Non-DNA binding, dominant-negative, human PPARγ mutations cause lipodystrophic insulin resistance

1The nuclear receptor (NR) peroxisome proliferator-activated receptor γ (PPARγ) is a ligand-inducible transcription factor that is essential for adipocyte differentiation. Alternative splicing and differential promoter usage generates two protein isoforms: PPARγ2, expressed from a single γ2 promoter, contains an additional 28 amino-terminal amino acids and is nearly adipose-specific; PPARγ1, whose expression can be regulated by multiple (γ1, γ3, γ4) promoters, is more ubiquitously distributed. In addition to adipogenesis, PPARγ also plays an important role in adipocyte lipid metabolism, regulating target genes (lipoprotein lipase, fatty-acid transport protein, aquaporin) that mediate triglyceride hydrolysis and fatty acid and glycerol uptake, together with genes (acylCoA synthetase, PEPCK, glycerol kinase) involved in fatty acid re-esterification and lipid storage. The thiazolidinedione (TZD) class of antidiabetic agents are synthetic, high-affinity PPARγ ligands and putative endogenous activators include fatty acids, eicosanoids, and prostaglandin derivatives as well as undefined ligands produced during adipocyte differentiation.

2The most common population genetic variant of PPARγ is a polymorphism replacing alanine for proline at codon 12 (Pro12Ala) in PPARγ2, with a meta-analysis of association studies showing that the Pro allele confers a modest but significant increase in diabetes risk. The discovery that PPARγ is a target for TZDs, which act by enhancing tissue insulin sensitivity, prompted screening of a cohort of subjects with severe insulin resistance, with identification of two missense PPARγ mutations (P467L, V290M) in unrelated cases. Functional studies showed that these mutant receptors retain DNA binding but exhibit significant impairment of transcriptional activation and coactivator recruitment in response to different ligands, due to the mutations destabilizing the carboxyterminal α helix of PPARγ, which mediates these functions. Consonant with heterozygosity in affected subjects and dominant inheritance in one kindred, the P467L and V290M mutant receptors inhibited the transcriptional activity of wild-type (WT) PPARγ in a dominant-negative manner. Subsequently, two further heterozygous mutations in the ligand binding domain (LBD) of PPARγ (R425C; F388L) have been described, with recognition that in addition to insulin resistance the phenotype also includes a stereotyped pattern of partial lipodystrophy (PLD).

3Following this, we described several individuals who were heterozygous for a frameshift/premature stop codon mutation, ([A553ΔAAAiT]fs.185[stop186]-hereafter abbreviated to FSX) in the DNA binding domain (DBD) of PPARγ, with this truncation mutant lacking DNA binding, transcriptional, and dominant-negative activity. Significantly, heterozygosity for the FSX mutation alone was not associated with insulin resistance, but individuals who were doubly heterozygous, with an additional defect in an unrelated gene encoding the muscle-specific regulatory subunit of protein phosphatase 1 (PPP1R3A), exhibited severe insulin resistance. Heterozygosity for a single nucleotide substitution in the promoter of human PPARγ4 leading to its altered expression in vitro has been associated with PLD and insulin resistance in one family, but the authors did not exclude the possibility of interaction with a defect at a second genetic locus to produce this phenotype.

4Here, we describe the identification of five heterozygous human PPARγ mutations (C114R, C131Y, C162W, R357X, [A935ΔC]fs.312[stop315]-hereafter abbreviated to FS315X) not associated with a PPP1R3A gene defect, in unrelated cases of lipodystrophic insulin resistance and show that these mutants inhibit WT receptor action via a non-DNA binding, dominant-negative mechanism.

5The case histories (see the Supplemental Data available with this article online) and characterization (Table 1) of index subjects (S1–S5) harboring PPARγ mutations indicate many of the features associated with previously described cases. All subjects showed marked fasting hyperinsulinaemia (Table 1) with acanthosis nigricans in a subset (S3, S4, S5), denoting severe insulin resistance; total body fat was reduced in all individuals, and imaging indicated a stereotyped pattern of partial lipodystrophy affecting gluteal (Figure S1) and peripheral limb depots; hepatic steatosis and marked dyslipidaemia (raised triglycerides, low high-density lipoprotein cholesterol [HDL-C]) with secondary complications (cutaneous eruptive xanthomata S3, S4; pancreatitis S5) were features of all cases; several individuals (S2, S3, S5) exhibited early-onset hypertension.

6We sequenced the γ4 promoter, coding exons and splice junctions of PPARG and identified heterozygous, missense mutations in the DBD (S1–S3), or premature stop mutations in the LBD (S4, S5) of the receptor in index cases. PPARG has also been sequenced by us in 215 additional subjects, comprising 93 patients from our severe insulin resistance cohort, 48 CEPH individuals of European descent and 27 Europid, hyperinsulinaemic participants in the Ely study, and 47 controls from four different ethnic groups, or sequenced by others in 24 African and 23 CEPH European individuals (Seattle SNPs project, http://pga.gs.washington.edu), and other than the Pro12Ala polymorphism neither these or other mutations have been identified. We also sequenced PPP1R3A in each proband and identified no mutations or polymorphisms, excluding a second genetic defect at this locus as described previously.

7Heterozygosity for PPARγ mutations in a parent and grandparent of S3 and a parent of S5 segregated with phenotype, constituting a dominant inheritance pattern in two families; one sibling of S2 with dyslipidaemia and insulin resistance was heterozygous for the PPARγ mutation whereas another genetically unaffected sibling was biochemically normal; the ascertainable family members of S1 were unaffected and normal and no relatives of S4 could be contacted (Figure 1B).

8Three missense mutations involve highly conserved cysteine residues within (C114R, C131Y, C162W) the DBD and two further nonsense (R357X) or frameshift/premature stop (FS315X) mutations truncate the receptor within the central part of its LBD (Figure 1A), predicting loss-of-function of the mutant proteins. We therefore characterised and compared the properties of these PPARγ mutants with the FSX mutant described previously.

9The receptor mutants exhibited negligible transcriptional activity, lacking constitutive basal activity noted previously with WT PPARγ as well as any response to rosiglitazone, a TZD receptor agonist (Figure 1C). Such complete loss of function was similar to the FSX mutant and might be anticipated with analogous truncation mutants (FS315X, R357X) not possessing the transactivation (AF2) domain at the receptor carboxyterminus (Figure 1A), but the lack of function with DBD mutants (C114R, C131Y, C162W), prompted further investigation of their DNA binding properties.

10PPARγ heterodimerizes with the retinoid X receptor (RXR) and this complex has been shown to bind a DNA response element (PPARE), consisting of a direct repeat (DR1) of the consensus sequence (AGGTCA) separated by a single nucleotide and a recent study has suggested that the stringency of PPARγ binding to some response elements is relatively relaxed, not needing complete integrity of its DBD. A range of previously documented or predicted PPAREs from known target genes were therefore tested in electrophoretic mobility shift assays and both DBD and LBD truncation receptor mutants showed negligible heterodimeric binding (Figure 1D). To examine interaction of mutant receptors with RXR, we coexpressed VP16-full length PPARγ fusions with Gal4DBD-RXR in a mammalian two-hybrid assay. In keeping with preservation of the dimerization interface within their intact LBD (Figure 1A), the DBD mutants interacted readily whereas the FS315X, R357X, and FSX mutants lacking this interface failed to be recruited to Gal4-RXR (Figure S2). It was therefore conceivable that the DBD mutants could be recruited indirectly to a PPARE by binding RXR, or conversely, that the LBD truncation mutants might bind a PPARE monomerically as has been documented with the thyroid hormone receptor (TR). However, unlike WT receptor, VP16-full length, mutant PPARγ fusions were unable to activate a PPARE-containing reporter gene (Figure S3), indicating that like FSX, these PPARγ mutants do not bind DNA directly or indirectly.

11The intracellular localization of WT PPARγ is predominantly nuclear and, analogous to steroid/thyroid hormone receptors, may be dependent on a putative nuclear localisation signal (NLS) located between its DBD and LBD (Figure 1A). Studies of GFP- PPARγ fusions showed that, in keeping with preservation of the putative NLS, both DBD and LBD truncation mutants localized to the nucleus comparably to WT, whereas the FSX truncation mutant, which lacks this sequence, remained cytoplasmic similar to GFP alone (Figure 1E).

12We next examined whether the PPARγ mutants might also retain the ability to interact with transcriptional coactivators. Steroid receptor coactivator-1 (SRC1/NCoA1) and PPARγ binding protein/thyroid receptor-associated protein 220 (PBP/TRAP220) interact directly with the AF2 domain of PPARγ, with the latter cofactor being required for receptor-mediated adipogenesis. Consistent with preservation of their AF2 domains, protein-protein interaction assays showed ligand-dependent binding of SRC1 or TRAP220 to the DBD mutants, but no specific interaction with FSX or LBD truncation mutants, which lack this region (Figure 1F). Conversely, we hypothesized that the PPARγ LBD truncation mutants would retain the ability to recruit coactivators, which can interact with receptor independently of its AF2 domain. PPARγ coactivator-1 (PGC1), which augments receptor action in fat cells, can bind PPARγ via its DBD and hinge region (αα 128-229); PDIP, isolated in a two hybrid assay using the DBD/hinge region of PPARγ, is a coactivator that also enhances PPARα activity. Both PGC1 and PDIP bound WT or FS315X and R357X LBD truncation mutants in protein-protein interaction assays, whereas the FSX mutant showed negligible interaction (Figure 1G).

13The observation that these PPARγ mutants translocate to the nucleus and interact with coactivators raised the possibility that they might interfere with WT receptor signaling. The murine adipocyte P2 (aP2) gene is a classical target of PPARγ action and the human homolog (FABP4) is similarly responsive. When coexpressed with WT PPARγ at equivalent levels in 3T3-L1 adipocytes, the DBD and LBD mutants blocked WT receptor mediated activation of the human aP2/FABP4 gene promoter comparably to an artificial, dominant-negative PPARγ mutant (AF2) described previously, whereas FSX lacked dominant-negative inhibitory activity (Figure 2A).

14We wished to determine whether such divergent dominant-negative inhibition by these PPARγ mutants versus FSX might operate in vivo. PPARγ is highly expressed in immature dendritic cells (IDCs) derived from primary human blood monocytes and mediates marked receptor responsiveness, with strong ligand-dependent induction of aP2 expression in these cells. Induction of aP2/FABP4 expression in IDCs containing DBD or LBD PPARγ mutations was severely attenuated compared to responses in control cells from either normal individuals (WT) or from subjects (IR) with comparable insulin resistance without a PPARγ gene defect. Significantly, aP2 induction in FSX mutation-containing cells was comparable to responses from control subjects (Figure 2B). We examined other PPARγ target genes, identified from extensive microarray profiling of normal IDCs (I.S. and L.N., unpublished data) and found that responses to PPARγ agonist in DBD and LBD truncation mutation-containing cells were markedly attenuated whereas FSX mutation-containing cells exhibited responses that were either similar or only slightly reduced compared to WT cells (Figure 2C). PPARγ mRNA levels in control and mutation-containing primary cells were similar (data not shown), suggesting that differential responsiveness was not due to altered receptor expression. Furthermore, PPARγ mRNA from both WT and R357X alleles was expressed in mutation-containing IDCs (Figure 2D), indicating that the R357X transcript is not subject to nonsense-mediated decay and both WT and R357X mutant PPARγ proteins were also expressed in these cells (Figure 2E).

15Finally, we determined whether dominant-negative inhibition by a non-DNA binding PPARγ mutant could interfere with a receptor-mediated biological process. Compared to control, WT PPARγ or GFP adenovirus-transduced human preadipocyte cells, both cellular differentiation (Figure 3A) and aP2 gene expression (Figure 3B) in cells transduced with C114R mutant PPARγ adenovirus were significantly reduced.

16We have shown previously that dominant-negative inhibition by PPARγ mutants (P467L, V290M), is mediated by repression of target genes by DNA-bound mutant receptors, analogous to mechanisms of other mutant nuclear receptors (e.g., the v-erbA oncogene, TRβ mutants in Resistance to Thyroid Hormone, PZLF-RARα fusion proteins in promyelocytic leukaemia). In contrast, the missense DBD and LBD truncation mutants identified here are unable to bind DNA, yet can inhibit WT PPARγ action, suggesting a different mechanism of transcriptional interference. Competition for shared cofactors by NRs was postulated to explain mutual antagonism of progesterone and estrogen receptor signaling and the subsequent observation that SRC1, a shared coactivator, could relieve such “squelching”, validated this hypothesis. Ligand-activated NRs have been shown to inhibit either their own function or that of heterologous receptors by limiting the availability of coactivators that are recruited to their transactivation domains. Our observations indicate that non-DNA binding, dominant-negative PPARγ mutants can recruit coactivators, suggesting an analogous cofactor sequestration mechanism for thereby restricting WT receptor function. Evidence suggests that similar mechanisms operate to inhibit PPAR signaling in other contexts: analogous to our natural DBD mutants, others have generated artificial, dominant-negative, PPARγ DBD mutants, which block either adipogenesis or neural stem cell differentiation; γORF4 is a newly identified human PPARγ splice variant with a truncated LBD (αα273), which has dominant-negative activity and is selectively overexpressed in colorectal neoplasia; a dominant-negative PPARα splice variant with a truncated LBD (αα 174), is expressed in human tissues including liver. Interestingly, heterozygous, non-DNA binding mutations in some nuclear receptors do not mediate a phenotype: mutations in the DBD of VDR only cause vitamin D resistance in the homozygous state; a “knock-in” mutation in the DBD of murine TRβ does not produce thyroid hormone resistance. Possibly due to its pivotal role in regulating transcription of genes mediating both adipocyte formation and function, we suggest that PPARγ signaling may be particularly sensitive to interference via the postulated “squelching” mechanism, with deleterious metabolic consequences. A corollary of this may be that even modest enhancement of normal receptor activity in key tissues could be beneficial, supporting attempts to develop partial or tissue-specific PPARγ agonists.

17The PPP1R3A (exons 1-4) and PPARγ (exons 1-6, B and promoter region of PPARγ4) genes were amplified using specific primers (available upon request) and sequenced as decribed previously.

18Full length WT and mutant PPARγ1 cDNAs were cloned in pGEX4T (Amersham Pharmacia Biotech), pCMX-VP16 (kind gift from R. Evans), pSG424 and pEGFP-C1 (Clontech), to yield GST-PPARγ1, VP16-PPARγ1, Gal4DBD-PPARγ and GFP-PPARγ1 fusions respectively.

19Electrophoretic mobility shift assays (EMSA) were performed as described with different natural PPAREs: aP2, derived by alignment of human and murine promoter sequences; Adiponectin: ACoABP; mCPT1,; LXRα,; CAP1,; LPL,; ACoAOx,; ACoAOx.

20293EBNA cells, cultured in DMEM/10%FCS were transfected with Lipofectamine2000- or calcium phosphate-mediated in 96- or 24-well plates respectively and assayed for luciferase and β-galactosidase activity as described following 36 hr with or without ligand. 3T3-L1 adipocyte cells were cultured and transfected with Lipofectamine2000 in 24-well plates as described above.

21293EBNA cells, grown on glass well slides were transfected using Lipofectamine 2000 with 1μg of EGFP-PPARγ1 fusions, fixed with 4% paraformaldehyde, mounted using vectashield and fluorescence was visualized by digital microscopy.

22With ethical approval, monocytes were harvested from peripheral blood by Ficoll gradient centrifugation and immunomagnetic cell separation using anti-CD14-conjugated microbeads (VarioMACS; Miltenyi Biotec), resuspended in 6-well plates at a density of 1.5 × 106 cells/ml and cultured in RPMI 1640 plus 10% FBS containing 800U/ml GM-CSF (Leucomax) and 500U/ml IL-4 (Peprotech) to generate IDCs as described with or without exposure to ligand for 24 hr.

23100ng of total RNA from IDCs, isolated using TRIZOL (Invitrogen), was reverse transcribed and analyzed by Taqman quantitative real-time PCR (qPCR) as described. The sequences of primers and probes are available upon request.

24Taqman qPCR low density arrays (TLDA) were used to quantify the expression of multiple target genes in IDCs, according to the manufacturer's instructions.

25To obtain cDNA, RNA was reverse transcribed using a High Capacity cDNA Archive kit (Applied Biosystems). The following commercially available Taqman assays (Applied Biosystems) were used: ADRP/ADFP (Hs00605340_m1), APOC1 (Hs00155790_m1), CLDN1 (Hs00221623_m1), aP2/FABP4 (Hs00609791_m1), CLECSF5 (Hs00183780_m1), CD1E (Hs00229421_m1), MYO1B (Hs00362654_m1), IL1R2 (Hs00174759_m1), OAS1 (Hs00242943_m1), p30 (Hs00396457_m1), cyclophilinA/PPIA (Hs99999904_m1). The comparative Ct method was used to quantify transcripts and normalize to cyclophilinA expression levels, which did not vary with ligand treatment. Thereafter, data were further normalized to expression levels in ligand-treated WT IDC samples using GeneSpring 7.2 software (Agilent).

26PPARγ cDNAs were amplified from WT or R357X mutation-containing IDCs by RT-PCR using forward (CTCCTTGATGAATAAAGATGGGG) and reverse (ATGTCTTCAATGGGCTTCACAT) primers, the PCR products were digested with Cac8I enzyme (New England Biolabs) and analyzed by agarose gel electrophoresis.

27IDCs, harvested from 200ml of peripheral blood, were lysed in RIPA buffer containing a protease inhibitor cocktail (Roche) and cell supernatants immunoprecipitated using a mouse monoclonal anti-PPARγ antibody (K8713, Perseus Proteomics) and analyzed by SDS-PAGE. Western blotting was carried out using a rabbit polyclonal anti-PPARγ antibody (H-100, Santa Cruz Biotechnology).

28Recombinant type 5 adenoviruses (Ad5) expressing GFP alone or with either WT or C114R mutant PPARγ1 were generated using the AdEasy Vector System (Quantum Biotechnologies, Montreal), amplified and purified as described. 6-well plates of Chub-S7 human preadipocyte cells were cultured and infected with 2x107 pfu/well of recombinant virus 24 hr prior to differentiation in the presence of 100nM rosiglitazone as described. Comparable infection efficiency was verified by fluorescence microscopy with subsequent qPCR analysis on days 0, 3, 5 and 7. Fully differentiated cells were fixed and stained with Oil Red-O as described.