Concerted Action of Aldehyde Dehydrogenases Influences Depot-Specific Fat Formation

Autocrine retinoic acid production by aldehyde dehydrogenase-1 enzymes establishes differences between fat depots through regulation of transcription factors ZFP423 and PPARγ.

Abstract

Vitamin A metabolite retinoic acid (RA) regulates life-sustaining differentiation processes and metabolic homeostasis. The aldehyde dehydrogenase-1 (Aldh1) family of enzymes (Aldh1a1, a2, and a3) catalyzes RA production from retinaldehyde and thereby controls concentrations of this transcriptionally active metabolite. The hierarchy of Aldh1 functions in adipose tissue has not been elucidated. We hypothesized that Aldh1 enzymes produce endogenous RA and regulate adipogenesis and fat formation in a fat depot-specific manner. We demonstrate that adipogenesis in vitro is accompanied by RA production generated primarily by Aldh1a1. In Aldh1a1-deficient adipocytes, adipogenesis is impaired compared with wild-type adipocytes due to markedly reduced expression of PPARγ regulated through zinc-finger protein 423 (ZFP423)-dependent mechanisms. These effects were recovered to some extent either by RA stimulation or overexpression of any of the Aldh1 enzymes in Aldh1a1^−/− cells arguing that Aldh1a1 plays a dominant role in autocrine RA production. In vivo studies in C57/BL6 and Aldh1a1^−/− mice on a regular diet revealed that multiple Aldh1 enzymes regulate differences in the formation of sc and visceral fat. In Aldh1a1^−/− mice, visceral fat essentially lacked all Aldh1 expression. This loss of RA-producing enzymes was accompanied by 70% decreased expression of ZFP423, PPARγ, and Fabp4 in visceral fat of Aldh1a1^−/− vs. wild-type mice and by the predominant loss of visceral fat. Subcutaneous fat of Aldh1a1^−/− mice expressed Aldh1a3 for RA production that was sufficient to maintain expression of ZFP423 and PPARγ and sc fat mass. Our data suggest a paradigm for regulation of fat depots through the concerted action of Aldh1 enzymes that establish RA-dependent tandem regulation of transcription factors ZFP423 and PPARγ in a depot-specific manner.

Vitamin A and its metabolites regulate embryogenesis and differentiation of various cells in adults, including adipocytes (1, 2). These functions are mediated by the three major vitamin A metabolites: retinol, retinaldehyde, and retinoic acid (RA) (3, 4). These metabolites are ligands for nuclear RA receptors (RAR), and exert specific transcriptional effects (1, 2, 4, 5). Among retinoids, RA has the highest affinity for binding to RAR (6). Binding of RA activates RAR heterodimerization with retinoid X receptor (RXR), releases transcriptional corepressors, and recruits transcriptional coactivators (3). Heterodimers of RAR with RXR bind to a RA response element (RARE) and regulate transcription of target genes (3). During adipogenesis, stimulation with RA represses the transcription factors CCAAT/enhancer-binding proteins (C/EBP)-α and -β by an RAR-dependent mechanism that prevents peroxisome proliferator-activated receptorn γ (PPARγ) expression and inhibits adipogenesis (2, 7). In contrast, in preadipocytes expressing PPARγ, RA stimulation enhances adipogenesis (2, 7). Retinaldehyde can act as a PPARγ inhibitor and suppress adipogenesis throughout differentiation (8). Although the role of retinoids in transcriptional control of adipogenesis is widely recognized, the mechanisms of retinaldehyde and RA production during the course of adipogenesis are far less clear.

Retinaldehyde is generated from retinol by alcohol dehydrogenases (Adh) and retinol dehydrogenases (Rdh) (1, 9). RA is produced solely from retinaldehyde by the cytosolic aldehyde dehydrogenase-1 (Aldh1, also known as Raldh) family of enzymes (1). Thereby, Aldh1 enzymes control concentrations of both retinaldehyde and RA (1). The expression of vitamin A-metabolizing enzymes is under spatiotemporal control and varies among different tissues (1, 9). Both adipocytes and fat tissue express retinaldehyde-generating Adh1 and RA-generating Aldh1a1 (Raldh1) enzymes (8). In the absence of Aldh1a1, both adipogenesis in vitro and diet-induced fat formation in vivo are markedly impaired (8). These findings in Aldh1a1^−/− adipocytes and mice were unforeseen, given the redundant function of the other members of the Aldh1 family (Aldh1a2 and a3) with respect to retinaldehyde oxidation and RA production. Compared with other enzymes in the Aldh1 family, Aldh1a1 has many unique characteristics. Aldh1a1 is expressed in abundance in many tissues, where it has been shown to generate RA in vivo (10, 11); however, this enzyme can potentially use other aldehyde substrates (12). Aldh1a2 and a3 enzymes are much more specific for retinaldehyde as a substrate (13). Aldh1a2 and a3 have distinctly regulated expression patterns and functions in specific tissues (1). For example, Aldh1a2 is a critical enzyme for RA generation in the embryonic mesoderm (14), whereas Aldh1a3 is critical for RA generation in the embryonic eye (15) as well as for regulation of insulin and glucagon from pancreatic islets in diabetic mice (13). A combination of specific and redundant functions among Aldh1 enzymes could potentially influence adipogenesis and fat formation. Here, we provide evidence that RA generation for adipogenesis is supported predominantly by the Aldh1a1 enzyme, whereas in vivo, different fat formation in visceral and sc depots in mice and humans result in part from different expression and concerted action of all Aldh1 enzymes.

Results

Aldh1a1 is the major enzyme expressed during adipocyte differentiation in vitro

Vitamin A metabolism is dependent on a number of vitamin A-metabolizing enzymes (Fig. 1A). We analyzed expression of RA-generating enzymes in nondifferentiated and differentiated 3T3-L1 fibroblasts to examine their effects on adipocyte differentiation (Fig. 1B). Among efficient RA-generating enzymes, Aldh1a1 was the only enzyme significantly induced during 3T3-L1 differentiation (201%, P < 0.035); moreover, Aldh1a1 had the highest expression among Aldh1 enzymes (Fig. 1B). Aldh1a1 expression increased proportionally to adipocyte differentiation as indicated by a significant correlation between Fabp4 and Aldh1a1 expression (P < 0.001, Pearson correlation; Fig. 1B, inset). The mitochondrial Aldh2 expression was not altered during adipogenesis in 3T3-L1 fibroblasts (Supplemental Fig. 1A, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). Aldh2 and Aldh8A1 are known to be inefficient at retinaldehyde oxidation and were not studied here (1). Stimulation of mature adipocytes (differentiated for 6 d) with RA markedly inhibited Aldh1a1 expression in a concentration-dependent manner, whereas expression of Aldh1a2 and a3 was not altered (Fig. 1C). This experiment suggests a potential feedback mechanism in response to administration of high RA concentrations in adipocytes that is mediated primarily by Aldh1a1. Because expression profiles suggested a possible Aldh1a1 role in adipogenesis, our next studies dissected the role of this enzyme in RA production and in adipogenesis and analyzed possible redundancies in functions of other members of the Aldh1 gene family in vitro.

Fig. 1.

Aldh1a1 expression is increased after adipogenesis in 3T3-L1 preadipocytes. A, Schematics of the major enzymes participating in vitamin A metabolism. Rald is generated from retinol by Adh (Adh1, Adh3, and Adh4) and Rdh (Rdh1 and Rdh10); the last are members of the short-chain dehydrogenase/reductase (Sdr) family. The other 15 Sdr members of this family have reduced activity toward all-trans retinol isomers and are not studied here (reviewed in Ref. 33). RA is produced solely from retinaldehyde by the cytosolic Aldh1 family of enzymes (Aldh1a1, Aldh1a2, and Aldh1a3). This family is also known as the retinaldehyde dehydrogenase family, and alias names of these enzymes are Raldh1, Raldh2, and Raldh3. Excess RA is oxidized primarily by cytochrome P450, family 26, subfamily B, polypeptide 1 (Cyp26B1) and by Cyp26A1. B, The expression of RA-generating enzymes in nondifferentiated (Non-D) and differentiated 3T3-L1 adipocytes (6 d of differentiation). In all studies, mRNA expression was measured by TaqMan. Data were normalized using 18S as an endogenous control. Relative expression was calculated based on the comparative cycle threshold (Ct) method. *, P < 0.05 (Mann Whitney U test, n = 3, mean ± sd). The inset shows a correlation between mRNA expression of Fabp4 and Aldh1a1 in differentiation experiments (n = 9, Pearson correlation test, P < 0.001). C, 3T3-L1 fibroblasts were differentiated for 6 d. These adipocytes were stimulated with the indicated RA concentrations for 12 h, and mRNA was then isolated and analyzed for Aldh1 expression by TaqMan as in C. *, P < 0.007 (Mann Whitney U test, n = 3, mean ± sd).

Aldh1a1 deficiency impairs RA production and adipogenesis in vitro

To elucidate time-dependent endogenous production of nuclear receptor ligands in adipogenesis, previous studies employed sensitive and specific reporter assays in permanently transfected 3T3-L1 fibroblasts (16). Using a similar approach, we created the RARE luciferase reporter wild-type (WT) and Aldh1a1^−/− [knockout (KO)] fibroblast cell lines RARE_WT and RARE_Aldh1a1KO, respectively. Adipogenesis in these cell lines was induced by a standard protocol (8, 17). Kinetics of RARE activation during adipogenesis are shown in Fig. 2A. Differentiation in RARE_WT fibroblasts was accompanied by RARE activation that reached a maximum (384%) at d 7 of differentiation compared with nondifferentiated RARE_WT fibroblasts. The RARE activation suggested that endogenous RA was generated during adipogenesis. In RARE_Aldh1a1KO fibroblasts, differentiation was accompanied by a minor (165%) increase in RARE activation compared with nondifferentiated RARE_Aldh1a1KO fibroblasts. At all time points, RARE activation by endogenous RA generation was significantly lower in RARE_Aldh1a1KO than in RARE_WT fibroblasts. Importantly, stimulation of RARE_Aldh1a1KO and RARE_WT fibroblasts with RA or TTNPB (4-[(E)-2-(5,6,7,8-Tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl]benzoic acid), a synthetic RAR ligand, yielded similar activation in these cell lines (Fig. 2A, inset). Therefore, decreased RARE activation in RARE_Aldh1a1KO fibroblasts might result from impaired endogenous RA production, consistent with the loss of Aldh1a1, the predominantly expressed enzyme of the Aldh1 family.

Fig. 2.

Aldh1a1 and Aldh1a3 generate RA and regulate PPARγ mRNA expression. A, RARE_WT (○) and RARE_{Aldh1a1−/−} (RARE_Aldh1a1KO, ●) fibroblast cell lines were derived from fibroblasts stably transfected with lentiviral RARE reporter vector. Differentiation was induced using a standard differentiation procedure. Cells were harvested every day and assayed for luciferase activity, whereas the other cell set was used to measure mRNA (A). Data are shown as a ratio of luciferase activity on the indicated day to that seen on d 0 in the same cell line. All data are shown as mean ± sd (n = 4). *, Statistically different compared with d 0; #, statistically different between RARE_WT and RARE_{Aldh1a1−/−} on the same day of differentiation (both P < 0.05, Mann Whitney U test). The inset shows RARE_WT (white bar) and RARE_{Aldh1a1−/−} (RARE_KO, black bar) fibroblasts stimulated with all-trans RA (0.1 μm) for 3 h. Luciferase activity was not statistically different in these fibroblasts. B, mRNA expression of PPARγ and other adipogenic markers (data not shown) was measured on d 6 of differentiation in RARE_WT and RARE_{Aldh1a1−/−} fibroblasts. (Mann Whitney U test, n = 3; P < 0.001). C, RARE_{Aldh1a1−/−} fibroblasts were transiently transfected with full-length control vector (Co), human Aldh1a1 (a1+), Aldh1a3 (a3+), or both of these vectors (a1+a3+) as described in Materials and Methods. After transfection, differentiation was induced by standard procedure. Overexpression of human Aldh1a1 and Aldh1a3 is shown in Supplemental Fig. 1, C and D. Luciferase activity was assayed on d 6 of differentiation. Data are shown as fold induction compared with activity in RARE_{Aldh1a1−/−} fibroblasts transfected with a control vector (black bar). *, Significant difference compared with RARE_Aldh1a1KO fibroblasts transfected with control vector (Mann Whitney U test, n = 3; P < 0.01). D, mRNA samples from these cells were used to measure expression of mouse and human Aldh1a1, a2, and a3 as well as PPARγ. The summarized expression levels of all Aldh1 enzymes correlate with PPARγ expression (ANOVA test, n = 12; P < 0.016). E–G, RARE_{Aldh1a1−/−} fibroblasts were differentiated in the presence of 10 nm all-trans RA for 8 d. RA (10 nm) was added 3 h after the differentiation mix and in the medium every 48 h. E, PPARγ mRNA expression was significantly increased in RARE_{Aldh1a1−/−} fibroblasts stimulated with RA (Mann Whitney U test, mean ± sd, three independent experiments, P < 0.025). F, Significantly increased ratio of Fabp4 to Pref1 expression (Mann Whitney U test, mean ± sd, three independent experiments, P < 0.05) in RARE_{Aldh1a1−/−} fibroblasts stimulated with RA compared with those treated with vehicle (Veh, ethanol/dimethyl sulfoxide, 50%). G, Representative oil red O staining of neutral lipids in these differentiated RARE_{Aldh1a1−/−} fibroblasts stimulated with vehicle and RA.

RARE_Aldh1a1KO fibroblasts had impaired adipogenesis, resulting in the markedly lower expression of PPARγ than in RARE_WT fibroblasts (Fig. 2B). Overall suppression of adipogenesis in Aldh1a1^−/− fibroblasts has been previously reported (8) and assessed by an increased Pref1 to Fabp4 expression ratio (Supplemental Fig. 1B). To elucidate whether the changes in PPARγ are attributed to RA generation, we performed rescue experiments by overexpression of human Aldh1a1 and Aldh1a3 enzymes or their combination. We took advantage of the similar catalytic function of human Aldh1 enzymes in RA generation combined with their distinct sequence to distinguish them from the endogenous mouse Aldh1 enzymes. We did not overexpress Aldh1a2 because this gene was expressed at negligible levels in both nondifferentiated and differentiated 3T3-L1 fibroblasts. Transient transfection resulted in a marked, significant increase in human Aldh1a1 and Aldh1a3 expression, whereas the expression of endogenous mouse Aldh1a2 and -a3 was below detection limits (Supplemental Fig. 1, C and D).

Overexpression of Aldh1a1, Aldh1a3, or both led to a significant increase in RARE-luciferase activity (from 240 to 260% compared with control) in Aldh1a1^−/− fibroblasts (Fig. 2C). Overexpression of Aldh1a1, Aldh1a3, or both resulted in significantly increased PPARγ expression that was correlated with the expression levels of all Aldh1 enzymes and was not related to the expression of a specific member of this family (Fig. 2D). Of note, cells with combined Aldh1a1 and Aldh1a3 expressed significantly more Aldh1a1 than Aldh1a3 (Supplemental Fig. 1, C and D); however, the artificial expression was below physiological expression levels of Aldh1a1 and Aldh1a3 seen in 3T3-L1 adipocytes (Fig. 1B).

To elucidate whether Aldh1a1 effects were dependent on RA production, we stimulated Aldh1a1^−/− fibroblasts with low concentrations of RA (10–100 nm) during differentiation. In agreement with the role of Aldh1a1 in RA production, RA stimulation helped to rescue, in part, PPARγ expression (Fig. 2E) and differentiation (Fig. 2, F and G) in Aldh1a1^−/− fibroblasts after 8 d of differentiation. The increase in differentiation was characterized by an increased ratio of adipogenesis/preadipogenesis markers Fabp4/Pref-1 (Fig. 2F) and formation of adipocytes with large lipid droplets (Fig. 2G) seen in Aldh1a1^−/− fibroblasts stimulated with RA. These changes were associated with significantly increased expression of lipogenic enzymes, such as fatty acid synthase (Supplemental Fig. 2A), whereas expression of adiponectin or C/EBPα was not changed (Supplemental Fig. 2, B and C). Both administered RA or RA produced by overexpression of Aldh1 enzymes can partially regain PPARγ expression; thus, the impact of Aldh1a1 on adipogenesis is linked, at least in part, to its dominant role in RA production during adipogenesis.

Consistent with the expression profile, the impact of Aldh1a3 on adipogenesis in 3T3-L1 fibroblasts was minor (Fig. 3). We studied the effects of the Aldh1a3 expression profile in two representative cell lines stably transfected with scrambled (3T3-L1_shCo) or Aldh1a3 short hairpin RNA (shRNA) containing control or specific shRNA, which is shown in Supplemental Fig. 3, A and B. A scrambled shRNA control cell line principally expressed Aldh1 enzymes in a similar fashion as nontransfected 3T3-L1 fibroblasts (Supplemental Fig. 2A). As expected, stable transfection of 3T3-L1 fibroblasts with Aldh1a3 shRNA (3T3-L1_sha3) decreased Aldh1a3 expression that was significantly different in differentiated 3T3-L1_sha3 vs. 3T3-L1_shCo fibroblasts (Supplemental Fig. 2A), but Aldh1a1 remained the major Aldh1 enzyme expressed in all differentiated 3T3-L1 clones. Aldh1a2 expression was decreased in nondifferentiated 3T3-L1_sha3 and remained the same in differentiated 3T3-L1_sha3 compared with control cells (Supplemental Fig. 3B). The small decrease in Aldh1a3 expression in 3T3-L1_sha3 vs. 3T3-L1_shCo did not influence expression of PPARγ or other adipogenesis markers, including Fabp4, C/EBPα, and C/EBPβ, and adiponectin (Fig. 3, A–D, and data not shown). In contrast, the expression of preadipocyte marker Pref1 and visceral adipocyte marker visfatin was significantly increased in differentiated 3T3-L1_sha3 vs. 3T3-L1_shCo fibroblasts (Fig. 3, E and F), suggesting an Aldh1a3 role in the regulation of preadipocyte responses and some depot-specific adipokine production. Importantly, the effects on PPARγ expression and overall adipogenesis were dependent on RA production and could be recovered to a significant extent by expression of any member of Aldh1 family.

Fig. 3.

Decreased expression of Aldh1a3 influences Pref1 and visfatin expression without altering major adipogenic transcription factor expression. 3T3-L1 fibroblast cell lines were stably transfected with scrambled (scr) shRNA or Aldh1a3 (a3) shRNA and differentiated for 6 d. The expression levels of Aldh1 enzymes in these cells are shown in Supplemental Fig. 2, A and B. A–F, mRNA expression of the adipogenic markers PPARγ (A), Fabp4 (B), C/EBPβ (C), and adiponectin (Adipoq, D) as well as markers of preadipocytes, preadipocyte factor-1 (Pref1, E), and visceral fat, visfatin (F), in differentiated Aldh1a3 shRNA 3T3-L1 adipocytes (black bars) compared with differentiated scrambled shRNA 3T3-L1 adipocytes (white bars). P value indicates differences in differentiated Aldh1a3 shRNA vs. scrambled 3T3-L1 adipocytes (Mann Whitney U test, n = 3, mean ± sd).

Aldh1 enzymes and RA regulate PPARγ expression by mechanisms involving zinc-finger protein 423 (ZFP423) expression

Two major transcriptional pathways, C/EBP and ZFP423, have been shown to induce PPARγ expression during adipogenesis (2, 18). In our studies on rescuing PPARγ expression and adipogenesis in Aldh1a1^−/− cells with Aldh1a1 and/or Aldh1a3 expression (Fig. 2D), we found a significant correlation between Aldh1 and ZFP423 expression levels, whereas expression of transcription factors C/EBPβ (Fig. 4A) and C/EBPα (data not shown, coefficient of determination of a linear regression, R² = 0.05, not significant) was not influenced by the expression of Aldh1 enzymes. To strengthen the link between expression of Aldh1 enzymes and ZFP423, we analyzed expression in differentiated fibroblasts (adipocytes) that were stably transfected with shRNA. The WT adipocytes expressed WT levels of Aldh1 enzymes (high Aldh1a1/low Aldh1a2 and a3), the Aa1⁻ adipocytes were deficient in Aldh1a1 and other Aldh1 enzymes, and Aa1⁻a2⁺a3⁺ adipocytes were deficient in Aldh1a1 but expressed stable increased levels of Aldh1a2 and a3 enzymes (Supplemental Fig. 3C). ZFP423 expression was markedly decreased (99.9%) in Aa1⁻ adipocytes compared with WT adipocytes (WT adipocytes, 100%) and recovered, partially, in Aa1⁻a2⁺a3⁺ adipocytes expressing Aldh1a2 and a3 enzymes (Fig. 4B). In these adipocytes, the expression of all Aldh1 also correlated with ZFP423 (R² = 0.75; P < 0.000, ANOVA). In contrast, C/EBPβ expression was increased in Aa1⁻ adipocytes (158%) but not altered in Aa1⁻a2⁺a3⁺ (104%) compared with WT adipocytes. Similarly, C/EBPα expression was also marginally influenced by Aldh1 expression (Supplemental Fig. 3D), in contrast to the expression of lipogenic genes Fabp4 and adiponectin (Supplemental Fig. 3, E and F). Finally, administration of RA during adipogenesis rescued both PPARγ (Fig. 2E) and ZFP423 expression in a dose-dependent manner in Aldh1a1^−/− fibroblasts (Fig. 4C). C/EBPβ expression at these nanomolar RA concentrations was not changed. Together these experiments suggest that intrinsic RA production in adipocytes promotes adipogenesis through the expression of PPARγ that is mediated primarily by ZFP423.

Fig. 4.

Aldh1 enzymes and RA stimulation induce ZFP423 expression. A, RARE_{Aldh1a1−/−} fibroblasts were transiently transfected with human full-length control vector (Co), Aldh1a1 (a1+), Aldh1a3 (a3+), or both of these vectors (a1+a3+) and differentiated for 6 d as described in Fig. 2, C and D. Along with Aldh1 and PPARγ expression, mRNA samples from these cells were used to measure expression of ZFP423, C/EBPα, and C/EBPβ expression. The summarized expression levels of all Aldh1 enzymes correlate with ZFP423 but not C/EBPβ expression (ANOVA test, n = 12, n.s., not significant). C/EBPα did not correlate with Aldh1 (ANOVA test, n = 12; R² = 0.05, n.s.). B, WT and Aldh1a1^−/− fibroblast cell lines were stably transfected with scrambled and Aldh1a2 and -a3 shRNA to obtain clones deficient in all Aldh1 enzymes (Aa−) as well as those expressing no Aldh1a1 but high Aldh1a2 and a3 levels (Aa−a2+a3+) upon differentiation. The expression levels of Aldh1 enzymes in these cells are shown in Supplemental Fig. 2C. mRNA expression of ZFP423 and C/EBPβ was measured in cells differentiated for 6 d. Asterisks indicate significant differences as compared with WT adipocytes (A_WT). #, Significant differences between Aa− and Aa−a2+a3+ (Mann Whitney U test, n = 3, mean ± sd). PPARγ and ZFP423 had similar expression patterns in these cells (data not shown). C/EBPα was modestly reduced (40%) in both Aa⁻ and Aa⁻a2⁺a3⁺ clones compared with A_WT (data not shown). C, ZFP423 and C/EBPβ expression in RARE_{Aldh1a1−/−} fibroblasts that were differentiated in the absence and in the presence of different RA concentrations (indicated in the figure) for 8 d as described in Fig. 2E. P values indicate significant increase in ZFP423 expression in cells stimulated with RA compared with vehicle (Veh, ethanol) (Mann Whitney U test, n = 3, mean ± sd).

Fat depot-specific Aldh1 expression underlies the distinct expression of transcription factors

The difference in PPARγ expression is a distinguishing characteristic between sc and visceral adipose tissues (19), but mechanisms establishing this specific expression pattern are unknown. We compared Aldh1 expression in sc and visceral tissue of seven C57/BL6 WT and six Aldh1a1^−/− females fed a regular chow diet (Fig. 5, A and B). Aldh1a1 was predominantly expressed in both sc (Fig. 5A) and visceral (Fig. 5B) fat compared with other Aldh1 enzymes, although the expression of Aldh1a1 and a3 was lower in visceral than in sc fat. In Aldh1a1^−/− mice, the Aldh1a3 enzyme was expressed in sc fat (Fig. 5A), whereas visceral fat did not express any Aldh1 enzymes (Fig. 5B). The decrease in RA-generating enzymes in sc fat of Aldh1a1^−/− mice did not influence ZFP423 expression levels (Fig. 5C); however, in visceral fat, a marked 70% decrease in ZFP423 was observed in Aldh1a1^−/− compared with WT mice (Fig. 5D). In these fat tissues, ZFP423 expression correlated with the PPARγ expression (Fig. 5E). Accordingly, whereas PPARγ and Fabp4 underwent moderate or no change in expression in sc fat of Aldh1a1^−/− mice (Fig. 5F), expression of both PPARγ and Fabp4 was markedly (70%) reduced in visceral fat of these mice as compared with WT mice (Fig. 5G).

Fig. 5.

Fat depot-specific expression of Aldh1a3 and Aldh1a1 enzymes manifests as decreased ZFP423, PPARγ, and Fabp4 expression in visceral but not in sc fat of Aldh1a1^−/− mice. mRNA was isolated from sc and perigonadal visceral fat of WT (n = 7) and Aldh1a1^−/− (n = 6) female mice fed regular chow. All data are shown as mean ± sd unless otherwise indicated. Statistical significance was examined by Mann Whitney U test. A, Expression of Aldh1 enzymes in sc fat. *, Significant decrease in Aldh1a1 and a3 expression in Aldh1a1^−/− compared with WT mice. B, Expression of Aldh1 enzymes in visceral fat. *, Significant decrease in Aldh1a1 and a3 expression in Aldh1a1^−/− compared with WT mice. #, Significant differences compared with Aldh1 expression in sc fat (A). C and D, ZFP423 expression in sc (C) and visceral (D) fat. *, Significant differences in gene expression in Aldh1a1^−/− compared with WT fat (same fat depot), P < 0.002 (Mann-Whitney U test); #, significant differences in gene expression in visceral compared with sc fat within the same genotype (paired t test, P < 0.001). E, Correlation between expression levels of ZFP423 and PPARγ (ANOVA test, n = 25; P < 0.002). F and G, PPARγ and Fabp4 expression in sc (F) and visceral (G) fat.

Noting the altered expression of Pparγ, the key transcriptional regulator of adipogenesis, in response to depot-specific RA production in Aldh1a1^−/− mice, we could also expect to change fat distribution. Aldh1a1^−/− mice weighed less than WT mice (Fig. 6A). We employed magnetic resonance imaging (MRI) to analyze fat distribution in WT and Aldh1a1^−/− mice. Mice were scanned throughout their length, and the resulting 69 cross-sectional images were analyzed for fat content (Fig. 6B). We observed significantly higher fat accumulation in visceral regions between sections 18–31 in WT compared with Aldh1a1^−/− mice (Fig. 6, B–D). In the other regions, fat accumulation varied among different animals and was not significantly different between genotypes. Representative perigonadal and pericardial sections in WT and Aldh1a1^−/− mice are shown in Fig. 6, C and D. The fat reduction in these areas was in agreement with some loss in PPARγ expression seen in sc fat.

Fig. 6.

Aldh1a1 deficiency influences fat formation in specific visceral regions. A, Whole-body weights are shown in WT (n = 10) and Aldh1a1^−/− (n = 9) female mice on regular chow (same as in Fig. 4). B, Analysis of fat area of 69 MRI cross-sectional images performed in frozen WT and Aldh1a1^−/− female mice from this study (n = 3 per group). *, Significant differences in fat areas in WT and Aldh1a1^−/− mice. Dashed lines show approximate position of significantly affected fat areas. C and D, Cross-sectional (two right panels of pericardial and peribladder regions) and coronal (left image) MRI images from representative WT (C) and Aldh1a1^−/− (D) mice. B, Bladder; H, heart; s, sc; v, visceral.

Fat depot-specific expression of Aldh1 enzymes in women

To gain insight into the relevance of Aldh1 enzymes in human fat biology, we analyzed expression of Aldh1 enzymes in sc and visceral (omental) fat isolated from four healthy women (Fig. 7). In these women, Aldh1a1 mRNA expression was the most abundant among the Aldh1 enzymes in both sc and omental fat. However, Aldh1a2 and a3 expression were significantly higher in omental fat than in sc fat. Therefore, in both humans and rodents, Aldh1 enzyme expression was fat depot-specific, but Aldh1a2 and a3 expression was higher in visceral fat than in sc fat.

Fig. 7.

Predominant expression of Aldh1a1 and dissimilar expression of Aldh1a2 and a3 enzymes in sc compared with omental fat in women. Relative mRNA expression of Aldh1 enzymes in sc (A) and omental (B) adipose tissue isolated from four women. Aldh1 expression is shown as mean ± sd. #, P < 0.03, differences between Aldh1 expression in sc and omental fat in each donor (paired t test).

Discussion

Although the etiology of visceral obesity is not fully understood, it is known that hormone and cytokine production in adipose tissue contributes to depot-specific fat formation (20–22). Here we show that the Aldh1 family of enzymes generating RA is an essential autocrine pathway in adipogenesis that regulates depot-specific fat formation by transcriptional mechanisms involving PPARγ activation through ZFP423.

The importance of vitamin A metabolism in adipogenesis and fat formation was established based on evidence that retinoids suppress adipogenesis and fat formation (2, 5, 23) as well as recent studies in animals deficient in single genes responsible for the generation of vitamin A metabolites (8, 24, 25). These studies reported somewhat paradoxical observations that RA supplementation (5, 23) and deficient RA generation in Aldh1a1^−/− mice both render resistance to diet-induced obesity (8). Our studies here take into account that endogenous retinoid production is a result of the concerted action of multiple enzyme families (1). Families of enzymes can produce similar retinoids, e.g. the Aldh1 enzyme family members, Aldh1a1, a2, and a3, can all use retinaldehyde for RA generation (1). Here we show that adipogenesis is accompanied by RA generation predominantly by Aldh1a1 (200% expression) (Fig. 1). Importantly, stimulation with RA specifically inhibits Aldh1a1 expression (Fig. 1C), suggesting a probable negative feedback mechanism regulating Aldh1a1 expression and RA generation. Inhibition of Aldh1a1 expression by RA also provides a possible explanation of the paradoxical suppression of fat formation seen in previous studies with Aldh1a1^−/− mice and obese rodents treated with RA (5, 8, 23). The Aldh1a1-dependent inhibition of adipogenesis is mediated by physiological RA (10–100 nm) concentrations. At higher RA concentrations (10 μm) used in previous seminal studies on adipogenesis inhibition (2, 7), RA induced an array of effects.

We show that Aldh1 expression is necessary for endogenous RA production (Fig. 2A). We measured RA during adipogenesis in sensitive and specific RARE-reporter 3T3-L1, WT, and Aldh1a1^−/− cell lines. Earlier, an analogous approach was used to demonstrate endogenous PPARγ ligand generation in the course of adipogenesis (16). It is generally accepted that RARE activation is due to increased RA generation, because RA has high specificity and affinity for RAR (1, 6, 14), and RARE reporters are frequently used to assay spatiotemporal RA production in embryogenesis (1). In our experiments, both WT and 3T3-L1 RARE preadipocytes exhibited increased RA generation during the course of adipogenesis, with maximum RA production at d 5 in WT and d 4 in 3T3-L1 adipocytes (Fig. 2A and data not shown for 3T3-L1). The Aldh1a1 enzyme was responsible for approximately 70% of generated RA, based on comparison of RARE_WT and RARE_Aldh1a1KO differentiated fibroblasts. The major Aldh1a1 contribution to RA production was consistant with predominant expression of this enzyme during differentiation (Fig. 1C). Nevertheless, Aldh1a1 and Aldh1a3 appear to be redundant with respect to RA generation; hence, overexpression of either of these enzymes can partially recover RA generation (Fig. 2C). This redundant function of Aldh1a1 and Aldh1a3 is also manifested in vivo in sc fat of Aldh1a1^−/− mice. In the absence of Aldh1a1 expression, the remaining Aldh1a3 partially supported sc fat formation (Fig. 5, A and B), whereas the loss of all Aldh1 enzymes in visceral fat markedly impaired formation of this fat depot in Aldh1a1^−/− mice (Fig. 7B). Specific expression of different Aldh1 enzymes in sc and visceral depots offers a therapeutic opportunity to regulate RA production in a tissue-specific fashion.

Our data provide evidence that production of RA by Aldh1a1 and other Aldh1 enzymes was causatively linked to regulation of PPARγ expression. Aldh1a1^−/− fibroblasts expressed 29-fold reduced PPARγ levels compared with WT adipocytes and failed to differentiate (Fig. 2B). Overexpression of one or a combination of Aldh1 enzymes significantly increased PPARγ in differentiated Aldh1a1^−/− fibroblasts (Fig. 2D). RA was similarly effective in rescuing PPARγ expression (Fig. 2E) and adipogenesis in Aldh1a1^−/− fibroblasts, arguing that RA is a central mediator of Aldh1-dependent responses in adipocytes. In the absence of Aldh1 enzymes, retinaldehyde spared from conversion to RA (8, 11), can potentially contribute to the suppression of PPARγ activation as an inhibitor of PPARγ and its heterodimeric partner RXR (8). Given the master role of PPARγ in the transcriptional regulation of adipogenesis, Aldh1 enzymes could be considered as a key autocrine pathway in adipogenesis acting as a PPARγ switch.

The mechanisms regulating PPARγ expression in adipogenesis are not completely understood, although recent studies demonstrate the essentiality of transcription factor ZFP423 in this process (18). It has been shown that ZFP423 overexpression is sufficient to rescue adipogenesis through PPARγ in NIH3T3 fibroblast lines that lack PPARγ and resist differentiation (18). The means by which ZFP423 is regulated during adipogenesis remained unclear. Here we show that RA produced by Aldh1 enzymes is a critical mediator of ZFP423 expression. ZFP423 expression levels were reduced by 99% in Aldh1a1^−/− compared with WT fibroblasts; however, ZFP423 expression could be rescued by either expression of any Aldh1 enzymes or RA stimulation (Figs. 2 and 4). In consonance with these loss- and gain-of function studies, ZFP423 and PPARγ expression correlated in vivo in visceral and sc fat of WT and Aldh1a1^−/− mice. Our results suggested that Aldh1 enzymes regulate PPARγ expression predominantly through ZFP423-dependent mechanisms. Given the importance of ZFP423 regulation in adipogenesis, more studies are needed in the future to elucidate mechanistic aspects of ZFP423 dependence on RA.

The Aldh1a1-dependent inhibition of adipogenesis is mediated by physiological RA (10–100 nm) concentrations. Previous studies have demonstrated that RA at higher concentrations (10 μm) acts in adipogenesis through RAR interaction with C/EBPα and/or C/EBPβ transcription factors that block PPARγ expression and activity (2, 18). It is plausible that at these concentrations, RA activates different pathways that inhibit differentiation independent of the induction of ZFP423 and PPARγ or before expression of these transcription factors. The potential mechanisms can involve transcriptional RA-dependent pathways, including RARγ autoregulation (26) and interaction with C/EBP as well as direct protein modifications through retinoylation (27). In our studies, expression of the Aldh1a1 enzyme was associated with a modest induction C/EBPα and reduction of C/EBPβ expression levels; however, these effects could not be rescued by other Aldh1 enzymes. These unique responses to Aldh1a1 may indicate specific roles of the Aldh1a1 enzyme in both adipogenesis and fat-depot differences. In addition to the common function of all Aldh1 enzymes of using retinaldehyde for RA generation, each member of the Aldh1 family may have a specific function. Aldh1a1 can use different aldehyde substrates, including lipid aldehydes and acetaldehydes (11). Other Aldh1 enzymes also exerted distinct effects. For example, decreased Aldh1a3 expression, but not Aldh1a1 (data not shown), in vitro was associated with increased expression of visfatin. This adipokine mediates insulin responses mainly in visceral fat. Therefore, Aldh1a3 can potentially regulate endocrine responses in visceral fat. More future studies would provide insight into the distinct roles of Aldh1a1, a2, or a3 in the regulation of cytokines specific for visceral or sc fat and the mechanisms of their actions.

These transcriptional effects in response to RA production by Aldh1 enzymes could have important implications for the regulation of fat depots in vivo. Many differences between visceral and sc depots stem from the different PPARγ expression in these tissues. Our studies in Aldh1a1^−/− mice shed light on the mechanisms that establish depot-specific transcription patterns. In visceral adipose tissue, ZFP423, PPARγ, and Fabp4 expression was decreased by 70% in Aldh1a1^−/− compared with WT mice (Fig. 5, A and B). In sc fat, PPARγ was reduced by only 40%, and the expression of both inducing factor ZFP423 and gene Fabp4 was similar in Aldh1a1^−/− and WT mice. These differences in PPARγ expression in Aldh1a1^−/− mice could be attributed to Aldh1a3 expression in sc but not visceral fat. Aldh1a3 helped to maintain some RA production and was capable of inducing both ZFP423 and PPARγ in vitro and could be expected to exert similar effects in sc fat. These data suggest that concerted Aldh1 function in vivo may include regulation of transcription in a fat depot-specific manner.

Our findings raise questions as to whether Aldh1 enzymes act in a species-specific manner. We found substantial differences between Aldh1 expression patterns in visceral and omental fat isolated from female mice and women, respectively (Fig. 7). The visceral fat in mice and omental fat in humans may originate from different progenitors that may be predisposed to express dissimilar amounts and types of Aldh1. Another possibility could be differences in cholesterol/fat content in food consumed by humans and rodents. Cholesterol can directly regulate expression of Aldh1a1 and a2 enzymes through transcriptional activation of sterol regulatory element-binding protein 1c transcription factor (28). In our studies, mice received a regular chow diet, but high-fat diets increase expression of Aldh1a1 and a2, at least in the liver (28). Intriguingly, fat accumulation in specific depots, insulin resistance, lipolysis, lipogenesis, response to high-fat feeding, and other adipose tissue characteristics vary across different mouse strains and species (29). It remains to be investigated whether different expression patterns of Aldh1 enzymes contribute to distinct characteristics of fat tissue depots.

Importantly, Aldh1a1, which regulates fat formation, was abundantly expressed in both human and mouse fat depots. Aldh1a1 was the major enzyme for RA production and regulation of adipogenesis by tandem expression of ZFP423 and PPARγ. Even so, only concerted action of Aldh1 enzymes had diverged fat accumulation and transcription responses in visceral and sc fat depots. Visceral obesity in humans increases the risk of premature death related to the development of type 2 diabetes, certain cancers, and cardiovascular disease, whereas sc fat produces a variety of cytokines with antiinflammatory and insulin-sensitizing properties (20, 30, 31). Our data outline Aldh1-dependent mechanisms that preferentially reduce visceral fat and can potentially lead to the development of preventive measures and treatment of visceral obesity.

Materials and Methods

Reagents

We purchased reagents from Sigma-Aldrich (St. Louis, MO) and cell culture media from Invitrogen (Carlsbad, CA) unless otherwise indicated; Cignal Lenti inducible RARE reporter vectors were purchased from SA Biosciences/QIAGEN (Valencia, CA).

Human subjects

mRNA was isolated from fat tissue, which was obtained from four healthy women who underwent surgery as kidney donors and had given informed consent. The protocol was approved by the Mayo Clinic Institutional Review Board for Human Research. Subjects were 41 ± 2 yr of age. The subjects' mean body mass index was 30.3 ± 0.4 kg/m² (more information in supplemental data).

Animal studies and MRI

Generation and characterization of Aldh1a1^−/− mice, including metabolic responses in high-fat-fed mice, have been previously described (1, 8, 32). All experimental protocols were approved by the Institutional Animal Care and Use Committee. Age-matched (13 months old) and sex-matched Aldh1a1^−/− (nine females) and WT (10 females) mice were fed regular chow (Research Diets Inc., New Brunswick, NJ). Water and food was ad libitum. Weight was measured monthly. The absence of glucose intolerance in both genotypes fed regular chow was confirmed by glucose tolerance testing (data not shown). Glucose tolerance tests were performed in overnight fasted mice by ip injection of a single 25% dextrose injection (0.004 ml/g body weight) using a glucometer for measurements [Accu-Chek Advantage (Roche Diagnostics Corporation, Indianapolis, IN) or OneTouch Ultra (LifeScan, Milpitas, CA)]. Body fat distribution was measured by MRI using a 9.4-T Bruker BioSpin wide-bore scanner. Multi-slice multi-echo sequence (repetition time 1400 msec; echo time 12 msec; flip angle 180°; matrix 128 × 128; four averages) was used to acquire a total of 80 1-mm-thick images per mouse scanning from the tail to the head. ParaVision version 4.0 software was used for MR acquisition and reconstruction. All image processing and analysis was performed in OsiriX software (The Osirix Foundation, Geneva, Switzerland). Representative coronal whole-body images were obtained using the maximal intensity projection technique. Visceral (perigonadal) and sc fat pads were dissected and analyzed for mRNA expression.

Cell culture

Murine 3T3-L1 preadipocytes were cultured, maintained, and differentiated using standard procedures (8, 17). Preadipocyte cell lines were derived from embryonic fibroblasts of WT and Aldh1a1^−/− as previously described (8, 17). In all preadipocyte cultures, differentiation was induced (d 0) with a standard differentiation mixture of 3-isobutyl-1-methylxanthine (0.5 mm), dexamethasone (1 μm), and insulin (1.7 μm) in DMEM containing 10% fetal bovine serum. For differentiation into adipocytes, cells were maintained for 7 d in DMEM medium containing 10% fetal bovine serum and insulin (1.7 μm), which was replaced every 48 h.

Oil red O staining

Adipocytes were fixed in 10% formalin in PBS for 1 h, washed with PBS, and then stained with 60% of oil red O solution (5 mg/ml oil red O in isopropanol) for 15 min and hematoxylin and eosin as described before (8).

Analysis of mRNA

mRNA was isolated from adipose tissue and adipocyte cultures according to the manufacturer's instructions (QIAGEN). For semiquantitative analysis of expression, cDNA was prepared from purified mRNA and analyzed using 7900HT Fast Real-Time PCR System and TaqMan fluorogenic detection system (Applied Biosystems, Foster City, CA). Validated primers were also purchased from Applied Biosystems. Reference sequences, gene names, and aliases are described in Supplemental Table 1. Comparative real-time PCR was performed in triplicate, including no-template controls. The mRNA expression of the genes of interest was compared with 18S expression levels. Occasionally, we used TATA-box binding protein as a control in genes expressing at low levels. Relative expression was calculated using the comparative cycle threshold (Ct) method. All data shown in the figures are based on the Ct method normalized to 18S.

Transfection studies

Stably transfected RARE-reporter clones

We derived stably transfected 3T3-L1, WT (A_WT), and Aldh1a1^−/− (A_Aldh1a1KO) cell lines according to the manufacturer's instructions (SA Biosciences). Briefly, after reaching 70% confluence, cells were transfected with Cignal Lenti RARE-LUC reporter suspension (25 multiplicity of infection per 10⁴ cells) in the presence of Polybrene (Millipore, Billerica, MA) transduction reagent in serum-free MEM. After 3 h, cells were supplemented with 10% heat-inactivated calf serum. Stable clones were selected and derived from single cells selected with puromycin (0.75–1.5 mg/ml; Invitrogen). We measured luciferase activity using a dual-luciferase reporter assay (Promega, Madison, WI).

Clones stably transfected with shRNA

3T3-L1 preadipocytes were transfected with shRNA sequences corresponding to nonspecific (scrambled) siRNA, Aldh1a2 siRNA, or Aldh1a3 siRNA gene silencer sequences (Santa Cruz Biotechnology, Santa Cruz, CA). shRNA lentiviral particles contained three to five expression constructs each, encoding target-specific 19–25 nucleotides (plus hairpin) shRNA to knockdown gene expression after transduction.

Transfection and selection of specific clones were performed according to the manufacturer's instructions. Briefly, preadipocytes were plated at 85% confluence in six-well plates. Three hours later, plating medium was replaced with OptiMEM, and cells were transfected with a transduction mixture (200 μl/well). Transduction mixture was comprised of 10% viral particles (4 × 10⁻⁸ infectious units of virus) and 3% Polybrene (Millipore) solutions in Opti-MEM, which were incubated for 15 min before the addition to cells. 3T3-L1 preadipocytes were incubated overnight and selected with puromycin (1.0 mg/ml; Invitrogen). Clones were derived from a single transfected cell tested for the presence of all Aldh1 genes before and after differentiation using a TaqMan assay.

Transiently transfected Aldh1a1^−/− (A_Aldh1a1KO) preadipocytes

RARE_AldhaA1KO preadipocytes were transfected with human full-length Aldh1a1 and Aldh1a3 vectors (Origene, Rockville, MD). Cells were transfected with FugeneHD (Roche) in OptiMEM according to manufacturer's instructions. After 12 h, transfection medium was replaced with growth medium. The next day, the cells were differentiated by a standard procedure for 6 d (17). Differentiated cells were then harvested and assayed for luciferase activity according to Promega manufacturer's instructions. Another set of similarly treated cells was used for mRNA expression analysis.

Statistical analysis

Data are shown as mean ± sd or mean ± se of experiments that were performed at least in triplicate. Group comparisons were performed using Mann Whitney U test unless otherwise indicated, and correlations were examined by Pearson test.