Coordinated Upregulation of Oxidative Pathways and Downregulation of Lipid Biosynthesis Underlie Obesity Resistance in Perilipin Knockout Mice: A Microarray Gene Expression Profile

Two groups of 6- to 10-week-old male mice were used in this study with six perilipin knockout (plin^−/−) mice in one group and six littermate wild-type (plin^+/+) mice in the other. All animal protocols were approved by the Institutional Animal Care and Use Committee at Baylor College of Medicine.

Five different tissues—WAT, liver, heart, skeletal muscle, and kidney—were isolated from each of the 12 mice, and total RNA was extracted from them using the Trizol (Invitrogen, Carlsbad, CA) reagent. To decrease variability introduced by individual biological differences, we made two sets of RNA. In the first set, equal amounts of total RNA were pooled from three mice per tissue for each group (pooled set). In the second set, we used the individual total RNA extracted from each tissue of the remaining three mice in each group (nonpooled set). We thus had four samples for each tissue per group (one pooled and three nonpooled). For the plin^−/− WAT and heart, we had only three samples (one pooled and two nonpooled). These were then used to generate biotinylated cRNA according to the protocols provided by Affymetrix (Santa Clara, CA).

A hybridization mixture containing 15 μg of the labeled cRNA was hybridized to Affymetrix MG-U74A-v2 chips, which were then washed, scanned, and analyzed. The chips were evaluated and subjected to global scaling as described elsewhere (9). The background and the scaled noise of each chip were averaged, and the means and SD of all expression intensities were calculated for the two groups (plin^−/− versus plin^+/+). The results obtained from these five tissues were analyzed with the Affymetrix Genechip MAS V.5.0 and with dChip V.1.2 (10) software. For the perfect-match/mismatch (PM/MM) model (10), we applied four filters to screen for significantly altered gene expression between these two groups: 1) 0.2 < SD/mean < 10; 2) experiment-KO/basal-WT (E/B) or B/E >1.2, 3) E-B or B-E >255, where 255 was the average scaled noise (9), and 4) P < 0.05. To extend the reliable detection limit of expression in WAT, we also used the PM-only model (dChip V.1.2) using the same filters, as the PM-only model is immune to the adverse effect of the cross-hybridization experienced by some of the MM probes (online appendix 1 [available at http://diabetes.diabetesjournals.org]). Only results that showed a statistically significant difference between plin^−/− and wild-type transcripts by this microarray analysis are presented. An alternative quantification method using fluorescence-based one-step RT-PCR (see below) applied to selected transcripts (online appendix 2) produced data that were concordant with that obtained by microarray analysis.

Quantitative real time RT-PCR reactions were carried out for selected genes using gene-specific primers either with SYBR Green or with specific fluorescent probes. Pooled and individual total RNA samples from the WAT of the plin^−/− and littermate plin^+/+ mice was used as a template. The SYBR Green reactions were performed using the ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) or DNA Engine Opticon (MJ Research, Reno, NV) and LightCycler-RNA Master SYBR Green I kit (Roche, Mannheim, Germany) and specific primers (sequences available on request) for each of the selected genes (online appendix 2). After initial RT temperature of 61°C for 20 min, denaturation for 2 min at 95°C followed by 40 cycles of denaturation at 95°C, primer annealing at 55°C and extension at 72°C, each for 22 s, were carried out with fluorescence detection done in the extension phase of each cycle. A melting curve analysis was performed to confirm the specificity of the amplification. The results were quantified after normalizing to transferrin in each of the samples. A similar reaction was carried out using specific fluorescent probes as described earlier (11) for selected genes, and the results were normalized to cyclophilin in each of the samples.

Genes were annotated using public and Affymetrix databases. To detect metabolic pathways we also used MAPP-Finder V.1.0 and GenMAPP V.1.0 software (12,13).

Of the 12,488 genes and expressed sequence tags represented on the MG-U74A v.2 oligonucleotide array, transcripts were deemed to be significantly up- or downregulated in the plin^−/− mice (Fig. 1, Table 1) when they met the criteria defined in gene expression assays and data analysis. This analysis was performed for each of the tissues studied.

As expected, WAT displayed the greatest changes in gene expression (Table 1). Expression patterns of genes for entire metabolic pathways were altered in plin^−/− mice. The major pathways that were affected include β-oxidation, the Krebs cycle, and the electron transport chain (Tables 2 and 3, Fig. 2). Many genes involved in transcription and translation regulation were also downregulated in plin^−/− WAT, accounting for 33% of all its downregulated transcripts. All of the genes that are significantly changed in plin^−/− WAT are listed in online appendix 3.

The rate-limiting step for β-oxidation of long-chain fatty acids is their transport across the inner mitochondrial membrane via the carnitine palmitoyl transferase (Cpt) and carnitine/acylcarnitine translocase (Cac) system. Both Cpt2 and Cac genes were significantly upregulated in the plin^−/− mice. The transcripts of two of the three enzymes involved in β-oxidation, namely, 3-ketoacyl-CoA thiolase B and enoyl CoA hydratase, were also significantly upregulated in the WAT of plin^−/− mice. The mRNA for 2,4-dienoyl CoA reductase 1, which is involved in the oxidation of unsaturated fatty enoyl-CoA esters (that have double bonds in both odd- and even-numbered positions before their further degradation via β-oxidation), was also upregulated significantly in the plin^−/− mice. In kidney, where it has the strongest expression, the gene for the organic cation/carnitine transporter 2 (Octn2 or Slc22a5) was also upregulated. This transporter is involved in an active sodium-dependent high-affinity cellular uptake of carnitine. Thus, in the WAT of plin^−/− mice, there is a coordinated upregulation of multiple genes regulating fatty acid oxidation (Fig. 2).

Multiple transcripts of fatty acid/cholesterol biosynthesis-related pathways were downregulated in WAT; a few of these transcripts were also downregulated in muscle, kidney, heart, and liver (Table 2). In WAT, the transcript for ATP citrate-lyase, the primary enzyme responsible for the synthesis of cytosolic acetyl-CoA, which plays a central role in de novo lipid biosynthesis, was decreased. Transcripts for members of the Scd family, which are involved in a rate-limiting step in the synthesis of monounsaturated fatty acids (see below), were also significantly decreased in WAT of the plin^−/− mice (Fig. 3).

The mRNAs for fatty-acyl CoA ligase long chain 4 and for long-chain fatty-acyl elongase were also decreased in the plin^−/− mice. The former activates long-chain fatty acids for the synthesis of cellular lipids that preferentially use arachidonate and eicosapentanoate as substrates, and the latter is involved in converting palmitic to stearic acid (14).

The mRNAs for a number of sterol biosynthetic enzymes were downregulated. These included the transcripts for isopentenyl diphosphate δ isomerase, which participates in the isoprenoid biosynthetic pathway, giving end products that include steroid hormones, bile acids, and cholesterol; squalene epoxidase, a potential target for lipid/cholesterol lowering therapies (15–17) that catalyzes the first oxygenation step in sterol biosynthesis; and lanosterol 14-α-demethylase (Cyp51), also a target in cholesterol therapy (18–20), which catalyzes the removal of the 14 α-methyl group from lanosterol, a critical step in the production of ergosterol in the cholesterol biosynthetic pathway (Fig. 3).

In the kidney, the gene for hydroxysteroid dehydrogenase 10, an oxidoreductase participating in lipid metabolism, was downregulated. In skeletal muscle, the transcript for oxysterol binding protein-like 1A was also downregulated. Oxysterol binding protein is an intracellular transporter of oxygenated derivatives of cholesterol from lysosomes to the nucleus, where they downregulate genes for the LDL receptor. In the liver, the mRNAs for steroid 21-hydroxylase and steroid sulfatase, which are involved in the conversion of sulfated steroid precursors, were also significantly downregulated (Table 2).

A notable finding in plin^−/− WAT was the coordinated downregulation of the stearoyl coenzyme A desaturase (Scd) genes Scd1 and Scd2 (Fig. 3, online appendices 4 and 5). Scd1 mRNA was expressed at a high level in WAT of wild-type mice and was significantly downregulated in plin^−/− mice.

The Scd family, also known as δ-9-desaturase in humans and asebia in mice, is composed of iron-containing enzymes that catalyze a rate-limiting step in the synthesis of monounsaturated fatty acids by introducing the first double bond, mainly in oleate and palmitoleate. Scd expression is modulated by dietary factors (e.g., polyunsaturated fatty acids, cholesterol, vitamin A [21–23]), hormonal signals (e.g., insulin, glucagon [24]), environmental factors (e.g., temperature changes, thiazolinediones, metals, alcohol, phenolic compounds [25–28]), peroxisomal proliferators (29), and developmental processes (30). Altered Scd activity has been observed in a wide range of disorders, including diabetes, atherosclerosis, cancer, obesity, and viral infections (25,31,32).

The plin^−/− mice are characterized by a ∼60% decrease in total body lipid content but have the same body weight as their wild-type littermates, despite an ∼30% increase in caloric intake (7). They also exhibit increased O₂ consumption, reflecting an increased metabolic rate compared with wild-type mice. A similar phenotype is evident in mice with Scd1 deficiency, occurring either for a natural mutation, as in the asebia mice (33), or as a result of gene targeting, as in the scd1^−/− mice (34). They exhibit an ∼50% decrease in adiposity with a normal body weight and an ∼25% increase in their food intake, accompanied by an increase in their O₂ consumption. Furthermore, like the plin^−/− mice, the scd1^−/− mice show an upregulation of the fatty acid oxidation genes with a concomitant downregulation of the fatty acid synthesis genes. These parallel observations suggest that the downstream effects in both of these models, such as the relative or absolute increase in saturated fatty acids in the tissue, concomitant with a decrease of unsaturated fatty acids in the absence of Scd1 (scd1^−/− mice), or with a decrease in expression of both Scd1 and Scd2 in the plin^−/− mice (Table 2), may contribute to some of the observed changes in these models. However, other changes in gene expression profile in the plin^−/− mice and not reported in the scd1^−/− mouse point to additional regulatory mechanisms that come into play in the plin^−/− mice.

Sterol regulatory element-binding protein-1 (SREBP-1) has been postulated to be a regulator of Scd (35–37). In this study, we note a significant alteration in the expression of multiple members of the Scd gene family, all without a significant change in the expression of SREBP-1. Thus, it is likely that this downregulation of the Scd genes in the plin^−/− mice occurs either downstream of SREBP-1 or via a separate regulatory pathway. SREBP-1-independent regulation of the expression of Scd1 by cholesterol has been reported previously (38,39). The transcripts for Scd1 and Scd2 together with a Scd1-like gene all were significantly downregulated in the WAT, and that for Scd1 in muscle and Scd2 in kidney were significantly downregulated in plin^−/− mice (Table 2).

Transcript alignment by hierarchical clustering is useful to identify genes coordinately regulated and selectively expressed among different tissues. Hierarchical clustering (dChip V 1.2) shows that in plin^−/− WAT, transcripts for Scd1 and Scd2, were present in the same cluster (online appendix 4). Transcripts for ATP citrate lyase, a major factor determining the level of cytosolic acetyl CoA (see below), also occurred in the same cluster, suggesting a coordinated downregulation for these genes. It is interesting that a similar clustering of these genes was seen in the leptin-regulated genes studied in leptin-treated wild-type WAT (40). Hierarchical clustering of five different tissues (WAT, liver, heart, muscle, and kidney) from both plin^−/− and wild-type revealed the following genes as coordinately regulated by this analysis in relation to Scd: long-chain fatty acyl elongase clustered with Scd2, whereas pyruvate dehydrogenase kinase 1, aldehyde dehydrogenase 1A4, and eukaryotic translation initiation factor 1A clustered with Scd1, all being significantly downregulated in WAT of plin^−/− mice (online appendix 5).

The major portion of the energy from the oxidation of carbohydrates and fatty acids comes from the generation of reducing equivalents in the Krebs cycle and the subsequent generation of ATP via redox reactions of the electron transport chain. The Krebs cycle, fatty acid β-oxidation, and the electron transport-mediated ATP generation occur inside the mitochondria, whereas glycolysis, lipolysis, and fatty acid synthesis are cytosolic events. The mitochondrial membrane is impermeable to acetyl CoA, the common intermediate through which both glucose and fatty acids are fed into the Krebs cycle. For glucose to be converted into acetyl CoA, it is first converted to pyruvate in the cytosol through glycolysis. Pyruvate is transported and converted into acetyl CoA in the mitochondria by the pyruvate dehydrogenase complex, a key mitochondrial membrane-bound multisubunit enzyme.

The transcript for dihydrolipoamide S-acetyltransferase, the E2 component of the pyruvate dehydrogenase, was downregulated in plin^−/− mice. This downregulation would be expected to inhibit the conversion of pyruvate into acetyl CoA and direct pyruvate to either oxaloacetate by pyruvate carboxylase or to lactate. However, there also was a decrease in the transcript of pyruvate dehydrogenase kinase 1 (Pdk1), which normally inhibits Pdh by phosphorylating the E1a component of the Pdh complex. In this context, it is important to note that the key enzyme in determining fuel selection has been postulated to be Pdk4 (41–43), whereas the role of Pdk1 in this process is unclear. Although Pdk4 transcript level was not changed significantly, its activity in the plin^−/− mice is expected to be increased, as NADH is reportedly a strong activator of this enzyme (41), and intramitochondrial NADH levels are expected to be elevated in these mice as a result of the increased β-oxidation.

In addition to the above, transcripts for enzymes of the Krebs cycle, such as isocitrate dehydrogenase [NAD⁺]γ, succinate-CoA ligase Gdp-forming α, succinate dehydrogenase, and malate dehydrogenase, all were upregulated in plin^−/− mice (Fig. 2, online appendix 6). These four steps represent four of the five irreversible steps of the Krebs cycle; they catalyze reactions that generate reducing equivalents (NADH, FADH2) and GTP. Their upregulation suggests that the flux through the Krebs cycle is maintained at a higher level to keep up with the activated lipolysis and β-oxidation in the plin^−/− mice.

In the plin^−/− mice, with the increase in acetyl CoA generated from an activated β-oxidation, a rise in intramitochondrial citrate (the first intermediate generated in the Krebs cycle) would be expected. This increase would result in an increase in cytoplasmic citrate, because citrate is transported across the mitochondrial membrane via the tricarboxylate (citrate) carrier. Cytoplasmic citrate plays an important role in the regulation of both carbohydrate and lipid metabolism. Increased cytoplasmic citrate allosterically inhibits the rate-limiting step of glycolysis catalyzed by phosphofructo kinase-1, which is further inhibited by an increase in the cellular ATP/ADP ratio in these animals, as would be expected by the increased fatty acid oxidation and the activated Krebs cycle.

The citrate-malate shuttle is involved in transporting acetyl CoA, generated in the mitochondria, to the cytoplasm, where it acts as a precursor to malonyl CoA, the substrate of fatty acid synthesis. Once in the cytosol, citrate is broken down into oxaloacetate and acetyl CoA by ATP-citrate lyase, and the transcript for this enzyme was downregulated in the plin^−/− mice. Hence, although intramitochondrial acetyl CoA levels may be increased in the plin^−/− mice, this is not likely to be reflected in the cytoplasm, because of this downregulation. As a consequence, there would be a decrease in substrate available for acetyl CoA carboxylase, which generates malonyl CoA from acetyl CoA and with it a decreased precursor pool for fatty acid synthesis. In addition, the decrease in malonyl CoA would decrease the inhibition on Cpt and further stimulate an increase in β-oxidation. Furthermore, the higher levels of saturated fatty acids that may result from the decrease in Scd expression have been postulated to inhibit acetyl CoA carboxylase (34,44), further decreasing malonyl CoA and increasing β-oxidation.

The mitochondrial electron chain is composed of five complexes, each consisting of multiple subunits. In the plin^−/− mouse WAT, there was an upregulation of transcripts for at least one subunit of each of the five complexes of the electron transport chain (Table 3, Fig. 2, online appendix 6). The coordinated increase in these transcripts is in keeping with the demands of the increased flux of reducing equivalents generated by activated β-oxidation and the Krebs cycle. The consumption of molecular oxygen for oxidative metabolism mediated by the activation of the electron transport chain accounts for part of the increase in O₂ consumption observed in the plin^−/− mice. However, these changes in WAT probably do not account for all of the increase in O₂ consumption seen in plin^−/− mice (11 and 20% while on regular diet and high-fat diets, respectively) (7), as these changes were not reflected in skeletal muscle, which normally contributes to a larger portion of the total body O₂ consumption.

Although we observed a 25% increase in the transcript level for peroxisome proliferator-activated receptor-α (PPAR-α) in this microarray analysis, this difference between plin^−/− and wild-type WAT did not reach statistical significance. As PPAR-α regulates many of the genes involved in fatty acid oxidation and thus is a possible candidate to mediate many of the changes observed in plin^−/− mice, we studied the expression of PPAR-α in these mice using real-time quantitative RT-PCR as a complementary approach. PPAR-α was seen to be significantly increased by fourfold in the plin^−/− mice as compared with the wild-type controls (online appendix 2). A few other well-studied genes involved in lipid and carbohydrate metabolism, including leptin, resistin, and adiponectin, are notable for the lack of a significant difference in expression between plin^−/− and wild-type mice.

Aquaporin 7, a water/glycerol channel in WAT, was upregulated in plin^−/− mice. The PPAR-α target gene fasting-induced adipocyte factor, a novel endocrine signal involved in the regulation of lipid metabolism (34,45), was upregulated in plin^−/− mice, as was observed in Scd1^−/− mice (34). We also found that in the WAT of plin^−/− mice, the transcripts for 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 1 and triosephosphate isomerase were upregulated. Whereas in the WAT of plin^−/− mice there was an upregulation of the transcripts for prostacyclin synthase (involved in prostacyclin synthesis) and that for 5-lipoxygenase activating protein (involved in leukotriene synthesis), there was a downregulation of the transcripts for phospholipase A2 and phospholipase A2 activating protein, both involved in the generation of arachidonate, the precursor substrate for all of the eicosanoids.

We observed in plin^−/− WAT an increased expression of transcripts normally thought to be associated with leukocytes (Table 3). Among them are transcripts for NADPH-dependent electron transporters (46), together with the colony-stimulating factor 1 and 2 receptors, the transcription factor Pu1, recently reported as uniquely expressed in the stromal fraction isolated from wild-type and ob/ob adipose tissue (47), F4/80, and Mac-1β (48). Recently, the conversion of preadipocytes into macrophages has been demonstrated (48), whereas on the flip side, macrophages, which accumulate lipid and become foam cells in atherosclerotic plaques, have been found expressing adipophilin (49) and perilipin transcripts (50), pointing to the expression of common genes and regulatory pathways in these different cell types. However, these have to be interpreted cautiously as the WAT also has other nonadipose components, especially a vascular tissue component, which may contribute to a disproportionately higher leukocyte associated transcripts in the plin^−/− mice, as a result of their lower relative adipose content. An annotated list of the genes for NADPH-dependent electron transporters is presented in online appendix 6. Transcripts of genes involved in heme synthesis, such as heme oxygenase (decycling) 1 and δ aminolevulinic acid synthase 1 and 2, were also increased in the plin^−/− WAT.

We have previously shown that perilipin gene ablation leads to constitutively activated lipolysis, resulting in an increase in free fatty acids and glycerol in WAT (7). We now demonstrate mechanisms through which the plin^−/− mice adapt to this altered metabolism by activating mechanisms to dispose of these lipolytic products through upregulation of oxidative catabolic pathways and downregulation of lipid/sterol synthetic pathways. This coordinated program of changes in oxidative catabolism and pathways of energy expenditure thus would underlie the obesity resistance observed in these mice.

Although perilipin has not been previously implicated in any transcriptional regulation, either directly or indirectly, we show that the disruption of perilipin leads to extensive changes in gene expression in the WAT. Because the constitutively activated lipolysis is the primary consequence of the lack of perilipin, the resultant increase in intracellular fatty acids and their subsequent oxidation products would be the most likely candidates to mediate many of the observed changes in these mice. Although the exact mechanism of nutrient sensing remains unclear, it is likely that these lipid intermediates may act as ligands, co-factors, or co-activators for transcription factors that are the effectors of the alterations observed in the global gene expression. Indeed, the observation that all of these metabolic pathways, with multiple components of each pathway, are coordinately changed argues strongly for a set of transcriptional factors or co-activators as mediators for some of the observed changes. It is interesting that the increased expression levels of PPAR-α observed in the plin^−/− mice offer one potential mechanism for a large number of these coordinated changes in gene expression. However, other transcriptional factors and co-activators may also play a role, although candidate factors as other PPARs, nuclear respiratory factors, and PPAR-γ coactivator-1 were not significantly changed in the plin^−/− mice. We further note that the absence of perilipin in the WAT also leads to concomitant changes in gene expression in other tissues where perilipin is not normally expressed. This reflects an intricate cross-talk between tissues. This cross-talk between tissues is likely to be mediated via the lipid ligands themselves or via secreted molecules such as adipocyte-derived cytokines and hormones. The identification of the mediator(s) for this cross-talk, however, awaits further study.

This work was supported by a grant (HL-51586) from the National Institutes of Health to L.C. and by a grant (72570) from CONACyT, Mexico, to F.C.C. L.C. was also supported by the Betty Rutherford Chair in Diabetes Research at St. Luke’s Episcopal Hospital and Baylor College of Medicine.

We thank Arturo Panduro for encouraging F.C.C. to pursue this work; Rainer Lanz for the use of the ABI Prism sequence detection system for the real-time PCR experiments; Benny Chang, Ke Ma, Susan Hilsenbeck, Patrick Younan, and Darin Tessier for technical advice; and Tracy Lynn Duncan and Leslie J. Wu for help in preparing the manuscript.