Regulation of Apolipoprotein M Gene Expression by MODY3 Gene Hepatocyte Nuclear Factor-1α: Haploinsufficiency Is Associated With Reduced Serum Apolipoprotein M Levels

Hnf-1α null mice and their heterozygous littermates were kept in a sterile micro isolator and maintained in the Laboratory of Animal Research Center, a pathogen-free animal facility at Rockefeller University (17). The animals were maintained on a 12-h light/dark cycle and fed a standard rodent diet.

The apoM cDNA was amplified by PCR and cloned into pBS and expression vector pGEX-4T3 using primers GST apoM1 5′-CGGGATCCCGAGCGTTCCACCAAGTTTGGGCAGCTCTG-3′ and GST apoM2 5′-CCGCTCGAGCGGGTGAAGTCACTGGTCACTTGC-3′. The reporter plasmid pGL2-papoM was generated by cloning a 1,055-bp fragment of the mouse apoM promoter into pGL2-luciferase. The reporter plasmid pGL2papoMmut is identical to pGL2-papoM except for a mutated HNF-1 binding site at position −103 to −88 (GAAGGGGCGTATTAGCA).

The mouse apoM cDNA was cloned into the Nde/BamHI sites of expression vector PET3a (Novagen), before transformation into Escherichia coli Bl21(DE3). Protein expression was induced by 0.1 mmol/l isopropyl-β-d-thiogalactoside for 3 h at 37°C. Cells were harvested and washed in buffer A (50 mmol/l Tris, 100 mmol/l NaCl, 5% glycerol, 1 mmol/l DTT, 0.02% Triton X [pH 7.6]) and resuspended in buffer A before sonication. The sonicated cells were centrifuged at 11,000 rpm for 30 min. The pellet was washed three times in buffer A, once with 0.5 mol/l guanidine HCl, and twice with 1 mol/l guanidine HCl. The protein was eluted from the inclusion bodies in 3 mol/l guanidine HCl solution and centrifuged for 2 h at 40,000 rpm. The supernatant containing the recombinant apoM protein was dialyzed against dialysis buffer (0.1 mol/l NaHC0₃, 0.5 mol/l NaCl [pH 8.5]) for 12–16 h.

Two 20–amino acid peptides (apoM113–132: TEGRPDMKTELFSSSCPGGI and apoM140–159: GYQRFLLYNRSPHPPEKCVE) were synthesized and processed to >90% purity, conjugated to KLH, and used for immunization of rabbits (Bethyl Laboratories, Montgomery, TX). Antisera were affinity purified and tested by enzyme-linked immunosorbent assay and Western blotting. Affinity-purified antiserum apoM140–159 was used for all studies.

Total protein extracts (10 μg) of livers and kidney from wild-type, heterozygous, and null Hnf-1α mice were separated by SDS-PAGE (12.5%) and transferred onto a nitrocellulose membrane (Schleicher & Schuell) by electroblotting (30 min, 15 V). ApoM protein was detected with anti–apoM140–159 antiserum (1:1,000) and goat anti-rabbit IgGs conjugated to horseradish peroxidase (1:10,000) in Tris-buffered saline supplemented with 5% nonfat dry milk. For visualization, the Renaissance chemoluminescence substrate (NEN) was used.

Semiquantitative RT-PCR was performed as described previously (19). PCR synthesis for each primer pair was quantified at 15, 20, 25, and 30 cycles in a test reaction to ensure that the quantitative PCR amplification was in the linear range. The primer sequences apoM-F 5′-GTGCCCCGGAAGTGGACATACC-3′ and apoM-R 5′-AGCGGGCAGGCCTCTTGATTC-3′ amplify a 275-bp fragment and were used for semiquantitative RT-PCR analysis.

Electrophoretic mobility shift assay (EMSA) analysis was performed as described previously (16). Nuclear cell extracts were incubated with ³²P-labeled ds-stranded oligonucleotide probes containing the wild-type or mutant HNF-1 binding sites in the apoM promoter (sequences: 5′-AGGTTGAAGTTACTTATTAGCAGGT-3′ and 5′-AGGTTGAAGGGGCTTATTAGCAGGT-3′, respectively). Competition analysis was performed using a probe with the HNF-1 site in the insulin growth factor binding protein (sequence: 5′-ACTGGTTAATGATTGGCATGC-3′) (15). Supershifts were carried out with an anti–HNF-1α antibody (Geneka Biotechnology, Montreal, Canada).

Livers were dounce homogenized 15 times in 100:1 (vol:vol) of buffer A (10 mmol/l KCl, 1.5 mmol/l MgCl₂, 10 mmol/l HEPES [pH 7.9], 1 mmol/l DTT, 1 mmol/l NaVO₄, 1× complete TM protease inhibitor [Boehringer Mannheim]). After centrifugation at 2,000g for 10 min at 4°C, the pellet was resuspended in 4 vol of the same buffer, dounced 10 times, and centrifuged again at 2,000g. The nuclei-containing pellet was resuspended in 2 vol of buffer B (420 mmol/l NaCl, 10 mmol/l KCl, 20 mmol/l HEPES [pH 7.9], 20% glycerol, 1 mmol/l DTT, 1 mmol/l NaV0₄, 1× complete TM protease inhibitor) and extracted for 30 min at 4°C on a shaking rotor. After centrifugation at 16,000g, the supernatant was diluted 10-fold in buffer C (10 mmol/l KCl, 20 mmol/l HEPES [pH 7.9], 20% glycerol, 1 mmol/l DTT, 1 mmol/l NaV0₄, 1× complete TM protease inhibitor) and centrifuged for 15 min at 16,000g. The supernatant was loaded onto a Biomax-50 filter unit (Millipore) and centrifuged for 2 h at 4,000g at 4°C.

HepG2 cells were grown on Dulbecco’s modified Eagle’s medium supplemented with 15% FCS. A modified calcium phosphate precipitation procedure was used for transient transfections, as described (18). A total of 0.5 ml of the precipitate, containing 1 μg of CMV-LacZ, 0.5 μg of luciferase reporter plasmid, and the indicated amount of expression vectors and carrier DNA (up to 10 μg), was added per 60-mm dish. Luciferase was normalized for transfection efficiency by the corresponding β-galactosidase activity.

Nine diabetic patients with previously identified mutations in the HNF-1α gene (E129T, P291fsinsC, E132K, R159W, W267R) were identified from the U.K. MODY collection. HNF-4α Q268X subjects of the RW pedigree were previously identified by sequencing analysis (5). Nine of these diabetic subjects with HNF-4α mutations were included in the study. Control subjects were spouses of patients who attended the Exeter (U.K.) diabetes clinic. Study groups were matched for sex, age, and BMI, and diabetic groups were matched for HbA_1c. The clinical characteristics of the four study groups are shown in Table 1.

Plasma apoM levels were determined by immunoblotting using a standard control serum. Plasmas were diluted 1:100 in PBS, and 10 μl of diluted plasma was analyzed on 12.5% denaturing acrylamide gels. After immunoblotting, apoM was detected using anti–apoM140-159 antibody (1:1,000) and goat anti-rabbit IgGs conjugated to horseradish peroxidase (1:10,000) in Tris-buffered saline supplemented with 5% nonfat dry milk. Bands were visualized using the Renaissance chemoluminescence substrate (NEN). A standard, consisting of an equal mixture of plasma of all normal subjects, was used in each blot as a reference. All samples were loaded in duplicate, and measurements were repeated twice. Band intensities on autoradiographs were analyzed by densitometry, and the mean value of two measurements was compared with the standard. Concentrations of apoAI were determined with immunoturbidimetric assays from Roche Diagnostics using a Hitachi 917 autoanalyzer.

Results are given as mean ± SE. The statistical analysis was performed using Microsoft Excel 98 statistical package. The t test was used to assess statistical significance, and the null hypothesis was rejected at the 0.05 level.

To identify genes that are regulated by Hnf-1α, we studied gene expression profiles using oligonucleotide microarrays in livers of Hnf-1α^−/− mice and wild-type littermates. The analysis of this dataset (see www.nature.com/ng/journal/v27/n4/extref/ng0401–375-S1.pdf) revealed that apoM gene expression was ∼25-fold increased in Hnf-1α^+/+ compared with Hnf-1α^−/− mice. We analyzed the mRNA of apoM by Northern blotting in multiple tissues and found expression to be restricted to kidney and liver, a finding that is consistent with an earlier report (20) (Fig. 1A). Steady-state mRNA levels of apoM were essentially absent in liver and kidney of mutant Hnf-1α mice (Fig. 1B). In contrast, the expression of apoA1 and apoC1, two lipoproteins that are major components of the HDL particle, was similar in mutant and wild-type mice. We also analyzed apoM protein levels in plasma of wild-type, Hnf-1α^+/−, and Hnf-1α^−/− mice. Anti-apoM antibodies were generated against a peptide encoding amino acid residues 140–152 of the apoM protein, which are completely conserved between mouse and human. Immunoblot analysis of plasma of wild-type and mutant mice showed that apoM protein was absent in plasma of Hnf-1α^−/− mice and protein levels were reduced ∼50% in Hnf-1α^+/− compared with wild-type littermates (133 ± 13 vs. 71 ± 2.6; P = 0.009; Fig. 2A). These data indicated that Hnf-1α^−/− mice have a specific defect in apoM gene expression and that gene dosage of Hnf-1α correlates with apoM serum levels in mice.

We have previously shown that Hnf-1α^−/− mice exhibit hypercholesterolemia and that cholesterol accumulates mainly in large buoyant HDLc particles (16). To show that the reduction in apoM expression is not secondary to hypercholesterolemia, we examined the apoM distribution in plasma lipoproteins of wild-type mice, in animals that lack the LDL receptor (Ldlr^−/−; Ldlr^−/− mice accumulate cholesterol predominantly in LDL particles), or in Hnf-1α (Hnf-1α^−/−) mice (21). Fasting plasma lipoprotein composition was analyzed after fast-protein liquid chromatography fractioning. Figure 2B shows that the cholesterol in the plasma of Ldlr^−/− mice was mainly carried in LDL particles but also present in VLDL and HDL fractions. In contrast, cholesterol in the plasma of Hnf-1α^−/− mice was almost exclusively carried by the HDL fraction and an abnormal fraction with intermediate buoyancy between HDL and LDL peaks. ApoM expression was only detected in the HDL peak of wild-type and Ldlr^−/− mice but absent from the HDL and HDLc fraction of Hnf-1α^−/− animals (Fig. 2C; data not shown), demonstrating that apoM is a component of the HDL particle and that apoM expression is independent of plasma cholesterol levels and defects in cholesterol metabolism.

The in vivo data suggested that apoM expression is significantly reduced in the liver and the kidney of Hnf-1α^−/− mice compared with wild-type littermates, suggesting that Hnf-1α might be a transcriptional activator of the apoM gene. We therefore analyzed the promoter regions of human and mouse apoM and identified a highly conserved putative HNF-1 binding site at position −88 to −103 (Fig. 3A). We cloned a 1,045-bp apoM promoter containing the putative HNF-1 binding site and cloned it upstream of the luciferase gene to generate plasmid pGL2-papoM. Cotransfection of this construct with an HNF-1α expression vector in Cos7 cells (that do not express HNF-1α) led to an ∼28-fold dose-dependent activation of the apoM promoter (Fig. 3B). Transfection of pGL2-papoM into HepG2 cells that express endogenous HNF-1α also led to an induction of luciferase expression. To test whether the HNF-1 binding site in the upstream regulatory region was responsible for HNF-1–dependent transactivation, we mutated this site by site-specific mutagenesis. The mutated promoter-construct pGL2-papoMmut was transfected into HepG2 cells, and luciferase activity was compared with cells that were transfected with pGL2-papoM. Luciferase gene expression was completely abolished in cells that contained pGL2-papoMmut, indicating that this site is required for transcriptional activation of the apoM gene in vitro (Fig. 3C).

EMSAs were performed to investigate whether HNF-1α can bind to the putative HNF-1 binding site in the apoM promoter. Nuclear extracts were prepared from livers of wild-type mice, incubated with a ³²P-labeled oligonucleotide that contained the HNF-1 binding site, and subjected to electrophoretic shift analysis (EMSA; Fig. 3D). A binding activity could be detected (lane 2) and competed using an unlabeled excess of “cold” HNF-1α binding oligonucleotide (lane 3) but not with an oligonucleotide containing the mutated HNF-1 binding site (lane 4). Furthermore, a supershift of the complex was observed after preincubation of the liver nuclear extract with a monospecific anti–HNF-1α antiserum (lane 5). This supershift was not generated with a control (anti–STAT-1) antibody (lane 6). Together, these data suggest that HNF-1α can bind to a highly conserved HNF-1 binding site in the apoM promoter and activate transcription.

Because apoM gene expression and serum apoM levels are critically dependent on Hnf-1α function in mice, it might be anticipated that loss of function by HNF-1α mutations in humans could manifest by decreased serum apoM concentrations. To test this hypothesis, we first studied whether apoM could be detected in plasma. The complete coding sequence of apoM was cloned into bacterial expression vector PET3a and expressed as an insoluble protein in inclusion bodies (Fig. 4A). Recombinant protein was refolded as described in research design and methods, and the concentration was determined. We then measured the concentration of apoM protein in plasma by preparing a standard curve of recombinant protein using semiquantitative immunoblotting techniques (Fig. 4B). Protein concentrations of apoM in human plasma were ∼0.1 g/l. We noticed a small difference in molecular weight of the native and recombinant protein, indicating that apoM may be modified posttranslationally. We next compared the expression levels of apoM protein in plasma of control subjects and HNF-1α–deficient and HNF-4α–deficient diabetic subjects (MODY3 and MODY1, respectively). Study groups were composed of nine diabetic patients with HNF-1α mutation (HNF1α^+/−), nine diabetic subjects with HNF-4α mutation (HNF4α^+/−), and nine normal control subjects (HNF1α^+/+). These subjects were matched with respect to ethnicity, sex, age, and BMI. The clinical characteristics of the different study groups are shown in Table 1.

Of the MODY3 patients, four subjects had mutations in the B-domain (HNF-1αP129T, R159W, E132K) that is responsible for a strong cooperativity in the binding of two monomers to a palindromic HNF-1 binding site. Four subjects had an insertion mutation in the transactivation domain (P291fsinsC) that leads to a truncated protein and exhibits moderate dominant negative activity in vitro (13). One subject had a mutation in a highly conserved residue of helix 3 of the DNA recognition homeodomain (W267R). Subjects with mutations in HNF-4α (MODY1) were from the RW pedigree. In the RW family, diabetes is caused by Q268X mutation in the HNF-4α gene (5). This mutation results in a truncated HNF-4α protein with no transcriptional activity (22). Serum apoM protein levels were measured by immunoblotting and by comparing the data with a standard (control) serum (Fig. 4C). Subjects with HNF-1α mutations had an ∼36% reduction in apoM levels compared with control subjects (1.0 vs. 0.65; P = 0.02). In contrast, HDL and serum apoA1 levels were similar in both study groups, suggesting that reduced apoM levels are caused by HNF-1α haploinsufficiency and are not due to changes in HDL concentrations (Fig. 5). We were concerned that hyperglycemia may contribute to the changes in apoM concentrations because MODY3 subjects had significantly higher HbA_1c levels than normal control subjects (7.2 ± 3.9 and 5.5 ± 0.2%, respectively). We therefore measured apoM levels in diabetic patients (HbA_1c 7.2 ± 1.8%) with HNF-4α haploinsufficiency (MODY1). Serum apoM levels of subjects with MODY1 were similar than in the control group and significantly higher than in MODY3 patients (Fig. 5). This result indicated that HNF-1α haploinsufficiency rather than hyperglycemia is responsible for the decrease in plasma apoM levels.

HNF-1α is a transcription factor that has an important role as a key regulator of pancreatic islet, liver, and kidney gene expression. Genetic studies have shown that mutations in HNF-1α result in an autosomal-dominant form of type 2 diabetes characterized by defects in pancreatic β-cell function and glucose-stimulated insulin release (14,23,24). In the kidney, HNF-1α has been shown to be a key regulator of the glucose transporter SGLT2, and haploinsufficiency of HNF-1α is associated with a low renal threshold for glucose that can lead to early glycosuria, even in the absence of high blood glucose levels (25,26). Liver function has not been thoroughly assessed in patients with HNF-1α mutations.

We recently performed a genetic screening to identify genes that are regulated by HNF-1α. We used livers from wild-type and Hnf-1α–deficient animals and performed a large-scale gene expression analysis using oligonucleotide expression arrays (16). In this study, we identified an expressed sequence tag that encodes apoM and that was highly regulated in livers of mutant animals. We show that Hnf-1α is required for apoM expression and that Hnf-1α directly activates apoM transcription by binding to a conserved binding site in the apoM promoter.

ApoM is a novel lipoprotein that is expressed in the liver and the kidneys and is predominantly present in HDL in plasma (20). We have recently shown that Hnf-1α–deficient mice develop hypercholesterolemia as a result of a defect in bile acid transport, increased bile acid and cholesterol synthesis, and impaired HDL metabolism. The increased plasma cholesterol in Hnf-1α^−/− mice resides predominantly in large buoyant HDLc that may be caused by reduced activity of hepatic lipase and increased expression of HDL cholesterol esterifying enzyme lecithin: cholesterol acyl transferase (16). The biological function of apoM is currently unknown; however, the absence of apoM in HDL particles of mutant Hnf-1 mice may suggest that it contributes to the defect in HDL cholesterol in these animals. ApoM is distantly related to the lipocalin superfamily. Most members of this family have the ability to bind and transport hydrophobic molecules in the plasma (27). It is possible that apoM serves as a specific function in the metabolism and transfer of cholesterol (and/or other bioactive molecules) in the blood. ApoM is unusual in having its hydrophobic signal peptide retained in the mature protein, where it presumably serves as a phospholipid anchor. The synthesis of apoM therefore may be closely linked to the HDL assembly and secretion from the liver.

ApoM has previously been suggested to be a minor component of the HDL, VLDL, and LDL lipoprotein classes compared with the major apolipoproteins (20). In our studies, we found apoM exclusively associated with HDL particles. Furthermore, to assess the plasma concentration of apoM compared with other lipoproteins, we generated recombinant apoM in bacteria. Using recombinant apoM as a standard, we determined that the range of apoM concentrations in human plasma is 0.05–0.15 g/l. Serum concentrations of other HDL-containing apolipoproteins such as apoCs and apoE range between 0.02 and 0.1 g/l, and apoA1 levels range between 1.1 and 2.2 g/l. This suggests that apoM is a major component of the HDL particle.

Our data demonstrate that HNF-1α haploinsufficiency is associated with low/normal apoM plasma levels. We have shown previously that MODY1 is characterized by low serum levels of triglycerides (Table 1), lipoprotein(a), apoAII, and apoCIII (28). Thus, concentrations of various lipoproteins in serum may aid in the differential diagnosis between MODY1 and MODY3. Subjects with early-onset type 2 diabetes, positive family history consistent with autosomal-dominant inheritance, negative anti-islet antibodies, early glucosuria, sensitivity to sulfonylureas, and low serum apoM levels should be suspected of having MODY3. Genetic testing for mutations in the HNF-1α and HNF-4α genes, as well as other genes, should be carried out for definite identification of various MODY subtypes.

This work was supported by Clinical Research Center Grant M01-RR00102 from the National Center for Research Resources at the National Institutes of Health (NIH) to Rockefeller University and M01-RR00042 to the University of Michigan, NIH Grants R01DK55033-01 (to M.S.) and DK-20572 (to the Michigan Diabetes Research and Training Center), and Medical Scientist Training Program Grant GM07739 (to D.Q.S.), Deutsche Forschungsgemeinschaft (to S.M.R. and C.W.). E.R.P holds a Wellcome Trust Clinical Training Fellowship (GR065686MA), and A.T.H. holds a Wellcome Trust Career Leave Fellowship.