The effect of obesity and aging on NAD⁺/Sirtuin metabolism transcription and DNA methylation in subcutaneous adipose tissue of monozygotic twin pairs discordant for BMI

Abstract

Introduction

The expression and/or activity of sirtuins (SIRTs), nicotinamide adenine dinucleotide (NAD⁺)-dependent enzymes that regulate cellular energy metabolism, is decreased in obesity and in aging in animal models. However, the impact of obesity compared to aging on NAD⁺/SIRT expression in human white adipose tissue (AT) remains unexplored. Here, we unravel the effects of obesity and aging on the expression of NAD⁺/SIRT pathway and their associated genes in subcutaneous AT of identical twin pairs discordant for weight, in two age groups.

Methods

We examined 49 monozygotic twin pairs discordant for BMI (within-pair difference in BMI ≥ 2.5 kg/m², with mean BMIs 25.6 kg/m² (leaner) and 30.8 kg/m² (heavier), aged 22–38 and 56–69 years. Detailed phenotyping included body composition, insulin resistance (oral glucose tolerance test) and plasma lipids and inflammation markers. RNA sequencing and DNA methylation analyses in AT identified differentially expressed and methylated NAD⁺/SIRT pathway genes in obesity and aging, with linear mixed models linking gene expression to metabolic features.

Results

SIRT5 and NAD⁺ biosynthetic genes were downregulated in AT in both obesity and aging. Obesity was characterized by downregulation of AT NAD⁺/SIRT genes, and NAD⁺/SIRT regulated mitochondrial oxidative metabolism genes, and upregulation of stress markers. Aging showed a downregulation of AT PARPs, except upregulation for PARP1, a main consumer of NAD⁺. Mitochondrial metabolism and glycolysis genes were linked to corresponding DNA methylation. Downregulation of NAD⁺/SIRT genes correlated with increased adiposity, insulin resistance, inflammation, and dyslipidemia.

Conclusion

Impaired NAD⁺/SIRT metabolism in AT may play a key role in obesity- and aging-related diseases. Both conditions are characterized by downregulation of NAD⁺/SIRT pathway genes, correlating with increased adiposity, insulin resistance, inflammation, and dyslipidemia. Obesity uniquely disrupts expression of NAD⁺/SIRT regulated mitochondrial genes, while aging is characterized by altered PARP expression, particularly increased PARP1, likely exacerbating metabolic dysfunction in AT.

Introduction

Obesity, a major public health challenge, has increased globally, with now more than 2.5 billion people living with overweight or obesity [1]. Simultaneously, the world population is aging at a rapid pace [2]. Both obesity and aging are associated with increased risk of type 2 diabetes (T2DM), cardiovascular diseases, dyslipidemia and mortality [3]. However, whether the molecular pathogenesis of disease development in obesity is like that of aging remains unclear.

Obesity accelerates aging [4] and reduces healthy lifespan by as much as 7–9 years [5]. A proposed shared mechanism between obesity and aging is a disruption in sirtuin (SIRT) and nicotinamide adenine dinucleotide (NAD⁺) metabolism [6]. SIRTs, a family of seven enzyme-encoding proteins, function as deacetylases, desuccinylases, demalonylases, and mono-ADP ribosyltransferases, playing crucial roles in regulating metabolic and stress-response pathways [7]. Their localization within the cell varies: SIRT1 and SIRT6 are nuclear, SIRT2 is cytoplasmic, and SIRT3, SIRT4, and SIRT5 are mitochondrial. SIRTs influence the activity of key metabolic regulators, including TP53, PGC-1α, and FOXO. SIRT enzymes rely on NAD⁺ for their activity, competing with other NAD⁺ consumers like poly(ADP-ribose) polymerases (PARPs), which are involved in DNA repair, chromatin remodeling and transcriptional regulation [8]. NAD⁺ can be synthesized from tryptophan or nicotinic acid via the de novo and Preiss-Handler pathways, respectively, though the de novo pathway is inactive in AT (white adipose tissue). The main NAD⁺ biosynthesis pathway in AT is the salvage pathway, which recycles nicotinamide [9]. In animal models, NAD⁺ homeostasis is disrupted in conditions like obesity, diabetes, inflammation, aging and cancer in several tissues, including AT [9, 10].

Subcutaneous AT is a key organ regulating whole-body metabolic homeostasis [11]. In humans, AT mRNA expression of SIRTs has been shown to decrease in obesity [12–16]. Weight loss, whether diet- [13] or surgery-based [17], induces increases in SIRT1 [13, 17] and SIRT3 [17] expression in AT. NAD⁺ biosynthesis genes, such as NAMPT, NMNAT2 and NRK1, in AT are also negatively affected by obesity [14] and improve after weight loss [13]. In addition, PARP enzyme activity in human AT increases in obesity [13, 14] and decreases after weight loss [13].

While the above studies demonstrate that AT NAD⁺/SIRT metabolism is disrupted in obesity, the role of aging in regulating SIRTs in humans, particularly in AT, remains poorly studied. In the liver, both the gene expression and the enzyme activities of SIRT1 and SIRT3 are suppressed in aging [18]. In preclinical and human studies, high blood SIRT6 expression was positively associated with longevity [19]. Although AT is proposed to be one of the key functional tissues connecting SIRTs and NAD⁺ metabolism to longevity [20] no data exist on NAD⁺/SIRT metabolism in aging human AT.

We hypothesize that obesity and aging may have similar effects on NAD⁺/SIRT metabolism.

To test this, our aim was to determine: (1) the extent to which obesity and aging in humans are associated with NAD⁺/SIRT metabolism-related gene expression profiles in AT, (2) whether these processes involve epigenetic modifications, and (3) how these gene expression differences correlate with clinical features. By studying monozygotic twin pairs discordant for BMI, we controlled for genetic and shared environmental factors, isolating the effects of obesity. Additionally, comparing transcript profiles of younger versus older twin individuals allowed us to assess the impact of aging.

Materials and methods

Participants

The twin pairs included in this study were recruited from population-based longitudinal studies, FinnTwin16 (n = 2839 pairs), FinnTwin12 (n = 2578 pairs) and the Older Finnish Twin Cohort (n = 2932 pairs) [21–23]. Recruitment was made based on responses to questions regarding weight and height, to identify monozygotic pairs with at least a 2.5 kg/m² intrapair difference in BMI. Forty-nine such BMI-discordant monozygotic twin pairs (27 of which were female) were examined in detail in this study. The twins, aged 22–69 years, were divided into younger (22–38 years, 27 pairs, 17 female) and older groups (56–69 years, 22 pairs, 10 female). Eight pairs were discordant for T2DM (1 younger, 7 older), with the heavier twin affected, and 4 older pairs were concordant for T2DM. The study was approved by the Ethics Committee of the Hospital District of Helsinki and Uusimaa, and all participants provided written informed consent, adhering to the Declaration of Helsinki.

Study design

We compared the twins in two different settings to study the NAD⁺/SIRT metabolism gene expression alterations upon obesity and aging. First, we compared the co-twins to each other to analyze the effects of obesity. Secondly, we compared the younger twin individuals as a group to the older twins to analyze the effects of aging on NAD⁺/SIRT pathway gene expression. We then compared the results from the obesity and aging analyses to determine which genes were differentially regulated in obesity and aging. In the significantly differentially expressed genes in either obesity or aging, we also examined how the gene expression was associated with methylation and with clinical characteristics in twin individuals.

Anthropometric and lifestyle measurements

We measured height and weight in light clothing, body composition by whole-body dual-energy X-ray absorptiometry (Lunar Prodigy, software version 8.8, Madison, WI, USA) and in the young twins, the amount of abdominal AT and intra-abdominal AT by magnetic resonance imaging and liver fat by magnetic resonance spectroscopy [11]. Physical activity was assessed using the Baecke questionnaire [24] and smoking as described earlier in [25].

Clinical chemistry

We took fasting plasma samples for measurement of lipids, liver enzymes, and high-sensitivity C-reactive protein [11], whereafter we performed a standard 4-point (0, 30, 60, 120 min) oral glucose tolerance test with a 75 g glucose load for calculation of the homeostatic model assessment-insulin resistance index and Matsuda-index from the glucose and insulin measurements [11].

AT biopsy collection, transcriptomics and DNA methylation of NAD⁺/SIRT metabolism genes

We collected the AT biopsies during a fasting state, under local anesthesia (lidocaine) from superficial subcutaneous abdominal AT, using a surgical technique or through a needle biopsy and processed as explained previously [11]. We used a part of the fresh AT sample, digested with collagenase, for measurement and calculation of adipocyte volume and another frozen part to extract RNA and DNA using the AllPrep RNA, DNA, miRNA Universal Kit (QIAGEN, Nordic, Solletuna, Sweden) with a DNase I (QIAGEN) digestion [11]. We performed the RNA sequencing using the Illumina HiSeq2000 platform and the DNA methylome was obtained using MethylationEPIC BeadChip arrays (Illumina) [11, 26]. DNA methylation data were preprocessed and normalized as previously described [27].

The genes from NAD⁺/SIRT-related pathways were compiled using IPA (QIAGEN, Nordic, Solletuna, Sweden) and Gene Ontology GO (http://geneontology.org/), using the search terms “SIRT pathway” and “NAD⁺/SIRT metabolism”, respectively. The list was extended by several genes based on literature research [10, 13, 28–33]. The final compilation for the RNA expression analysis contained 168 genes, presented in Table S1. For the DNA methylation analyses, these genes had a total of 4351 differentially methylated CpG sites (as per Illumina annotation). For result visualization (Fig. 1), genes were grouped according to their biological roles as indicated by KEGG pathways, STRING [34] and previous literature.

Fig. 1. Venn diagram of differentially expressed NAD⁺/SIRT pathway genes of the heavier versus leaner co-twins and the older versus the younger twin pairs comparisons.

Differentially expressed genes (p < 0.05) are shown. Color coding indicates log2 fold change with gradients of blue illustrating downregulation and gradients of red illustrating upregulation in the previously stated comparisons. Biological processes have been grouped and the lines connect genes that are biologically related to each other. The bolded gray circle around the gene name in obesity or aging means FDR < 0.05. For the genes in the middle of the Venn diagram bolding on the left side means significance in obesity, whereas right side means significance in aging. A heatmap of the significantly differentially expressed genes been attached below the Venn diagram.

Statistical and bioinformatics methods

All anthropometric and clinical parameters are expressed as mean ± standard deviation for variables with normal distribution, and as median (interquartile range) for variables with non-normal distribution (based on the Shapiro–Wilk test). To determine the differences in clinical variables between co-twins, we used a paired Wilcoxon signed-rank test while the differences between the younger and the older groups were assessed using linear mixed modeling, in R software, with two-tailed p < 0.05 for significance. Group comparisons for sex, smoking status and diabetes were made using the Chi-Squared test.

We performed differential gene expression analyses (between leaner and heavier co-twins, and younger and older groups) using the R package Limma (Voom) [35]. We first investigated differences between BMI-discordant co-twins. We next investigated differences between the age groups in individual twins while considering the relatedness of the co-twins. Statistical models were adjusted for sex, diabetes status, smoking and technical factors. The leaner to heavier comparison was additionally adjusted for age and the younger to older comparison was additionally adjusted for BMI. The 168 genes were all protein-coding and had at least ten counts per gene transcript in 70% of the samples. Multiple testing correction was done by the Benjamini and Hochberg method with FDR < 0.05 considered statistically significant.

Next, we examined whether the expression of the differentially expressed genes from the heavier-leaner and younger-older comparisons was associated with the CpG methylation levels of the respective genes. We performed linear mixed modeling (R-Lmer) in the twins considered as individuals, with twinship as a random effect and sex, age, diabetes status and smoking status as fixed effects, with FDR < 0.05 for significance threshold. For each gene, we also calculated the ratios of FDR-significant CpG sites to all CpG sites related to that gene to evaluate the biological significance of the associations.

Finally, we examined whether the expression of the differentially expressed genes from the heavier-leaner and younger-older comparisons was associated with clinical features. We performed linear mixed modeling (R-Lmer) in individual twins with twinship as a random effect and adjusted the model for sex, age, diabetes status, and smoking status. FDR < 0.05 was considered significant.

Results

Differences in clinical characteristics in obesity and in aging

In order to have a targeted study approach for the effects of obesity and aging we selected 168 genes related to metabolism in AT (see “Materials and methods”). As per study design, the leaner and heavier co-twins were highly discordant for BMI (within-pair difference mean 5.2 kg/m²) and all measures of adiposity. The heavier twins had 65% more AT, twice the amount of intra-abdominal fat, 360% more liver fat and 35% larger adipocyte volume than their leaner cotwins (Table 1). Metabolically, the heavier twins exhibited greater insulin resistance, dyslipidemia, and elevated high-sensitivity C-reactive protein levels, indicating increased low-grade inflammation.

Table 1.

Anthropometric and clinical parameters of the study participants, stratified by weight (leaner vs. heavier co-twins in BMI discordant pairs, and by age group (young vs. old individuals).

Variable	Lean (n = 49)	Heavy (n = 49)	p value	Young (n = 54)	Old (n = 44)	p value
Age (years)	45.7 ± 18.0	45.7 ± 18.0	–	30.1 ± 4.67	64.8 ± 3.83	<0.001
Sex (% (n male individuals))	45 (22)	45 (22)	–	37 (20)	54 (24)	0.081
Current smoker (% (n))	27 (13)	16 (8)	0.037	28 (15)	14 (6)	0.141
T2DM (% (n))	8.2 (4)	24 (12)	0.002	2 (1)	34 (15)	<0.001
BMI (kg/m2)	25.6 (23.0–28.0)	30.8 (27.5–34.7)	<0.001	27.5 (24.3–31.5)	28.5 (25.8–31.7)	0.608
Fat percentage (%)	33.1 ± 8.97	40.9 ± 7.31	<0.001	36.6 ± 9.57	37.6 ± 8.38	0.643
Subcutaneous adipose tissue (cm3)	3300 (2450–4670)	5702 (4470–7730)	<0.001	4660 (3300–6440)	3210 (2910–3680)	0.297
Intra-abdominal adipose tissue (cm3)	606 (358–1080)	1400 (848–2310)	<0.001	834 (532–1560)	2340 (1960–2630)	0.037
Liver fat (%)	0.58 (0.41–0.89)	1.5 (0.68–6.3)	<0.001	0.80 (0.51–3.6)	1.5 (1.2–2.2)	0.684
Cell volume (pl)	424 (308–563)	574 (453–802)	<0.001	446 (355–528)	631 (461–799)	<0.001
Total cholesterol (mmol/l)	4.7 (4.1–5.2)	4.7 (4.1–5.6)	0.516	4.4 (3.9–4.8)	5.2 (4.4–5.9)	<0.001
LDL (mmol/l)	2.8 (2.2–3.3)	2.9 (2.5–3.8)	0.132	2.6 (2.2–3.0)	3.2 (2.6–4.1)	0.001
HDL (mmol/l)	1.6 (1.3–1.9)	1.4 (1.1–1.7)	<0.001	1.4 (1.2–1.7)	1.5 (1.3–1.9)	0.171
Triglycerides (mmol/l)	0.92 (0.68–1.1)	1.1 (0.93–1.3)	0.003	0.99 (0.73–1.3)	1.0 (0.83–1.2)	0.429
ALT (IU/l)	21 (16–29)	23 (17–33)	0.157	23 (16–28)	21 (17–35)	0.251
GGT (IU/l)	18 (11–21)	21 (13–30)	0.008	19 (11–26)	24 (21–25)	0.984
Hs-CRP (mg/l)	1.0 (0.65–3.0)	2.3 (0.88–4.5)	0.007	1.4 (0.73–4.1)	1.6 (0.68–3.7)	0.526
Fasting glucose (mmol/l)	5.5 (5.0–5.7)	5.8 (5.2–6.0)	0.015	5.2 (4.9–5.6)	6.0 (5.5–6.2)	0.001
Fasting insulin (mU/l)	4.8 (3.4–6.8)	7.9 (5.7–12)	<0.001	5.5 (3.4–8.7)	6.8 (5.4–12.5)	0.015
HOMA Index	1.1 (0.71–1.7)	2.0 (1.5–3.1)	<0.001	1.1 (0.82–2.0)	1.9 (1.4–3.6)	0.001
Matsuda Index	7.2 (4.8–9.5)	4.1 (2.7–5.2)	<0.001	6.7 (4.2–9.6)	4.5 (2.6–5.8)	<0.001
Physical activity (Total Baecke)	8.4 (7.0–9.2)	7.9 (6.9–9.0)	0.398	8.6 (7.1–9.3)	7.9 (6.8–8.8)	0.082

Data are based on 26 twin pairs. Data are shown as mean ± SD (normally distributed variables), median (interquartile range, for skewed variables). p values were obtained using paired Wilcoxon rank-sum test for lean-heavy comparisons, linear mixed modeling for young-old comparisons and Pearson’s Chi-Squared test for categorical variables.

BMI body mass index, LDL low-density lipoprotein, HDL high-density lipoprotein, ALT alanine aminotransferase, AST aspartate transaminase, GT glutamyl transferase, Hs-CRP high-sensitivity C-reactive protein, HOMA Index homeostatic model assessment for insulin resistance, BP blood pressure.

In the younger versus older twin comparisons, the average age difference was 34.7 years, but the age groups had similar median BMIs (Table 1). The median BMI for the younger group was 27.5 kg/m² and for the older group 28.5 kg/m², with no statistically significant difference. However, the older participants had larger adipocytes, more intra-abdominal AT, a higher total, and LDL cholesterol, and were more insulin resistant than the younger study participants. The proportion of males was 37% in the younger and 54% in the older group (p = 0.126).

Both obesity and aging were associated with downregulation of NAD⁺ metabolism genes in AT

In total, we identified 73 FDR-significant differentially expressed genes in the obesity and aging comparisons (Table S1). Most of the genes were specific to either only obesity (i.e., 53 genes out of the 73 differentially expressed genes) or only aging (15 genes) (Fig. 1), but some genes were observed as significantly differentially expressed in both conditions (5 genes). Key shared genes downregulated (FDR < 0.05) in obesity and in aging were SIRT5 and NRK1 (NAD⁺/SIRT metabolism), PARP12 (PARP family), H6PD (glucose metabolism), and ACADL (mitochondrial beta oxidation).

In both heavier co-twins and older individuals many differentially expressed genes of the NAD⁺ metabolism pathway that were downregulated in AT belonged to the Salvage and the Preiss-Handler Pathways (Fig. 2). Specifically, NMNAT2 (Log2FC = −0.36, FDR = 0.0069), NMNAT3 (Log2FC = −0.20, FDR = 0.0020) and NRK1 (Log2FC = −0.17, FDR = 0.018) showed significant downregulation in obesity (Table S1), while NADSYN1 (Log2FC = −0.28, FDR = 0.0222) and NRK1 (Log2FC = −0.29, FDR = 0.0251) were downregulated in aging (Table S1). The rate-limiting gene NAMPT was not significantly differentially expressed in either condition.

Fig. 2. Overview of differentially expressed genes in the NAD⁺ biosynthetic pathways in AT.

The leaner versus heavier co-twin comparison (i.e., obesity) is shown on the left, while the younger versus older twin comparison (i.e., aging) is shown on the right. Red color illustrates upregulation in gene expression, while blue color illustrates downregulation in gene expression. Dark red or blue colors indicate differentially expressed genes (FDR < 0.05).

Both obesity and aging are characterized with AT SIRT5 downregulation

SIRT gene expression was condition-specific, except for SIRT5, which was significantly downregulated in both heavier co-twins (Log2FC = −0.19, FDR = 0.0003) (Table S1) and older individuals (Log2FC = −0.23, FDR = 0.0144) (Table S1) compared to their respective controls (leaner co-twins and younger individuals) (Fig. 1 and Table S1). We found that specific to obesity, AT SIRT1 (Log2FC = −0.30, FDR = 0.0023) and SIRT3 (Log2FC = −0.12, FDR = 0.0129) were downregulated and SIRT6 was upregulated (Log2FC = 0.15, FDR = 0.0089) in the heavier co-twins, but none of these differences were seen in aging.

AT PARP gene expression differences were associated with both obesity and aging

PARPs, like SIRTs, require NAD⁺ as a substrate and they are considered as one of the main NAD⁺ consumers in cells [10]. We observed that the expression of PARP1, the most abundant PARP isoform, was similar in AT in leaner and heavier co-twins but was upregulated in older compared to younger participants (Log2FC = 0.11, FDR = 0.0319) (Fig. 1 and Table S1). PARP12 was upregulated in obesity and PARP16 downregulated. Most other PARPs (PARP8, PARP10, PARP12, PARP14) were consistently downregulated in AT in aging. In addition, CD38, another key NAD⁺-degrading enzyme, was upregulated in AT in aging, but not in obesity.

Upregulation of glycolysis and downregulation of mitochondrial energy metabolism-related genes was a characteristic of obesity, but not aging

Given the important role of NAD⁺/SIRT pathways on cellular energy metabolism [6], we next concentrated on genes encoding for glycolysis proteins as well as those regulating mitochondrial oxidative pathways that have been associated with NAD⁺/SIRT before. Both obesity and aging showed an upregulation of AT H6PD (Log2FC = 0.11, FDR = 0.0058, and Log2FC = 0.18, FDR = 0.0164, respectively) in the early step of glycolysis, but thereafter, the glycolysis-related genes were mostly only associated (up- or downregulated) with obesity (Fig. 1 and Table S1). In heavier co-twins, central glycolysis genes were all upregulated in AT, but in aging, the genes (ENO2, PFKM) that were significantly different between the younger and older cohorts were downregulated (Table S1). Further, we found that many AT genes responsible for the mitochondrial respiratory complexes I–V and fatty acid beta oxidation were downregulated in heavier co-twins (Fig. 1 and Table S1). Intriguingly, none of these differences were seen in the comparison between younger and older individuals.

Our findings also revealed a significant downregulation of AT NAD⁺/SIRT regulated genes involved in fatty acid metabolism (PPARa and PPARγ, PGC-1α, Log2FC = −0.40 to −0.21, FDR = < 0.0001 to 0.0009) in heavier co-twins (Fig. 1 and Table S1). These expression differences were not seen in the comparisons between younger and older individuals.

AT cellular stress and apoptosis genes upregulated in obesity but downregulated in aging

SIRTs assist in stress resistance, for example, via activating p53 (TP53) regulation [36]. The expression of AT TP53 was upregulated in heavier co-twins (Log2FC = 0.13, FDR = 0.0225) (Fig. 1 and Table S1), but not in older twins (Table S1). The genes controlled by TP53 that were differentially expressed and upregulated in heavier co-twins and covered a wide range of cellular processes, including autophagy (e.g., BECN1) and DNA repair and response to oxidative stress (e.g., E2F1). In older twins, there was downregulation of chromatin regulatory genes (e.g., GCN5) and genes associated with apoptosis, inflammation, and autophagy (e.g., STK11) in AT (Fig. 1 and Table S1).

AT gene expression and DNA methylation were associated in both obesity and aging

Since both obesity and aging are related to changes in DNA methylation, we next analyzed the associations between the significantly (FDR < 0.05) differentially expressed AT SIRT target genes and their corresponding DNA methylation sites in individual twins. NAD⁺ has been previously suggested to regulate DNA methylation via gene activation [37]. Of the 73 significantly differentially expressed AT genes in the obesity and aging comparisons, 24 genes had their expression associated with their DNA methylation with FDR < 0.05 (Table S2). Genes with the highest proportion of significant methylation sites included NDUFS2 (mitochondrial complex 1), LDHD (lactate dehydrogenase), GPD1L (glycerol-3-phosphate dehydrogenase) genes with 52%, 38%, and 27% of the total CpG sites being significantly associated with the respective gene’s expression (Table S3). Additionally, the main mitochondrial biogenesis and fatty acid oxidation regulator PGC-1α had 18% of its total CpG sites significantly associated with the PGC-1α gene expression. The top 10 genes also included GPI (19%) and PFKB (14%) with their functions in glycolysis.

AT NAD⁺/SIRT gene expression is associated with metabolic health

Finally, we analyzed the relationship between gene expression and clinical parameters in individual twins. As shown in Fig. 3 and Table S4, SIRT1 and SIRT3, NAD⁺ metabolism genes (e.g., NMNAT2 and NMNAT3), and mitochondrial oxidative phosphorylation genes associated negatively with AT amount, intra-abdominal fat amount and liver fat percentage. Furthermore, they positively associated with markers of good metabolic health, i.e., HDL, the Matsuda index and the physical activity Baecke index. The reverse was true for PARP1. The SIRT-target gene TP53 associated positively with intra-abdominal and liver fat. These results may imply that NAD⁺/SIRT and mitochondrial oxidative metabolism genes, as well as DNA damage response and stress tolerance in AT closely associate with whole-body metabolism.

Fig. 3. Associations between AT NAD⁺/SIRT pathway gene expression and clinical parameters.

Results of the linear mixed modeling between significant NAD⁺/SIRT metabolism gene expressions and clinical parameters. Color of the cell denotes standardized regression coefficients and represents the SD change in clinical measurements in relation to the SD change in gene (predictor variable) expression (blue, negative association; red, positive association), and asterisks mark p < 0.05. Genes have been grouped according to biological processes.

Discussion

We investigated the gene expression and methylation profiles of NAD⁺/SIRT pathway and its controlled processes in AT, uncovering their critical associations with obesity, aging, and related metabolic dysfunctions. Our findings reveal a consistent downregulation of SIRT5, and NAD⁺/SIRT metabolism-related genes in both obesity and aging, building on prior obesity research [13, 14, 38] while offering a novel insight into aging human AT. Notably, our study highlights key distinctions in AT between these two conditions: obesity was characterized by a downregulation of NAD⁺/SIRT regulated gene expression, whereas aging exhibited a uniform downregulation of all PARPs except PARP1, a unique pattern not observed in obesity. Furthermore, several mitochondrial oxidative metabolism and glycolysis-related genes were linked to DNA methylation and displayed significant associations with metabolic parameters, underscoring their pivotal role in the metabolic health impacts of both obesity and aging.

Among the SIRTs, SIRT5 was the only one consistently downregulated in AT in both obesity and aging, corroborating previous reports of obesity-related reductions [14] and extending this finding to aging. In murine studies, upregulation of SIRT5 has been associated with AT metabolic flexibility and subcutaneous AT browning by enhancing C/EBPβ and UCP1 expression [39] and by inducing brown adipocyte differentiation and activation of browning genes [40]. The role of SIRT5 in humans remains to be further studied. Other sirtuins showed significant alterations in AT in obesity: SIRT1 and SIRT3 were downregulated, while SIRT6 was upregulated. Lower SIRT1 and SIRT3 expression in human AT in obesity has been associated with metabolic AT dysfunction, including impaired mitochondrial function, fatty acid oxidation and insulin sensitivity, along with increased inflammation [12, 14, 17]. Low SIRT1 activity may also impair adipogenesis, as shown in mice [41]. Consistent with this, we observed downregulation of key lipid metabolism regulator genes (PGC-1α, PPARa, PPARγ and PCK1) in obesity, while these changes were absent in aging. However, adipocyte hypertrophy, a hallmark of dysfunctional AT, was observed in both obesity and aging in our study, consistent with previous findings [42]. Interestingly, SIRT6, known for its roles in genomic stability and inflammation reduction, was upregulated in AT in obesity despite prior reports of decreased protein levels in AT in obesity in humans and animal models [43]. In contrast, no differences in SIRT6 expression were observed between age groups, despite its association with longevity in animal models [44]. These findings highlight the need for further research into the roles of SIRTs in obesity- and aging-related processes in human AT.

Among the PARPs, PARP1 was upregulated in aging, unlike the other PARPs. PARP1 activity has been shown to increase upon aging in preclinical studies in worms and in non-adipose tissues in mice [45]. PARP1 has been shown both to promote and protect from aging in animal models [46, 47]. PARP1 participates in genomic stability by regulating the cell cycle and chromatin structure. It also participates in DNA repair and telomere maintenance. However, PARP1 has also been shown to increase inflammation in animal models and participate in the development of neurodegenerative diseases in humans [46, 47]. Thus, as discussed previously [46, 47], this dual nature makes interpreting PARP1 upregulation in aged adipose tissue complex, as it might represent a compensatory response to maintain genomic stability, but it could also contribute to increased inflammatory processes. Therefore, further mechanistic studies are needed to understand this finding.

In obesity, the NAD⁺/SIRT-regulated genes involved in the TCA cycle and mitochondrial respiratory complexes were downregulated in AT, consistent with prior studies [11, 13, 14]. However, these changes were not observed in AT in aging. While mitochondrial function has been proposed as a major regulator of aging in AT [48], our results suggest that reduced gene expression of mitochondrial oxidative metabolism in human AT is primarily linked to obesity, not aging. Moreover, we found that the NAD⁺/SIRT-regulated glycolysis genes were upregulated in AT in obesity.

Obesity, but not aging, was associated with upregulation of AT NAD⁺/SIRT-regulated genes promoting cellular stress responses and downregulation of genes inhibiting these responses. SIRT-mediated upregulation of TP53, a gene linked to various processes such as oxidative stress, DNA damage, and apoptosis, is commonly associated with the development of metabolic diseases [49].

SIRTs, as epigenetic regulators, play crucial roles in modulating cellular metabolism [50]. Our exploratory results showed significant associations between the expression of NAD⁺/SIRT metabolism genes and their corresponding DNA methylation. Notably, genes related to mitochondrial metabolism and glycolysis showed significant gene expression-methylation links. In our earlier work, we also found differentially methylated CpG sites associated with gene expression in AT from BMI-discordant twin pairs [51]. In both the present and the previous study, no consistent relationships were observed between hypermethylation of gene bodies or hypomethylation of promoters and gene upregulation. Interestingly, insulin resistance has been proposed to result partly from epigenetic regulation of mitochondrial metabolism- and glycolysis-related genes, with SIRTs as key mediators [52]. Obesity-related epigenetic age acceleration in blood has been suggested to occur partly due to insulin resistance [53]. Moreover, obesity has been shown to accelerate epigenetic aging of metabolically active tissues such as visceral AT [54].

Finally, we associated the NAD⁺/SIRT metabolism genes that were significantly differentially expressed in AT either in obesity or in aging, with clinical parameters, using individual twins. Previously, it has been shown that sex affects gene expression in AT, such as adipogenesis, oxidative phosphorylation, and fatty acid metabolism [55], but we did not observe sex differences in key NAD⁺/SIRT metabolism genes. We found that in addition to SIRT1 and SIRT3, mitochondrial oxidative metabolism genes correlated negatively with adiposity measures (including intra-abdominal and liver fat), insulin resistance, dyslipidemia, and inflammation. Consistent with these findings, high AT SIRT expression has been associated with favorable outcomes, including improved glucose tolerance, lower fat percentage, and healthier lipid profiles [56–58]. Conversely, AT PARP1 expression was positively associated with markers of liver fat accumulation and insulin resistance. Collectively, these findings highlight the central role of AT SIRT1, SIRT3, and mitochondrial function in the development of obesity-related diseases [59], while also implicating PARP1 in the progression of metabolic dysfunction.

The strength of this study lies in the use of a unique, deeply phenotyped dataset of monozygotic twin pairs, enabling the investigation of NAD⁺/SIRT metabolism in obesity while controlling for genetic and early environmental factors. This study also addresses aging effects across two distinct age groups. As most NAD⁺/SIRT metabolism studies have focused on animal models, our findings provide valuable insights into human AT and its relevance to metabolic health. However, the study is limited by the lack of data on protein levels, enzymatic activities, and NAD⁺ metabolites. As gene expression may not capture functional activity, these findings warrant cautious interpretation. Unfortunately, there was not enough study material to perform these functional measurements. Another limitation of our study is its cross-sectional nature, which prohibits causal inferences. Although we can exclude genetic and shared early environmental factors from the obesity associations, residual confounding is possible. Finally, while sex-related differences in AT gene expression are known to exist [55], no such differences were observed in this study.

Conclusions

Our study demonstrates that both obesity and aging are associated with lower expression of SIRT5 and genes related to NAD⁺/SIRT metabolism in AT. Obesity uniquely exhibited decreased expression of NAD⁺/SIRT regulated mitochondrial oxidative metabolism genes and upregulated cellular stress-related genes, whereas aging showed distinct downregulation of most PARP isoforms, except for higher PARP1 expression. Expression of several NAD⁺/SIRT-regulated genes linked to mitochondrial oxidative metabolism and glycolysis was associated with DNA methylation. Our association analyses suggest that in AT higher expression of SIRT1, SIRT3, and mitochondrial genes plays a protective role against obesity-related diseases, while PARP1 expression may contribute to metabolic dysfunction. Future mechanistic studies are needed to clarify the complex role of NAD⁺/SIRT metabolism in obesity- and aging-related diseases, particularly in relation to mitochondrial function, cellular stress, and metabolic health.

Author contributions

Conceptualization was done by KHP, BvdK, and MM. Writing was done by HL, KHP, BvdK, SH, and MM. The investigation and data collection were performed by KHP and SH. PP generated the RNA-seq data, and MA, AK, SHTL, and PP participated in RNA-seq analyses. MO generated the DNA methylation data, and AHe participated in the DNA methylation data analyses. The formal statistical and bioinformatics analyses were conducted by MM and HL. The visualization was done by HL with the help of BvdK, MM, and SH. AHe, JL, JK, and P-HG participated in the imaging of the twins. HL, BvdK, MM, SH, MO, TT, EP, and KHP contributed to data interpretation. All authors contributed to the manuscript revision, read, and approved the final manuscript. The work was supervised by KHP, SH, BvdK, and MM.

Funding

This study was supported by the Finnish Medical Foundation (HAKL, KHP, SH); the Finnish Medical Association (Finska Läkaresällskapet) (HAKL); Orion Research Foundation sr (BWvdK, SH); Paulo Foundation (SH, KHP); Paavo Nurmi Foundation (SH), Helsinki University Hospital (SH, KHP); Novo Nordisk Foundation #NNF24SA0090438 (BWvdK), NNF23SA0083953 (SH) and NNF25OC0100827 (SH), NNFLMSA898RSYR, NNF24OC0091683, NNF20OC0060547, NNF17OC0027232, and NNF10OC1013354 (KHP); Research Council Finland 361956 and 338417 (SH); 315589 and 345867 (TT); 297908 and 328685 (MO); 286359, 314455, 335445, and Profi6 336449 (EP); 265240, 263278, 308248, 312073, 352792 (JKa); 266286, 272376, 314383 and 335443 (KHP); the Finnish Diabetes Research Foundation (KHP, SH); European Foundation for the Study on Diabetes EFSD (SH), HHMI Gilliam Fellowship (MM); National Institute of Health (T32HG002536) (MM), Gyllenberg Foundation (KHP); Finnish Foundation for Cardiovascular Research (KHP); the Sigrid Jusélius Foundation (MO, JKa, KHP); Government Research Funds (KHP, SH); the University of Helsinki (KHP).

Data availability

RNA sequencing data are part of the ‘Twin Study’ and are deposited with the Biobank of the Finnish Institute for Health and Welfare (https://thl.fi/en/web/thl-biobank/for-researchers/sample-collections/twin-study) with the identification number THLBB2021_001. For details on accessing the data, see https://thl.fi/en/web/thl-biobank/for-researchers/application-process. All bona fide researchers can apply for the data.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41366-025-02007-w.