Variation in the Adiponutrin Gene Influences Its Expression and Associates With Obesity

Obese, nondiabetic Scandinavian Caucasian subjects (234 in total; 63 men and 171 women, median age 42 years [33–52], and BMI 40.3 kg/m² [35.5–45.3]) from the obesity outpatient clinic at the University Hospital MAS in south of Sweden were randomly and individually matched for age (±5 years) and sex with nondiabetic, normal-weight subjects participating in the Botnia study (63 men and 171 women; median age 43 years [34–53], and BMI 22.6 kg/m² [21.3–23.8]) (18–20). Diabetes status was based on a previous diagnosis, on-going hypoglycemic medication, and fasting plasma glucose according to World Health Organization criteria. Biopsies of subcutaneous and visceral abdominal fat for mRNA expression studies were obtained from 36 nondiabetic, obese patients undergoing bariatric surgery (4 men and 32 women, median age 36 years [29–48], BMI 41.6 kg/m² [37.7–45.1], and waist circumference 118 cm [106–129]) and from 5 nondiabetic, nonobese patients undergoing abdominal elective surgery (2 men and 3 women, mean age 72 years [51–77], and BMI 26 kg/m² [24–29]); all of these subjects were fasted overnight but were in a stable metabolic state (19). DNA was available from 32 of these obese subjects, and data for homeostasis model assessment (HOMA) calculation was available for 23 subjects. Needle biopsies of abdominal subcutaneous adipose tissue for lipolysis studies were obtained from 130 obese patients (33 men and 97 women, median age 38 years [29–48], and BMI 39.4 kg/m² [35.3–43.5]) and 100 lean subjects (38 men and 62 women, median age 36 years [30–47], and BMI 23.4 kg/m² [21.5–24.2]). They were consequently recruited either by local advertisement or from the obesity clinic at the hospital to study the influence of genetic variance on adipocyte lipolysis regulation. They were all healthy and, except for obesity, free of metabolic disease. All were living in the Stockholm area and at least second generation Scandinavian. None was completely sedentary or involved in athletic performances. None had undergone a slimming effort or experienced a change (>1 kg) in body weight during the last 6 months before study, according to self-report. DNA was available for all patients included in the lipolysis study. All patients gave their written consent, and the local ethics committees approved the study.

DNA was extracted from whole blood by conventional methods as previously described or by using QiaGen MiniPrep (QiaGen, Hilden, Germany) at the DNA/RNA Genotyping Lab, SWEGENE Resource Center for Profiling Polygenic Disease, Lund University, Malmö University Hospital (Malmö, Sweden). Seven single nucleotide polymorphisms (SNPs) were chosen from the National Center for Biotechnology Information and HapMap databases (http://www.ncbi.nlm.nih.gov and http://www.hapmap.org, respectively) positioned ∼3–5 kb apart starting in the 5′ promoter region upstream of the transcription start in the 23.8-kb-long gene coding for ADPN to the untranslated region 3′ of the gene (Fig. 3). Two additional SNPs were chosen for genotyping after analyzing HapMap data using the Tagger program (predicting tagSNPs, SNPs that could explain most part of the variation in a gene) (Paul de Bakker; http://www.broad.mit.edu/mpg/tagger/). Sequences for primers and probes will be given on request. The chosen SNPs were genotyped using Assay by Design from Applied Biosystems (Foster City, CA) for TaqMan allelic discrimination performed with ABI 7900HT in 5-μl reaction volumes with 5 ng DNA and Universal TaqMan 2× PCR MasterMix according to manufacturer’s recommendations (Applied Biosystems). Success rate for all genotyping was >99%, and there was no deviation from Hardy-Weinberg equilibrium (Supplement 1, which is detailed in the online appendix [available at http://diabetes.diabetesjournals.org]).

Total RNA from biopsies was extracted using TRI REAGENT according to manufacturer’s recommendations (Sigma, St. Louis, MO). Preadipocytes of human origin were differentiated and treated as previously described (19,21). The adipocytes were treated for 4 h with glucose (1, 5, and 25 mmol/l), palmitate (C16:0) (0.2, 0.5, and 1.6 mmol/l), and oleate (C18:1) (0.2, 0.5, and 1.6 mmol/l) in presence or absence of 1 nmol/l insulin, duplicates for each condition. RNA was extracted using RNeasy mini kit (QiaGen). cDNA was synthesized using Superscript II RNase H⁻ Reverse Transcriptase and random hexamer primers (Life Technologies, Frederick, MD). Samples were analyzed with real-time PCR using the ABI 7900HT sequence detection system in 10-μl reaction volumes with 50 ng cDNA and Universal TaqMan 2× PCR MasterMix according to manufacturer’s recommendations (Applied Biosystems). All samples were run in triplicate, and data were calculated using the standard curve method and expressed as a ratio to the endogenous control cyclophilin A (arbitrary units). Cyclophilin A was chosen because this housekeeping gene transcript has been shown to be stable in microarray studies comparing expression in adipose tissue between obese and nonobese subjects (22). Primers and probe for ADPN (NM_025225) were designed using Primer Express software (Applied Biosystems): forward primer, 5′-ATTTGCAACTTGCTACCCATTAGG-3′; reverse primer, 5′-CATGTCACCAGTCTCTGGACAATC-3′; and probe, 5′-(FAM)TAATGTCTTATGTAATGCTGCCCTGTACCCTGC(TAMRA)-3′ (MWG-Biotech, Edsberg, Germany).

Adipose samples (1–2 g) were obtained by needle biopsy from the abdominal subcutaneous area under local anesthesia. The adipose tissue was collagenase treated, and isolated fat cells were collected and subject to lipolysis experiments as previously described (23,24).

Allele frequencies were compared using the McNemar test and genotype frequencies using χ² statistics. We used permutation testing (10.000 permutations) to correct for multiple testing. Wilcoxon’s signed rank test was used for paired comparisons, Mann-Whitney U test was used for unpaired comparisons, and Spearman correlation was used for estimating relationships between variables. Kruskal-Wallis was used for testing the difference in expression between genotype groups. ANOVA was used to test for difference in expression in the adipocyte cultures when comparing more than two treatments. A P value <0.05 was considered statistically significant. Data are presented as median with interquartile range (25th–75th percentile) in parentheses or as means ± SE. All statistical calculations were performed using Number Cruncher Statistical Systems 2000 software (NCSS, Kaysville, UT).

We examined the ADPN mRNA expression in adipose tissue biopsies obtained from fasted nonobese or obese subjects undergoing cholecystectomy or bariatric surgery (Fig. 1A). ADPN mRNA expression levels were significantly higher in obese compared with nonobese humans in both subcutaneous (2.3- ± 0.2-fold, P = 0.026) and visceral adipose tissue (2.9- ± 0.3-fold, P = 0.006). The expression levels in visceral fat were significantly higher than in subcutaneous fat in the obese subjects (P = 0.027). Visceral and subcutaneous ADPN mRNA levels were correlated to each other (R = 0.46, P = 0.003; n = 41). Insulin resistance characterized the obese participants in the biopsy study (HOMA 2.91 [2.27–4.52], median age 35 years [29–49], and BMI 41.4 [37.6–45.2], n = 23). These 23 subjects were not different from the other participants in the biopsy study concerning age, sex, and BMI (data not shown). Visceral but not subcutaneous ADPN mRNA expression was inversely correlated to fasting plasma insulin (R = −0.47, P = 0.02, n = 24; Fig. 1B) and HOMA (R = −0.43, P = 0.04, n = 23). Accordingly, visceral ADPN mRNA was significantly lower (1.81 [0.97–2.37] vs. 2.73 [2.00–4.71], P = 0.03) in subjects with HOMA values above median (>2.91) compared with subjects with HOMA below median (<2.91), respectively. There was no correlation between BMI and either subcutaneous (R = 0.02, P = 0.9) or visceral (R = −0.08, P = 0.6) ADPN mRNA expression. The nonobese subjects were significantly older than the obese subjects, but we found no correlation between ADPN mRNA expression and age (data not shown).

We next examined the regulation of ADPN mRNA expression in human adipocytes in vitro. The circulating levels of glucose, free fatty acids, and insulin are all elevated in the postprandial state, which resembles an insulin-resistant state. To differentiate the effect of carbohydrates and fatty acids from that of insulin, we treated human cultured adipocytes with various concentrations of glucose, palmitate, and oleate in the presence or absence of insulin for 4 h (Fig. 2). Increasing the glucose concentration alone did not affect ADPN mRNA expression significantly. Adding insulin together with glucose resulted in increased expression (P < 0.05) with a significant trend toward increased expression with increasing glucose concentrations (ANOVA, P = 0.025; Fig. 2A). By contrast, inclusion of free fatty acids had no effect on ADPN mRNA expression (ANOVA, P = 0.78 and 0.84 for the effect of insulin in the presence of palmitate and oleate, respectively; Fig. 2B and C). Time-course experiments performed in three independent cell cultures at 5 mmol/l glucose suggested that the effect of insulin on ADPN mRNA is rapid, seen within 1 h, and transient, returning to basal levels within 6 h (data not shown).

Because both disease manifestation and gene expression may be influenced by genetic variation, we performed a case-control association study to examine whether variants in the adiponutrin gene was associated with obesity. Nine SNPs in the adiponutrin gene were genotyped in an obesity case-control cohort matched for age and sex (Fig. 3). All SNPs were in Hardy-Weinberg equilibrium, and they were equally distributed among males and females (data not shown). Two SNPs showed significant association with obesity: the tagSNP rs2072907, located in intron 5 (P_Allele = 0.0013, P_Genotype = 0.0068), and rs1010022, located in intron 7 (P_Allele = 0.0023, P_Genotype = 0.0056). The common alleles of both SNPs were overrepresented in obese compared with lean subjects (Tables 1 and 2,1,2). The difference in allele frequency between obese and lean subjects was significant after correcting for multiple testing (by permutation) for both rs2072907 (P = 0.015) and rs1010022 (P = 0.021). A third SNP, rs2076211, located in intron 5, also showed allelic and genotype association with obesity, the common allele being overrepresented in obese subjects (P_Allele = 0.016, P_Genotype = 0.046). However, the difference in allele frequency was not significant after correcting for multiple testing (P = 0.13). SNP rs738409 (Ile148Met), located in exon 3, showed significant difference in allele but not genotype frequency between obese and lean subjects, the common allele being overrepresented in obese subjects (P_Allele = 0.023, P_Genotype = 0.12). But again, the difference in allele frequency was not sustained after correcting for multiple testing (P = 0.19). The remaining five SNPs (rs9626055, located in the 5′ region upstream of the putative ADPN mRNA transcription initiation site; rs2076212, a second coding SNP located in exon 2 leading to an amino acid substitution (Gly115Cys); rs5764034, located in intron 3; the tagSNP rs3810622, located in intron 8; and the 3′-untranslated region SNP, rs2294919) showed no significant association with obesity (Tables 1 and 2).

Haplotype analysis was performed using the Haploview software (25). Haplotype blocks were determined by the method based on D′ values (26). Two haplotype blocks, the first consisting of rs738409 and rs5764034, and the second consisting of rs2076211, rs2072907, rs1010022, rs3810622, and rs2294919, showed strong linkage disequilibrium, with high pairwise magnitude of linkage disequilibrium (D′ > 0.8) and a logarithms of odds score >2.0. The linkage disequilibrium between the two obesity-associated SNPs rs2072907 and rs1010022 is very high (r² = 0.96 and D′ = 1.00 [0.98–1.00]). Two of the haplotypes, one in the first block (rs738409G/rs5764034C) and the other in the second block (rs2076211T/rs2072907C/rs1010022G/rs3810622T/rs2294919C), were significantly more common among the control subjects compared with the obese cases (P = 0.023 and 0.012, respectively). The significance of these findings disappeared when adjusting for multiple testing (Supplement 2, which is detailed in the online appendix).

The obesity-associated ADPN SNPs are located in introns that may contain regulatory elements for gene expression (27). We therefore genotyped the participants of the biopsy study and re-examined ADPN mRNA according to ADNP genotype (Fig. 4). None of the biopsy study subjects were homozygous for the less common variant alleles. Homozygous carriers of the obesity-associated alleles for SNPs rs2072907 and rs1010022 had significantly lower ADPN mRNA expression in subcutaneous tissue (P = 0.043 and 0.018, respectively; Fig. 4A and B). There was no significant difference in visceral expression between genotype carriers. The decrease in subcutaneous compared with visceral ADPN mRNA levels noted in all obese subjects (Fig. 1A) was restricted to homozygous carriers of the risk alleles for SNPs rs2072907 (P = 0.020) and rs1010022 (P = 0.023; Fig. 4A and B).

Because ADPN mRNA expression differed between carriers of the obesity-associated SNPs and because increased gene expression may indicate increased protein levels and function, we next explored whether adipocytes from different genotype carriers behave differently with respect to in vitro lipolysis. As shown in Table 3, nonstimulated lipolysis was significantly higher in obese carriers of the obesity-associated SNPs (P = 0.025 for both rs1010022 and rs2072907). No such differences were observed between nonobese genotype carriers. There were no differences in maximal lipolysis in response to isoprenaline or norepinephrine (Table 3) nor any differences concerning half-maximal doses for these lipolytic agents or lipid content per cell between different genotype carriers (data not shown).

In this study, we found that two common variants in the gene encoding adiponutrin, a novel adipocyte lipase/transacylase, show association to obesity. ADPN mRNA expression levels were increased in adipose tissue of obese subjects, and the degree of upregulation was influenced by ADPN genotype and, in the visceral adipose tissue, by insulin sensitivity. In vitro experiments using human adipocytes confirmed the nutritional influence on ADPN mRNA regulation and suggest that the upregulation of ADPN expression seen in the presence of carbohydrates requires insulin. The obesity-associated ADPN mRNA increase may in other words be secondary to the hyperinsulinaemia that often characterize obesity. These findings suggest a possible protective effect of adiponutrin during obesity. Obesity-associated ADPN SNPs were associated with lower ADPN mRNA expression and increased basal adipocyte lipolysis, suggesting that the predominant role of adiponutrin in obese subjects may be anabolic lipogenic, i.e., reflecting the transacylase activity, rather than catabolic lipolytic.

Until recently, HSL was the only known triacylglycerol lipase in the adipocyte (4,28). Surprisingly, HSL knockout mice retained a lean phenotype, residual lipase activity, and accumulation of diacylglycerol rather than triacylglycerol, suggesting the existence of important complimentary lipases (29–32). Adiponutrin (identical to calcium-independent phospholipase A2 ε [iPLA2ε]), along with two other recently discovered lipases, desnutrin (identical to adipose triglyceride lipase or iPLA2ζ) and GS-2 (identical to iPLA2η), are prime candidates for this lipase activity (5–8). An updated view of adipocyte lipolysis is emerging in which one or more of these proteins are responsible for the first step of lipolysis, the triacylglycerol hydrolysis step; HSL is responsible for the second diacylglycerol hydrolysis step; and monoglyceride lipase is responsible for the final step (4). Although the three complimentary lipases show similarities, such as their predominant or exclusive expression in adipose tissue and induction of expression during the early stages of adipocyte differentiation, their respective genes seem to be differentially regulated, and it is possible that they serve different functions. For example, the mRNA expression of the gene encoding iPLA2ζ (PNPLA2) is transiently induced during fasting, as would be expected for an enzyme involved in energy release, whereas that of ADPN is decreased. Upon refeeding, ADPN mRNA expression is then strongly induced by carbohydrates, shown either by a high-carbohydrate diet given to rats or by exposing 3T3-L1 mouse adipocytes to glucose (5,7,9–12). In addition, although ADPN mRNA is increased in rodent obesity, the expression of PNPLA2 seems to be downregulated (5,7). Although ADPN mRNA expression increased with increasing concentrations of glucose in our cultured human adipocytes, a marked increase in the expression of ADPN was not seen until after adding insulin. This further differentiates the regulation of ADPN from that of PNPLA2, which is downregulated by insulin in mouse 3T3-L1 cells (33). Meanwhile, the inclusion of the fatty acids palmitate and oleate had no effect on ADPN mRNA expression analogous to the lack of effect of a high-fat diet on Adpn expression in rats (9). Adipose tissue Adpn mRNA levels in mice follow the serum levels of insulin, but in contrast to our findings, insulin itself only had a minor effect on Adpn mRNA expression in 3T3-L1 cells (5). It is thus possible that adiponutrin gene expression is subject to species-specific regulation. Importantly, however, 3T3-L1 mouse adipocytes differ in many ways from the primary human adipocytes used here (21).

We conclude from these experiments that the effect of carbohydrates on ADPN mRNA expression is dependent on insulin. To date, the promoter of the adiponutrin gene has not been analyzed, but it is likely that insulin responsive transcription factor(s) plays an important role. For example, Baulande et al. (5) have shown that the expression of adiponutrin in 3T3-L1 cells and rat adipose tissue follows the same pattern as the expression of adipocyte determination and differentiation factor–1/sterol regulatory element-binding protein–1c, a transcription factor that mediates important insulin effects (34) both during 3T3-L1 adipocyte differentiation and in the white adipose tissue of rats during feeding, fasting, and refeeding conditions (5).

In vivo, obese subjects with decreased insulin sensitivity had decreased ADPN expression, at least in the visceral adipose tissue. The limited number of participants in this study makes the interpretation of correlations suggestive, but our results are supported by findings in the first human report on adiponutrin. Liu et al. (13) found significantly higher insulin sensitivity in subjects with increased expression of ADPN mRNA in subcutaneous adipose tissue and an inverse correlation between ADPN mRNA and the level of glycemia and/or indirect measures of insulin resistance. We also found increased ADPN mRNA levels in both subcutaneous and visceral adipose tissue from obese subjects. Although a similar trend observed by Liu et al. (13) in subcutaneous adipose tissue failed to reach statistical significance (visceral adipose tissue was not available), they did find a 36 and 58% decrease in ADPN mRNA expression after caloric restriction after 2 and 21 days, respectively.

Genome-wide scans from both humans and rats have shown association and linkage to the region harboring the gene encoding adiponutrin with type 2 diabetes, plasma insulin levels, and body weight (3,14–17). Because no previous study of the role of genetic variation in this gene on obesity-related variables has been performed, we selected nine SNPs covering the entire gene. We found two intronic SNPs showing association to obesity. SNPs are not independent of each other, and historical recombination results in patterns of SNPs inherited together in the form of haplotypes (35). Haplotypes constructed from data available from the public database HapMap showed high recombination in the 5′ end of ADPN, whereas SNPs in the 3′ end are inherited together in a large haplotype block (http://www.hapmap.org). Our present haplotype analysis shows the same structure. Haplotypes containing the nonobese alleles (rs738409G, rs2072907C, and rs1010022G) were more common among the lean control subjects, but the significance of these findings disappeared when adjusting for multiple testing. By necessity, constructing haplotypes will create subgroups and decrease the statistical power of the analysis. Until replicated in other populations, these results should be viewed with caution. TagSNPs were predicted to capture all or most of the variation in the large 3′ haplotype (Paul de Bakker; http://www.broad.mit.edu/mpg/tagger/). The haplotype contained two intronic tagSNPs (rs2072907 and rs3810622) of which only rs2072907 was associated with obesity. It was also the only obesity-associated tagSNP that showed significant differences in adipose tissue ADPN mRNA levels and adipocyte lipolysis, supporting the contribution of genetic variation in ADNP to obesity. Although our study populations are genetically homogenous and carefully matched and phenotyped, our findings need to be confirmed in larger populations. One must also consider the possibility that the SNPs described here are nonfunctional and only in linkage disequilibrium with the causal variant(s).

Our findings concerning insulin regulation of ADPN mRNA expression are analogous to the strong enhancing effect of insulin and glucose on lipogenic enzymes such as fatty acid synthase and acetyl-CoA reductase (36). In fact, every aspect of ADPN/Adpn expression is reminiscent of a lipogenic enzyme (37). It was therefore interesting to note that basal lipolysis was increased in adipocytes from obese carriers of ADPN SNPs associated with obesity and low mRNA expression, possibly reflecting a lack of adiponutrin transacylase activity. A transacylase uses mono- and diacylglycerols as acyl donors/acceptors instead of using glycerol phosphate, dihydroxyacetone phosphate, or monoacylglycerol as initial acyl acceptors in the acyl-CoA–dependent enzymatic synthesis of triglycerides (6,38). It should be pointed out that such an acyl-CoA–independent process has been described in intestinal mucosa cells but that its existence and relative contribution in human adipocytes is largely unknown (4,6,38). These data provide a feasible mechanism by which the adiponutrin gene is associated with obesity. It is possible that the upregulation of ADPN, and hence the ability to increase triacylglycerol synthesis through transacylation, is an appropriate response aimed at compensating for the imbalance in triglyceride homeostasis. Such a protective mechanism would reduce the risk of systemic increases in free fatty acids and their consequences. Depending on the concomitant lipase activity of adiponutrin, other non-HSL lipases, and HSL itself, they may form a futile cycle to dissipate an increased caloric load (28). In obese, insulin-resistant subjects or in carriers of the associated ADPN risk alleles in whom the expression is insufficiently increased, the net result would be an increased basal rate of lipolysis. The finding that basal lipolysis was increased in obese but not in nonobese carriers of the risk allele further strengthens the functionality of this association. Importantly, there were no apparent differences in adipocyte cell size or lipid content per cell between genotype carriers. Such differences may otherwise have biased the results, because it is known that obese subjects have larger adipocytes that show increased basal lipolysis (24). The fact that the ADPN alleles that were associated with obesity were also associated with decreased adipose tissue ADPN expression and increased nonstimulated adipocyte lipolysis suggests that it may be worthwhile to examine involvement in phenotypes associated with obesity rather than obesity per se. Certainly, the possible role of adiponutrin in the pathogenesis of the metabolic syndrome and type 2 diabetes should be subject to future investigation.

In conclusion, genetic variation in ADPN is associated with obesity, ADPN mRNA expression levels, and basal adipocyte lipolysis. The gene is upregulated in obese subjects, but the upregulation is dependent on both genotype and insulin. The regulation of ADPN is reminiscent of a protein involved in lipogenesis rather than lipolysis. We therefore hypothesize that this is the main action of adiponutrin under conditions of energy excess such as in obesity. The relative contribution of the lipase and transacylase activities of adiponutrin requires further investigation. This is particularly urgent concerning the transacylation pathway of triacylglycerol synthesis in adipocytes. This is a completely new concept that may turn out to be important both in the understanding and treatment of obesity-related conditions such as type 2 diabetes and cardiovascular disease.

This investigation has received funding from the Swedish Research Council, Novo Nordisk Foundation, the Crafoord Foundation, Malmö University Hospital Foundation, the Albert Påhlsson Foundation, the Lundberg Foundation, the Diabetes Association in Malmö, Region Skåne, ALF, the Magnus Bergvall Foundation, the Fredrik and Ingrid Thurings Foundation, the Borgströms Foundation, the Lars Hjerta Foundation, and the Diabetes Program Lund University.

We are greatly indebted to the study subjects for their participation.