Differential association between macroH2A1.1 and macroH2A1.2 protein levels and BMI in human adipose tissue.
In various cell types, macroH2A1.1 displays an anti-proliferation role, while macroH2A1.2 has an anti-differentiation and pro-proliferative role [14–17, 19, 20]. Adult adipocytes are considered terminally differentiated cells. Here, we employed human visceral adipose tissue from obese subjects versus mildly overweight subjects to study the correlation between macroH2A1 isoform protein levels and body weight. Visceral adipose tissue biopsies were excised from patients with body mass index (BMI) ranging from 25 to 40, while they underwent abdominal surgery. Patient characteristics (sex, age, pathology) are described in Additional file 1: Supplementary Table I (Supplementary Material). Consistent with macroH2A1.2 being expressed at low levels in differentiated tissues, immunoblotting analysis showed that it was barely detectable at high exposure in human adipose tissue (Fig. 1a). Conversely, macroH2A1.1 isoform was found expressed, with a trend toward higher levels in subjects with high BMI (30–40) compared to subjects with lower BMI (25–26, mildly overweight) (Fig. 1a). Consistent with human data, confocal immunofluorescence analysis of white adipocytes in adipose tissue from wild-type mice revealed strong nuclear positivity (green) of macroH2A1.1 and absence of macroH2A1.2 isoform (Fig. 1b, left). Immunoblotting analysis confirmed a very weak expression of macroH2A1.2 compared to macroH2A1.1 in the adipose tissue (Fig. 1b, right).
MacroH2A1.2 transgenic (Tg) mice are leaner independently of food intake and energy expenditure
As human adult adipose tissue seems devoid of macroH2A1.2, we sought to study the effect or reintroducing the protein by systemic transgenic overexpression. We used a fusion plasmid (pCX-MH2A/EGFP) consisting of a CAG promoter, a chimeric cDNA encoding mouse macroH2A1.2 (GenBank accession number AF171080) and a fusion polypeptide with EGFP at the C terminus of macroH2A1.2 to generate macroH2A1.2–EGFP transgenic (Tg) mouse lines by DNA microinjection into pronuclear stage embryos [26] (Fig. 2a). Tg mice were healthy and fertile, and green fluorescence could be detected neonatally throughout the body [26] (Fig. 2b). The fusion protein could be easily detected in nuclear extracts of internal organs such as the kidney and liver by immunoblot analysis: A band of 42 kDa was detected in both wild-type and Tg mice using an anti-macroH2A1 antibody, whereas a band of 67 kDa was detected only in Tg organs, consistent with the expected size of the macroH2A1.2–EGFP and detected with an anti-GFP antibody [26] (Fig. 2c). Endogenous macroH2A1.1 expression in these organs was not changed upon overexpression of macroH2A1.2–EGFP transgene (data not shown). In the visceral adipose tissue, macroH2A1.2 immunopositivity was detected by confocal microscopy in the Tg, but it was absent in wild-type mice (Additional file 2: Figure S1). Quantitative EchoMRI/CT scan showed that Tg mice were in average ~5% shorter than wild type (Fig. 3a, p < 0.0001), with ~20% less lean mass and ~fivefold lower fat mass (Fig. 3b, p < 0.0001). Consistently, macroH2A1.2 Tg mice were also macroscopically leaner under a standard (chow) diet (Fig. 3c, upper panels). Similarly, when fed an obesogenic (12 weeks, 60% energy from lard [22]) high fat (HF) diet, macroH2A1.2 Tg mice appeared leaner and protected from fat induced-increased adiposity to the naked eye (Fig. 3c, lower panels). Accordingly, body weight of age-matched Tg mice was strikingly lower than wild-type mice both under a chow diet (26.4 ± 0.9 versus 33.9 ± 1.1, p < 0.0001) and under a HF diet (36.3 ± 2.1 vs 44.9 ± 1.2, p < 0.0001) (Fig. 3d). Of note, the weight of macroH2A1.2 Tg mice fed an obesogenic HF diet was not statistically different than the baseline one of wild-type mice fed a chow diet (36.3 ± 2.1 versus 33.9 ± 1.1, ns). These variations in body weight were mirrored by profound differences in the visceral adipose fat ratio as determined by EchoMRI/CT scan analysis (Fig. 3e): macroH2A1.2 Tg mice displayed more than threefold lower visceral adiposity compared to wild type fed a chow diet (8.6 ± 3.7 versus 26.5 ± 2, n = 8, p < 0.0001). Consistently, under HF diet Tg mice accumulated significantly less total visceral fat than wild-type mice (Fig. 3e). These large differences in body weight and adiposity could not be explained by a decrease in food intake in Tg animals, as both genotypes were found to eat the same amounts (Fig. 3f). The respiratory exchange ratio (RER) is the ratio between the amount of CO2 produced and O2 consumed by breathing. Measuring this ratio can be used for estimating which fuel (carbohydrate or fat) is being metabolized to supply the body with energy. We observed significant increase in basal RER from 0.87 ± 0.03 in wild-type mice to 0.97 ± 0.01 in macroH2A1.2 Tg mice (p < 0.05) (Fig. 3g), reflecting a switch from an energy consumption consisting of a mix of carbohydrates and fat to an energy consumption indicative of carbohydrate being the predominant fuel source, in Tg animals. Upon a HF diet, wild-type animals had a decreased RER compared to chow diet fed littermates (0.78 ± 0.04, p < 0.05) indicating fat as the predominant fuel source, while Tg mice displayed a RER, 0.84 ± 0.01, similar to baseline wild-type levels, i.e., reflecting energy consumption combining carbohydrates and fat (Fig. 3g). IGF-1 blood levels were similar in wild-type and Tg mice indicating that differences in body weight and size were independent of IGF-1 (Fig. 3h). Overall, these data demonstrate that macroH2A1.2 Tg have reduced total and visceral fat depots and an increased energy expenditure from carbohydrates.
MacroH2A1.2 Tg mice are more glucose tolerant and insulin sensitive, and display smaller pancreatic islets
In comparison with wild-type mice, an improvement in metabolic health is thus observed in Tg mice; we therefore also investigated the ability of Tg mice to respond to a glucose or insulin challenge. Basal glucose levels were considerably lower in Tg versus wild-type mice, fed a chow (124.9 ± 11.4 versus 147.25 ± 8.86, p < 0.05) or a HF diet (155.67 ± 10.6 versus 190 ± 11.3, p < 0.0001) (Fig. 4a, b). In glucose tolerance tests (GTT), glucose levels remained significantly lower in macroH2A1.2 Tg mice at every time point, compared to wild-type littermates, both upon a chow or a HF diet (Fig. 4a). Insulin tolerance tests (ITT) showed that the insulin-mediated decrease in glycemia was much more pronounced and statistically significant in Tg mice versus wild-type mice at every time measured (p < 0.0001 at 15, 30, 45, 60, 120 min time points) upon a chow diet (Fig. 4b). Upon a HF diet, statistical differences were observed between macroH2A1.2 Tg and wild-type mice only after 30 min (Fig. 4b). To gain insight into the mechanism by which systemic glucose tolerance is improved in chow and HF diet fed macroH2A1.2 Tg mice, we characterized insulin-induced AKT signaling in the skeletal muscle, liver and adipose tissues under insulin-stimulated conditions (0.75 U kg − 1 body weight, injected 15 min before killing) (Additional file 3: Figure S2). AKT phosphorylation (Ser473) was increased in insulin-responsive peripheral tissues of macroH2A1.2 Tg mice fed either a chow or a HF diet compared with wild-type controls (Additional file 3: Figure S2). Circulating insulin levels did not differ between genotypes under a chow diet, and they were found increased to the same extent upon HF diet (Fig. 4c). However, pancreatic islets were strikingly smaller in macroH2A1.2 Tg compared to wild-type mice (Fig. 4d, left panels); semi-quantitative double immunofluorescence confocal analysis in pancreatic islets showed a similar insulin content in beta cells in both genotypes, with a not significant trend toward a decreased glucagon content in alpha cells in Tg mice (Fig. 4d, right panels). Hence, macroH2A1.2 Tg mice are more glucose tolerant and insulin sensitive, and display smaller pancreatic islets.
MacroH2A1.2 Tg mice have decreased hepatic and pancreatic fat content and inflammation upon a HF diet
Obesity is almost invariably associated with NAFLD and to non-alcoholic fatty pancreas disease (NAFPD), two interrelated conditions characterized by parenchymal triglyceride accumulation and inflammation. NAFLD and NAFPD are risk factors to develop cirrhosis and cancer [1, 27]. We sought to analyze the lipid content in the liver of macroH2A1.2 Tg versus wild-type mice: H&E staining revealed evident differences, with a total protection from lipid accumulation in Tg compared to wild-type animals that developed instead extensive mixed micro/macro/vesicular NAFLD upon HF diet (Fig. 5a, left panels). Accordingly, NAFLD and lobular inflammation scores were highest in livers of wild-type mice fed a HF diet; conversely, NAFLD score of Tg animals upon HF diet was similar to the one of wild-type mice fed a chow diet (Fig. 5a, right panels). H&E staining of pancreata revealed a marked decrease in fat infiltration in Tg mice versus wild-type mice (Fig. 5b, left panels). Histological analysis revealed also that upon a chow diet, wild-type mice contained ~5% of intralobular fat, while Tg mice were devoid of it; upon a HF diet, Tg mice accumulated only <5% of intralobular fat, while pancreata from wild-type mice displayed ~25% intralobular, ~1–2% interlobular fat accumulation and ~2 of Mathur score (Fig. 5b, right panels). MacroH2A1.2 Tg mice are thus protected against HF-induced NAFLD and NAFPD, consistent with their protection from obesity.
macroH2A1.2 counteracts adipogenesis in vivo and in vitro
Our data indicate that macroH2A1.2 Tg mice do not display important differences in food intake as compared to wild-type mice. We thus hypothesized that the striking protection from obesogenic diet-induced increase in body weight and adiposity in these animals could be attributed to an inhibition of adipogenesis. Adipocytes are the major storage site for fat in the form of Tgs, and this can be accomplished in two different ways, by expanding the available adipose cells or by recruiting new fat cells upon differentiation. Histological analyses of visceral white adipose tissues revealed a ~60% decrease in adipocyte size (area) in Tg mice versus wild-type mice even upon a chow diet (Fig. 6a). Upon a HF diet, Tg animals displayed adipocytes on average ~35% smaller than wild type (Fig. 6a). Variations in the circulating concentration of adipose tissue-secreted adipokine leptin in the four mice groups correlated well with the size of body fat stores (Fig. 6b). To explore the molecular mechanisms that might be involved in the decreased adiposity of macroH2A1.2 Tg mice, we used a gene array profiling the expression of 84 key genes involved in white adipose tissue adipogenesis, including hormones, adipokines, enzymes and transcription factors. Using a twofold cutoff difference in mRNA expression upon HF diet in either of the two groups of animals, we identified 20 genes, 18 of which were oppositely regulated in the adipose tissue of macroH2A1.2 Tg mice compared to wild-type mice (Fig. 6c, Additional file 1: Supplemental Table III). Pro-differentiation and pro-adipogenic genes ACACB, AGT, FASN, RETN and SLC2A4 were significantly upregulated in wild-type adipose tissue and downregulated in macroH2A1.2 Tg adipose tissue when mice were fed a HF diet (Fig. 6c). Conversely, anti-adipogenic genes E2F1, EGR2, JUN, LMNA, anti-inflammatory genes SIRT1, SIRT2, thermogenic gene UCP1 and proliferation-regulating genes ANGPT2, CCND1, CDKN1A, CDKN1B, were upregulated in macroH2A1.2 Tg adipose tissue and downregulated in wild-type adipose tissue when mice were fed a HF diet (Fig. 6c). BMP7 and WNT5B were found two–threefold upregulated in the adipose tissue of both genotypes upon a HF diet (Fig. 6c, d). To understand whether macroH2A1.2 could affect the differentiation of pre-adipocytes into mature adipocytes, we used the well-established murine 3T3-L1 cell model. Stable expression of GFP, macroH2A1.2, and its sister splicing variant macroH2A1.1, in 3T3-L1 pre-adipocytes was achieved by lentiviral transduction as previously described [12], and differentiation into mature adipocytes was obtained through a 15 days long protocol based on the sequential addition of dexamethasone, IBMX and insulin [28] (Fig. 7a). Expression of macroH2A1.1 transgene did not have an effect on endogenous macroH2A1.1 protein levels that were stable along differentiation (day 1–5–15, Additional file 4: Figure S3). Conversely, expression of macroH2A1.1 and macroH2A1.2 transgenes led to markedly decreased levels of endogenous macroH2A1.2 at all time points of the differentiation protocol, compared to GFP-expressing control cells (Additional file 4: Figure S3). Of note, in GFP-expressing control cells macroH2A1.2 endogenous levels decreased during the differentiation of pre-adipocytes into adypocytes (day 5 and day 15 compared to day 1, Additional file 4: Figure S3), which is consistent to the low levels of macroH2A1.2 in adult human and mouse VAT (Fig. 1). At the end point of the protocol, mature 3T3-L1 adipocytes were analyzed for lipid content using Oil Red O (ORO) staining: We found that, compared to GFP-overexpressing cells, macroH2A1.2-GFP-overexpressing cells displayed a ~2.5-fold decrease in lipid content, while macroH2A1.1-overexpressing cells accumulated ~1.7-fold more lipids (Fig. 7b). MacroH2A1.2-dependent mRNA upregulation of RETN, E2F1 and EGR2, together with downregulation of FASN, was observed in 3T3-L1 adipocytes, mirroring the in vivo data (Additional file 5: Figure S4). In vitro and in vivo data collectively suggest that macroH2A1.2 might impair adipocyte differentiation while, in the 3T3-L1 model, overexpression of macroH2A1.1 leads to augmented lipid accumulation. Interestingly, macroH2A1.1 has been consistently reported to be highly expressed and have an anti-proliferative action, while macroH2A1.2 was expressed at low levels, in differentiated cells [14]. Consistently, confocal immunofluorescence analysis of adult heart tissue, which is slowly or not proliferating and possesses scarce regenerative capacity, in WT mice fed a chow diet revealed predominant expression of macroH2A1.1 and not of macroH2A1.2 (Additional file 6: Figure S5). In contrast, under the same conditions the adult mouse liver, which is a regenerative organ mainly due to the high proliferation rate of hepatocytes, both macroH2A1.1 and macroH2A1.2 are expressed (Additional file 6: Figure S5).
Genome occupancy of macroH2A1.2 display minor changes upon adipocyte differentiation
We sought to analyze if changes in chromatin occupancy macroH2A1.2 might play a role in transcriptional changes associated with adipocyte differentiation in vitro. We used ChIP-Seq, using a ChIP grade anti-GFP antibody, to isolate and sequence genomic regions associated with macroH2A1.2-GFP chimeric protein. No antibody was used as a negative control to subtract aspecific peak distribution. The data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE85796 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE85796). Both pre- and post-3T3-L1 adipocyte differentiation, macroH2A1.2-associated reads displayed a bell shape distribution with a peak at about −7000–8000 bp upstream of TSS (Additional file 7: Figure S6). Similar binding affinity of macroH2A1.2 histone was observed in mature adipocytes and in pre-adipocytes, as determined by sequence read tag density (Fig. 8).
Binding sites were subsequently grouped by gene section, i.e., 3′ or 5′ untranslated region (UTR), coding sequence, intergenic, intron, TTS, non-coding and promoter-TSS (Fig. 8b). The frequency of occupancy showed that macroH2A1.2 binding was enriched in intergenic and intron regions, with no significant differences between pre- and post-differentiation conditions (Fig. 8b, upper panels). Similarly, filtering binding sites to exclude intergenic and intron regions, in order to highlight exons, TTS, 5′UTR, noncoding, promoter-TSS and 3′ UTR, did not evidence differences in binding patterns (Fig. 8b, lower panels). Analysis of genome occupancy of the pro- and anti-adipogenic genes (ACACB, AGT, FASN, RETN, SLC2A4, E2F1, EGR2, JUN, LMNA, SIRT1, SIRT2, UCP1, ANGPT2, CCND1, CDKN1A and CDKN1B) under the control of macroH2A1.2 in vivo (Fig. 6c) identified enriched regions (binding sites) with a false discovery rate (FDR) <0.01 only in 4 genes (UCP1, CDKN1A, FASN, Slc2a4), which were though inconsistent between biological duplicates and were not significantly different between 3T3-L1 pre-adipocytes and in 3T3-L1 mature adipocytes (Fig. 9 and data not shown). To gain a more general view, macroH2A1.2-bound genes in 3T3-L1 cells in ChIP-Seq were represented by Circos plots (Fig. 10a). Zooming to represent magnifications of example chromosomes 5, 7 and X uncovered that with the exception of few distinct regions macroH2A1.2 significant genomic binding patterns in pre-adipocytes and in mature adipocytes were very similar (Fig. 10b–d). Our genomic analysis thus showed that the process of differentiation induces modest changes in macroH2A1.2 genome occupancy that may not account for its transcriptional effects in fat cells.
We then hypothesized that macroH2A1.2 could modulate transcription through the recruitment/modulation of the activity of transcription factors (TF). We thus examined the binding region sequences within the two datasets (3T3-L1 pre- and post-differentiation) to search overrepresented binding motifs for TF that might play a role in mediating macroH2A1.2-dependent effects. To this aim, we used PscanChIP tool that, starting from a collection of genomic regions, evaluates both motif enrichment and positional bias within them. Interestingly, in both pre-adipocytes and in mature adipocytes paired box 4 (PAX4) ranked as the most enriched TF presenting a binding matrix among macroH2A1.2-binding regions (Fig. 11a). GATA-binding sites were overrepresented in pre-adipocytes (Fig. 11a). In addition to the common matrices enriched in both pre- and post-differentiation adipocytes (including the master TF regulator of adipogenesis PPARγ), we observed several matrices specifically overrepresented in either pre-adipocytes (NFIC:TLX1, PKNOX1, TGIF2, TGIF1, ESR1, GATA3, PKNOX2, ZBTB18, TFEC and BHLHE41) or mature adipocytes (RXRA::VDR, Gmeb1, Prrx2, FEV, SP4, PROP1, NEUROG2, TEAD4, TP53 and SCRT2) (Fig. 11b). Of note, anti-adipogenic RXRA::VDR and TP53 were overrepresented specifically in mature macroH2A1.2-overexpressing adipocytes. Several of the TFs emerging from the analysis are known to bind to homeobox genes (NFIC:TLX1, PKNOX1, TGIF1, TGIF2, PKNOX2 in pre-adipocytes; Prrx2 and PROP1 in mature adipocytes) (Fig. 11b). Other binding matrices identified TF that function in neural/muscular/bone developmental programs (ZBTB18, TFEC, BHLHE41 in pre-adipocytes, and GMEB1, SP4, NEUROG2, TEAD4 and SCRT2 in mature adipocytes), likely associating to transcriptional repression neighboring macroH2A1.2 binding in adipose cells (Fig. 11b). These data demonstrate that, despite minor changes in genomic occupancy, macroH2A1.2 might associate with different TF-binding sites upon adipocyte differentiation, suggesting a potential transcriptional mechanism.