The sentrin-conjugating enzyme mUbc9 interacts with GLUT4 and GLUT1 glucose transporters and regulates transporter levels in skeletal muscle cells

As a bait for screening with the two-hybrid system, the vector pLexA-GLUT4 COOH-t expressing a fusion protein composed of amino acids 467–509 of rat GLUT4 linked to the DNA-binding domain of the yeast LexA transcriptional activator was constructed by PCR amplification of rat GLUT4 cDNA (provided by G. I. Bell, University of Chicago, Chicago) and subsequent subcloning into the yeast vector pBTM116 (22). The L40 yeast strain of Saccharomyces cerevisiae was transformed with pLexA-GLUT4 COOH-t by the lithium acetate protocol (23). Yeast expressing the pLexA-GLUT4 COOH-t fusion then were transformed with a library expressing cDNAs derived from mouse TA1 adipocytes as fusions to the VP16 activation domain (provided by S. M. Hollenberg, Oregon Health Sciences University, Portland, OR). The TA1 library was constructed by using cultures with a greater than 70% adipocyte population, contains fragments of 0.35–0.7 kb, and includes approximately 5 × 10⁶ transformants (22). A total of 27 × 10⁶ transformants were screened; approximately 200 colonies showed activation of the yeast reporter genes, and 80 colonies were selected that showed dependence on the LexA-GLUT4 COOH-t fusion protein for activation of reporter genes by using a LexA-lamin fusion as a negative control. Plasmids from library clones subsequently were isolated from the yeast, transferred to Escherichia coli, and sequenced, and the GenBank/European Molecular Biology Laboratory databases were screened for similar sequences by using the GCG program fasta.

3T3-L1 fibroblasts (American Type Culture Collection) were grown and differentiated into adipocytes as described (5). L6 rat skeletal muscle cells were cultured in Eagle's minimum essential medium supplemented with 10% donor calf bovine serum and nonessential amino acids in a 5% CO₂ atmosphere at 37°C. Total lysates and subcellular fractions from 3T3-L1 adipocytes or L6 myoblasts were prepared as described (24–26).

For overexpression of mUbc9 in L6 cells, the full-length mUbc9 sequence was cloned into the mammalian expression vector pcDNA3.1 (Invitrogen). A negative mutant form of mUbc9 containing Ala instead of Cys⁹³ was generated by PCR-based site-directed mutagenesis by using the QuickChange site-directed mutagenesis kit (Stratagene). L6 cells were transfected at 85% confluence by using Lipofectamine (Life Technologies, Gaithersburg, MD). L6 myoblasts stably overexpressing GLUT4 were generated similarly by transfection with pCR3.1 containing the full-length rat GLUT4 cDNA sequence.

GLUT4 and GLUT1 mRNA levels were determined by quantitative reverse transcription–PCR of total cell RNA by using the forward primer 5′-AGA GTG CCT GAA ACC AGA GG-3′ and the reverse primer 5′-CAG TCA GTC ATT CTC ATC TGG C-3′ for amplification of GLUT4 mRNA and the forward primer 5′-GAG TGC CTG AAA CCA GAG G-3′ and the reverse primer 5′-CTC ACA CTT GGG AGT CA-3′ for amplification of GLUT1 mRNA. Amplification products were electrophoresed on 1% agarose gels and stained with ethidium bromide.

To generate a GST fusion protein containing mUbc9 (GST-mUbc9), the full-length mUbc9 coding sequence was cloned into the pGEX-3X plasmid (Amersham Pharmacia), expressed in the E. coli strain DH5α, and purified according to the manufacturer's instructions. To generate hexahistidine-tagged DHFR (6xHis-DHFR) fusion proteins containing glucose transporter fragments, the GLUT4 COOH-terminal fragment (residues 467–509), the GLUT1 COOH-terminal fragment (residues 451–492), and a shorter fragment of the GLUT4 COOH-terminal region (residues 478–509) were generated by PCR amplification and cloned into pQE-40 (Qiagen). Plasmid constructs were expressed in the E. coli strain M15[pREP4] and purified according to the manufacturer's instructions.

To study the association between GST-mUbc9 and transporters from L6 myoblasts overexpressing GLUT4, GST-mUbc9 first was immobilized onto glutathione-Sepharose 4B beads (Amersham Pharmacia) for 1 h at 4°C in 500 μl of TNE buffer (120 mM NaCl/1 mM EDTA/1 mM PMSF/5 μg/ml leupeptin/10 mM Tris·HCl, pH 7.6); the beads (50 μl) were washed twice with 1 ml of TNE buffer at 4°C and resuspended in 150 μl of ice-cold TNE buffer. The immobilized fusion proteins were incubated for 16 h at 4°C with 150 μl of solubilized membranes (solubilization buffer: 1% Triton X-100/0.5% sodium deoxycholate/1% SDS/0.5% BSA; protein concentration, ≈0.5 mg/ml). The beads then were washed four times with 1 ml of TNE buffer at 4°C, and the fusion proteins were eluted by incubation with 80 μl of glutathione elution buffer (10 mM reduced glutathione/50 mM Tris⋅HCl, pH 8.0) for 30 min at 20°C. Eluates were resolved by SDS/PAGE and subjected to immunoblotting with anti-GLUT4, anti-GLUT1, or anti-GLUT3 antibodies.

The association between GST-mUbc9 and recombinant glucose transporter fragments fused to the 6xHis-DHFR was studied in vitro according to previously reported protocols (27).

For immunoprecipitation of glucose transporters, membranes were solubilized in PBS containing 1% Triton X-100, 0.5% sodium deoxycholate, 1% SDS, and 0.5% BSA (for GLUT4) or 2% C12E9, 1 μM aprotinin, 1 μM pepstatin, and 1 μM leupeptin (for GLUT1). Collection of immunocomplexes, SDS/PAGE, and immunoblotting were performed as described previously (24), and the proteins were visualized by enhanced chemiluminescence (ECL) or by incubation with ¹²⁵I-protein A followed by autoradiography. Quantification was performed by densitometric analysis by using optilab image analysis software (Graftek SA, Mirmande, France).

To identify GLUT-interactive proteins, a cDNA library derived from TA1 adipocytes was screened with the COOH-terminal domain of rat GLUT4 (residues 467–509) by using a LexA/VP16-based two-hybrid system. The majority of the cDNAs isolated with this method encoded five independent inserts derived from a single gene and fused to VP16 in the same reading frame at different positions in the 5′ untranslated region. Sequence analysis of the longest clone revealed a cDNA with a 5′ untranslated region of 48 bp and an ORF of 474 bp, encoding a protein of 158 aa with a calculated M_r of 18,000. Comparison with sequences in GenBank/European Molecular Biology Laboratory databases demonstrated 100% identity to mUbc9 (20, 28), a recently identified mammalian homologue of Ubc9, which first was cloned as a putative member of the E2 ubiquitin-conjugating enzyme family in yeast (29).

mUbc9 was detected as a 20-kDa protein in extracts from multiple rat tissues, including skeletal muscle, heart, liver, and kidney, and in a number of cultured cell lines by immunoblotting with anti-mUbc9 antibodies (not shown). In 3T3-L1 adipocytes, mUbc9 was found to be most abundant in high- and low-density microsomal membranes, less abundant in nuclei/mitochondria and plasma membranes, and almost absent in the cytosol, thus overlapping with GLUT4 and GLUT1 glucose transporters in its distribution in these insulin-sensitive cells (Fig. 1). Preferential localization of mUbc9 in intracellular membranes also was observed in L6 skeletal muscle myoblasts (Fig. 1) and terminally differentiated myotubes (data not shown).

To investigate potential mUbc9 regulatory effects on glucose transporters, mUbc9 or a mUbc9 mutant with substitution of Cys⁹³, in the active site of the enzyme, with Ala (mUbc9-Ala⁹³) was overexpressed in L6 skeletal muscle cells. The mUbc9-Ala⁹³ mutant was used as a catalytically inactive dominant negative because it is unable to form sentrin conjugates (30). Multiple clones of L6 skeletal muscle myoblasts stably transfected with pCR3.1 plasmids encoding mUbc9 or mUbc9-Ala⁹³ were obtained, and three independent clones of each were selected that overexpressed the mUbc9 and mUbc9-Ala⁹³ proteins, respectively, by 2.5- to 3-fold in comparison with nontransfected (wild-type) L6 myoblasts (Fig. 2A). Multiple control cell lines were isolated after transfecting cells with the pCR3.1 plasmid lacking an insert (Neo), and levels of mUbc9 protein in Neo control lines were not significantly different from those observed in the nontransfected L6 myoblasts (Fig. 2A).

In mUbc9-overexpressing myoblasts, the cellular content of the GLUT4 protein was found to undergo a marked, 8-fold increase in comparison with Neo or wild-type controls (Fig. 2B Top, P < 0.05). The GLUT4 transporters in mUbc9 myoblasts showed insulin-regulated translocation from intracellular membranes to plasma membranes (Fig. 2C Bottom). Overexpression of catalytically inactive mUbc9-Ala⁹³ did not modify significantly the characteristic low levels of GLUT4 protein in the L6 myoblasts. From these data, it can be concluded that overexpression of mUbc9, but not mutant mUbc9-Ala⁹³, leads to an increased abundance of functional GLUT4 transporters in L6 cells.

In contrast to the effects of mUbc9 on GLUT4, the total cellular content of the GLUT1 transporter protein was reduced by 65% in myoblasts overexpressing mUbc9 and increased by 35% after transfection with the mUbc9-Ala⁹³ mutant in comparison with Neo or wild-type control cells (Fig. 2B Middle, P < 0.05). The opposite effects of mUbc9 on the GLUT4 and GLUT1 proteins were specific for these transporters, as evidenced by unaltered levels of GLUT3, an additional transporter isoform present in L6 myoblasts, after overexpression of either mUbc9 or mUbc9-Ala⁹³ (Fig. 2B Bottom). As shown in Fig. 2D, GLUT4 and GLUT1 mRNA levels were not different in mUbc9, mUbc9-Ala⁹³, and control myoblasts (GLUT4/cyclophyllin mRNA ratio was 0.571 ± 0.006, 0.572 ± 0.007, 0.568 ± 0.002, and 0.567 ± 0.010, and GLUT1/cyclophyllin mRNA ratio was 0.569 ± 0.021, 0.572 ± 0.007, 0.570 ± 0.001, and 0.561 ± 0.007 in wild-type, Neo, mUbc9, and mUbc9-Ala⁹³ myoblasts, respectively, not statistically significant). Thus, the changes in the amounts of the two transporter proteins in these cells do not result from altered expression of the GLUT4 and GLUT1 genes, but likely reflect distinct posttranscriptional effects on the two transporter isoforms.

The opposite regulation of GLUT4 and GLUT1 by mUbc9 results in significant changes in glucose transport. Overexpression of mUbc9 resulted in markedly reduced basal transport in each of the individual mUbc9 clones, with a mean 60% decrease in comparison with wild-type or Neo myoblasts (Fig. 3, P < 0.05). The fold stimulation of glucose transport by insulin was enhanced significantly in mUbc9 myoblasts, with a mean 3.3-fold transport increase as compared with 1.8-fold in wild-type and Neo controls (Fig. 3, P < 0.05). By contrast, overexpression of the catalytically inactive mUbc9-Ala⁹³ protein resulted in increased basal transport (65% increase, P < 0.05, Fig. 3) and a 43% reduction of the fold stimulation of glucose uptake by insulin (P < 0.05, Fig. 3).

The basis for the interaction between mUbc9 and glucose transporters was studied further in vitro. To test whether purified mUbc9 could bind directly to glucose transporters expressed in mammalian cells, a full-length recombinant mUbc9 protein fused to GST was immobilized on glutathione-Sepharose beads and incubated with solubilized intracellular or plasma membranes of L6 myoblasts overexpressing GLUT4. As shown in Fig. 4A, the GST-mUbc9 fusion bound to GLUT4 and GLUT1 transporters. By contrast, no association was observed between the transporters and GST alone or between mUbc9 and GLUT3 transporters.

To map the transporter domain that binds to mUbc9, the direct association of mUbc9 with recombinant COOH-terminal fragments of GLUT4 and GLUT1 produced as fusions with DHFR was evaluated. GST-mUbc9 bound to DHFR fusions containing the COOH-terminal domains of GLUT4 and GLUT1, but not to DHFR alone, and no association was observed between GST and any of the DHFR fusions (Fig. 4B). The ability of mUbc9 to associate with both GLUT4 and GLUT1 at the COOH terminus suggested that mUbc9 binds to a homologous amino acid region in the two transporters. Because residues 467–477 of rat GLUT4 and residues 451–461 of rat GLUT1 are 82% identical (Fig. 4C), representing a candidate binding site for mUbc9, the association of mUbc9 with a shortened GLUT4 COOH-terminal fragment lacking residues 467–477 was evaluated and found to be greatly reduced (Fig. 4C). By contrast, mutation of Leu⁴⁸⁹ and Leu⁴⁹⁰, distal to amino acids 467–477 of rat GLUT4, did not affect the interaction with mUbc9 (Fig. 4C). Furthermore, a synthetic peptide corresponding to residues 467–477 of rat GLUT4 inhibited the association of mUbc9 with the GLUT4 and GLUT1 COOH-terminal fusion proteins, including the GLUT4 fragment with mutation of Leu⁴⁸⁹ and Leu⁴⁹⁰, in a dose-dependent manner (Fig. 4C). A peptide of the same composition with a scrambled sequence (Pro-Arg-Arg-Thr-Glu-Arg-Asp-Phe-Val-Thr-Gly) had no effect on these associations (not shown). Thus, mUbc9 interacts with GLUT4 (and probably GLUT1) by binding to a previously unrecognized functional domain involving the initial 11 aa of the transporter COOH terminus.

Sentrin, the specific substrate for the conjugating activity of mUbc9 (31, 32), recently has been renamed sentrin-1, after the identification of two additional polypeptides sharing higher homology with sentrin than with ubiquitin (33). To test whether GLUT4 and GLUT1 are covalently modified by attachment of sentrin-1, solubilized membrane fractions from basal and insulin-treated 3T3-L1 adipocytes were subjected to immunoprecipitation with anti-GLUT4 or anti-GLUT1 antibodies, and the immunoprecipitates were analyzed by immunoblotting with antibodies to sentrin-1. A 90-kDa protein was detected by anti-sentrin-1 antibodies in both anti-GLUT4 and anti-GLUT1 immunoprecipitates, but not in immunoprecipitates obtained with unrelated antibodies (Fig. 5). The 90-kDa protein was found in low-density microsome, high-density microsome, and plasma membrane fractions from basal and insulin-treated cells (Fig. 5). Because sentrin-1 has a molecular mass of 18 kDa, this 90-kDa protein is consistent with the addition of two sentrin-1 molecules to glucose transporters. To demonstrate the sentrin-conjugated transporters in Fig. 5, 9-fold greater amounts of the glucose transporter immunoprecipitates were immunoblotted with anti-sentrin-1 as compared with antiglucose transporter antibodies. Because anti-GLUT4, anti-GLUT1, and anti-sentrin-1 antibodies exhibited similar blotting efficiencies for their respective fusion proteins (data not shown), the immunoprecipitation data suggest that a relatively small percentage of glucose transporters are conjugated to sentrin in 3T3-L1 adipocytes.