UBASH3A Mediates Risk for Type 1 Diabetes Through Inhibition of T-Cell Receptor–Induced NF-Kb Signaling

Diabetes Volume 66, July 2017 2033

UBASH3A Mediates Risk for Type 1 Diabetes Through Inhibition of T-Cell Receptor–Induced NF-kB Signaling

Yan Ge,1,2 Taylor K. Paisie,1,2,3 Jeremy R.B. Newman,2,4 Lauren M. McIntyre,2,4 and Patrick Concannon1,2

Diabetes 2017;66:2033–2043 | https://doi.org/10.2337/db16-1023

Although over 40 type 1 diabetes (T1D) risk loci have been study focuses on one such locus, on chromosome 21q22.3, mapped in humans, the causative genes and variants for containing two genes, TMPRSS3 and UBASH3A,ofwhich GENETICS/GENOMES/PROTEOMICS/METABOLOMICS T1D are largely unknown. Here, we investigated a candi- the latter is generally considered the most likely candidate date gene in the 21q22.3 risk locus—UBASH3A,whichis for this T1D risk locus (2–9). Besides T1D, single nucleotide primarily expressed in T cells where it is thought to play a polymorphisms (SNPs) in the 21q22.3 chromosomal region largely redundant role. Genetic variants in UBASH3A have are associated with several other autoimmune diseases, sug- been shown to be associated with several autoimmune gesting that this locus plays a broad role in autoimmunity diseases in addition to T1D. However, the molecular mech- (10–12). anism underlying these genetic associations is unresolved. UBASH3A (also known as STS-2, TULA,andCLIP4)is Our study reveals a previously unrecognized role of UBA- expressed primarily in T cells (13) and encodes a protein SH3A in human T cells: UBASH3A attenuates the NF-kB called ubiquitin-associated and SH3 domain–containing A signal transduction upon T-cell receptor (TCR) stimulation (UBASH3A). The UBASH3A protein has three functional by specifically suppressing the activation of the IkBki- nase complex. We identify novel interactions of UBASH3A domains: the N-terminal UBA (ubiquitin-associated), SH3 with nondegradative polyubiquitin chains, TAK1 and (SRC homology 3), and the COOH-terminal histidine phos- NEMO, suggesting that UBASH3A regulates the NF-kB phatase (also referred to as phosphoglycerate mutase-like signaling pathway by an ubiquitin-dependent mechanism. [PGM]) domains. It has been shown that the UBA domain Finally, we show that risk alleles at rs11203203 and binds to monoubiquitin (Ub); the SH3 domain interacts with rs80054410, two T1D-associated variants in UBASH3A, CBL—an E3 ubiquitin ligase—and dynamin; and the PGM increase UBASH3A expression in human primary CD4+ domain mediates self-dimerization (14–18). UBASH3A has T cells upon TCR stimulation, inhibiting NF-kB signaling four identified ubiquitination sites at lysine residues 15, via its effects on the IkB kinase complex and resulting in 202, 309, and 358. Monoubiquitination at Lys 202 causes reduced IL2 gene expression. UBASH3A to adopt a closed conformation, which prevents the binding of the UBA domain to substrates in trans (17). UBASH3A has a paralogue, UBASH3B (also known as Type 1 diabetes (T1D) is an autoimmune disease arising from STS-1 and TULA-2), which shares the same domain structure the destruction of the insulin-producing pancreatic b-cells. as UBASH3A. UBASH3B differs from UBASH3A in several T1D is a common, complex disease with multiple genetic significant ways. UBASH3B is ubiquitously expressed and and environmental risk factors. Although genome-wide has not been associated with any autoimmune or immune- association studies have discovered over 40 chromosomal mediated disorder in genome-wide association studies. regions where there is significant statistical evidence of UBASH3B displays significant protein tyrosine phosphatase association with T1D (1), the causative genes and variants activity both in vitro and in vivo and suppresses T-cell re- located in most of these regions have yet to be identified ceptor (TCR) signaling by dephosphorylating ZAP-70 and and their mechanisms of action determined. The current Syk (19–23). In contrast, UBASH3A exhibits very weak,

1Department of Pathology, Immunology and Laboratory Medicine, University of This article contains Supplementary Data online at http://diabetes Florida, Gainesville, FL .diabetesjournals.org/lookup/suppl/doi:10.2337/db16-1023/-/DC1. 2 Genetics Institute, University of Florida, Gainesville, FL © 2017 by the American Diabetes Association. Readers may use this article as 3 Genetics & Genomics Graduate Program, University of Florida, Gainesville, FL long as the work is properly cited, the use is educational and not for profit, and the 4 Department of Molecular Genetics & Microbiology,UniversityofFlorida,Gainesville,FL work is not altered. More information is available at http://www.diabetesjournals Corresponding author: Patrick Concannon, patcon@ufl.edu. .org/content/license. Received 22 August 2016 and accepted 20 March 2017. 2034 Novel Role for UBASH3A in T1D Susceptibility Diabetes Volume 66, July 2017 possibly acid-dependent, phosphatase activity in vitro; Stimulation and Lysis of Cells in vivo, knockout of the murine homolog of UBASH3A For stimulation longer than 30 min, cells were stimu- results in only a modest increase in phosphorylation of lated with 5 mg/mL plate-bound anti-CD3 (clone OKT3; ZAP-70 (19,24). BioLegend) with or without 5 mg/mL soluble anti-CD28 Mice lacking either Ubash3a or Ubash3b alone, or in antibody (clone CD28.2; BioLegend). For stimulation shorter combination, exhibit no overt defects without immune chal- than 30 min, cells were starved of serum for 4 h, and then 2/2 2/2 lenge (13). However, T cells from Ubash3a Ubash3b incubated for 30 min on ice with 10 mg/mL soluble anti- double-knockout mice are hyperresponsive to TCR stim- CD3 and 10 mg/mL soluble anti-CD28. Next, 10 mg/mL ulation compared with T cells from wild-type (WT) mice, goat anti-mouse IgG (SouthernBiotech) was added to the 2/2 2/2 whereasTcellsfromUbash3a and Ubash3b single- cells, followed by incubation at 37°C for the indicated pe- knockout mice display only a modest increase in prolif- riods of time. Mock stimulation was performed with only eration (19). A similar hierarchical response is seen in cell culture medium. the trinitrobenzene sulfonic acid–induced colitis model, Whole-cell lysates were extracted as previously described where knockout of either Ubash3a or Ubash3b increases (27). For some experiments as indicated in the figure leg- both inflammation and T-cell responses, but the ends, whole-cell lysates were extracted with cell lysis buffer 2/2 2/2 Ubash3a Ubash3b double-knockout mice display a containing 25 mmol/L HEPES, pH 7.0, 150 mmol/L NaCl, more severe phenotype than either of the single-knockout 0.5% NP-40, 1 mmol/L EDTA, protease and phosphatase mice (25). These findings suggest that Ubash3a, in combi- inhibitors, and 5 mmol/L N-ethylmaleimide (NEM), a deu- nation with Ubash3b, acts to inhibit T-cell activation and biquitinase inhibitor. function, albeit by an as-yet-unresolved mechanism. Nuclear extracts were prepared using the NE-PER Nuclear In the current study, we define the roles of UBASH3A and Cytoplasmic Extraction Reagents (Thermo Fisher Scien- and its genetic variants in T1D. Our study reveals novel tific), according to the manufacturer’sprotocol. interactions between UBASH3A, TAK1, and NEMO, which regulate TCR-induced NF-kB signaling. T1D risk alleles Transfection of HEK293T Cells in UBASH3A are shown to be associated with increased AcDNAofthefull-lengthtranscriptofUBASH3A was cloned UBASH3A expression and decreased IL2 expression in acti- into the pcDNA3.1 vector (Thermo Fisher Scientific). Expres- vated human primary CD4+ T cells. sion constructs encoding WT (Addgene plasmid #17608), lysine-48 (K48)-only (Addgene plasmid #17605), and lysine-63 (K63)-only Ub (Addgene plasmid #17606) tagged with hem- RESEARCH DESIGN AND METHODS agglutinin (HA) were gifts from Ted Dawson (Johns Hopkins Sample Information University) (28). HEK293T cells were transfected using the Frozen viable peripheral blood mononuclear cells (PBMCs) X-tremeGENE HP DNA Transfection Reagent (Roche). from healthy subjects of European ancestry were obtained Coimmunoprecipitation and Immunoblotting from the Type 1 Diabetes Genetics Consortium (T1DGC) Coimmunoprecipitation and immunoblotting were performed UBASH3A and from STEMCELL Technologies. genotyping as previously described (27), and antibodies used for these data used in this study were either obtained from T1DGC experiments are provided in Supplementary Table 1. or generated by PCR and Sanger sequencing. All biospeci- mens and data were represented by only nonidentifying Quantitative PCR codes. This study was approved by the University of Florida Frozen PBMCs from healthy subjects were thawed, and Institutional Review Board. primary CD4+ T cells were negatively selected using the Human CD4+ T Cell Isolation kit and LS MACS columns Generation of UBASH3A2/2 and (Miltenyi). Cells were stimulated as described above for 6 h, UBASH3A-Overexpressing Cell Clones and then total RNA was extracted using the RNeasy Plus To knockout UBASH3A in Jurkat cells, a CRISPR construct Mini kit (QIAGEN). First-strand cDNA was synthesized us- targeting exon 2 of the UBASH3A gene was generated (26) ing oligo(dT)20 primer and the iScript Select cDNA Synthesis using the guide sequence 59-CACGGGGAGGAAGACGGC kit (Bio-Rad). PCRs containing SYBR Green I were performed GG-39 and the pSpCas9n(BB)-2A-Puro plasmid (Addgene on a LightCycler 480 II real-time PCR instrument (Roche). All plasmid #48141, a gift from Feng Zhang, Massachusetts samples were tested in duplicate, and CT values were gener- Institute of Technology). To overexpress UBASH3A in ated by the second derivative maximum method provided Jurkat cells, a cDNA of the full-length, predominantly by the Roche software. Relative gene expression levels were DCT expressed transcript of UBASH3A was cloned into the calculated using the 2 method, where DCT =meanCT pEF-DEST51 vector (Thermo Fisher Scientific). The CRISPR (GAPDH) – mean CT (gene of interest). The primers used and pEF-DEST51 constructs were delivered into Jurkat cells for the assay are provided in Supplementary Table 2. by electroporation. Cell clones were obtained by limiting 2/2 dilution. UBASH3A clones were screened by PCR and IL-2 ELISA Sanger sequencing, and UBASH3A-overexpressing clones Cells were stimulated as described above for 24 h, and were identified by immunoblotting with anti-V5. the Human IL-2 ELISA Max Deluxe kit (BioLegend) diabetes.diabetesjournals.org Ge and Associates 2035 was used to measure IL-2 production in the culture supernatants.

Glutathione S-Transferase Pull-Down Assays cDNA sequences of the UBA (residues 20–60), SH3 (residues 241–300), and PGM (residues 316–623) domains of UBASH3A were cloned into the pGEX-6P-1 vector (GE Healthcare). Glutathione S-transferase (GST)-tagged UBA, SH3, and PGM domains of UBASH3A were expressed in Escherichia coli and puriﬁed. Whole-cell lysate from unstimulated Jurkat cells was precleared by incubation with Glutathione Sepharose 4B resin for 2 h at 4°C. One milligram of the precleared lysate was incubated for 2 h at 4°C with 20 mL of packed, fresh Glu- tathione Sepharose 4B resin, and with one of the following at the same molar concentration: 11.6 mg GST-tagged UBA, 12.6 mg GST-tagged SH3, or 22.9 mg GST-tagged PGM domain of UBASH3A and 10 mgGST.Nolysatecontrols contained all the same ingredients as the experimental re- actions except for the precleared lysate. Eluates were extracted from the resin using LDS sample buffer and analyzed by immunoblotting.

Binding of UBASH3A to Ubiquitin Oligomers Five hundred micrograms of whole-cell lysate from HEK293T cells expressing V5-tagged UBASH3A was incubated with 0.5 mg of anti-V5 antibody for 4 h at 4°C. Next, 20 mL Protein G Dynabeads were added to the lysate, followed by 1-h incubation at 4°C. The beads were washed Figure 1—Generation and characterization of Jurkat-derived clones four times with the cell lysis buffer, and the supernatant deficient in or overexpressing UBASH3A. A: Two Jurkat-derived UBASH3A2/2 was discarded. The beads were resuspended in the cell lysis clones, 2.1D6 and 2.1F7, were generated by CRISPR targeting exon 2 of the UBASH3A gene. Partial exon 2 sequences of fi 2/2 buffer, and 2.5 mgofaspeci c type of ubiquitin oligomers parental Jurkat cells and the two UBASH3A clones are shown. was added to the beads, K48-linked Ub oligomers (K482–7), Insertions and deletions are underscored and italicized. B:Whole- K63-linked Ub oligomers (K63 – ), or methionine-1 (M1)- cell lysates from 2.1D6, 2.1F7, and Jurkat cells were immunopre- 2 7 – linked tetra-ubiquitin oligomers (M1 ) (Boston Biochem). cipitated with either anti-UBASH3A (lanes 1 3) or IgG (lane 4). The 4 immunoprecipitates were subjected to immunoblotting with a different After 2-h incubation at 4°C and four washes of the beads, anti-UBASH3A antibody. UBASH3A and its monoubiquitinated form the captured Ub oligomers and proteins were eluted using are indicated by arrow and asterisk, respectively. C: Immunoblot anal- LDS sample buffer containing 50 mmol/L DTT. The eluates ysis of whole-cell lysates from 2.1D6, 2.1F7, and Jurkat cells using anti- UBASH3B. D: Whole-cell lysates from three Jurkat-derived clones were subjected to immunoblotting with a polyclonal anti- overexpressing UBASH3A with a COOH-terminal V5 tag were sub- ubiquitin antibody. jected to immunoblotting with anti-V5 and subsequently with anti– g-tubulin after stripping. The blots in B–D are representative of two Statistical Analyses independent experiments. KO, knockout. The Prism software v. 6.0 (GraphPad) was used to perform Student t tests and linear regression analysis and to calcu- late Pearson correlation coefficient.

a COOH-terminal V5-tagged version of the 623-amino– RESULTS acid, major isoform of UBASH3A in humans was expressed UBASH3A Downregulates IL-2 Production in WT Jurkat cells, and three clones were generated by To explore the function of UBASH3A in humans, Jurkat limiting dilution—1F6, 2F5, and 5E4. The expression levels lymphoblastic T cells null for UBASH3A were generated of V5-tagged UBASH3A and its monoubiquitinated form using CRISPR/Cas9. Two clones, 2.1D6 and 2.1F7, with were comparable in 1F6 and 2F5 cells but higher than those distinct frameshifting mutations in exon 2 of the UBASH3A in 5E4 cells (Fig. 1D). gene on both alleles, were obtained by limiting dilution To examine the effect of modulating UBASH3A expression 2/2 (Fig. 1A). Full-length UBASH3A protein could not be de- levels on T-cell activation and function, the UBASH3A tected in the whole-cell lysates from 2.1D6 and 2.1F7 cells clones and the UBASH3A-overexpressing clones were stim- (Fig. 1B), whereas normal levels of UBASH3B could (Fig. ulated for 24 h with either anti-CD3 alone or anti-CD3 plus 1C), conﬁrming the speciﬁcity of the knockouts. In parallel, anti-CD28. IL-2 production, with either stimulation regimen, 2036 Novel Role for UBASH3A in T1D Susceptibility Diabetes Volume 66, July 2017

2/2 was two- to eightfold higher in UBASH3A 2.1D6 and 5E4 clone had the lowest level of exogenous UBASH3A 2.1F7 cells compared with parental Jurkat cells (P , expression and produced the largest quantity of IL-2 0.0001) (Fig. 2A). Overexpression of UBASH3A resulted in (Figs. 1D and 2B). 2/2 a comparably signiﬁcant attenuation of IL-2 production Quantiﬁcation of IL2 transcripts in the UBASH3A upon stimulation (Fig. 2B). A dose effect was apparent clones revealed the same inverse correlation with UBASH3A among the three UBASH3A-overexpressing clones: the expression as observed for IL-2 secretion (Fig. 2C). Thus,

Figure 2—IL-2 production in Jurkat cells and in Jurkat-derived clones deficient in or overexpressing UBASH3A. A and B: ELISA measurements of IL-2 in the culture supernatants of Jurkat cells and of UBASH3A2/2 (A)andUBASH3A-overexpressing (B) clones after 24-h stimulation with either anti-CD3 alone or anti-CD3 plus anti-CD28. The dashed lines indicate the ELISA detection limit (i.e., 7.8 pg/mL), and data points below the dashed lines represent values extrapolated from the IL-2 ELISA standard curve. The data are pooled from four ELISA experiments. C:Relative 2/2 mRNA levels of IL2 in Jurkat cells and in UBASH3A 2.1D6 and 2.1F7 cells. The cells were stimulated for 6 h as described in RESEARCH DESIGN AND METHODS. The data are pooled from two quantitative PCR experiments. Each data point in A–C represents an individual measurement of one sample. Numbers of samples (n) are shown. The mean and SEM values are indicated by solid lines and error bars, respectively. Unpaired two- tailed Student t tests were performed to assess statistical significance: *P = 0.02, ***0.0001 # P < 0.0002, ****P < 0.0001. diabetes.diabetesjournals.org Ge and Associates 2037 the effect of modulating UBASH3A expression on IL-2 pro- encoding critical cytokines and chemokines, such as duction in Jurkat cells is largely attributable to the effect on IL2, in T cells (29,30). IL2 transcription. To determine whether the observed effect of UBASH3A on IL2 transcription is mediated through NF-kB signaling, UBASH3A Suppresses TCR-Induced NF-kB Signaling nuclear extracts were prepared and immunoblotted for The NF-kB family of transcription factors plays a central NF-kB p65 (also known as RelA). The baseline levels of 2/2 role in innate and adaptive immune responses by regulating nuclear NF-kB p65 were similar in UBASH3A 2.1F7 the proinflammatory and antiapoptotic gene transcription cells, UBASH3A-overexpressing 2F5 cells, and Jurkat cells programs. In unstimulated T cells, NF-kB dimers are (Fig. 3A). However, upon 6-h anti-CD3 plus anti-CD28 2/2 retained in an inactive state in the cytoplasm bound to stimulation, UBASH3A 2.1F7 cells had increased level IkB proteins. Upon TCR engagement, the IkBkinase of NF-kB p65 in the nucleus relative to Jurkat cells, whereas (IKK) complex, consisting of IKKa and IKKb (two kinases) 2F5 cells appeared to have less (Fig. 3A and Supplementary and the regulatory component NEMO, is phosphorylated Table 3). These findings suggest that the effect of UBASH3A and activated and, in turn, phosphorylates IkBproteins on IL-2 production is mediated by NF-kB and that UBASH3A leading to their degradation in proteasomes. Once released acts upstream of the nuclear translocation of NF-kB. from the binding of IkB proteins, cytosolic NF-kBdimers To assess the effect of manipulating UBASH3A levels on translocate to the nucleus, bind specific DNA sequences, the activity of the IKK complex, the amount of active, pho- and promote transcription of target genes, including those sphorylated IKKa and IKKb was measured after TCR/CD28

Figure 3—UBASH3A inhibits TCR-induced NF-kB signaling. A: Nuclear extracts were prepared from Jurkat, UBASH3A2/2 2.1F7, and UBASH3A-overexpressing 2F5 cells after mock stimulation or stimulation (Stim) with anti-CD3 plus anti-CD28 for 6 h and then subjected to immunoblotting with anti-NF-kB p65 followed by anti-ﬁbrillarin (as loading control) after stripping. B–E: Jurkat and UBASH3A2/2 2.1F7 cells were mock stimulated or stimulated with anti-CD3 plus anti-CD28 for the indicated periods of time, and whole-cell lysates from the cells were analyzed. B: Immunoblotting with an antibody recognizing both phospho-IKKa (Ser176/180) and phospho-IKKb (Ser177/181) and then, after stripping, with anti-IKKa/b and anti–g-tubulin sequentially. C and D: Immunoblotting with anti-IkBa (C)oranti–phospho-PKCu (Thr538) (D) and subsequently with anti–g-tubulin after stripping. E: The lysates, which contained 5 mmol/L NEM, were immunoprecipitated with anti- phosphotyrosine or IgG, and the immunoprecipitates were subjected to immunoblotting with anti-TAK1. The lysates were also directly analyzed by immunoblotting with anti-TAK1 and subsequently with anti–g-tubulin after stripping. The blots in A–E are representative of three independent experiments. KO, knockout. 2038 Novel Role for UBASH3A in T1D Susceptibility Diabetes Volume 66, July 2017

2/2 costimulation. Upon activation, UBASH3A 2.1F7 cells synthetic Ub oligomers, including K48-pUb (2–7 Ub), K63- had increased levels of phosphorylated IKKa/b compared pUb (2–7 Ub), and M1-linked tetra-ubiquitin (4 Ub), was with Jurkat cells (Fig. 3B and Supplementary Table 3). Con- assessed by immunoblotting with an anti-ubiquitin anti- sistent with this observation, the degradation of IkBa body. V5-tagged UBASH3A preferentially bound to the 2/2 was accelerated in UBASH3A 2.1F7 cells compared with K63-linked pUb chains (K63-linked chains with 2 and Jurkat cells, either in the absence of stimulation or after 3 ubiquitin moieties were detected with longer exposure anti-CD3 plus anti-CD28 stimulation for various periods of of the blot shown in Fig. 4A) and the M1-linked tetra-Ub time ranging from 15 min to 6 h (Fig. 3C and Supplemen- chain (Fig. 4A and B). However, no binding to K48-linked tary Table 3). pUb chains was observed even with extended exposure of Knocking out UBASH3A did not markedly alter the acti- the blot shown in Fig. 4A (data not shown). vation of PKCu and TAK1 upon TCR/CD28 costimulation, TodeterminewhattypeofpUbchainscanbeconjugated as similar amounts of phosphorylated PKCu and phosphor- to UBASH3A, HEK293T cells were cotransfected with 2/2 ylated TAK1 were observed in UBASH3A 2.1F7 cells constructs encoding V5-tagged UBASH3A and one of the compared with parental Jurkat cells (Fig. 3D and E and following HA-tagged Ub—WT Ub, K48-only Ub, or K63- Supplementary Table 3). PKCu is selectively required for only Ub. The latter two types of Ub had all lysine residues the TCR-induced NF-kB signaling (31) and is activated via mutated except for the indicated lysine residue, thus forcing PLCg1 and diacylglycerol as a result of TCR ligation, as well polyubiquitin linkage via the indicated lysine residue. Im- as via costimulatory signals from CD28 and PI3K activation. munoblotting with anti-V5 showed that all three types of PKCu phosphorylates the membrane-associated protein Ub chains could be conjugated to UBASH3A (Fig. 4C). Less CARMA1, which in turn recruits BCL10 and MALT1, form- K48-polyubiquitinated UBASH3A than K63-polyubiquiti- ing the CBM signaling complex. The active CBM complex nated UBASH3A was detected in the whole-cell lysates leads to the phosphorylation and activation of TAK1, a from the transfected HEK293T cells (Fig. 4C), but this dif- kinase that phosphorylates IKKa and IKKb and hence ference might result from the slightly lower expression level activates the IKK complex (29,32). The increased activity of of the exogenous K48-only Ub compared with that of the 2/2 the IKK complex in TCR-stimulated UBASH3A cells, K63-only Ub (Fig. 4D). compared with Jurkat cells, without comparable increases in the activation of the upstream signaling molecules PKCu UBASH3A Interacts With TAK1 and NEMO via and TAK1 suggests that UBASH3A specifically suppresses Its SH3 Domain the activation of the IKK complex by a previously unrecog- UBASH3A did not noticeably affect the polyubiquitination of nized mechanism. NEMO and TRAF6 upon TCR/CD28 costimulation, as similar amounts of polyubiquitinated NEMO and polyubiquiti- 2/2 UBASH3A Both Binds to and Is Modified nated TRAF6 were observed in Jurkat cells, UBASH3A by K63-Mediated Ubiquitin Chains 2.1F7 cells, and UBASH3A-overexpressing 2F5 cells (Supple- Polyubiquitination plays a key role in propagating TCR- mentary Fig. 1). UBASH3A could suppress the activation of induced NF-kB signaling, acting both as a signal for protein the IKK complex upon TCR stimulation by interacting with turnover and as a scaffold for protein complex assembly. subunits of the IKK and/or the TAK1 complex, either di- Upon TCR stimulation, IkBproteinsareconjugatedwith rectly or by binding to the Ub chains that are associated K48-linked polyubiquitin (pUb) chains and then degraded with those molecules. UBASH3A coimmunoprecipitated via the ubiquitin-proteasome pathway. MALT1 of the CBM with TAK1 in Jurkat cells with or without anti-CD3 plus complex binds to TRAF2 and TRAF6, leading to TRAF2/6 anti-CD28 stimulation (Fig. 5A). Interaction between V5- oligomerization and activation of the E3 ubiquitin ligase tagged UBASH3A and NEMO was detected by coimmuno- activity of TRAF2/6. TRAF2/6 then catalyzes K63-linked precipitation in 2F5 cells, with or without anti-CD3 plus polyubiquitination, which facilitates the activation of the anti-CD28 stimulation (Fig. 5B). TAK1 complex—consisting of TAK1, TAB1, and TAB2/3 To map the interaction of UBASH3A with TAK1 and —and the IKK complex. In addition, MALT1, BCL10, NEMO, the UBA, SH3, and PGM domains of UBASH3A TRAF6, and NEMO are all polyubiquitinated upon TCR werefusedwithGSTandexpressedandpurified from E. coli stimulation, which promotes the NF-kB signaling (32,33). (Supplementary Fig. 2). GST pull-down assays were carried The importance of polyubiquitination in transducing out using whole-cell lysates from unstimulated Jurkat cells TCR-induced NF-kB signaling and the presence of the func- and each of the purified GST-tagged domains of UBASH3A. tional ubiquitin-binding UBA domain in UBASH3A raise the The UBASH3A SH3 domain was able to pull down both possibility that UBASH3A might inhibit the NF-kB signal- TAK1 and NEMO (Fig. 5C). ing pathway by either regulating or responding to polyubiquitination events (15–17). To determine the specificity of Two T1D-Associated SNPs in UBASH3A Regulate the UBASH3A UBA domain for ubiquitin linkages, V5- UBASH3A and IL2 Expression in Human Primary tagged UBASH3A was expressed in HEK293T cells and CD4+ T Cells then immunoprecipitated using Protein G Dynabeads cou- Our studies indicate that modulating UBASH3A expression pled with anti-V5. The binding of V5-tagged UBASH3A to levels in Jurkat cells results in considerable changes in both diabetes.diabetesjournals.org Ge and Associates 2039

Figure 4—Binding of UBASH3A to ubiquitin oligomers and ubiquitination of UBASH3A. A: Whole-cell lysates from HEK293T cells expressing V5-tagged UBASH3A (lanes 1–4) or V5 (lanes 5–8) were incubated with Protein G Dynabeads coupled with anti-V5. In lane 9, no lysate was incubated with Protein G Dynabeads coupled with IgG. After washing, the Dynabeads were incubated with equal amounts of K48-pUb (2–7Ub), K63-pUb (2–7 Ub), or M1-linked tetra-ubiquitin (4 Ub) chains. The ubiquitin oligomerscapturedbytheDynabeadsweredetectedbyimmuno- blotting with anti-ubiquitin and subsequently with anti-UBASH3A after stripping. B: Two micrograms of each of the input ubiquitin oligomers used in A was subjected to immunoblotting with the same anti-ubiquitin antibody used in A. C and D:Inlanes1–3, HEK293T cells were cotransfected with constructs encoding V5-tagged UBASH3A and the indicated HA-tagged ubiquitin—WT, K48-only, and K63-only ubiquitin. In lanes 4 and 5, HEK293T cells were transfected with the construct encoding V5-tagged UBASH3A or the construct encoding HA-tagged WT ubiquitin, and whole-cell lysates containing 5 mmol/L NEM were subjected to immunoblotting with anti-V5 and subsequently with anti–g-tubulin after stripping in C and with anti-HA in D. All the blots in A–D are representative of two independent experiments.

TCR-induced NF-kB signaling and IL-2 production. To ex- 541T.C)—two highly correlated (r2 =1)SNPsinintron tend these results to primary human T cells, the effects of 4ofUBASH3A—as credible causative variants associated T1D-associated SNPs on UBASH3A and IL2 expression were with T1D with high statistical significance (9). The minor explored. A fine-mapping study using ImmunoChip identi- allele of each of these SNPs confers risk for T1D and occurs fied rs11203203 (c.554–365G.A) and rs80054410 (c.554– at a frequency of approximately 0.3 in the HapMap CEU 2040 Novel Role for UBASH3A in T1D Susceptibility Diabetes Volume 66, July 2017

Figure 5—UBASH3A interacts with TAK1 and NEMO. A:JurkatandUBASH3A2/2 2.1F7 cells were mock stimulated or stimulated (Stim) with anti-CD3 plus anti-CD28 for 10 min, and whole-cell lysates from the cells, which contained 5 mmol/L NEM, were immunoprecipitated with either anti-TAK1 or IgG. The immunoprecipitates were subjected to immunoblotting with anti-UBASH3A and subsequently with anti-TAK1 after stripping. B: UBASH3A-overexpressing 2F5 cells were mock stimulated or stimulated with anti-CD3 plus anti-CD28 for 15 min, and whole- cell lysates from the cells, which contained 5 mmol/L NEM, were immunoprecipitated with either anti-NEMO or IgG. The immunoprecipitates and the lysates were subjected to immunoblotting with anti-V5 and anti-NEMO, respectively. C:GSTpull-downassaysusingGST-taggedUBA,SH3, and PGM domains of UBASH3A with and without whole-cell lysate from unstimulated Jurkat cells. The pull-down products and the Jurkat lysate were subjected to immunoblotting with anti-TAK1 and anti-NEMO. Arrows indicate isoforms of TAK1. All the blots in A–C are representative of two to three independent experiments. KO, knockout.

population. Previous studies showed that rs11203203 and at rs11203203 and in subjects homozygous for the minor rs80054410 are located in a putative enhancer or super- risk allele (A) of rs11203203 were significantly higher than enhancer region in human primary CD4+ T cells (9,34). that in subjects homozygous for the major allele (G) (Fig. To delineate the functions of rs11203203 and 6B). To isolate the effects of rs11203203 from other loci, rs80054410, CD4+ T cells were purified by negative selec- allele-specific expression was assayed using mRNA-seq data tion from PBMCs from healthy subjects of European ances- on CD4+ T cells from 38 T1D subjects of European ancestry. try with different genotypes at these two SNPs (Fig. 6A, The risk allele of the noncoding SNP rs11203203 led to genotypes at rs80054410 are completely correlated with imbalanced expression of the alleles of rs2277798, a coding those at rs11203203). Quantitative PCR was conducted to SNP in exon 1 of UBASH3A (Supplementary Fig. 3). In stim- measure the relative amounts of UBASH3A and IL2 tran- ulated CD4+ T cells, subjects homozygous for the minor, scripts in CD4+ T cells that were stimulated with anti-CD3 risk allele of rs11203203 had significantly more UBASH3A plus anti-CD28, or mock stimulated, for 6 h. In stimulated mRNA and less IL2 mRNA, compared with subjects homo- CD4+ T cells, the relative mRNA level of IL2 was inversely zygous for the major allele of rs11203203 (Fig. 6B). correlated with that of UBASH3A (Pearson r = 20.39, two- tailed P = 0.03) (Fig. 6A), consistent with the finding in Jurkat cells that UBASH3A downregulated IL-2 production DISCUSSION upon TCR stimulation. In mock-stimulated CD4+ Tcells,the We previously identified a statistically significant associa- relative mRNA levels of UBASH3A in subjects heterozygous tion between SNPs in the chromosome 21q22.3 region and diabetes.diabetesjournals.org Ge and Associates 2041

only very weak phosphatase activity relative to UBASH3B, and there is no clear difference in the specificity of the phosphatase domains of the two proteins that would ex- plain the exclusivity of the observed genetic associations of multiple autoimmune diseases with UBASH3A.Thisledus to explore the function and specificity of the UBA and SH3 domains of UBASH3A in T cells. We found that UBASH3A preferentially binds to the nondegradative K63- and M1-linked Ub chains but not detectably to the degradative K48-linked Ub chains and that UBASH3A downregulates TCR-induced NF-kBsignalingby specifically suppressing the activation of the IKK complex. These results suggest that UBASH3A acts on the NF-kB signaling pathway through a mechanism involving ubiquitin- mediated protein complex formation rather than protein turnover. Our identification of two novel binding partners for UBASH3A, TAK1 and NEMO, lends further support to this idea as both proteins play key roles in the NF-kB signaling pathway. NEMO binds to M1- and K63-linked Ub chains (35,36) and to K63/M1-linked hybrid Ub chains (37). These interactions of NEMO with nondegradative Ub chains play critical roles in the NF-kB signaling pathway as unanchored K63-linked pUb chains can directly activate TAK1 and the IKK complex in vitro, possibly by inducing complex formation and trans-autophosphorylation (38), and residues of NEMO that are necessary for binding M1-linked linear ubiquitin chains are required for NF-kB activation by TNF-a and other agonists (36). We propose that besides its limited role as a phosphatase, UBASH3A specifically inhibits the activation of the IKK complex by competing with NEMO for the binding of K63-linked and/or M1-linked Ub chains and/or by altering the conformation of the IKK complex. The 21q22.3 locus is associated with T1D (P =1.23 2 10 15), although with a modest odds ratio of 1.16 (9). There are multiple SNPs, including rs11203203 and rs80054410, proximal to, or within, the transcription unit of UBASH3A that display statistically significant associations with T1D, and these SNPs may have independent, Figure 6—rs11203203 and rs80054410 regulate the expression of UBASH3A and IL2 in human primary CD4+ T cells. A and B:Human or even opposing, effects on the disease risk via different primary CD4+ T cells were negatively selected from PBMCs from mechanisms (9). Indeed, a large number of cis-eQTL (ex- healthy subjects and then stimulated with anti-CD3 plus anti-CD28 pression quantitative trait loci) SNPs affecting UBASH3A or with culture medium for 6 h. The relative mRNA levels of UBASH3A + IL2 expression in human primary CD4 T cells have been iden- and are shown. Each data point represents one subject, and the fi data are pooled from six quantitative PCR experiments. In A,thebest- ti ed (39). Controlling for the effects of these multiple fit line generated by linear regression is represented by a solid line, and variants considerably reduces the number of subjects avail- the 95% CI bands of the best-fit line are indicated by dashed curves. able for any one genotype. Thus, in our study, we focused Genotypes at rs11203203 and rs80054410 are shown. In B, the mean and SEM values are indicated by solid lines and error bars, respec- only on genotypes at rs11203203 and rs80054410, without tively. Unpaired one-tailed Student t tests were performed to assess controlling for the genotypes at other T1D-associated SNPs statistical significance. in UBASH3A. We controlled generally for genetic back- ground by using only healthy subjects of European ancestry; it is possible that in T1D subjects, risk alleles at other loci, and possibly the disease itself, might amplify the effects of T1D in a linkage study in T1D multiplex families (2). Fine- rs11203203 and rs80054410 on NF-kB signaling. Never- mapping further localized the association, pointing toward theless, we were able to identify a modest, but statistically UBASH3A as a candidate for the causal gene in the region significant, effect of genotype at rs11203203 on UBASH3A (9). Most studies of UBASH3A have focused on its potential transcript levels and a corresponding change in IL2 tran- regulatory role as a phosphatase; however, UBASH3A has script levels in activated human primary CD4+ Tcells.These 2042 Novel Role for UBASH3A in T1D Susceptibility Diabetes Volume 66, July 2017 observations in primary cells were completely consistent 8. Frederiksen BN, Steck AK, Kroehl M, et al. Evidence of stage- and age- with the results we obtained when we artificially manipu- related heterogeneity of non-HLA SNPs and risk of islet autoimmunity and type 1 lated UBASH3A levels in Jurkat cells by knockout or over- diabetes: the diabetes autoimmunity study in the young. Clin Dev Immunol 2013; expression. 2013:417657 9.Onengut-GumuscuS,ChenW-M,BurrenO,etal.;Type1DiabetesGenetics In summary, our study reveals a previously unrecognized Consortium. Fine mapping of type 1 diabetes susceptibility loci and evidence for role for UBASH3A in human T cells: UBASH3A inhibits IL2 colocalization of causal variants with lymphoid gene enhancers. Nat Genet 2015;47: the TCR-induced NF-kB signaling pathway and hence 381–386 fi expression, speci cally by suppressing the activation of the 10. Jin Y, Birlea SA, Fain PR, et al. Variant of TYR and autoimmunity susceptibility IKK complex through interactions with TAK1 and NEMO loci in generalized vitiligo. N Engl J Med 2010;362:1686–1697 and specific nondegradative ubiquitin chains. These func- 11. Zhernakova A, Stahl EA, Trynka G, et al. Meta-analysis of genome-wide as- tions of UBASH3A are enhanced by the risk alleles of two sociation studies in celiac disease and rheumatoid arthritis identifies fourteen non- previously identified T1D-associated SNPs in UBASH3A, HLA shared loci. PLoS Genet 2011;7:e1002004 indicating that a reduced capacity of CD4+ T cells to express 12. Diaz-Gallo L-M, Sánchez E, Ortego-Centeno N, et al. Evidence of new risk IL2, and possibly other critical genes regulated by NF-kB, in genetic factor to systemic lupus erythematosus: the UBASH3A gene. PLoS One 2013; response to TCR stimulation contributes to T1D risk in a 8:e60646 UBASH3A-dependent manner. 13. Carpino N, Turner S, Mekala D, et al. Regulation of ZAP-70 activation and TCR signaling by two related proteins, Sts-1 and Sts-2. Immunity 2004;20:37–46 14. Wattenhofer M, Shibuya K, Kudoh J, et al. Isolation and characterization of the Acknowledgments. This research uses resources from the T1DGC, a UBASH3A gene on 21q22.3 encoding a potential nuclear protein with a novel – collaborative clinical study sponsored by the National Institute of Diabetes and combination of domains. Hum Genet 2001;108:140 147 Digestive and Kidney Diseases, National Institute of Allergy and Infectious Diseases, 15. Feshchenko EA, Smirnova EV, Swaminathan G, et al. TULA: an SH3- and UBA- – National Human Genome Research Institute, Eunice Kennedy Shriver National Institute containing protein that binds to c-Cbl and ubiquitin. Oncogene 2004;23:4690 of Child Health and Human Development, and JDRF. The authors thank Matthew 4706 Mika (University of Virginia) for reagents, Jocyndra A. Wright (University of Florida) for 16. Kowanetz K, Crosetto N, Haglund K, Schmidt MHH, Heldin C-H, Dikic I. Sup- technical assistance, and David A. Ostrov (University of Florida) for help with puri- pressors of T-cell receptor signaling Sts-1 and Sts-2 bind to Cbl and inhibit endo- – fication of GST-fusion proteins. cytosis of receptor tyrosine kinases. J Biol Chem 2004;279:32786 32795 Funding. This study was supported by National Institute of Diabetes and Digestive 17. Hoeller D, Crosetto N, Blagoev B, et al. Regulation of ubiquitin-binding proteins – and Kidney Diseases grants DK106718, DK046635, and DK085678 (to P.C.) and by by monoubiquitination. Nat Cell Biol 2006;8:163 169 JDRF Advanced Postdoctoral Fellowship award 3-APF-2016-177-A-N (to Y.G.). 18. Bertelsen V, Breen K, Sandvig K, Stang E, Madshus IH. The Cbl-interacting – Duality of Interest. No potential conflicts of interest relevant to this article protein TULA inhibits dynamin-dependent endocytosis. Exp Cell Res 2007;313:1696 were reported. 1709 Author Contributions. Y.G. and P.C. conceived the study, interpreted the 19. San Luis B, Sondgeroth B, Nassar N, Carpino N. Sts-2 is a phosphatase that data, and wrote the manuscript. Y.G. and T.K.P. performed the experiments. Y.G. negatively regulates zeta-associated protein (ZAP)-70 and T cell receptor signaling analyzed the data. J.R.B.N. and L.M.M. aligned and analyzed the mRNA-seq data. pathways. J Biol Chem 2011;286:15943–15954 P.C. is the guarantor of this work and, as such, had full access to all the data in the 20. Mikhailik A, Ford B, Keller J, Chen Y, Nassar N, Carpino N. A phosphatase study and takes responsibility for the integrity of the data and the accuracy of the activity of Sts-1 contributes to the suppression of TCR signaling. Mol Cell 2007;27: data analysis. 486–497 Prior Presentation. Parts of this study were presented in abstract form at the 21. Agrawal R, Carpino N, Tsygankov A. TULA proteins regulate activity of the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 protein tyrosine kinase Syk. J Cell Biochem 2008;104:953–964 June 2017. 22. Chen X, Ren L, Kim S, et al. Determination of the substrate specificity of protein-tyrosine phosphatase TULA-2 and identification of Syk as a TULA-2 sub- References strate. J Biol Chem 2010;285:31268–31276 1. Concannon P, Rich SS, Nepom GT. Genetics of type 1A diabetes. N Engl J Med 23. Thomas DH, Getz TM, Newman TN, et al. A novel histidine tyrosine phos- 2009;360:1646–1654 phatase, TULA-2, associates with Syk and negatively regulates GPVI signaling in 2. Concannon P, Onengut-Gumuscu S, Todd JA, et al.; Type 1 Diabetes Genetics platelets. Blood 2010;116:2570–2578 Consortium. A human type 1 diabetes susceptibility locus maps to chromosome 24. Chen Y, Jakoncic J, Carpino N, Nassar N. Structural and functional charac- 21q22.3. Diabetes 2008;57:2858–2861 terization of the 2H-phosphatase domain of Sts-2 reveals an acid-dependent 3. Smyth DJ, Plagnol V, Walker NM, et al. Shared and distinct genetic variants in phosphatase activity. Biochemistry 2009;48:1681–1690 type 1 diabetes and celiac disease. N Engl J Med 2008;359:2767–2777 25. Newman TN, Liverani E, Ivanova E, et al. Members of the novel UBASH3/STS/ 4. Barrett JC, Clayton DG, Concannon P, et al.; Type 1 Diabetes Genetics Con- TULA family of cellular regulators suppress T-cell-driven inflammatory responses sortium. Genome-wide association study and meta-analysis find that over 40 loci in vivo. Immunol Cell Biol 2014;92:837–850 affect risk of type 1 diabetes. Nat Genet 2009;41:703–707 26. RanFA,HsuPD,WrightJ,AgarwalaV,ScottDA,ZhangF.Genomeengineering 5. Grant SFA, Qu H-Q, Bradfield JP, et al.; DCCT/EDIC Research Group. Follow-up using the CRISPR-Cas9 system. Nat Protoc 2013;8:2281–2308 analysis of genome-wide association data identifies novel loci for type 1 diabetes. 27. Ge Y, Onengut-Gumuscu S, Quinlan AR, et al. Targeted deep sequencing in Diabetes 2009;58:290–295 multiple-affected sibships of European ancestry identifies rare deleterious variants in 6. Plagnol V, Howson JMM, Smyth DJ, et al.; Type 1 Diabetes Genetics PTPN22 that confer risk for type 1 diabetes. Diabetes 2016;65:794–802 Consortium. Genome-wide association analysis of autoantibody positivity in type 28. Lim KL, Chew KCM, Tan JMM, et al. Parkin mediates nonclassical, 1 diabetes cases. PLoS Genet 2011;7:e1002216 proteasomal-independent ubiquitination of synphilin-1: implications for Lewy body 7. Johnson K, Wong R, Barriga KJ, et al. rs11203203 is associated with type 1 formation. J Neurosci 2005;25:2002–2009 diabetes risk in population pre-screened for high-risk HLA-DR,DQ genotypes. Pediatr 29.HaydenMS,GhoshS.NF-kB, the first quarter-century: remarkable progress Diabetes 2012;13:611–615 and outstanding questions. Genes Dev 2012;26:203–234 diabetes.diabetesjournals.org Ge and Associates 2043

30. Pahl HL. Activators and target genes of Rel/NF-kBtranscriptionfactors.On- 35. Lo Y-C, Lin S-C, Rospigliosi CC, et al. Structural basis for recognition of diu- cogene 1999;18:6853–6866 biquitins by NEMO. Mol Cell 2009;33:602–615 31. Sun Z, Arendt CW, Ellmeier W, et al. PKC-u is required for TCR-induced 36. Rahighi S, Ikeda F, Kawasaki M, et al. Speciﬁc recognition of linear ubiquitin NF-kB activation in mature but not immature T lymphocytes. Nature 2000;404: chains by NEMO is important for NF-kappaB activation. Cell 2009;136:1098–1109 402–407 37. Emmerich CH, Ordureau A, Strickson S, et al. Activation of the canonical IKK 32. Chen ZJ. Ubiquitination in signaling to and activation of IKK. Immunol Rev 2012; complex by K63/M1-linked hybrid ubiquitin chains. Proc Natl Acad Sci U S A 2013; 246:95–106 110:15247–15252 33. Wertz IE, Dixit VM. Signaling to NF-kB: regulation by ubiquitination. Cold Spring 38. Xia Z-P, Sun L, Chen X, et al. Direct activation of protein kinases by unanchored Harb Perspect Biol 2010;2:a003350 polyubiquitin chains. Nature 2009;461:114–119 34. Hnisz D, Abraham BJ, Lee TI, et al. Super-enhancers in the control of cell 39. Raj T, Rothamel K, Mostafavi S, et al. Polarization of the effects of autoimmune identity and disease. Cell 2013;155:934–947 and neurodegenerative risk alleles in leukocytes. Science 2014;344:519–523