A M55V polymorphism in a novel SUMO gene (SUMO-4) differentially activates heat shock transcription factors and is associated with susceptibility to type I diabetes mellitus.

Three SUMO (small ubiquitin-related modifier) genes have been identified in humans, which tag proteins to modulate subcellular localization and/or enhance protein stability and activity. We report the identification of a novel intronless SUMO gene, SUMO-4, that encodes a 95-amino acid protein having an 86% amino acid homology with SUMO-2. In contrast to SUMO-2, which is highly expressed in all of the tissues examined, SUMO-4 mRNA was detected mainly in the kidney. A single nucleotide polymorphism was detected in SUMO-4, substituting a highly conserved methionine with a valine residue (M55V). In HepG2 (liver carcinoma) cells transiently transfected with SUMO-4 expression vectors, Met-55 was associated with the elevated levels of activated heat shock factor transcription factors as compared with Val-55, whereas the levels of NF-kappaB were suppressed to an identical degree. The SUMO-4M (Met) variant is associated with type I diabetes mellitus susceptibility in families (p = 4.0 x 10(-4)), suggesting that it may be involved in the pathogenesis of type I diabetes.

SUMO-1 1 is involved in the post-translational modification of cellular proteins and regulates various cellular processes such as nuclear transport, oncogenesis, stress response, inflammation, and the response to viral infection (1). SUMO-1 is related to ubiquitin, and although both share only an 18% sequence homology, they have a similar three-dimensional structure (1). In contrast to ubiquitin, which tags proteins for degradation, SUMO-1 seems to enhance protein stability and modulate the subcellular localization. The list of mammalian proteins, which are known to be sumoylated, is growing continually and includes RAN GTPase-activating protein 1 (nuclear import) (2), IB␣ (NF-B inhibition) (3), c-Jun (transcription factor), p53 (tumor suppressor) (4), GLUT1 and GLUT4 (glucose transport) (5), heat shock transcription factors 1 and 2 (HSF1 and 2) (6, 7), and FAS/APO-1 (cell death domain/ apoptosis) (8).
SUMO-1 is expressed initially as an inactive precursor of 101 amino acids. A C-terminal proteolytic event exposes two glycine residues at the C terminus that are involved in the formation of a peptide bond with the ⑀-amino group of a lysine residue of a target protein. The activating enzymes E1 (AOS1 and UBA2) and the conjugating enzyme E2 (UBC9) are required for sumoylation in vitro (1). In vivo, an additional enzyme E3 is required for efficient sumoylation (9). The E2 enzyme, UBC9, is specific for sumoylation, and recently, two additional SUMO family members have been identified based on the interaction with UBC9, SUMO-2, and SUMO-3 (10,11). SUMO-1 shares a 48% identity with SUMO-2 and a 46% with SUMO-3. Very little is known regarding the functions of SUMO-2 and SUMO-3. However, SUMO-2 and SUMO-3 contain a consensus SUMO modification site (Fig. 1), which enables self-sumoylation and the formation of polymeric chains (11), in contrast to SUMO-1. Thus, SUMO-2 and SUMO-3 may have different roles in cellular functions.
Here, we identify and characterize a novel fourth SUMO gene (SUMO-4) whose protein product is 87% homologous with SUMO-2 and report a unique polymorphism not encoded in any of the other SUMO genes (M55V). The SUMO-4M (Met variant) activates HSFs to a larger extent and is associated with susceptibility to type I diabetes (p ϭ 4.0 ϫ 10 Ϫ4 ). SUMO-4M may be involved in the pathogenesis of type I diabetes.

EXPERIMENTAL PROCEDURES
RT-PCR-Total RNA was purchased from Ambion (Austin, TX), and RT-PCR performed using the Omniscript RT-PCR kit from Qiagen (Valencia, CA). Total RNA (0.5 g) was analyzed in the first strand reverse transcriptase reaction using a primer that includes poly(dT 19 )-CATTTTAAAC found in SUMO-2 mRNA and deduced from the poly(A) site in SUMO-4 genomic DNA by following directions from Qiagen. The primer was removed from the reaction mixture by purification using a Cycle-Pure kit from Omega Bio-Tek (Doraville, GA). The purified cDNA (10 l to one-third of the total) was amplified in 25 l of PCR reactions containing dNTPs, 1.25 M betaine (Sigma), and primers for SUMO-4F1 (forward) (5Ј-GCAATATGCTTGTGTACACATAC-3Ј) and either SUMO-4R1 (reverse) (5Ј-CACAGAAGAAGTCAAGACTGAG-3Ј) or SUMO-2R1 (5Ј-CAAGGAAGGAGTCAAGACTGA-3Ј). The PCR reaction was "hotstarted" by adding Taq polymerase (PGC Scientific, Gaithersburg, MD) at 95°C. The SUMO-2R1 and SUMO-4R1 primers have three mismatches, which were sufficient to obtain allele-specific PCR using the conditions described. After 35 cycles of PCR using an annealing temperature of 55°C, 5 l of the PCR product was analyzed on a 1.5% agarose gel to identify the expected 466-bp DNA fragments. Control experiments were performed in which equivalent amounts of RNA were amplified without the prior reverse transcriptase step to demonstrate that RT-PCR products were generated from RNA and not from contaminating DNA. Nucleotide sequence analysis at our Child Health Research Center Sequencing Core at Baylor College of Medicine was used to confirm the specificities of the SUMO-2 and SUMO-4 RT-PCR products, respectively.
Genotyping-Genotyping was done using a 3-step primer extension assay (12). First, a defined region was amplified using 30 cycles of PCR consisting of 1 min at 94°C (denaturation), 1 min at 55°C (annealing), and 1 min at 72°C (extension). PCR using SUMO-4F1 and SUMO-R2 primers (see RT-PCR above) resulted in a PCR product of 466 bp. In the next step, dNTPs and primers remaining from the PCR were removed by enzymatic digestion using shrimp alkaline phosphatase and exonuclease I (each from Amersham Biosciences). In the final step, the single base extension (SBE) reaction using an oligonucleotide primer resides directly adjacent to the nucleotide in question. The SBE primer was 5Ј-GAACCACGGGGATTGTCA-3Ј. The SBE reaction contained the extension primer, all four dideoxy dNTPs (Amersham Biosciences), Taq polymerase, and the purified PCR product. The SBE assay was done using 49 cycles of denaturation at 96°C for 30 s, annealing at 55°C for 30 s, and extension at 60°C for 30 s. After the SBE primer was extended with the nucleotide analogue, the resulting products were analyzed using ion-pair reverse-phase high pressure liquid chromatography with the WAVE System (Transgenomic, Inc., Omaha, NE).
Recombinant Clones-Vectors pHSE-Luc, pNF-B-Luc, pTAL-Luc, and pCMV-Myc were purchased from Clontech (Palo Alto, CA). Renilla luciferase reporter phRL-null vector was purchased from Promega (Madison, WI), and HSF2 was purchased from the American Type Culture Collection (Manassas, VA). pET-15b vector was purchased from Novagen (Madison, WI). SUMO-1, SUMO-2, SUMO-3, SUMO-4M, and SUMO-4V constructs were made to express active recombinant proteins ending in the terminal di-glycine residues ( Fig. 1, additional C-terminal residues are cleaved normally in vivo). UBC9 was made to the fulllength protein. All of the recombinant proteins were expressed in pET-15b vector. This vector allows the expression of recombinant proteins containing a His tag at the N terminus of the recombinant protein.
The recombinant proteins were propagated in BL21(DE3)-transformed cells, and recombinant proteins were purified utilizing the His-bind resin (Ni 2ϩ ) as described by the manufacturer (Novagen). Purified proteins were dialyzed against 50 mM Tris-Cl, pH 7.5, with the exception of SUMO-4, which was dialyzed against 25 mM MES, pH 6.0, and 75 mM NaCl. The sequence of all of the recombinant clones was verified by nucleotide sequence analysis. RT-PCR from 293 human embryonic kidney (HEK) cells (Ambion) was used as a source of RNA to make cDNA in all of the constructs. The HEK cells are heterozygous and express mRNA for both SUMO-4M and SUMO-4V. For SUMO-4, the primer SUMO-4F2-GCCATATGGCCAACGAAAAGCC containing a NdeI site and primer SUMO-4R2-TCAACCTCCCGTAGGCTGTTG, which terminates the SUMO-4 amino acid sequence by two amino acids, were utilized. The RT-PCR products were ligated into TA-cloning vector pCR2.1 (Invitrogen) and transformed into XL1-Blue Escherichia coli cells. A 328-bp restriction endonuclease fragment created by the digestion of recombinant clones with the restriction endonucleases, NdeI (site in primer) and BamHI (site in vector), was subcloned into the NdeI/BamHI site of pET-15b. RT-PCR products of SUMO-2 were made using primers SUMO-2F2-GCCATATGGCCGACGAAAAGCC (NdeI site in primer) and SUMO-2R2-CCGGATCCTCAACCTCCCGTCTGCT (BamHI site in primer) and were ligated into pCR2.1. A NdeI/BamHI restriction endonuclease fragment from recombinant SUMO-2/pCR2.1 was subcloned into the NdeI/BamHI site of pET-15b. UBC9, SUMO-1, and SUMO-3 cDNA were made by RT-PCR using primers that did not contain NdeI or BamHI restriction endonuclease sites: UBC9F1-CG-AGGGACTTTGAACATGTC; UBC9R1-CTGCGATGCCACAAGGT; SUMO-1F-GGTGAAGCCACCGTCATC; SUMO-1R-GAATATCTAAA-CTGTTGAATGACCC; SUMO-3F-CGCTCGGAAGCCATGT; and SUMO-3R-CTGCCAGGCTGCTCTCCG. An aliquot of the above reactions was re-amplified by PCR using primers containing NdeI and BamHI sites: UBC9F2-CGAGGGACTTTGCATATGTC; UBC9-R2-CAAGGTGGATCCTTATGAGG; SUMO-1F2-CCA-CCGTCCATA-TGTCTGACCAG; SUMO-1R2-AAAGAATATCGGATCCGTTGACTAA-CCCCCC; SUMO-3F2-CGCTCGGAACATATGTCCG; and SUMO-3-R2-CAGGCTGGGATCCGGCTAACCTCC. The secondary PCR products were digested with NdeI and BamHI and were subcloned directly into the NdeI/BamHI site of pET-15b. For subcloning into a mammalian expression vector, pCMV-Myc (Clontech, Palo Alto, CA), the recombinant SUMO-4/pET vectors were altered by 3 nucleotide changes to create a EcoRI site, in-frame with the first Met of both SUMO-4M and SUMO-4V. The QuikChange site-directed mutagenesis kit utilizing PfuTurbo DNA polymerase and primers 5Ј-GTGCCGCGCGGAAT-TCATATG-3Ј and its reverse complement were used per the manufacturer's instructions (Stratagene, La Jolla, CA). An EcoRI/BamHI double digest resulted in a restriction endonuclease fragment, which was subcloned into the EcoRI/BglII sites of pCMV-Myc. HSF2 plasmid was amplified with primers HSF2NDEF-CGCCGCGTTACATAT-GAAGCAGAG and HSF2BAMR-CTGTTTTATCGGATCCTTCTGAG-CAC, which resulted in a 410-bp PCR product containing NdeI and BamHI sites and included the complete DNA-binding domain (DBD) of HSF2. After digestion with NdeI and BamHI, the sequence was subcloned into the NdeI and BamHI sites of pET-15b. The recombinant HSF2-DBD/pET vector was altered to create a EcoRI site, in-frame with the first Met of HSF2, and the EcoRI/BamHI restriction endonuclease fragment was subcloned into the EcoRI/BglII sites of pCMV-Myc as described above and is called pCMV-Myc-HSF2-DBD82K. The lysine 82 was mutated to arginine using the primer 5Ј-GACTCTGGAATTGTA-AGGCAAGAAAGAGATG-3Ј and its reverse complement as described above and is called pCMV-Myc-HSF2-DBD82R.
Antibodies-The SUMO-4-specific antibodies were prepared by the following protocol. The antiserum was made in rabbits immunized with the peptide-CEPRGLSVKQIRFRFG (SUMO-4V) conjugated to keyhole limpet hemocyanin carrier protein through the N-terminal Cys of the peptide (Sigma Genosys). The IgG fraction of the antiserum (4 ml/run) was purified using the Ag antibody purification kit from Pierce. SUMO-2 peptide-CERQGLSMRQIRFRFD made by Sigma Genosys was bound to Sepharose using the Sulfolink kit from Pierce (Rockford, IL). The IgG was fractionated over the SUMO-2 peptide column using the Pierce protocol and was tested for specificity against recombinant proteins for SUMO-1, SUMO-2, SUMO-3, SUMO-4M, and SUMO-4V by Western blotting. The SUMO-4 antibody was used at a dilution of 1:1000 and detected only SUMO-4M and SUMO-4V at approximately equal signal intensities. Monoclonal antibody for c-Myc (Clontech, Palo Alto, CA) was used at a 1:2000 dilution.
Transfections-HepG2 were grown at 37°C and 5% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum, 1% penicillin-streptomycin, and 2 mM glutamine. Cells were seeded into 6-well plates and grown for 24 h. The cells were transfected at ϳ75% confluency with 20 g of DNA/well of pCMV-Myc-SUMO-4 constructs using a standard calcium phosphate protocol (13). After 6 h, the cells were shocked in 15% glycerol for 90 s and allowed to recover in Dulbecco's modified Eagle's medium with 10% fetal calf serum. After a 48-h incubation, the wells were washed with ice-cold phosphate-buffered saline and scraped and the cells were concentrated by centrifugation. Nuclear or cytoplasmic extracts were prepared as described previously (14). The pellet was suspended in buffer 1 (200 l of 10 mM Tris-HCl, pH 7.6, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT) for 30 min, cells were lysed by passage through a 22-guage needle, and nuclei were pelleted by centrifugation. The supernatant (cytoplasmic extract) was stored at Ϫ20°C, and the nuclei were suspended in buffer 2 (50 l of 20 mM Tris-HCl, pH 7.6, 25% sucrose, 400 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM DTT) for 60 min. The nuclear extract was recovered from the supernatant after centrifugation and stored at Ϫ20°C. Protein concentration was measured using the Bradford assay as described by the manufacturer (Bio-Rad). 20 g of nuclear and cytoplasmic extracts were analyzed by Western blotting using c-Myc primary antibody. For co-transfections of pCMV-Myc-SUMO-4/pCMV-Myc-DBD-HSF2 constructs, 1 g of each plasmid was transfected in 24-well plates as described above and whole cell extracts (WCEs) were made by dissolving the cells directly into 100 l of 1ϫ NuPAGE lithium dodecyl sulfate sample buffer containing reducing agent (Invitrogen). 20 l of WCE was analyzed by Western blotting using the c-Myc primary antibody.
Western Blotting-Nuclear, cytoplasmic, or WCEs were resolved on 10% SDS-PAGE (NuPAGE) and transferred onto nitrocellulose. Blots were blocked in 5% nonfat dry milk and probed with primary antibodies followed by anti-rabbit-horseradish peroxidase or anti-mouse-horseradish peroxidase secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Bands were visualized using ECL chemiluminescent kit (Amersham Biosciences) and autoradiography (Kodak X-Omat AR film).
Luciferase Assay-Transfections for luciferase reporter assays were done as described above using 2.2 g of transfected DNA in 24-well plates. The amount of DNA transfected was 1 g of test plasmid (either the SUMO-4M or SUMO-4V construct or the control pCMV-Myc lacking the SUMO-4 sequences), 1 g of either firefly luciferase reporter vector (pHSE-Luc or pNF-B-Luc) or the pTAL-Luc control vector lacking any response element, and 0.2 g of Renilla vector. The Renilla vector (phRL-null) served as an internal control for transfection efficiency. Dual-luciferase assays were done using a kit from Promega (Madison, WI) as described previously (13). The amount of firefly luciferase activity was normalized to that of the control Renilla luciferase activity. To determine the specific effect of the heat shock element (HSE) and NF-B response elements without SUMO-4 overexpression in HepG2 cells, the combination of pCMV-Myc control/pTAL-Luc/phRL-null vec-tor was transfected to determine background luciferase activity. The combination of pCMV-Myc control/pHSE-Luc/phRL-null or pCMV-Myc control/pNF-B-Luc/phRL-null were transfected next into HepG2 cells, and the ratio of firefly to Renilla luciferase was compared with the background levels determined above. This comparison provided a relative amount of active transcription factors in the resting HepG2 cells that could bind to the HSE or NF-B response elements, respectively. SUMO-4M or SUMO-4V/pHSE-Luc or pNF-B-luc/phRL-null combinations were transfected into HepG2 cells to determine the effect of SUMO-4M or SUMO-4V overexpression on the specific response elements. The effect of overexpression of SUMO-4 on reporter gene expression was compared with the levels of reporter expression lacking SUMO-4 calculated above. To study the effect of oxidative stress, H 2 O 2 (Sigma) was added to media to a final concentration of 2 or 4 mM during the final 18 h of transfection. 4 -8 transfections were done for each combination of constructs, and the results were plotted as the mean Ϯ S.E. The student's t test was used to determine the significance of the findings.
In Vitro Sumoylation Assay-HepG2 cells (ϳ3 ϫ 10 5 ) were concentrated by centrifugation and suspended in 100 l of solution 3 (50 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 1 mM DTT, 12.5 mM MgCl 2 , 20% glycerol, 0.1 M KCl, 1% Triton X-100) (14). After 5 min of incubation on ice, the lysed cells were centrifuged for 5 min at 10,000 rpm and the supernatant was collected. The whole cell extract was used in the in vitro sumoylation assay. Formation of the thioester adducts between the recombinant E2 enzyme UBC9 and SUMO-4M or SUMO-4V was initiated essentially as described previously (15). Reactions contained whole cell extracts from HepG2 (7 g), and recombinant SUMO-4M or SUMO-4V (2 g) and UBC9 (5 g) were incubated in 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 5 mM ATP, and 2 mM DTT (30 l of total reaction volume). Control reactions lacking recombinant UBC9 or including 100 mM DTT also were prepared. Reactions were incubated for 90 min at 30°C and then stopped by adding 5 l of NuPAGE lithium dodecyl sulfate buffer and heated to 100°C for 4 min. Samples (15 l) were resolved on 10% SDS-PAGE, transferred onto nitrocellulose, and analyzed by Western blotting using SUMO-4-specific primary and antirabbit-horseradish peroxidase secondary antibodies as described above.
Genetic Analyses-The chromosome 6q25 region has been shown to be associated with type I diabetes susceptibility (IDDM5) by linkage studies (16). SUMO-4 is located in this region. The Tsp test (17) was used to access linkage disequilibrium between the SUMO-4 marker alleles and disease, because this test provides a valid 2 test for linkage disequilibrium in the presence of linkage. Tsp examines whether a heterozygous parent transmits the same or different marker alleles to each of their affected children (sibling pairs).

Tissue Specificity of an mRNA Encoded by a Novel SUMO
Gene on Chromosome 6q25-A SUMO-related nucleotide sequence is located within intron 6 of the TAB2 (mitogen-activating protein kinase kinase kinase 7-interacting protein-2, MAPK7IP2) gene on chromosome 6q25 (TAB2 DNA NCBI accession number AL031133 and smt3-like protein NCBI accession number CAA20019). In contrast to SUMO-1, SUMO-2, and SUMO-3, the SUMO-related DNA sequence does not contain any introns; however, it contains an open reading frame of 95 amino acids, a putative polyadenylation site, and a consen-sus ribosome-binding site (Kozak sequence) for correct translation initiation. Moreover, the predicted amino acid sequence is similar to that of SUMO-1, SUMO-2, and SUMO-3 with the largest sequence homology to SUMO-2 ( Fig. 1, 87% amino acid homology). We have named this "putative" protein SUMO-4 based on the strong sequence homologies.
To determine whether an mRNA specific to SUMO-4 is expressed, we performed RT-PCR from total RNA of various tissues. In contrast to SUMO-2, which is highly expressed in all of the tissues examined, SUMO-4 mRNA was detected mainly in kidney and HEK cells (Fig. 2). A nucleotide sequence analysis of the RT-PCR products confirmed their identities as SUMO-2 and SUMO-4, respectively.
Amino Acid Polymorphism in SUMO-4 -A SNP was identified previously in this gene (rs237025, chromosome 6 position 149656820; www.ncbi.nlm.nih.gov/). The C/T polymorphism is in the antisense strand and encodes methionine 55 (ATG) and valine 55 (GTG), respectively. Indeed, DNA isolated from HEK cells is heterozygous for the A/G DNA polymorphism. To determine the frequency of the SNP, we typed 115 unrelated Caucasian individuals. The allele frequencies of A and G were 0.49 and 0.51, respectively. Met-55 is conserved in SUMO 1, SUMO-2, and SUMO-3 proteins in humans (Fig. 1) as well as in SUMO-related sequences in the following organisms: mouse; Xenopus; Drosophila, and yeast (18). The high degree of conservation of Met-55 between SUMO proteins within humans and among different species suggests that the M55V substitution may affect the structure and function of SUMO proteins.
In Vitro Sumoylation-To determine whether SUMO-4 utilizes the E2 enzyme, UBC9, we tested recombinant SUMO-4M and SUMO-4V proteins in an in vitro sumoylation assay containing whole cell extract from HepG2 cells containing or lacking recombinant UBC9 (Fig. 3, lanes 1-6). High molecular weight (HMW) bands with molecular masses greater than 60 kDa were identified in Western blots for both SUMO-4M and SUMO-4V proteins (Fig. 3, lanes 1 and 4, respectively) but were absent when recombinant UBC9 was omitted from the reaction (Fig. 3, lanes 2 and 5). The HMW bands were also absent when recombinant UBC9 plus 100 mM DTT was present (Fig. 3, lanes  3 and 6). DTT (100 mM) inhibits the formation of the thioester bonds and therefore provides additional evidence that the HMW bands are the result of sumoylation (Fig. 3, lanes 1 and  4). UBC9-dependent SUMO-4 bands of 26 and 33 kDa are also present but in relatively lesser quantities than the HMW bands (Fig. 3, lane 1 and very weak in lane 4). These complexes are of the correct size for a SUMO-4 dimer and a SUMO-4⅐UBC9 complex, respectively (Fig. 3).
Transfection Studies in HepG2 Cells-To test SUMO-4 overexpression in vivo, we used SUMO-4M and SUMO-4V constructs tagged with c-Myc in mammalian expression vectors containing the CMV promoter that were transfected into HepG2 cells (Fig. 3, lanes 7-10). In cytoplasmic extracts prepared from SUMO-4M and SUMO-4V transfections, most of the protein detected in Western blots was unconjugated free SUMO-4 (Fig. 3, lanes 8 and 10). In contrast, nuclear extracts from these experiments demonstrated mostly multiple bands greater than 50 kDa (Fig. 3, lanes 7 and 9). These results suggest that SUMO-4 targets multiple proteins, and these are found primarily in the nucleus. Furthermore, the SUMO-4M and SUMO-4V expression products detected by the c-Myc tag appear to be equally present in the HepG2 cells (Fig. 3, lanes  7-10), demonstrating that the most abundant sumoylated proteins in HepG2 cells are not sumoylated differentially by the Met and Val variants.
Association of SUMO-4M with Susceptibility to Type I Diabetes Mellitus-Studies have shown the linkage of markers on chromosome 6q25 (IDDM5) with susceptibility to type I diabetes mellitus (16). However, the specific identity of IDDM5 has yet to be determined. Type I diabetes is characterized by selective ␤-cell destruction, resulting in an absolute requirement for exogenous insulin and a young albeit heterogeneous age of onset. The etiology and pathogenetic mechanisms of ␤-cell destruction are not understood completely, although an autoimmune process clearly is involved. The location of SUMO-4 in the IDDM5 region makes it an interesting "candidate gene" for the IDDM5 diabetes susceptibility. We typed the M55V DNA polymorphisms (A/G) in 478 families to determine whether there was a familial association with type I diabetes. The A polymorphism (encoding Met) was transmitted 57.1% of the time (467/818) to the diabetic offspring from heterozygous (A/G) parents (Tsp ϭ 4.0 ϫ 10 Ϫ4 ). For comparison, the A polymorphism was transmitted 46.2% of the time (78/169) to the unaffected offspring from the heterozygous parents. The transmission results in the unaffected siblings demonstrates that the association with type I diabetes susceptibility is not a general transmission bias. DISCUSSION We have characterized a novel SUMO gene that is related to SUMO-2 but whose mRNA is expressed in a limited number of tissues. We confirm that this novel protein belongs to the SUMO family, because its expression uses the SUMO-specific E2 enzyme UBC9 in vitro. This becomes the fourth member of the SUMO family (SUMO-4). We identified and characterized a polymorphism within this family, SUMO-4M55V. Although, the allelic frequency of SUMO-4V is 0.51 in the Caucasian population, the valine substitution has not been reported in any other human SUMO member (18). The highly conserved nature of this Met amino acid across various SUMO family members (Fig. 1) and across species (18) suggests that it may be biologically important for the function of the molecule.
We demonstrate an increase in HSE reporter expression due to SUMO-4M compared with SUMO-4V. However, the differences in reporter expression were specific for HSE and were not found for NF-B (Fig. 5) or activator protein 1 (data not shown). HSF1 molecules are not normally active in resting cells but become active after cells are exposed to various stress conditions (19). In contrast, HSF2 is activated by hemin in human K562 erythroleukemia cells and is active in mouse ES cells during early embryogenesis but is not activated by the classical stress stimuli (19). The difference in HSF-reporter expression due to SUMO-4M compared with SUMO-4V also was found under mild oxidative stress conditions (2 mM H 2 O 2 ), which resulted in a 2-fold activation of endogenous HSFs (Fig. 6). However, under severe oxidative stress (4 mM H 2 O 2 ), the increased levels of HSE-reporter levels of Met-55 compared with Val-55 was less evident (Fig. 6) and may be attributed to endogenous SUMO-1, SUMO-2, or SUMO-3, competing with SUMO-4 for activated HSF-1 substrate. Alternatively, the HSE reporter assay may no longer be in a linear range because of saturation of HSE binding sites in the reporter. It remains to be determined whether our differential activation of HSFs by SUMO-4M compared with SUMO-4V is due to HSF1 and/or HSF2. However, the roles of HSF1 and HSF2 may not be independent in stressed cells because interactions and cooperative effects have been detected in both of them (20).
We also have demonstrated that SUMO-4M is associated with type I diabetes susceptibility. SUMO-4 sumoylation of HSF1 and/or HSF2 may be involved directly in the type I diabetes pathogenesis through abnormal HSP expression in pancreatic ␤-cells. HSPs are "molecular chaperons" present during the assembly of proteins and take part in the processing and presentation of antigens. In the non-obese diabetic mouse, the onset of ␤-cell destruction is associated with spontaneous development of T-lymphocytes reactive to members of the 60-kDa heat shock protein (Hsp60). Autoimmune non-obese diabetes may involve molecular mimicry between epitopes of Hsp60 and a ␤-cell-specific molecule (21). Furthermore, in humans, autoantibodies are found in multiple autoimmune diseases, including type I diabetes and rheumatoid arthritis, which react with Hsp60 (22). Type I diabetes seems to respond favorably to therapeutic vaccination with a peptide of Hsp60 (23). The SUMO-4 association with the type I diabetes may be due to altered HSF regulation of Hsp60, although the specific mechanism of susceptibility remains to be determined.