5]	100 [84; 120]
HbA1c (%)	5.35 [5.1; 5.6]	5.7 [5.3; 6.35]	5.8 [5.35; 6.8]
Triglyceride (mg/dL)	64 [54.5; 104.25]	169 [106.5; 289.5]∗	102 [81.5; 114]
HDL cholesterol (mg/dL)	66 [60.75; 74.25]	46 [35; 50.5]∗	38 [27; 58]∗
Total cholesterol (mg/dL)	175.5 [159.5; 180.25]	206 [160.5; 234]	146 [117.5; 157.5]†
LDL cholesterol (mg/dL)	93 [83.75; 108]	135 [113.5; 168]∗	87 [50; 108.5]†
BUN (mg/dL)	8.5 [8; 11]	14 [11; 19]	74 [52; 104.5]∗
Cr (mg/dL)	0.65 [0.48; 0.7]	0.9 [0.6; 1.3]	10 [5.6; 11]∗†
AST (U/L)	14 [13.25; 15.75]	26 [20.5; 46.5]∗	23 [16; 41]
ALT (U/L)	12 [9.5; 15.5]	61 [22.5; 115]	9 [5; 50]
Uric acid (mg/dL)	4.1 [3.45; 4.9]	6.5 [6.25; 8.95]∗	5.1 [3.9; 6.8]
DXA parameters
Total body fat (%)	38.3 [31.8; 41.43]	43.6 [39.65; 49.25]	24.4 [18.1; 37.75]†
Trunk fat (%)	39.5 [31.55; 42.93]	48.2 [43.8; 52.15]	27.9 [19.7; 36.4]†
Android fat (%)	44.35 [33; 49.65]	50.6 [48.35; 58.1]	32.5 [22.1; 39.9]†
Gynoid fat (%)	47.35 [44.98; 50.45]	45.6 [41.95; 53.25]	29.4 [21.05; 42.45]
Arm fat (%)	36.6 [30.98; 43.63]	34.9 [33.2; 47.8]	24.5 [17; 43.65]
Android fat mass (kg)	1.46 [1.08; 1.89]	4.53 [3.60; 8.19]∗	1.35 [0.92; 2.02]†
Gynoid fat mass (kg)	3.72 [3.35; 5.03]	7.69 [7.23; 12.12]	1.90 [1.46; 3.44]†
Android/gynoid fat ratio§	0.371 [0.309; 0.422]	0.603 [0.495; 0.683]∗	0.664 [0.555; 0.693]∗

Data were analyzed using the Kruskal-Wallis one-way ANOVA with pairwise comparisons and are presented as median values and [quartiles]. Abbreviations: F, female; M, male; BMI, body mass index; BP, blood pressure; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; BUN, blood urea nitrogen; Cr, creatinine; AST, aspartate aminotransferase; ALT, alanine aminotransferase; DXA, dual-energy X-ray absorptiometry.

§: android/gynoid fat ratio was calculated via dividing android fat mass by gynoid fat mass.

∗p < 0.05 as compared with normal weight group.

†p < 0.05 as compared between the Obesity and Uremia groups.

The bold word refers to variables with p < 0.05 as compared with the indicated group.

2.2. HOTAIR overexpression remarkably inhibits in vitro abdominal adipogenesis

Since HOTAIR is located within the developmental HOXC gene cluster and functions as an epigenetic regulator [11,14], we hypothesized that its low expression in the abdominal/arm SAT might be an epigenetic memory mark for maintaining the biological function of regional adipose tissue. We conducted constitutive overexpression of HOTAIR using lentiviral transduction in immortalized human abdominal preadipocytes (Abd HOTAIR-OE cells), and an empty vector was used to generate Abd Control cells. The immortalized human abdominal preadipocyte cell line was derived from the abdominal SAT of a 50-year-old healthy male (BMI: 24.4 kg/m²) and has been shown to be a useful cellular model to evaluate depot-of-origin adipocyte biological function [23]. First, we examined HOTAIR expression between Abd HOTAIR-OE cells and Abd Control cells during in vitro adipogenic differentiation and confirmed that Abd HOTAIR-OE cells presented persistently elevated levels of HOTAIR throughout the entire course of adipogenesis (Figure 1C). Time-course photograph and quantification of Oil red O staining showed that the Abd Control cells could differentiate well with abundant accumulation of lipid droplets, whereas adipogenic differentiation was significantly suppressed in the Abd HOTAIR-OE cells (Figure 1D). To determine whether HOTAIR influences the transcription of known adipogenic genes, we performed real-time qPCR on CCAAT enhancer binding protein alpha (CEBP-α) and peroxisome proliferator-activated receptor gamma (PPAR-γ), which are master regulator transcription factors for determining adipogenesis. The expression levels of these two genes in Abd HOTAIR-OE cells were also substantially inhibited during the entire course of adipogenic differentiation (Figure 1E), indicating that the aberrant overexpression of HOTAIR largely impedes the normal process of in vitro abdominal adipogenesis.

2.3. HOTAIR overexpression during abdominal adipogenesis increases global CG methylation and maintains the methylation status of genes functionally enriched for neural development

The process of HOTAIR-mediated epigenetic regulation through chromatin modifications, such as modulating H3K27me3 and demethylating H3K4 by recruiting PRC2 and LSD1 to the target genes and repressing their expression, has been widely acknowledged [11,14,24,25]. However, HOTAIR is a fat-depot-specific long noncoding RNA, and it is still unknown whether HOTAIR can regulate DNA methylation, particularly in preadipocytes or differentiated adipocytes. To comprehensively understand HOTAIR-mediated epigenetic regulation via DNA methylation during adipose differentiation, we performed reduced representation bisulfite sequencing (RRBS) among Abd HOTAIR-OE and Abd Control cells to unbiasedly assess genome-wide DNA methylation on differentiation day 0 and day 14. The mapping statistics of RRBS data were analysed with a minimum 4x reading depth (Table 2). A heatmap of methylomes was plotted to visualise the methylation levels and the correlation among these methylomes. Clearly, the methylomes of Abd HOTAIR-OE and Control cells were separated into two groups (Supplementary Figure 2A), and the metagene plot of methylation levels shows that transcription start sites have a significant dip, potentially to allow the binding of transcription factors (Supplementary Figure 2B). Then, principal component analysis (PCA) and two-way ANOVA of global CG methylation obviously showed that HOTAIR overexpression was the predominant effect among these methylomes, and time (day 0 and day 14) had a secondary effect that mainly influenced the methylation levels of control cells (Figure 2A and Table 3). HOTAIR overexpression also led to the union state of the methylome, while the methylomes of D0 and D14 control cells were diverse (Figure 2A). Direct comparison of the methylation status between Abd HOTAIR-OE and Abd Control cells revealed that HOTAIR overexpression significantly increased global CG methylation levels, particularly at day 0 (Figure 2B). Notably, the process of adipogenesis increased global CG methylation in Abd Control cells but had no effect on Abd HOTAIR-OE cells (Figure 2B).

Table 2.

Mapping statistics of the RRBS data.

Sample name	Number of raw reads	Uniquely mapped reads	Mappability (%)	Deptha (X)
Day 0-HOTAIR-OE-1	267,058,158	151,599,008	56.77	28.38
Day 0-HOTAIR-OE-2	162,770,712	93,859,258	57.66	20.62
Day 0-control-1	137,163,220	71,441,736	52.09	16.22
Day 0-control-2	129,165,794	85,135,156	65.91	18.39
Day 14-HOTAIR-OE-1	125,431,522	78,341,008	62.46	15.69
Day 14-HOTAIR-OE-2	147,841,920	74,430,124	50.34	15.57
Day 14-control-1	158,658,448	90,837,542	57.25	18.01
Day 14-control-2	145,033,610	81,799,402	56.40	16.40

^aThe depth is counted by coverage of the RRBS-covered sites.

Figure 2.

Apply RRBS to analyze HOTAIR overexpression-mediated impact on genome-wide methylation during abdominal adipogenesis. (A) Principal component analysis (PCA) of methylome was performed among Abd HOTAIR-OE and Abd Control cells on adipogenic differentiation day 0 and day 14. (B) The global CG methylation levels of Abd HOTAIR-OE and Abd Control cells were compared. The global CG methylation between day 0-Control and day 14-Control also shows significant difference (∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001). (C) The number counts of hyper- and hypo- differentially methylated gene (DMG) in HOTAIR-OE cell show significant difference on both day 0 and day 14 and present an anti-correlation with time effect. (D, E) The genebody- and promotor- DMGs in Abd HOTAIR-OE on day 0 and day 14 were compared, and the methylation status was plotted.

Table 3.

Numbers of CG methylated regions that were significantly affected by HOTAIR and developmental time using two-way ANOVA.

	HOTAIR	Developmental time	Interaction
CG regionsa	5,289 (2,035)	–	–

The significant definition is p < 0.05 with Δ CG ≥ 20%.

^aParentheses indicate the affected genes.

To explore the effect of HOTAIR-OE-mediated DNA methylation at the gene level, we defined the HOTAIR-differentially methylated regions (DMRs) as △CG methylation levels greater than or equal to 25% and t-test p values less than 0.05 between Abd HOTAIR-OE and control cells. If the DMRs overlap with gene bodies or promoters, these genes are deemed HOTAIR-mediated differentially methylated genes (HOTAIR-DMGs). We found that there were more HOTAIR hyper-DMGs at day 0, but after adipogenic differentiation (day 14), more hypo-DMGs were observed (Figure 2C). This time effect-related anti-correlation on HOTAIR-DMGs was mainly caused by the increase in global methylation in the control cells at day 14, as shown in Figure 2B. Furthermore, we examined the status of HOTAIR overexpression-mediated DMGs between differentiation day 0 and day 14, and it is intriguing that the majority (more than 84%) of gene body and promoter DMGs (both hyper- and hypo-) could maintain their methylation status (Figure 2D,E), suggesting that the methylation of these genes is constant and continuously affected by HOTAIR.

To understand the signalling pathways and molecular mechanisms associated with HOTAIR-mediated methylation during abdominal adipogenesis, we performed functional analysis using Ingenuity Pathway Analysis (IPA) software (https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis) [26] on 540 constantly hypermethylated DMGs and 400 constantly hypomethylated DMGs. The top 5 canonical pathways, molecular and cellular functions and physiological system development and function are shown in Supplementary Table 1A and 1B. Notably, both constantly hypermethylated and hypomethylated DMGs are functionally enriched for the same molecular and cellular functions, including “Cell-to-Cell Signalling and Interaction”, “Cellular Assembly and Organization”, “Cell Morphology”, “Cellular Function and Maintenance” and “Cellular Development”. Regarding physiological system development and function, both were mostly enriched for “Nervous System Development and Function”. These results suggest that the biological function of Abd HOTAIR-OE preadipocytes was significantly altered and possibly deviated towards neural lineage cells. Moreover, the top canonical pathways are also associated with neuron or adipose tissue biology. The constantly hypermethylated DMGs were functionally enriched for “GABA Receptor Signalling”, “Synaptogenesis Signalling Pathway” and “Endocannabinoid Neuronal Synapse Pathway”. Then, the constantly hypomethylated DMGs were functionally enriched for “Netrin Signalling”, “White Adipose Tissue Browning Pathway”, “Axonal Guidance Signalling” and “CREB signalling in Neurons”, indicating that the neurotransmitter signals within the Abd HOTAIR-OE preadipocytes were enhanced, which likely promoted adipose browning. Under the assumption that hypomethylated genes tend to be increased, we examined the predictive upstream and downstream signalling pathways regulated by the “White Adipose Tissue Browning Pathway”-associated constantly hypomethylated DMGs. Intriguingly, the functions of lipolysis and transdifferentiation of white adipose tissue to beige adipose tissue were predicted to be activated accompanied by inhibition of white adipocyte differentiation (Supplementary Figure 3).

Taken together, our data suggest that genes involved in neural development and white adipose tissue browning are regulated by HOTAIR overexpression-mediated DNA methylation and that this alteration of neurometabolism during abdominal adipocyte differentiation may contribute to the anti-adipogenic effect of this gene.

2.4. HOTAIR-mediated differentially expressed genes are commonly maintained during abdominal adipogenesis with functions enriched in cellular movement and the cell cycle

To unbiasedly evaluate how HOTAIR overexpression influences the transcriptome, we performed RNA sequencing on both Abd HOTAIR-OE and Control cells at day 0, day 4, and day 14 of adipogenic differentiation. PCA disclosure time was the primary effect on the transcriptome, while within the same day, HOTAIR overexpression had a constant influence on the transcriptome to separate HOTAIR-OE and Control cells into two groups (Figure 3A). Since Abd HOTAIR-OE cells differentiated poorly as compared to Abd Control cells on differentiation day 14, it is not a surprise that these two cells have very different global transcription at this time point (Figure 3A). However, the PCA results of the methylome and transcriptome are very different, indicating that although HOTAIR-OE unifies the methylome, most of the gene expression is affected by the development time, not by their methylation status. Two-way ANOVA also showed that most of the gene expression was affected by the development time or the interaction effect of HOTAIR overexpression and development time, whereas 8,462 genes were directly associated with the HOTAIR-mediated methylation level (Supplementary Table 2). To examine whether the global transcription levels are influenced by the methylation status of the gene body or promoter, we further plotted the correlation between the methylation level and expression rank (Supplementary Figure 4A). Clearly, regardless of HOTAIR status or development time, gene methylation and expression show anti-correlations in promoters, whereas in the gene body, the anti-correlation is more obvious in highly expressed genes. These results support that the genes are globally repressed by high DNA methylation levels, and part of their expression is likely mediated by HOTAIR during adipogenesis.

Figure 3.

Apply RNA-sequencing to analyze HOTAIR overexpression-mediated impact on whole transcriptome during abdominal adipogenesis. (A) Principal component analysis (PCA) of transcriptome was performed among Abd HOTAIR-OE and Abd Control cells on adipogenic differentiation day 0 and day 14. (B) The numbers of up-/down-regulated differentially expressed genes (DEGs) in HOTAIR-OE cell on differentiation day 0, 4, and 14 were presented. (C) The transcriptional changes of DEGs-day 0 in HOTAIR-OE cell were shown from day 0 to day 4 and day 14.

Then, we defined the HOTAIR-mediated differentially expressed genes (HOTAIR-DEGs) as fold change less than −2 (downregulated) or more than 2 (upregulated) between Abd HOTAIR-OE and Control cells. Overall, there were slightly more downregulated DEGs than upregulated DEGs on day 0 (527 versus 514) but more upregulated DEGs than downregulated DEGs on day 4 (792 versus 637) and day 14 (2,382 versus 2,032) (Figure 3B). To specifically evaluate the transcriptome influenced by HOTAIR overexpression – not the developmental time – we particularly focused on the genes that were significantly dysregulated at day 0 of differentiation, as the dysregulation of genes could be more strongly associated with HOTAIR-mediated epigenetic regulation without interfering with the time effect of adipogenesis at later stages. Therefore, we further plotted the heatmap of expression fold change to visualize the dynamic change of day 0 HOTAIR-DEGs during adipogenesis day 4 and day 14 (Figure 3C). Intriguingly, among the 1,041 HOTAIR-DEGs at day 0, 27% (280 of 1,041) genes were constantly upregulated, and 19.9% (207 of 1,041) maintained downregulation. In total, almost half (46.8%) of day 0 HOTAIR-DEGs maintained their regulated status on 4 and 14 (Supplementary Figure 4B), indicating that these genes are constantly regulated by HOTAIR throughout the entire course of adipogenesis.

We further performed IPA core analysis [26] on the 277 and 203 constantly upregulated and downregulated HOTAIR-DEGs to explore the signalling pathways and molecular mechanisms underpinning HOTAIR-mediated transcriptional regulation during abdominal adipogenesis. The top 5 canonical pathways, molecular and cellular functions and physiological system development and function are shown in Supplementary Table 3A and 3B. For the top canonical pathways in the constantly upregulated HOTAIR-DEGs, “Signalling by Rho Family GTPase” and “Regulation of Actin-based Motility by Rho” were predicted to be activated (positive Z-score), and “Rho GDI Signalling” as a downregulator of Rho family GTPase was shown to be inhibited (negative Z-score), indicating that the Rho proteins involved in intracellular actin dynamics were significantly activated. Additionally, “Cellular Movement” was listed as the top molecular and cellular function in the constantly upregulated DEGs (Supplementary Table 3A). Regarding the constantly downregulated HOTAIR-DEGs, “Cell Cycle” is listed at the top molecular and cellular functions. Additionally, analysis of the canonical pathways indicated that the “Kinetochore Metaphase Signalling Pathway” was inhibited (negative Z-score) and that “Cell Cycle: G2/M DNA Damage Checkpoint Regulation” was activated (positive Z-score), showing that mitosis of the cell cycle was suppressed (Supplementary Table 3B).

Overall, these data indicate that HOTAIR overexpression constantly regulates the transcriptome to enhance Rho protein-associated cytoskeletal dynamics and inhibit the mitotic progression of the cell cycle, which likely leads to HOTAIR-associated antiadipogenic effects during abdominal adipogenic differentiation.

2.5. Integrative analysis of HOTAIR-DEGs and HOTAIR-DMGs to identify downstream functional pathways and putative HOTAIR-regulated genes during abdominal adipogenesis

To examine whether the DEGs regulated by HOTAIR-OE are associated with DNA methylation, we analysed the corresponding promoter △ methylation levels of constantly up- and downregulated DEGs in HOTAIR-OE cells at day 0 and day 14 of adipogenesis (Figure 4A). Significantly, the methylation level of promoters of constantly downregulated DEGs increased, followed by their gene expression decreasing when HOTAIR was overexpressed. In contrast, the methylation level of promoters of constantly upregulated DEGs decreased, and their gene expression increased when HOTAIR was overexpressed. The constantly downregulated HOTAIR-DEGs had higher promoter CG methylation level changes than the constantly upregulated HOTAIR-DEGs, indicating that these HOTAIR-DEGs were epigenetically regulated by DNA methylation at both day 0 and day 14 and that these genes were repressed by their promoter CG methylation level. Together, these findings emphasize that CG methylation mediated by HOTAIR reversely regulates gene expression.

Figure 4.

Combine RNA-sequencing and RRBS to identify putative functional genes or pathways regulated by HOTAIR during abdominal adipogenesis. (A) The corresponding promoter △ methylation levels of constantly up- and down-regulated DEGs in HOTAIR-OE cell were plotted at day 0 and day 14. The statistical significance was assessed using Student's t-test. ∗∗∗p < 0.001 (B) Within the day 0 DMG and DEG matched genes, there are 10 genes continuously maintain their methylation and corresponding expression status (7 for hyper-methylated and down-regulated, 3 for hypo-methylated and up-regulated) and were considered as the putative HOTAIR-regulated genes.

Then, we performed IPA core analysis [26] on the 87 and 205 genes concurrently recognized as HOTAIR-DEGs and HOTAIR-DMGs at day 0 or day 14 to investigate the functional pathways and molecular mechanisms influenced by HOTAIR-associated epigenetic and transcriptional regulation during abdominal adipogenesis (Supplementary Table 4A and 4B). Notably, at day 0 and day 14, these genes were both functionally enriched for similar molecular and cellular functions, including “Cellular Assembly and Organization” and “Cellular Function and Maintenance”. In terms of physiological system development and function, both of them were mostly enriched for “Nervous system Development and Function”, which was the same as the enriched functions in HOTAIR-DMGs (Supplementary Table 1A and 1B). Additionally, some nervous system-associated signalling pathways are listed in the top canonical pathways, including “GABA Receptor Signalling” at day 0 and “Semaphorin Neuronal Repulsive Signalling Pathway” and “Synaptogenesis Signalling Pathway” at day 14 (Supplementary Table 4A and 4B). Again, this integrative analysis suggests that HOTAIR overexpression could modulate the cellular function of abdominal preadipocytes and likely increase their ability for adipose-neuron crosstalk or transdifferentiation.

To identify the genes most likely regulated by HOTAIR-mediated DNA methylation, we selected the day 0 HOTAIR-DMGs that were concurrently recognized as HOTAIR-DEGs with the correct corresponding direction, i.e., genes that were hypermethylated matched with downregulated genes and genes that were hypomethylated matched with upregulated genes. Then, we further pinpointed a total of 10 genes that continuously maintained their methylation and corresponding expression status at day 4 and day 14, as shown in Figure 4B. There were 7 hypermethylated and downregulated genes (B3GNT5, ENOX1, GRTP1, LINC00669, PITPNC1, SLITRK4, HOXA2) and 3 hypomethylated and upregulated genes (AMZ1, CACNB4, GSN). These putative HOTAIR epigenetically regulated genes may directly or indirectly influence abdominal adipogenesis.

Collectively, these results provide supportive evidence that HOTAIR overexpression during abdominal adipogenesis is associated with the anti-correlation between promoter CG methylation and gene expression and causes epigenetic and transcriptional alterations in the genes functionally enriched for cellular assembly and organization and nervous system development. Furthermore, 10 genes are epigenetically regulated by HOTAIR through DNA methylation during abdominal adipocyte differentiation. To prove their direct interactions with HOTAIR and potential functional roles in abdominal adipogenesis, we performed more experiments using abdominal preadipocytes for validation.

2.6. Validation of putative genes epigenetically suppressed by HOTAIR during abdominal adipogenesis

Among these 10 genes, 5 genes were selected for evaluation because their functions are associated with adipocyte differentiation, neuronal development or cellular assembly: PITPNC1 is involved in lipid metabolism [27]; SLITRK4 modulates synapse differentiation [28]; ENOX1 regulates plasma membrane electron transport pathways [29]; GRTP1 encodes a GTPase activator protein [30]; and gelsolin (GSN) is a key regulator of actin filament assembly [31]. We first measured their expression (PITPNC1, SLITRK4, ENOX1, GRTP1 and GSN) using real-time qPCR in Abd HOTAIR-OE and Control cells on day 0, 4, 7 and 14 of adipogenesis. Two genes (SLITRK4 and PITPNC1) were shown to be significantly suppressed throughout the whole course of adipogenesis in Abd HOTAIR-OE cells (Figure 5A). The other three genes (ENOX1, GRTP1 and GSN) presented compatible regulated transcriptional direction as the RNA-sequencing data but did not reach a significant difference in the whole course of differentiation (Supplementary Figure 5A). Therefore, we mainly focused on SLITRK4 and PITPNC1 to further explore HOTAIR-mediated chromatin modifications.

Figure 5.

Validate the putative functional genes suppressed by HOTAIR using real-time qPCR, RNA immunoprecipitation (RIP) and chromatin immunoprecipitation (ChIP). (A) The expression of SLITRK4 and PITPNC1 on Abd HOTAIR-OE and Control cells during adipogenesis D0, D4, D7 and D14 was measured with real-time qPCR (n = 3). (B) RNA immunoprecipitation was performed via using EZH2 antibody in the D0 Abd HOTAIR-OE and Abd Control cells to evaluate the differential EZH2 interaction with HOTAIR between two cells (n = 5–6). (C) H3K27me3 chromatin immunoprecipitations were performed in the D0 Abd HOTAIR-OE and Abd Control cells to assess the differential chromatin occupancy on SLITRK4, PITPNC1 and PPARG promoter regions between two cells (n = 3). Mouse IgG was used as the negative control. (D) IPA was performed on the constantly down-regulated DEGs in HOTAIR-OE cell and PPARG was listed as a predicted upstream inhibitor (Z-score −1.79), which leads to inhibition of several genes. All data are presented as means ± SEM. Statistical significance was assessed using independent Student's t-test to compare two cells at the same time point; two-way ANOVA to evaluate the effects of two factors (HOTAIR overexpression & time points) & Mann-Whitney U test to examine the differential %Input between two cells, ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

HOTAIR, as a long noncoding RNA with a secondary structure [32], has been shown to suppress its target genes by interacting with PRC2 at the 5′ end and LSD1 at the 3’ end to modulate H3K27 trimethylation (from H3K27me2 to H3K27me3) and H3K4 demethylation (from H3K4me2 to H3K4me), respectively [11]. To examine whether HOTAIR could physically interact with PRC2 in the immortalized preadipocyte cell line, we performed RNA immunoprecipitation of EZH2, which is the functional enzymatic component of PRC2 as a methyltransferase [33]. Notably, after HOTAIR overexpression, the interaction between HOTAIR and EZH2 protein was significantly enhanced (Figure 5B), indicating that lncRNA-HOTAIR indeed interacts with PRC2 in abdominal preadipocytes. Subsequently, we performed chromatin immunoprecipitation using H3K27me3 antibody and ran real-time qPCR using the designed primers binding to the promoter region of the gene of interest. In Abd HOTAIR-OE cells, the occupancy of H3K27me3 was significantly enriched in the promoter region of SLITRK4, and there was a marginally significant increase (p = 0.061) in the promoter region of PITPNC1 (Figure 5C), suggesting that HOTAIR-associated histone modifications work together with DNA hypermethylation to constantly suppress the transcription of SLITRK4 and PITPNC1.

Since the PPARG expression of Abd HOTAIR-OE cells was also significantly and constantly inhibited throughout the course of adipogenesis (Figure 1E), we further performed H3K27me3 chromatin immunoprecipitation by running real-time qPCR targeting the PPARG promoter. Interestingly, we also observed that the occupancy of H3K27me3 on the promoter region of PPARG was significantly enhanced after HOTAIR overexpression (Figure 5C). This result indicates that some constantly downregulated HOTAIR-DEGs (such as PPARG) might be predominantly controlled by histone modifications without having significantly differential methylation status. Moreover, IPA upstream analysis of the constantly downregulated HOTAIR-DEGs showed that PPARG was a predicted upstream inhibitor (Z-score −1.79), leading to inhibition of molecules involved in the cell cycle (CDKN2C), lipid droplet formation (PLIN4) and fatty acid metabolism (CD36 and LIPE) (Figure 5D). These PPARG-associated networks provide a schematic view to explain how HOTAIR overexpression could possibly inhibit abdominal adipogenesis and will be warranted for further investigation.

2.7. Increased expression of Abd SAT SLITRK4 and VAT PITPNC1 is driven by obesity and positively correlated with central body fat distribution

After identifying SLITRK4 and PITPNC1 as HOTAIR epigenetically suppressed genes, we evaluated their potential functional roles in regional adipose tissue, particularly in abdominal or visceral fat depots. Therefore, we measured the transcriptional expression of these two genes at various fat depots among normal-weight individuals, subjects with severe obesity and uraemic patients. The results revealed that SLITRK4 is specifically highly expressed in the abdominal SAT of patients with severe obesity. Additionally, VAT from severely obese patients presented significantly increased expression of PITPNC1 (Figure 6A). In particular, the abdominal SAT SLITRK4 and VAT PITPNC1 levels were both positively correlated with the obesity-driven waist-to-hip ratio or DXA-quantified android-to-gynoid fat mass ratio (Figure 6B). This observational correlation suggests that these two genes/molecules might be functionally important for central fat accumulation and deserves future exploration.

Figure 6.

SLITRK4 and PITPNC1 as HOTIAR suppressed genes were, respectively, differentially expressed in abdominal SAT and VAT in the status of obesity and positively correlated with central body fat distribution. (A) Paired arm subcutaneous adipose tissue (Arm SAT), abdominal subcutaneous adipose tissue (Abd SAT) and gluteal subcutaneous adipose tissue (Glut SAT) were collected from 6 healthy individuals during cosmetic surgery. Paired Abd SAT and visceral adipose tissue (VAT) were collected from 5 morbidly obese individuals during bariatric surgery. Arm SAT was also collected from 5 patients with end stage renal disease (ESRD) as preparing the arteriovenous shunt. SLITRK4 and PITPNC1 expression in different fat depots was measured with real-time qPCR. Data presented as means ± SEM. Statistical significance was assessed using independent Student's t-test, ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. (B) Body composition and regional adiposity were measured via Dual-energy X-ray Absorptiometry (DXA). Waist circumference and hip circumference were measured to the nearest 0.1 cm. Waist-to-hip ratio & android fat-to-gynoid fat ratio were used as two major surrogates to present the central body fat distribution. The correlation between central body fat distribution and Abd SAT SLITRK4 and VAT PITPNC1 expression was assessed with Spearman's correlation test, ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

3. Discussion

Our results revealed that the long noncoding RNA HOTAIR, flanking by HOXC11 and HOXC12, is differentially expressed in various adipose tissues and is highly expressed in gluteofemoral fat depot but maintains low levels in the arm and abdominal and visceral fat depots. This body position-specific HOTAIR expression follows similar rules as the HOX clusters. HOX genes are numbered 1 to 13 from the 3′ end to the 5’ end of the gene sequence in the chromosome [34] and present collinearity with the same order of their expression along the anterior-posterior body axis [35,36]. HOX genes are known to be functionally important to ensure the development of correct structures in the correct places of the body during embryogenesis. Therefore, we hypothesize that HOTAIR is an important epigenetic regulator for controlling or maintaining the biological function of regional adipose tissue and might be altered under metabolic or inflammatory conditions. Indeed, we observed that abdominal SAT HOTAIR expression was further inhibited under severe obesity and that arm SAT HOTAIR levels were significantly enhanced in uraemic patients. Using human immortalized abdominal preadipocytes, we revealed that HOTAIR overexpression significantly inhibited in vitro abdominal adipogenesis, consistent with our observation that HOTAIR levels were negatively correlated with abdominal adiposity. However, this is contrary to another report that ectopic HOTAIR overexpression was shown to enhance abdominal adipogenesis using primary abdominal preadipocytes [22]. The incompatible results might be due to different tissue contexts or experimental protocols. Nevertheless, considering that intrinsic HOTAIR expression in abdominal subcutaneous adipose tissue is low, particularly in severe obesity, our finding that aberrant HOTAIR overexpression remarkably suppresses abdominal adipogenesis appears more reasonable.

In our experiments, we conducted aberrant overexpression (around 200 to 400-fold) of HOTAIR in the immortalised abdominal preadipocyte and observed a strong anti-adipogenic effect during abdominal adipogenesis. However, there may be a concern whether the very high HOTAIR levels could interfere the normal process of adipogenesis. To examine this possibility, we also generated Glut Control and Glut HOTAIR-OE cells using the paired immortalised gluteal preadipocytes [23]. Notably, on adipogenesis day 14, the HOTAIR expression in Glut HOTAIR-OE cells also reached around 200-fold increase, but the Glut HOTAIR-OE cells could still differentiate adequately with similar PPARG expression comparing to that in Glut Control cells (Supplementary Figure 6). These data indicated that HOTAIR overexpression presents abdominal fat depot-specific anti-adipogenesis and strengthen the interests to explore the underpinned mechanisms in this study.

HOTAIR-mediated histone modifications have been widely approved, but they have mainly been studied in the field of cancers. Our results further demonstrate that HOTAIR-mediated epigenetic regulation is also present in adipose tissue. However, HOTAIR-mediated epigenetic regulation is not limited to histone modifications. Studies have shown that it can sponge many microRNAs to modulate multiple biological or pathological functions during embryonic development, cancer and human diseases [37]. It was also found to induce promoter hypermethylation with transcriptional silencing of tumour suppressing genes, such as PTEN [19]. HOTAIR is also involved in posttranslational modifications, such as protein degradation (ubiquitination) [38], NF-κB inflammatory pathways [39], and the process of DNA repair or the cell cycle [40]. In this study, we mainly focused on exploring HOTAIR-mediated global methylation and its links with the regulation of the whole transcriptome by conducting integrative analysis of RRBS and RNA sequencing. We found that HOTAIR overexpression has a major impact on the adipocyte CG methylome; yet, the development time has more effects on gene expression. Additionally, HOTAIR overexpression constantly and concurrently influenced the methylation and transcriptional status of genes functionally enriched for “Nervous System Development and Function” during abdominal adipogenesis (Supplementary Table 1A, 1B, 3B, 4A, 4B). More intriguingly, the constantly hypomethylated DMGs were further enriched for the “White Adipose Tissue Browning Pathway” (Supplementary Table 1B and Supplementary Figure 3). It seems that these genes associated with nervous system development might play crucial roles in regulating the adipocyte biology of abdominal fat depots and could be largely controlled by HOTAIR-mediated CG methylation changes. Recent studies have revealed that human adipose tissue-derived stem cells can be transdifferentiated into neuron-like cells in vitro [41]. Additionally, adipose tissue can release signalling factors to local sensory fibres that convey to the central nervous system to regulate systemic metabolism, and adipose tissue is innervated by sympathetic nerves that regulate lipolysis and thermogenesis [42,43]. Therefore, it will be worth examining in the future whether HOTAIR overexpression could prompt abdominal adipose tissue towards beiging or browning, especially under stimulation with β3-adrenergic receptor agonists or cold exposure.

Regarding the HOTAIR-DEGs, we found that 46.8% of D0 DEGs in Abd HOTAIR-OE cells could be maintained as constantly upregulated or downregulated throughout the course of abdominal adipogenesis. Additionally, the promoter △ methylation levels of constantly downregulated HOTAIR-DEGs were significantly higher than those in the constantly upregulated HOTAIR-DEGs at day 0 and day 14, indicating that the DEGs in Abd HOTAIR-OE cells were constantly regulated by the differential methylation levels even though the absolute levels of difference were less than 2% (Figure 4A). These HOTAIR-DEGs not only have an anti-correlation with CG methylation but also may be regulated by other HOTAIR-mediated epigenetic mechanisms, such as changes in nuclear architecture or 3D chromatin structure/chromosome positions [44], histone modifications [11] or microRNA sponging [37]^, and have not been globally examined in our cells. However, we analysed the overall impacts of HOTAIR overexpression on the transcriptome of abdominal preadipocytes and highlighted that the molecular and cellular functions of “Cellular Movement” and “Cell Cycle” were most enriched for the constantly upregulated HOTAIR-DEGs and downregulated HOTAIR-DEGs, respectively. Our findings are in line with previous reports that the actin cytoskeleton and Rho pathway have been shown to be key regulators of adipogenic differentiation [45], and cellular proliferation during initial adipogenic differentiation is important for increasing the number of cells committed to adipocytes [46].

By combining RRBS analysis and RNA sequencing, we identified 10 genes heavily controlled by DNA methylation-mediated transcriptional regulation that maintained the same direction of change throughout the whole course of adipogenesis. Among them, we further identified two genes (SLITRK4 and PITPNC1) that are regulated not only by HOTAIR-mediated changes in DNA methylation but also by histone modifications to maintain their constantly repressed expression during in vitro abdominal adipogenesis in Abd HOTAIR-OE cells. In addition to HOTAIR-mediated modifications of H3K27me3, we also examined its regulation of H3K4me2 by performing chromatin immunoprecipitation (Supplementary Figure 5B). Our results showed that the occupancies of H3K4me2 on the SLITRK4, PITPNC1 and PPARG promoter regions were significantly increased in Abd HOTAIR-OE cells, which is opposite to our prediction since the HOTAIR-LSD1 complex was supposed to specifically demethylate mono- and dimethylation of H3K4 [11]. One possible explanation is that another histone demethylase rather than LSD1 might interact with HOTAIR to induce demethylation of H3K4me3 in abdominal preadipocytes.

To extend our in vitro findings of HOTAIR-mediated epigenetic modification to the in vivo situation of adipose tissue, we further measured SLITRK4 and PITPNC1 levels among various fat depots. When we compared their expression with HOTAIR, only SLITRK4 showed a clear negative correlation among abdominal SATs between normal-weighted and obese subjects (Figure 1A and Figure 6A). Due to the incompletely matched results, we further examined two other genes PPARG and PCDH10 that had been shown to be suppressed by HOTAIR in our in vitro cellular model and the previous report [18], respectively. As a result, PCDH10 transcription was significantly enhanced in the obese abdominal SAT, whereas PPARG expression was significantly suppressed (Supplementary Figure 7A). The increased PCDH10 expression in the low HOTAIR-expressed obese abdominal SAT supports that HOTAIR mediated epigenetic regulation also exhibits in vivo. Regarding to the PPARG gene, PPARG expression has been shown to be inhibited in the obese abdominal SAT [47], that our findings are in line with this observation. Moreover, recent studies revealed that PPARG levels were remarkably suppressed under the status of insulin resistance [48] and that the cytokines released by pro-inflammatory macrophages can inhibit the transcription of PPARG [49]. Hence, it is likely that, in the obese abdominal SAT, HOTAIR is not the sole factor to regulate the gene expression. The other factors may include the inflammatory microenvironment, insulin resistance and interactions between the diverse cell types within adipose tissue (adipocytes, preadipocytes, macrophages, neutrophils, lymphocytes, and endothelial cells) could all influence the gene transcription, particular the adipogenesis or lipid metabolism associated genes, such as PPARG and PITPNC1. Even though, the potential functional roles of SLITRK4 and PITPNC1 in abdominal adipocyte biology were still supported by their dynamically increased expression during adipogenesis, as observed in the well-differentiated Abd Control cells (Figure 5A). Moreover, their levels in the abdominal or visceral fat depots were shown to be positively correlated with obesity-driven apple-shaped body figures. Therefore, SLITRK4 and PITPNC1 are proposed to play important roles in inducing obesity-associated central fat accumulation. To date, few studies have focused on these two genes, and their functional characteristics in central obesity or regional adipose biology have yet to be fully elucidated.

The SLITRK4 (SLIT and NTRK-like family, member 4) gene, located on chromosome X, encodes a transmembrane protein belonging to the Slitrk family. These family members contain two N-terminal leucine-rich repeat extracellular domains resembling those found in the axonal growth-controlling protein Slit, as well as a conserved intracellular C-terminal region similar to the neurotrophin receptor tropomyosin-related kinase (Trk). These proteins were named Slitrks due to their similarity with Slits and Trks [28]. SLITRK4 is highly expressed in the adrenal gland and cerebellum, whereas adipose tissue is also listed as the 3rd rank of tissue with high SLITRK4 expression based on Human Protein Atlas GTEx (Genotype-Tissue Expression) data (RNA-seq data on RSEMv1.2.22 (v7)). Genetic mutations in Slitrks have been associated with several neuropsychiatric disorders, such as obsessive-compulsive disorder [28], schizophrenia or Tourette syndrome [50], suggesting that Slitrks play prominent roles in normal nervous system development. Regarding SLITRK expression in regional adipose tissue, only one study found that SLITRK expression is higher in normal breast tissue than in breast carcinoma and is associated with high serum oestradiol levels in an age-, menopausal-, or hormone therapy-dependent manner [51]. Our study is the first to report that SLITRK4 was highly expressed in abdominal subcutaneous adipose tissue, positively correlated with central obesity, and epigenetically suppressed by HOTAIR during in vitro abdominal adipogenesis. These findings suggest that low levels of HOTAIR in abdominal subcutaneous adipose tissue are required for the increase in SLITRK4, which promotes the development of central fat accumulation. However, the detailed cellular and molecular mechanisms mediated by SLITRK4 in abdominal fat depot are still largely unknown.

The phosphatidylinositol transfer protein, cytoplasmic 1 (PITPNC1) gene, located on chromosome 17, encodes the PITPNC1 protein belonging to the phosphatidylinositol transfer protein (PITP) family. PITPs can mobilize phosphatidylinositol (PI) from the endoplasmic reticulum to the resident kinase for phosphorylation and are involved in multiple cellular processes, including cell signalling, membrane trafficking, cytoskeletal remodelling, and lipid metabolism [27]. PITPNC1 was ubiquitously expressed in multiple tissues, with adipose tissue ranked as the top tissue, followed by the basal ganglia, breast, lung, and cerebral cortex based on Human Protein Atlas GTEx (Genotype-Tissue Expression) data (RNA-seq data on RSEMv1.2.22 (v7)). In addition, the PITPNC1 protein is a soluble PITP and unique for its high binding affinity with phosphatidic acid (PA), but its functional significance has just started to be studied [27,52]. Recently, PITPNC1 was found to be highly expressed in omental metastatic lesions of gastric cancer and upregulated by omental adipocytes [53], indicating that it might influence the association between obesity and cancers. Moreover, GWAS identified the single nucleotide polymorphism (SNP): rs8866 as a PITPNC1 expression-associated risk locus for type 2 diabetes [54]. Here, we found that PITPNC1 was highly expressed in visceral (omental) adipose tissue, positively correlated with central obesity, and epigenetically suppressed by HOTAIR during in vitro abdominal adipogenesis. Therefore, our data propose a new genetic mechanism through the HOTAIR/PITPNC1 pathway in which relatively low HOTAIR expression in visceral fat depots might allow the dynamic increase in PITPNC1 to promote ectopic lipid accumulation with subsequent development of insulin resistance and type 2 diabetes. However, further in vivo animal or human genetic studies are still required to validate this hypothesis.

Our data showed that HOTAIR is likely a functional lncRNA for regulating abdominal adipogenesis. It is not surprising that other lncRNAs might also involve in the pathophysiological process of obesity. Liu et al. [55] had identified several lncRNAs differentially expressed in abdominal SATs between obese and non-obese children and that is correlated with central obesity. Among the differentially expressed lncRNAs that were reported, we specifically measured two children obesity-upregulated lncRNAs RP11-20G13.3 and LINC00968 and two children obesity-downregulated lncRNAs GYG2P1 and OLMALINC in our abdominal SATs of normal-weighted and severely obese adults. Intriguingly, we also identified LINC00968 was significantly increased and OLMALINC was considerably downregulated in the obese abdominal SAT. Whereas RP11-20G13.3 levels showed a non-significant trend of increase and there is no difference of GYG2P1 expression in the obese abdominal SAT (Supplementary Figure 7B). Overall, our experiments on revisiting these lncRNAs also support that these lncRNAs consistently showed clear changes of expression as they had been reported between obese and normal Abd SAT.

In conclusion, our study indicates that HOTAIR is an important epigenetic marker with relatively low expression in the ventral and anterior body areas. Abdominal aberrant overexpression of HOTAIR strongly suppresses adipogenesis via intricate DNA methylation, histone modification, and transcriptional regulation of genes associated with nervous system development and lipid metabolism, such as SLITRK4 and PITPNC1. Further understanding and targeting of these HOTAIR-downstream molecular pathways may lead to the discovery of novel therapies for ameliorating obesity-associated central fat accumulation and insulin resistance.

4. Materials and methods

4.1. Collection of paired adipose tissue samples

The criteria for inclusion were as follows: individuals aged 20–80 years received cosmetic surgery or elective operations, including bariatric surgery due to morbid obesity or arteriovenous shunting surgery due to end-stage renal disease. The exclusion criteria were pregnancy, current acute illness of cerebrovascular accident, myocardial infarction, heart failure, or psychiatric diseases. Paired adipose tissue samples were collected based on the accessibility of operations, and all participants signed informed consent before participating in this study with agreement to the use of relevant personal information on a confidential basis. Ethical approval was granted by the Tri-Service General Hospital Ethics Committee (TSGHIRB: 2-108-05-052).

4.2. Anthropometric measurements and dual-energy X-ray absorptiometry

Body weight was measured on a standard scale to the nearest 0.1 kg, and standing height was detected using a wall-mounted stadiometer to the nearest 0.1 cm as a barefoot while wearing light indoor clothing. Waist circumference was measured at the midway horizontal plane between the inferior margin of the last rib and the crest of the ilium, and hip circumference was measured at its widest point. Waist and hip circumferences were also detected to the nearest 0.1 cm. The waist-to-hip circumference ratio and body mass index (BMI) were calculated. BMI was calculated as weight in kilograms divided by the square of height in metres. Blood pressure measurement was performed twice in a sitting position after resting for 5 min with at least 1 min separation, and the average value was used as the data presentation. Whole and regional body adiposity was measured using dual-energy X-ray absorptiometry (DXA) as the standard protocol mentioned before [56]

4.3. Measurement of biochemical data

Venous blood samples were drawn under fasting status for at least 8 h. Serum levels of total cholesterol, triglycerides, high-density lipoprotein, low-density lipoprotein, blood urea nitrogen (BUN), creatinine, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and uric acid were measured on a Beckman Synchron LX20 analyser (LX20; Beckman Coulter, Brea, CA, USA). Fasting glucose was measured in the plasma using the glucose oxidase method on a Beckman Glucose Analyzer II (Beckman Instruments, Fullerton, CA, USA). HbA1c was measured using ion-exchange HPLC on an HLC-723G11 instrument.

4.4. Generation of constitutive HOTAIR overexpression abdominal preadipocytes

The immortalized abdominal preadipocyte cell line was given by the Metabolic Research Group in the Oxford Centre for Diabetes, Endocrinology and Metabolism [23]. For overexpression of HOTAIR, HOTAIR2 (NR_003716) from the LZRS-HOTAIR plasmid (Addgene plasmid #26110) was cloned into the pEGFP-Lv105 vector (Capital Biosciences). pEGFP-Lv105 empty vector was used as a control. Then, ViraPower™ lentiviral packaging mix (Invitrogen) was used to produce lentiviral particles in HEK293 cells. Immortalized abdominal preadipocytes were transduced with lentiviral particles in 8 μg/ml hexadimethrine bromide (Sigma)-containing growth medium for 20 h. Then, the virus-transduced cells were revived in normal growth medium for an additional 24–28 h with subsequent selection in a puromycin (1–2 μg/ml)-containing growth medium.

4.5. Culture and differentiation of immortalized preadipocytes

The reagent compositions of the growth medium and adipogenic differentiation medium were as previously described [23]. For adipogenesis, immortalized preadipocytes were cultured in growth medium first. Then, two days after full confluence (differentiation day 0), the cells were treated with adipogenic differentiation medium for 14 days as previously described [23].

4.6. Oil red O staining to evaluate the status of adipogenesis

Four percent paraformaldehyde was used to fix day 14 differentiated preadipocytes for 30 min, followed by incubation with 60% isopropanol for 5 min. Then, a 3 mg/ml solution of oil red O was prepared in isopropanol at a final concentration of 60% as the working solution and filtered (0.8 μm, Millipore, Watford, UK). The filtered working solution was added to the cells for 5 min at room temperature, and then, the cells were gently rinsed three times with PBS and photographed using an inverted phase contrast microscope with a camera.

4.7. Gene expression analysis

TRIzol reagent was used to extract total RNA from cells following the manufacturer's protocol. cDNA synthesis was performed using a cDNA synthesis kit (ABI). Quantitative real-time PCR was performed using SYBR green master mix (BIOLINE) in a 7500-FAST PCR machine (ABI). The ΔCT values of the target genes were normalized to the ΔCt of stably expressed reference transcripts (GAPDH) [57]. The primer sequences used here are listed in Supplementary Table 5.

4.8. Reduced representation bisulfite sequencing (RRBS)

4.8.1. Genomic DNA was extracted using the GenEluteTM mammalian genomic DNA miniprep kit

(Sigma), and RRBS libraries were constructed using the Ovation® RRBS Methyl-Seq with TrueMethyl® oxBS preparation kit. The streamlined workflow consists of five main steps: Mspl digestion, adaptor ligation, final repair, bisulfite conversion, and PCR amplification. The libraries are sequenced after PCR amplification. The reads were then aligned to the human hg19 reference genome and called the methylation level by using BS Seeker2. The uniquely mapped reads were retained for downstream analysis. To calculate the bisulfite conversion rates, the trimmed reads were also aligned to the λ phage genome. The bisulfite conversion rates of each sample were 100% minus the methylation level called by BS Seeker2. To generate genome-wide DNA methylation profiles, we calculated the methylation level for each covered cytosine in the genome. As bisulfite treatment converted unmethylated cytosines (Cs) to thymines (Ts), we estimated the methylation level at each cytosine by #C/(#C + #T), where #C is the number of methylated reads and #T is the number of unmethylated reads. The methylation level per cytosine serves as an estimate of the percentage of cells that are methylated at this cytosine. In this study, we included only cytosines that were covered by at least 4 reads for the analysis. To identify differentially methylated regions (DMRs), a nonoverlapping tiling window approach was used for genome-wide screening. Genomic regions of 500 bp containing more than 4 RRBS-covered CpG sites covered by at least 4 reads in all compared samples were surveyed. A region was deemed differentially methylated if it showed a ≥25% difference in methylation level between Abd HOTAIR-OE and Control cells in average methylation level and a t-test p value < 5%. A gene is considered differentially methylated (DMG) if it contains one or more DMRs in the promoter (−1500 bp to +500 bp of transcription start site) or gene body.

4.9. RNA-sequencing

Poly-T oligo-attached beads were used to purify and fragment mRNA for cDNA synthesis. Then, a single ‘A’ nucleotide was added to the 3′ end of ds cDNA, and multiple indexing adaptors were ligated to 5′ and 3′ of the ends of ds cDNA, followed by PCR amplification, library quality validation (Agilent 2100 Bioanalyzer and Real-Time PCR system) and Illumina NextSeq sequencing. mRNA reads were aligned to the human reference genome (hg19) using HISAT2 [58]. The annotation/normalization of the raw RNA-seq data was performed using Cuffdiff. The differentially expressed genes (DEGs) were defined as a t-test p value < 5% with fold change ≧ 2 and △FPKM ≧ 0.1 between Abd HOTAIR-OE and Control cells.

4.10. RNA immunoprecipitation (RIP)

An EZ-Nuclear RIP (Cross-Linked) kit (Cat#17-10521, Millipore) was used to perform RIP experiments. Formaldehyde (0.3%) was added to the cells (1 × 106/RIP) for 10 min to achieve in vivo cross linking. The cross-linked chromatin was incubated on ice and fragmented under frequent vortexing for 30 min. Chromatin supernatant (10%) was collected as the input before adding the following immunoprecipitation antibodies: anti-EZH2 (Cat# CS203195, Millipore) and mouse IgG (Cat# CS200621, Millipore). Magna ChIP Protein A/G Magnetic Beads (Cat# CS207374, Millipore) were utilized to bind the antibody/protein/RNA complex. RIP samples were washed, crosslinks were reversed, and RIP RNA was purified. cDNA was synthesized for real-time qPCR using SYBR Green. To assess the interaction between HOTAIR and EZH2, HOTAIR primers are listed in Supplementary Table 5.

4.11. Chromatin immuno precipitation (ChIP)

An EZ-Magna ChIP™ Chromatin Immunoprecipitation kit (Cat#17-10086, Millipore) was used to perform ChIP experiments. Then, 1% formaldehyde was added to the cells (3 × 10⁶/ChIP) for 10 min to achieve in vivo cross linking. Sonication (30 s ON/OFF for 10 min in a Bioruptor UCD-200, Diagenode) of the cross-linked chromatin was performed to generate <500 bp DNA fragments. Chromatin supernatant (1%) was collected as the input before adding the following immunoprecipitating antibodies: ChIPAb + H3K27me3 (Cat#17-622, Millipore) or mouse IgG (Cat#12-371B, Millipore). Magnetic Protein A/G Beads (Cat# CS204457, Millipore) were used to bind the antibody/antigen/DNA complex. ChIP samples were washed, crosslinks were reversed, and ChIP DNA was isolated as the template for real-time qPCR using SYBR Green. To assess the occupancy of H3K27me3 on the promoter regions of the SLITRK4, PITPNC1, and PPARG genes, the primers are listed in Supplementary Table 5.

Data availability

The sequence data obtained in this study have been submitted to the NCBI Gene Expression Omnibus (GEO) under GEO accession GSE186159.

Funding sources

This work was supported by research grants from Academia Sinica and the Ministry of Science and Technology (MOST 108-2314-B-016 -019 -MY3, MOST 108-2313-B-001 -013 -MY3, MOST 109-2314-B-016 -033, MOST 109-2313-B-001 -009 -MY3, MOST 110-2314-B-016 -022, MOST 110-2314-B-016 -002 -MY3) and Tri-Service General Hospital (TSGH-C04-109030, TSGH-C03-110024, TSGH-PH-E-109006, TSGH-PH-E-110012), and Taichung Armed Forces General hospital (TCAFGH-E-110051) in Taiwan.

Author contributions

FC Kuo and PY Chen conceived and designed the study; FC Kuo, CH Lee, KF Hsu, HY Yang, LW Wu, CH Lu and YJ Hsu curated the data; FC Kuo, YC Huang, MR Yen and PY Chen performed the data analysis and statistical analyses; FC Kuo and YC Huang wrote the original draft; and YC Huang, MR Yen, PY Chen and FC Kuo reviewed and edited the manuscript.

Conflict of interest

The authors declare no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article: