HFD feeding induces an early modulation of distinct pathways in visceral and subcutaneous adipose tissue
To reveal the early events that occur in WAT upon a HFD, and explain its pathological modifications associated with the development of inflammation, we analyzed the transcriptomic changes induced by a HFD in C57Bl6/J male mice after 1, 8 and 20 weeks (Fig. S1A). The mice fed with a HFD showed a significant difference in weight after only 1 week of diet (Fig. 1a), which correlated with an expansion of WAT (Fig. 1b). Interestingly, at 20 weeks, the amount of scWAT was greatly increased, whereas vWAT displayed a loss of weight with no significant differences between the control and HFD group. As expected, mice under a HFD showed increased serum levels of insulin, leptin and resistin (Fig. 1c) and developed inflammation mainly in vWAT at 8 weeks, as shown by the higher number of F4/80+ cells and crown-like structures compared to scWAT as well as by an increased expression of inflammatory markers such as Ccl2, Tnf, Il1b and Cxcl2 (Fig. 1d, e).
To get an overview of gene expression changes upon a HFD, we performed RNA-seq in both vWAT and scWAT at the different time points and applied the pathway analysis known as signaling pathway impact analysis (SPIA ) to the results. Two main pathways emerged as commonly deregulated by 1 week of HFD in both vWAT and scWAT, the ECM receptor interaction and focal adhesion pathways, reflected by the increased expression of genes coding for membrane proteins mainly related to collagen genes (Fig. 2a, c and Table S1).
In contrast, some pathways were differentially regulated exclusively in only one of the two tissues. Systemic Lupus Erythematosus (SLE) and Alcoholism pathways were deregulated only in vWAT. More particularly, 27 histone genes common to both these pathways and coding for all the canonical nucleosomal histones (H2A, H2B, H3, H4), showed increased expression in vWAT under the HFD condition (Fig. 2c and Table S2).
Conversely, in scWAT, three other pathways linked to neurodegenerative diseases appear deregulated at 1 week, i.e., Huntington, Parkinson and Alzheimer diseases. Interestingly, the group of genes shared between these three pathways, and down-regulated under the HFD condition only in scWAT, mainly belong to the electron transport chain (ETC) and mitochondrial function (Fig. 2c and Table S3).
At 8 weeks of HFD treatment, consistently with our and previous observations (Fig. 1d, e; [13, 42]), the deregulated pathways appearing in vWAT showed a clear inflammatory response, with a number of pathways linked to inflammation, including NF-KB and TNFα signaling (Fig. 2b). In contrast, no pathways linked to inflammation were present in scWAT after 8 weeks of diet.
Finally at 20 weeks, whereas vWAT remained highly affected by the inflammatory response (Fig. S2A), scWAT displayed a deregulation of SLE and Cell cycle pathways, which is maintained from 8 weeks onwards (Table S4A). As for vWAT at 1 week of HFD, the upregulated genes in SLE were mostly histone genes (Table S4B). In addition, two significant pathways related to inflammation appeared in scWAT (“Chemokine signaling pathways” and “Leukocyte transendothelial migration” pathways).
Taken together, these data show that vWAT and scWAT have, in the early phase, a distinct response to the diet, involving histone genes and genes linked to mitochondrial activity in vWAT and scWAT, respectively, followed by the onset of a strong inflammatory response only in vWAT.
scWAT undergoes loss of beige adipocytes in the early phase of overnutrition
As shown in Fig. 2, the early transcriptional changes occurring specifically in scWAT upon HFD are linked to the reduced expression of ETC-related genes. To validate this observation, we looked at the levels of some of the corresponding proteins by Western blot. Indeed, the decrease of complexes I, III, IV occurs also at the protein level in scWAT upon HFD (Fig. 3a). This decrease in mitochondrial proteins correlates with an overall reduction in mitochondrial mass, as supported by the decrease in mitochondrial DNA, which gives an estimation of mitochondrial abundance (Fig. 3b). Thus, the observed down-regulation of mitochondrial respiration can be due to a global reduction of mitochondrial number.
scWAT is known to harbor both white and beige adipocytes [30, 40], and, therefore, to have a higher mitochondrial mass compared to visceral fat. To investigate if tissue composition was altered in scWAT upon HFD, we performed histological analysis to quantify the number of beige adipocytes (UCP1+) in control and HFD-treated mice. As shown in Fig. 3c, scWAT contains many clusters of small UCP1+ cells with multi-locular lipid droplets, characteristic of beige adipocytes. These features are strongly reduced after 1 week of HFD. The reduction in UCP1+ cells also correlates with a drop in the number of small cells ranging up to 50 µm2 (Fig. 3d), suggesting that the small cells, which disappear upon a HFD, are possibly those contributing to the hypertrophic expansion of white UCP1− adipocytes. In contrast, we observe a clear shift in size, with a significant increase in the number of large cells.
These data suggest that beige UCP1+ adipocytes in scWAT lose their beige properties and contribute to the expansion of scWAT under a HFD.
HFD differentially regulates adipocyte progenitor proliferation and differentiation in vWAT and scWAT
The massive increase in histone gene transcription seen already at 1 week in vWAT, and observed later in both vWAT and scWAT at 8 weeks (Fig. 2), could be associated with increased cell proliferation. Consistent with this hypothesis, the expression levels of some known markers of proliferation, such as Mki67 and E2f1 (Fig. S2B), were increased only in vWAT after 1 week of HFD, while they were upregulated in both vWAT and scWAT at 8 and 20 weeks.
While proliferation signals observed from 8 weeks on in vWAT might be linked to proliferation of immune cells, which are infiltrating vWAT at this time point (Fig. 1d, e), we explored which cell population is already proliferating upon over-nutrition after 1 week. Interestingly, the expression levels of the Zfp423 gene, one of the most used markers for AP commitment to preadipocytes, is significantly increased in both tissues at 1 week (Fig. S2B),while it is only increased in scWAT at 8 weeks. Moreover, at 20 weeks Zfp423 expression is oppositely regulated, being downregulated in vWAT and upregulated and in scWAT. These observations suggest that the proliferating signals observed early in vWAT and at later time points in scWAT could be associated with AP clonal expansion, which represents the first step of their commitment.
To confirm the distinct timing of AP commitment regulation in response to a HFD in the different AT depots, we exposed mice to bromodeoxyuridine (BrdU) either during the first week or the seventh week of diet (Fig. S1b, c). FACS analysis performed using the classical markers for AP identification (CD31−: CD45−: CD29+: CD34+: Sca-1+) showed increased BrdU incorporation only in vWAT during the first week of HFD treatment (Fig. 4a), indicating a rapid increase of AP proliferation in this tissue. In contrast, mice exposed to BrdU later in the HFD treatment, revealed a strong burst of AP proliferation only in scWAT (Fig. 4b), confirming the hyperplastic growth of this tissue after 8 weeks of diet.
Taken together, these results suggest that the increase in histone gene transcription seen in vWAT by pathway analysis is associated with AP commitment at the early stage of over nutrition. Conversely, in scWAT, the increased transcription of histone genes, likely associated with AP commitment, is only observed upon 8 and 20 weeks of HFD treatment (Figs. 2b, S2A).
For a healthy expansion of adipose tissue, AP commitment must be followed by terminal differentiation. We, thus, explored the influence of HFD on adipocyte differentiation in the two adipose tissue depots by looking, first, at the expression of known markers of mature adipocytes, such as PPARγ and Fabp4. As shown in Fig. S2C, Pparg expression was reduced by HFD in both vWAT and scWAT. In contrast, Fabp4 is upregulated by HFD treatment in both tissues at 1 week (Fig. 4c), while it is oppositely regulated at 8 weeks, with increased levels in scWAT and reduced expression in vWAT. Its levels are even more decreased in vWAT at 20 weeks, in parallel to the shrinking of this tissue (Fig. 1b).
Such expression pattern argues against a significant increase of adipocyte differentiation in response to HFD in vWAT. To further investigate this aspect, we treated mice with BrdU during the first week of HFD and the percentage of BrdU+ nuclei in the mature adipocyte fraction was measured after 1 and 5 weeks. As expected, no BrdU-positive nuclei in the mature adipocyte fraction could be seen at 1 week in both tissues (Fig. 4d). However, a general increase in BrdU-positive nuclei was seen after 5 weeks, especially in vWAT, but with no significant differences between control and HFD groups. Thus, although HFD rapidly induces AP clonal expansion in vWAT (Fig. 4a), this is not followed by a proportional increase of adipocyte differentiation, suggesting that the maturation of these committed cells cannot be properly completed.
HFD feeding induces a different acetylation profile in scWAT and vWAT in genomic regions linked to lipid metabolism and cell fate
To explore the molecular mechanisms underlying the distinct temporal regulation of AP proliferation and differentiation in vWAT and scWAT in response to the diet, we looked at epigenetic changes affecting transcription. We focused on the ChIP-seq profile of two enhancer markers, H3K27 acetylation (active enhancers), and H3K4 monomethylation, together with that of RNA Pol II recruitment. Retrieving all genomic regions with significant changes in H3K27Ac signals in HFD mice compared to controls (p value < 0.05), either in vWAT or in scWAT at 8 weeks, identified a total of 5984 regions. The tag density of these regions was also analyzed in 1-week samples to identify early changes possibly responsible for the onset of inflammation in vWAT only. While most of the tested regions respond similarly to the HFD in scWAT and vWAT, the hierarchical clustering highlighted two interesting clusters (yellow and green square) with a distinct pattern in both tissues (Fig. 5a).
The regions of the first cluster (359 regions) were characterized by increased H3K27Ac in scWAT upon HFD, as opposed to reduced acetylation levels in vWAT. This behavior was already present at 1 week and was more marked after 8 weeks of diet. The second cluster (711 regions) presented a similar acetylation profile but the differences were less marked between 1 and 8 weeks in vWAT. Of note, the total tag density obtained for H3K27Ac and H3K4me1 on the regions of cluster 1 and 2 is very different between vWAT and scWAT under control conditions, with a much lower average tag density in scWAT (Figs. 5b, S3A). This suggests that these enhancers, which are differentiating the response to a HFD in the two depots, are physiologically more accessible for transcription in vWAT than in scWAT. Furthermore, in vWAT, we observed significant changes in RNAPol II recruitment, with an average tag density increased after 1 week of HFD, but reduced after 8 weeks. In parallel, H3K4me1 levels are significantly reduced at 1 week but unchanged at 8 weeks, suggesting that changes in this marker are happening mainly in the early phase of overfeeding. Finally, scWAT is oppositely regulated compared to vWAT only at the level of H3K27Ac tag density at 8 weeks; while, H3K4me1 and RNAPol II are not statistically different (Fig. 5b). Thus, the distinct pattern between vWAT and scWAT mainly stems from the level of H3K27Ac after 8 weeks in the regions of both cluster 1 and cluster 2 (Figs. 5b, S3A).
To gain insights into the putative biological functions of these chromatin regions, we annotated them to the closest gene from their center. This step provided a list of 323 genes from the first cluster and 674 genes from the second one. Gene Ontology annotations of the genes in the first cluster gave a strong enrichment of pathways linked to fatty acid metabolism (Fig. 5c). This was due to the presence of genes like Ppara, Hacd2, Psapl1, Scd4, Elovl6, Gpat2, Acsf2, Pla2g5, Alox12, Pck1, Degs1, Acsbg1, which are associated with sphingolipid and long-chain fatty acid metabolism. This cluster also contains developmental genes (Nrp2, Irx4, Mfng, Tshz2, Shroom3, Nat8f2, Ackr3, Celsr1, Ambra1, Fzd4, Sufu, Farp1, Slit3, Wnt4, Hoxa3, Slc26a8, Smoc1, Asb1, Etl4, Fgf1, Sik1, Ihh, Pitx2, Bmp8a, Zfp423) and some genes belonging to the Wnt pathway (Wnt4, Rspo1, Fzd4, Mmp7, Camk2b, Axin2). Interestingly, the second cluster is also enriched in terms of cell commitment and regulation of cell growth (Fig. 5d). Thus, the decreased H3K27Ac of these regions in vWAT but not in scWAT suggests that, upon HFD treatment, vWAT undergoes a reduction in fatty acid metabolism and, in line with our previous observations, a perturbation of cell growth and cell differentiation.
The presence of many genes of the Wnt pathway in both cluster 1 (Wnt4, Rspo1, Fzd4, Mmp7, Camk2b, Axin2) and cluster 2 (Usp34, Nle1, Prickle1, Gsk3b, Tcf7l2, Src, Nfkb1, Dlx5, Mitf, Zfp703, Tnks, Ctnnb1, Smad3) was particularly interesting as this signaling pathway is known to play a central role in adipocyte differentiation [3, 36]. We, thus, measured the expression of different Wnt genes, which have distinct properties with respect to adipogenesis. Interestingly, many Wnt genes are already differentially expressed in vWAT and scWAT in control mice, with decreased Wnt2 and Wnt10b and increased Wnt2b, Wnt5b and Wnt7b, in vWAT compared to scWAT (Fig. S3B). This suggests that the Wnt pathway and its action might be intrinsically different in the two tissues. Knowing that defects in adipogenesis can have an impact on the generation of hypertrophic adipocytes, we hypothesized a possible involvement of Wnt as a crucial player in the different response of vWAT and scWAT to a HFD. Consistent with this hypothesis, the expression of Wnt10b, which is a well-known negative regulator of adipogenesis [3, 4], increases in vWAT after 1 week of diet (Fig. 6a). This up-regulation is accompanied by reduced phosphorylation of β-catenin, further suggesting that the canonical Wnt pathway was likely activated only in vWAT after 1 week of HFD (Fig. 6b). Unphosphorylated β-catenin is less targeted for proteasomal degradation and, thus, free to access the nucleus and activate transcription of anti-adipogenic genes [21, 27].
These results suggest that Wnt10b and activation of the canonical Wnt/β-catenin pathway may be important for the early distinct responses of vWAT and scWAT to a HFD. Altogether, the transcriptomic changes occurring in vWAT may sustain a transient induction of AP commitment in vWAT concomitantly with impairment of the differentiation capacity of committed APs. Conversely, in scWAT, AP commitment is associated with increased expression of markers of mature adipocytes, meaning that no block of differentiation is present in this tissue.
Opposite modulation of the angiogenic capacity of vWAT and scWAT in response to HFD
The results obtained so far highlighted important biological features, but only a few related to metabolism. We, thus, attempted to investigate further the RNA-seq data with metaboGSE , which is a recently published algorithm relying on a genome-scale metabolic network (GSMN) to simulate the cellular metabolism. Albeit limited to metabolic reactions, this method emphasizes the low expressed genes and produces alternative gene set enrichment results, which are not necessarily identified by classical algorithms like GSEA and topGO. Along this approach, the RNA-seq datasets at 1 and 8 weeks were analyzed and 28 pairwise comparisons between the 8 conditions were investigated for 678 GO term-associated gene sets. These gene sets were selected as they were well supported by the GSMN and minimally redundant (see “Methods” details).
The overall analysis with the 28 contrasts showed that the difference between tissues was dominating over the diet and time-point variations (Fig. S4A). In line with the previous observations [46, 50], metaboGSE highlighted the thermogenic capacity of scWAT, confirming the appropriate functionality of the method.
When comparing the diet, a set of biological processes related to cGMP signaling, nitric oxide, hyaluronan biosynthesis, blood pressure regulation, and prostaglandin secretion appeared to be weakly but consistently enriched across the 4 pairwise comparisons. Of note, a dysregulation of the cGMP signaling was already described in adipose tissue from obese patients, in association with inflammation .
Genes belonging to these 5 main biological processes were systematically removed from the model to test their effect on the GSMN and to identify those with the highest impact (see “Methods”). The simulation highlighted the following five genes as most impactful: Nitric oxide synthase Nos2, Argininosuccinate synthase Ass1, Hexokinase-3 Hk3, GTP cyclohydrolase 1 Gch1, and 4-aminobutyrate aminotransferase Abat (see “Methods” details and Figs. S4B, S5). Among them, Ass1 and Nos2 are involved in the same metabolic pathway, with ASS1 being a key enzyme in the metabolism of arginine, providing l-arginine as substrate for NOS proteins. Nos2 and Ass1 were significantly upregulated in vWAT at 1 and 8 weeks, respectively; while, they were reduced at later time point (20 weeks). Instead, in scWAT, both genes were unchanged after 1 week, but increased at 8 and 20 weeks compared to the controls (Fig. 7a). Such expression profile, particularly that of Nos2, was similar to that of adipocyte commitment and differentiation markers (i.e., Zfp423 and Fabp4), when considering the tissue difference in response to the diet. This observation prompted us to search for other genes with similar expression profiles and unrelated to metabolism. To this aim, we performed on the full RNA-seq dataset a weighted gene co-expression network analysis (WGCNA) , which classifies all co-expressed genes into modules (Fig. 7b). In this analysis, Ass1 belonged to the antiquewhite module (1851 genes), whose regulation had the same positive direction in vWAT and scWAT at all time points. In contrast, Nos2 belonged to the Chocolate4 module that contained a total of 321 genes (Fig. 7c). Notably, at 8 weeks, the correlation of this module with the diet response was significant only in scWAT; while, at 20 weeks, genes in the module had an opposite regulation in vWAT and scWAT. To further explore the biological relevance of this set of co-expressed genes, we performed pathway enrichment analysis of all genes belonging to the Chocolate4 module. The resulting GO terms were highly enriched for pathways linked to angiogenesis (Fig. 7d). Consistently, the expression levels of Vegfa, a potent angiogenic factor, as well as Cdh5 and Pecam1, two endothelial cell markers, were strongly reduced in vWAT after 20 weeks of HFD, indicating a progressive impairment of angiogenesis in vWAT in response to the diet, as opposed to scWAT (Fig. 7e). Moreover, a number of the GO terms enriched in the Chocolate4 module, including “Tissue migration”, “Epithelium migration” and “Epithelial cell migration”, were shared with the pathway analysis performed on cluster 1 of the ChIPseq analysis. Of note, several genes belonging to the Wnt pathway, such as Ctnnbip1, Fzd8, Apc2, Ccdc88c, Wnt11, Tax1bp3, and Tmem88 were also listed in the Chocolate4 module, which suggests again a key role of this signaling pathway in differentiating the responses of scWAT and vWAT to a HFD, not only at early time points, but also in the chronic setting.