The D0 Ig-like Domain Plays a Central Role in the Stronger Binding of KIR3DL2 to B27 Free H Chain Dimers

This information is current as Hiroko Hatano, Jacqueline Shaw, Kaitlin Marquardt, of September 23, 2021. Zhiyong Zhang, Laurent Gauthier, Stephanie Chanteux, Benjamin Rossi, Demin Li, Julie Mitchell and Simon Kollnberger J Immunol 2015; 194:1591-1601; Prepublished online 12

January 2015; Downloaded from doi: 10.4049/jimmunol.1402214 http://www.jimmunol.org/content/194/4/1591

Supplementary http://www.jimmunol.org/content/suppl/2015/01/09/jimmunol.140221 http://www.jimmunol.org/ Material 4.DCSupplemental References This article cites 38 articles, 15 of which you can access for free at: http://www.jimmunol.org/content/194/4/1591.full#ref-list-1

Why The JI? Submit online. by guest on September 23, 2021 • Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2015 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

The D0 Ig-like Domain Plays a Central Role in the Stronger Binding of KIR3DL2 to B27 Free H Chain Dimers

Hiroko Hatano,* Jacqueline Shaw,* Kaitlin Marquardt,† Zhiyong Zhang,* Laurent Gauthier,‡ Stephanie Chanteux,‡ Benjamin Rossi,‡ Demin Li,x Julie Mitchell,† and Simon Kollnberger*

We proposed that the killer cell Ig-like KIR3DL2 binding more strongly to HLA-B27 (B27) b2-microglobulin free H chain (FHC) dimers than other HLA–class I molecules regulates lymphocyte function in arthritis and infection. We compared the function of B27 FHC dimers with other class I H chains and identified contact residues in KIR3DL2. B27 FHC dimers interacted functionally with KIR3DL2 on NK and reporter cells more strongly than did other class I FHCs. Mutagenesis identified key Downloaded from residues in the D0 and other Ig-like domains that were shared and distinct from KIR3DL1 for KIR3DL2 binding to B27 and other class I FHCs. We modeled B27 dimer binding to KIR3DL2 and compared experimental mutagenesis data with computational “hot spot” predictions. Modeling predicts that the stronger binding of B27 dimers to KIR3DL2 is mediated by nonsymmetrical complementary contacts of the D0 and D1 domains with the a1, a2, and a3 domains of both B27 H chains. In contrast, the D2 domain primarily contacts residues in the a2 domain of one B27 H chain. These findings provide novel insights about the molecular basis of KIR3DL2 binding to B27 and other ligands and suggest an important role for KIR3DL2–B27 interactions in http://www.jimmunol.org/ controlling the function of NK cells in B27+ individuals. The Journal of Immunology, 2015, 194: 1591–1601.

he HLA class I molecule HLA-B27 (B27) is associated roles in determining the strength and quality of immune responses with the development of a group of inflammatory arthritic in arthritis and infection (3–5). T disorders, collectively known as the spondyloarthritides The KIR family member KIR3DL2 is expressed on NK and minor (1). B27 is also positively associated with more favorable out- T cell subsets (6). KIR–HLA interactions have been implicated in comes for patients with HIVor hepatitis C viral infections (2). B27 immune responses against pathogens and in autoimmunity (7). interactions, including interactions with mem- Originally, KIR3DL2 was identified as a receptor for HLA-A3 by guest on September 23, 2021 bers of the killer cell Ig-like receptor (KIR) family, play important and HLA-A11 (8–10). Subsequent studies suggested that HLA-A3 and HLA-A11 are weak ligands for KIR3DL2 or that their in- teraction with KIR3DL2 is highly specific. HLA-A3 licenses *Botnar Research Centre, Nuffield Department of Rheumatological and Musculo- skeletal Sciences, University of Oxford, Oxford OX3 7LD, United Kingdom; KIR3DL2-expressing NK cells with poor effector function, and †Department of Biochemistry, University of Wisconsin, Madison, WI 53706; HLA-A3 binding to KIR3DL2 is promoted by a limited number of x ‡Innate Pharma, 13276 Marseille, France; and Nuffield Division of Clinical Labo- viral peptide epitopes (11, 12). However, the fact that KIR3DL2 is ratory Sciences, University of Oxford, Oxford OX3 7LD, United Kingdom a framework encoding $63 allelic variants suggests that ORCID: 0000-0002-7118-3401 (D.L.). there are other ligands (13). KIR3DL2 also binds to b2-micro- Received for publication September 2, 2014. Accepted for publication December 10, globulin (b2m) free H chain (FHC) forms of B27, including B27 2014. dimers (termed B272) and other HLA class I FHCs (14, 15). This work was supported by Arthritis Research UK (to S.K., J.S., and H.H.), by the National Institutes of Health Research (to S.K.), by the Japan Society for the Pro- KIR3DL2 and other three-domain KIRs contain three Ig-like motion of Science (to H.H.), Natural Science Foundation of Shandong Province domains (D0, D1, and D2), which together form the ligand- Grants ZR2010HQ034 and 81101605 and Shandong Research Award Fund for Out- binding domain (13). It is unclear exactly how these domains standing Young Scientist Grant BS2009YY008 (to Z.Z.), by Cancer Research UK (to D.L.), by the National Science Foundation (Grants NSF-CDI CMMI-0941013 and determine KIR3DL2 binding to ligand. Additionally, KIR3DL2 NSF-FRG DMS-1160360 [to J.M.]), and Computation and Informatics in Biology forms a disulfide-bonded dimer, presumably via two unpaired and Medicine Training Grant NIH NIGMS 5T15LM007359 (to K.M.) cysteines in the stem region (8). The contribution of KIR3DL2 H.H., J.S., D.L., Z.Z., J.M., and S.K. designed the study. H.H., J.S., Z.Z., L.G., S.C., dimerization to ligand binding has not been studied. The D0 do- B.R., and S.K. acquired experimental data. H.H., J.S., Z.Z., D.L., and S.K. analyzed data. K.M. and J.M. performed structural modeling and analysis. H.H., J.M., and S.K. main of KIR3DL1 enhances ligand interactions by binding shared wrote the manuscript. features of HLA class I (16, 17). This manifests in a weak affinity Address correspondence and reprint requests to Dr. Simon Kollnberger, Botnar Re- of KIR3DL1 for different class I HLAs in functional studies (18). search Centre, Nuffield Department of Rheumatological and Musculoskeletal Sci- This suggests that other three-domain KIRs, including KIR3DL2, ences, University of Oxford, Windmill Road, Oxford OX3 7LD, U.K. E-mail address: [email protected] could bind to shared features of HLA class I. The online version of this article contains supplemental material. KIR3DL2 binds more strongly to B27 b2m FHC forms, in- Abbreviations used in this article: 221, LCL.721.221; B27, HLA-B27; CA, calpha cluding B27 FHC dimers, than to other class I HLAs (19). The atomic coordinate; FHC, free H chain; HA, hemagglutinin; KIR, killer cell Ig-like stronger interactions of B27 FHC forms with KIR3DL2 promote receptor; b2m, b2-microglobulin; PDB, Data Bank; R10, RPMI 1640, 10% the survival of NK and CD4 T cells and could account for the FCS; RMSD, root-mean-square deviation; WT, wild-type. increased proportions of these cells in spondyloarthritis (19–21). Copyright Ó 2015 by The American Association of Immunologists, Inc. 0022-1767/15/$25.00 The stronger binding of B27 FHC dimer forms to KIR3DL2 also www.jimmunol.org/cgi/doi/10.4049/jimmunol.1402214 1592 ROLE OF THE D0 DOMAIN OF KIR3DL2 IN BINDING TO B27 DIMERS could account for the increased proportions of KIR3DL2+ CD4 KIR3DL2 and B27 FHC dimer modeling and molecular T cells in healthy B27+ individuals (20). The stronger binding of dynamics KIR3DL2 to B27 FHC dimers is dependent on cysteine 67–de- To create a model for the FHC HLA-B*2705 (FHC B27) dimer, we first pendent dimerization (19). removed b2m from the crystal structure from Protein Data Bank (PDB) KIR3DL2 binding to B27 FHC dimers is inhibited by the HLA 1OGT (30). This structure was relaxed by simulating the monomer for 10 class I H chain Ab HC10 and by other B27 H chain Abs (22, 23). ns using NAMD (31). This simulation used TIP3 explicit water, Langevin We reasoned that the strong binding of KIR3DL2 to B27 FHC integration with a fixed 310 K˚ temperature, and the CHARMM27 force field (32). Two copies of this relaxed monomer were used to create a dimer dimers reflects an innate ability of KIR3DL2 to bind weakly to by applying a 180˚ rotation about the long molecular axis, followed by other HLA class I FHCs. Thus, we compared the strength of a translation that created a disulfide bond between the two C67 residues. functional interactions of KIR3DL2 with B27 FHC dimers and This FHC dimer was then relaxed for 10 ns using the same protocol ap- other HLA class I H chains. We modeled B27 FHC dimer binding plied to the monomer. To create a model for the interaction of the B27 dimer with KIR3DL2, the to KIR3DL2 and set out to identify contact residues in KIR3DL2 program MODELER (33) was used to create a structure for KIR3DL2 via and B27 that are involved in this interaction by targeted muta- comparative modeling with nine related KIR receptors as folding tem- genesis and epitope mapping of blocking Abs. plates. We created interface models for the interaction between the B27 dimer and KIR3DL2 via two strategies. First, we reproduced the binding mode of KIR3DL1 as in PDB 3VH8 by structural alignment, finding that it Materials and Methods fit remarkably well into the assembled B27 dimer. To support the validity Abs and cell lines used in this study of this interface, we also applied unconstrained docking using HADDOCK (34). We did not enforce any constraints upon the docking, allowing Anti-KIR3DL2 Ab DX31 (IgG2a isotype) was a kind gift from Dr. Jo Phillips (DNAX, Palo Alto, CA). D0-specific (D0A-D0C, all IgG1 isotypes) HADDOCK to randomize its Ambiguous Interaction Restraints and refine Downloaded from and D2A-specific (IgG1) and D1A-specific (IgG1) anti-KIR3DL2 Abs were the best models found. This procedure produced 10 clusters of predictions, produced by Innate Pharma (Marseille, France). HLA-A, B, C class I of which the minimum energy cluster (cluster 3) contained representatives negative LCL.721.221 (221) cell lines were transfected with pRSVNeo with extremely low root-mean-square deviation (RMSD) relative to the Calpha atomic coordinates (CA) generated from a structural alignment. In constructs of HLA-B*3501, HLA-B*0702, and HLA-B*27:05 (24). The particular, the best such structure had a total CA-RMSD of 1.1 A˚ , and all 221 cells transfected with HLA-G1 in pcDNA3.1 were a gift from Kalle , ˚ Soderstrom (Kennedy Institute of Rheumatology, University of Oxford), structures in the cluster had interface-CA-RMSD 2.5 A (Supplemental Fig. 1). and 221 cells transfected with HLA-*A0301 were a gift from Veronique http://www.jimmunol.org/ Braud (Institut de Pharmacologie Moleculaire et Cellulaire, Valbonne, Using the structure for the B27 dimer complex with KIR3DL2 generated France). Functional-grade DX17 (IgG1), IgG1, and IgG2a isotype-control via alignment, we simulated the complex for 40 ns, according to the protocol used for other relaxations. The longer time was necessary to allow the D0 mAbs were from BioLegend. domain of KIR3DL2 to undergo the conformational changes needed to form Tetramer preparation, eGFP plasmid construct generation, contacts with the second B27 monomer. After 30 ns of simulation, the , ˚ and FACS staining structure of KIR3DL2 exhibited very little flexibility ( 1.0 A CA-RMSD between 25 and 30 ns). The complex was simulated an additional 10 ns for B27 dimer and HLA-A3 tetrameric complexes were prepared for staining of good measure; most interface residues were present in all 180 structures KIR-transfected 293 T cells, as previously described (25). eGFP plasmid sampled from the final 1 ns of the trajectory (Supplemental Table I). constructs for KIR3DL1/2 and DS1 and KIR2DL4, 2DS4, and 2DL5 and Predicted hot spots also were persistent within these structures, but this the D0 domain of KIR3DL2 were prepared as described (26). KIR3DL2 was less pronounced as a result of the lack of crystallographic precision by guest on September 23, 2021 mutants for Ab epitope mapping were generated from the eGFP fusion within the equilibrated ensemble. construct of KIR3DL2 using the Stratagene Quick Change . Alanine mutagenesis hot spot prediction KIR3DL2CD3« reporter and NK functional assays The KFC2 method (35) was used to predict alanine mutagenesis hot spots Jurkat reporter cells were generated by lentiviral transduction with (residues in which alanine mutation results in a change in binding free KIR3DL2-CD3ε lentiviruses, as described (19). Point or double mutations energy [DDG] of . 2 kcal/mol). This biophysical knowledge-based were made using the Stratagene Quick Change kit. method predicts hot spots in solved or modeled protein–protein interfaces. KIR3DL2+ and KIR3DL1+ NK cell lines were generated and main- tained as described (21). Supernatants for IFN-g or IL-2 ELISA (eBio- Results science) were harvested from KIR3DL2-expressing NK or KIR3DL2 B27 FHC dimers bind more strongly to KIR3DL2 than to reporter cells after 24 h of stimulation with 221 cells in RPMI 1640, 10% FCS (R10) medium, as described (19). KIR3DL1 In some experiments, IFN-g production by NK cells was determined by We reported previously that B27 FHC dimers bind to KIR3DL1 and intracellular cytokine staining with phycoerythrin-conjugated anti–IFN-g (BioLegend), using methods adapted from Bowness et al. (20). KIR3DL2, which have D0, D1, and D2 Ig-like domains (25). In For degranulation experiments, 200,000 NK cells were cultured at a 2:1 contrast, two-domain KIRs with D1 and D2 domains do not bind ratio with 221 cells in R10 medium. Viable degranulating CD107a- to B27 FHC dimers (19). We reasoned that unique properties of expressing NK cells were enumerated by flow cytometry after a 6-h the D0 domain of these KIRs might enable binding. Thus, we first stimulation, as described (27). Dead cells were excluded by staining with sought to determine to what extent B27 FHC dimers bind to other allophycocyanin-Cy7 LIVE/DEAD Stain (DEAD stain; Invitrogen). For NK lysis experiments, 40,000 NK cells were cultured at a 2:1 ratio with two- or three-domain KIRs with a D0 domain. B27 FHC dimers CFSE-labeled 221 cells for 6 h in R10 medium before staining with stained KIR3DL2, KIR3DL1, and 3DS1 but not KIR2DL4 and allophycocyanin-Cy7 LIVE/DEAD stain (DEAD stain) and anti-CD56 2DL5 293T transfectants (Fig. 1A). Binding of B27 FHC dimer V450 (BioLegend), using a protocol adapted from Fischer et al. (28). tetramers to three-domain KIRs was inhibited by the HLA class I Viable CFSE+CD562 target cells were enumerated by flow cytometry. In some experiments 221 cells were treated briefly with acid to generate H chain mAb HC10 and by anti-KIR3DL1 (DX9), anti-KIR3DL2 increased levels of HLA class I H chains, as described (29). (DX31), and anti-KIR3DS1 (Z27) Abs (Fig. 1A) (25). We next compared functional interactions of KIR3DL1- and Immunoprecipitations KIR3DL2-expressing NK cell lines with B27 FHCs on B27- KIR3DL2CD3ε2tagged or hemagglutinin (HA)-tagged KIR3DL2- transfected LCL.721.221 (221B27) cells. Previously, we dem- transduced Jurkat cells were surface stained with D1A or IgG1 mAb. onstrated the expression of B27 FHC dimers by 221B27 cells. We Immunoprecipitation of KIR3DL2 was performed as outlined (29). studied the effect of 221B27 target cells on NK cell function by flow Immunoprecipitates resolved by nonreducing or reducing SDS-PAGE and Western blots were probed with HRP-conjugated anti-HA (clone 16B12; cytometry, measuring degranulation by surface staining for CD107a Life Technologies UK) or anti-CD3ε (rabbit polyclonal; Novus Biologicals expression and IFN-g production by intracellular cytokine staining. Europe) using protocols adapted from Bird et al. (24). Compared with parental 221 cells, 221B27 cells consistently inhibited The Journal of Immunology 1593

FIGURE 1. B27 FHC dimers (B272) bind strongly to KIR3DL2 and more weakly to KIR3DL1 and 3DS1. (A) FACS staining with B27 dimer tetramers of KIR3DL1-eGFP, KIR3DL2-eGFP, and KIR3DS1-eGFP

(D0, D1, D2) or KIR2DL4-eGFP and KIR2DL5A-eGFP Downloaded from (D0, D2 domain organization) transfected 293T cells in the presence of HC10 or anti-KIR Abs. Anti-KIR Abs used for blocking were anti-KIR3DL1 (DX9), anti- KIR3DL2 (DX31), and anti-KIR3DS1 (Z27). (B)CD107a expression by KIR3DL2 NK cells (upper panels)and KIR3DL1 NK cells (lower panels) stimulated with 221 cells or 221B27 cells with HLA class I H chain (HC10) or http://www.jimmunol.org/ DX17 Abs or isotype-control Abs. Representative staining from one of three independent experiments with NK lines from three individuals. (C)IFN-g intracellular cytokine staining of KIR3DL2 NK cells (upper panels)and KIR3DL1 NK cells (lower panels) stimulated with 221 cells or 221B27 cells with HLA class I H chain (HC10) or DX17 Abs or isotype-control Abs. Representative staining from one of three independent experiments with

NK lines from three individuals. by guest on September 23, 2021

the expression of CD107a and IFN-g production by KIR3DL2- or of HLA class I 221 transfectants (29). Thus, we compared KIR3DL1-expressing NK cells (Fig. 1B, 1C). Blocking KIR3DL2 KIR3DL2 interactions with B27 class I H chains and other class I binding to B27 FHCs with the H chain Ab HC10 consistently reduced H chains on transfected 221 cells using KIR3DL2CD3ε-trans- the inhibition of CD107a and IFN-g production by KIR3DL2+ NK duced Jurkat reporter cells. cells, but it had no significant effect on the function of KIR3DL1+ NK B27-transfected 221B27 cells consistently stimulated 6-fold cells (Fig. 1B, 1C). In contrast, the anti-HLA class I Ab DX17 re- higher levels of IL-2 secretion compared with 221B7 cells and versed functional inhibition of both KIR3DL2 and KIR3DL1 NK other HLA class I–transfected 221 cells (Fig. 2A). This was de- cells by 221B27 target cells (Fig. 1B, 1C). spite comparable levels of HLA class I H chain expression by 221 cells transfected with HLA-B7 and B27, measured by FACS KIR3DL2 binds strongly to B27 FHC dimers and more weakly staining with HC10 (Supplemental Fig. 2A). We consistently to other HLA class I H chains observed greater IL-2 production by KIR3DL2 reporter cells The fact that KIR3DL2 binds to B27 FHC dimers suggested that stimulated with EBV-immortalized B27+ B cell lines compared this receptor might also interact with other class I H chains. We with stimulation with B272 B cell lines (Fig. 2A). HC10 inhibited showed previously that the HLA class I H chain Ab HC10 rec- KIR3DL2 reporter cell interactions with B27 and other class I ognizes B27 FHC dimers and other class I H chains on the surface HLAs on both transfected 221 cells and EBV-transformed B cell 1594 ROLE OF THE D0 DOMAIN OF KIR3DL2 IN BINDING TO B27 DIMERS Downloaded from http://www.jimmunol.org/ by guest on September 23, 2021 FIGURE 2. KIR3DL2 binds to B27 FHC dimers more strongly than to other HLA class I H chains. (A) IL-2 secretion by KIR3DL2CD3ε Jurkat reporter cells stimulated with 221B27, 221, 221B7, 221A3, or 221G transfectants in the presence of isotype control IgG2a or HC10 mAbs (left panel). IL-2 secretion by KIR3DL2CD3ε Jurkat reporter cells stimulated with B27+ and B272 EBV-transformed B cell lines in the presence of the indicated Abs (right panel). (B) IL-2 secretion by KIR3DL2CD3ε Jurkat reporter cells stimulated with class I–transfected 221 cells without (hatched bars) or with (open bars) acid treatment (left panel). IL-2 secretion by KIR3DL2CD3ε Jurkat reporter cells stimulated with B27+ and B272 EBV-transformed B cell lines without (hatched bars) or with (open bars) acid treatment (right panel). Data in (A) and (B) are mean 6 SEM and are representative data from one of three in- dependent experiments. (C) IFN-g secretion by KIR3DL2+ NK cells stimulated with parental 221 cells or 221-B27, 221-B7, 221-B35, 221-A3, or 221-G cells. Data are mean 6 SEM from three independent experiments with two NK cell lines. (D) Effect of anti-KIR3DL2 mAb DX31 and H chain Abs (HC10) or isotype (IgG2a) on IFN-g secretion by KIR3DL2+ NK cells stimulated with 221B7 or 221B27 cells. Data are mean 6 SEM from three independent experiments with two NK cell lines. (E) Representative FACS staining of CD107a expression by KIR3DL2+ NK cells stimulated with 221B27 cells or 221B7 cells with anti-KIR3DL2 (DX31), H chain (HC10), or isotype (IgG2a) mAbs. (F) Mean CD107a expression by KIR3DL2+ NK cells stimulated with 221 or B27, HLA-A3, HLA-B7, HLA-B35, or HLA-G expressing 221 (top panel). Mean CD107a expression by KIR3DL2+ NK cells stimulated with 221B27 or 221B7 cells with anti-KIR3DL2 (DX31), H chain (HC10), or isotype control (IgG2a) mAbs (middle and bottom panels). Data are mean 6 SEM from four independent experiments with two NK cell lines. Hatched bars indicate CD107a expression by NK cells stimulated with untreated cells, and open bars indicate CD107a expression by NK cells stimulated with acid-treated 221 cells. (G) Proportions of viable CFSE-labeled DEAD stain low (lower gate) 221 transfectants after a 6-h incubation with/without KIR3DL2+ NK cells (top row). Proportions of viable CFSE-labeled 221B27 cells after coculture with KIR3DL2+ NK cells with HLA class I H chain (HC10), KIR3DL2 (DX31), or isotype (IgG2a) mAb (rows 2–5). (H) Mean proportions of viable CFSE+ 221B27 cells, 221 cells, or 221 cells transfected with other class I HLAs after incubation with KIR3DL2+ NK cell lines (top row). Effect of the anti- KIR3DL2 (DX31) and H chain Abs (HC10) on the proportions of viable CFSE-labeled 221B27, 221B7, or 221G cells incubated with KIR3DL2+ NK cells (rows 2–4). Data are mean 6 SEM from three experiments with two NK cell lines. *p , 0.05, **p , 0.01, ***p , 0.005, ANOVA (paired measures test). lines (Fig. 2A). These results suggested that KIR3DL2 also could generated increased levels of HLA class I FHCs on transfected bind to HLA class I FHCs other than B27. The low-level inter- 221 cells and EBV-immortalized B cell lines by acid treatment action of reporter cells with untransfected 221 cells could result and studied KIR3DL2 reporter cell interactions. from KIR3DL2 binding to HLA-F and HLA-E H chains, which Acid treatment of HLA class I–transfected 221 cells increased are expressed by 721.221 cells (36). HLA-F H chains were re- staining with the H chain–specific Ab HC10 while decreasing ported to bind to KIR3DL2 (15). staining with W632 Ab, which recognizes b2m-associated class I We showed previously biochemically that acid treatment (Supplemental Fig. 2). Acid treatment of 221B27 cells only increases HC10-reactive cell surface FHCs (29). Because class I modestly increased IL-2 production by KIR3DL2 reporter cells. In FHCs stimulated lower levels of IL-2 compared with B27 FHCs, contrast, acid treatment of 221 cells transfected with other class I we increased FHC levels on 221 and EBV-transformed B cell lines HLAs increased KIR3DL2 reporter cell IL-2 (Fig. 2B). Acid to determine whether KIR3DL2 binding could be enhanced. We treatment of HLA class I 221 transfectants increased FACS The Journal of Immunology 1595 staining with the HC10 H chain Ab to levels observed on B27 KIR3DL2 with B27 FHC dimers. Thus, we mutated the two transfectants (Supplemental Fig. 2B). However, acid treatment did cysteines in the KIR3DL2 stem to alanine and studied the inter- not increase IL-2 production by KIR3DL2 reporter cells stimu- action of KIR3DL2 mutant reporter cells with B27 FHCs on lated with other HLA class I transfectants to the levels observed 221B27 cells. Reporter cells transduced with the cysteine mutant for 221B27 cells (Fig. 2B). of KIR3DL2 produced equivalent amounts of IL-2 as did wild- KIR3DL2 reporter cell interactions were enhanced further by acid type (WT) reporter cells when stimulated with 221B27 cells treatment of transformed B cell lines. KIR3DL2 reporter cell (Fig. 3A). The cysteine mutant of KIR3DL2 was expressed at interactions were enhanced to a greater extent with acid-treated B27+ equivalent levels as the WT receptor (Fig. 3B). The majority of B cell lines compared with acid-treated B272 B cell lines (Fig. 2B), KIR3DL2 receptor immunoprecipitated from the surface of WT despite comparable increases in the expression of H chains with acid KIR3DL2CD3ε cells and Jurkat T cells expressing full-length treatment in B27+ and B272 B cell lines (Supplemental Fig. 2B). KIR3DL2 was monomeric (Fig. 3C). These results are consis- B27 FHC dimers bind more strongly to KIR3DL2 on primary NK cells than to other HLA class I FHCs We next tested the functional interactions of different HLA class I H chains with primary NK cell lines expressing KIR3DL2. KIR3DL2+ NK cell lines produced less IFN-g in response to stimulation with 221B27 cells compared with stimulation with

221 cells expressing other class I HLAs (Fig. 2C). Functional Downloaded from inhibition of KIR3DL2–B27 FHC interactions with anti-KIR3DL2 (DX31) or anti-HLA class I H chain (HC10) Abs increased IFN-g secretion by KIR3DL2+ NK cells stimulated with 221B27, but not 221B7, cells (Fig. 2D). We then compared the effect of KIR3DL2 binding to B27 and

other class I H chains on NK cell degranulation by measuring NK http://www.jimmunol.org/ expression of CD107a. The 221B27 cells consistently inhibited CD107a expression by KIR3DL2+ NK cells more strongly than did 221 cells expressing other class I HLAs (Fig. 2E, 2F). Following stimulation with 221B27 cells, a mean of 20% of KIR3DL2+ NK cells expressed CD107a compared with 39, 30, 34, 38, and 38% of CD107a-expressing NK cells following stimulation with 221, 221- A3, 221-B7, 221-B35, and 221-G transfectants, respectively (Fig. 2E). Acid treatment of 221 transfectants to increase class I H chain levels had no effect on CD107a expression by KIR3DL2+ by guest on September 23, 2021 NK cells. Acid treatment of 221B27 cells did not affect 221B27 inhibition of CD107a expression by KIR3DL2+ NK cells. KIR3DL2+ NK cell degranulation by untreated and acid-treated 221B27 cells waspromotedbyanti-KIR3DL2(DX31)andHLAclassIHchain (HC10) mAbs (Fig. 2E, 2F). In contrast, these Abs had no effect on KIR3DL2+ NK cell degranulation stimulated with untreated and acid-treated HLA-B7 targets (Fig. 2E, 2F). We also investigated the effect of KIR3DL2 binding to HLA class I H chains on NK cell lysis of CFSE-labeled 221 target cells. CFSE-labeled 221B27 target cells were more resistant to lysis by KIR3DL2+ NK cells than were 221 target cells expressing other class I HLAs (Fig. 2G, 2H). After incubation with KIR3DL2+ NK cells, a mean of 22% of CFSE-labeled 221B27 cells remained resistant to lysis compared with 0.5, 8, 0.6, and 1% of 221, 221A3, 221B7, and 221G cells, respectively. The different 221 cells used showed similar rates of cell death over the time course of the assay (Supplemental Fig. 2C). KIR3DL2+ NK cell lysis of 221B27 cells was promoted by anti-KIR3DL2 (DX31) and HLA class I H chain (HC10) mAbs (Fig. 2G, 2H). In contrast, these Abs had no effect FIGURE 3. KIR3DL2 dimerization is not necessary for binding to on KIR3DL2+ NK cell lysis of HLA-B7– and HLA-G–expressing B27 FHC dimers. (A)IL-2productionbyWTKIR3DL2CD3eand targets (Fig. 2H). Although HC10 does not bind HLA-G, the lack C281AC315A (CYS-) KIR3DL2CD3e mutant reporter cells in response to of inhibition of NK cells stimulated with 221G with DX31 Ab stimulation with 221B27 cells. Representative data from one of three in- suggests that NK cell KIR3DL2 did not functionally interact with dependent experiments. (B) Representative FACS stain of WT and ε HLA-G on these cells in these assays. C281AC315A KIR3DL2CD3 reporter cells with the indicated KIR3DL2 Abs (DX31 and D1A). (C) Western blots of cell surface immunoprecipi- B27 FHC dimer binding to KIR3DL2 is independent of tates from KIR3DL2CD3ε reporter cells with the D1A Ab resolved by disulfide-mediated KIR3DL2 dimerization nonreducing (2DTT) or reducing (+DTT) SDS-PAGE probed with HRP- conjugated anti CD3ε (upper panels). Western blots of cell surface KIR3DL2 was shown to form disulfide-bonded dimers (8). We immunoprecipitates from Jurkat cells expressing HA-tagged KIR3DL2 sought to determine whether cysteine-dependent receptor dimer- with the D1A Ab resolved by nonreducing (2DTT) or reducing (+DTT) ization of KIR3DL2 might account for the stronger interaction of SDS-PAGE probed with HRP-conjugated anti-HA (lower panels). 1596 ROLE OF THE D0 DOMAIN OF KIR3DL2 IN BINDING TO B27 DIMERS

tent with KIR3DL2 dimerization not being required for binding to B27 FHC dimers. The D0 domain of KIR3DL2 plays a central role in binding to B27 FHC dimers and other HLA class I FHCs We sought to determine the possible contributions of the different Ig domains in KIR3DL2 for binding to B27 and other HLA class I FHCs. Thus, we determined the effect of domain-specific Abs on 221 cells expressing different class I HLAs (19). The epitopes of five KIR3DL2-specific Abs were mapped to the D0, D1, and D2 domains of KIR3DL2. D0A, D0C, and DX31 were shown to recognize the D0 domain by FACS staining of 293T cells transfected with a truncated KIR3DL2 construct encoding the D0 domain of KIR3DL2 (Fig. 4A, top panel). Mutagenesis of the D0 domain residues I60 and G62 and R78 and L82 in KIR3DL2 to the corresponding amino acids in KIR3DL1 (I60NG62S and R78HL82P) inhibited recognition by D0C and DX31 Abs, re- spectively (Fig. 4A, middle and bottom panels). P179TS181T and

W226A mutagenesis inhibited D1A and D2A Ab binding to Downloaded from KIR3DL2 (Fig. 4B). D0A, D0C, and DX31 Abs inhibited IL-2 production by KIR3DL2 reporter cells stimulated with B27-transfected cells (Fig. 4C). These D0 domain–specific Abs also inhibited IL-2 production by KIR3DL2 reporter cells stimulated with parental

221 cells or 221 cells transfected with HLA-A3 or other class I http://www.jimmunol.org/ HLAs (Fig. 4C). In contrast, the D1- and D2-specific Abs D1A and D2A did not affect IL-2 production by KIR3DL2 reporter cells stimulated with parental and transfected 221 cells (Fig. 4C). D0 domain–specific Abs inhibited KIR3DL2 reporter cell inter- actions with B27+ and B272 EBV-immortalized B cell lines (Fig. 4D). Identification of residues in KIR3DL2 involved in binding to B27 FHC dimers and other HLA class I ligands by guest on September 23, 2021 We next determined which amino acids in KIR3DL2 were involved in binding to B27 FHC dimers and other class I HLAs. Because KIR3DL1 binds more weakly to B27 FHC dimers than b2m- associated B27 (this study) (25), we reasoned that targeted mu- tagenesis of potential contact amino acids in KIR3DL2 that dif- fered from KIR3DL1 would affect binding. Thus, we mutated potential contact residues in KIR3DL2 to alanine or the corre- sponding amino acids in KIR3DL1. We also studied the effect of mutating potential contact residues that were conserved between KIR3DL1 and KIR3DL2 on binding. Two independent sets of mutants were produced. The expression levels of these mutants relative to WT KIR3DL2 reporters were determined by staining with two anti-KIR3DL2 Abs (D1A and DX31) (Fig. 5A, 5B). We first studied the effect of mutagenesis of residues in the D0 domain of KIR3DL2 on binding to HLA class I FHCs. F9A mu- tagenesis inhibited KIR3DL2 reporter stimulation by 221B27 cells and parental 221, 221A3, and 221B7 cells (Fig. 5C, 5D). R32A mutagenesis inhibited KIR3DL2 reporter interactions with 221B27 cells (Fig. 5C, 5D). R32 in the KIR3DL2 D0 domain appears to act FIGURE 4. Abs that target the D0 domain of KIR3DL2 inhibit binding to as a clamp for binding to b2m class I FHCs. Mutation of R32 to B27 and other HLA class I FHCs. (A) FACS staining of 293T cells transfected with the D0 domain of KIR3DL2 with D0A, D0C, DX31, and IgG1 and IgG2a the corresponding histidine residue (R32H) in KIR3DL1 inhibited Abs (top panels). FACS staining of the I60NG62S KIR3DL2CD3ε Jurkat binding to both 221B27 cells and other 221 transfectants (Fig. 5C). reporter mutant cells with D0C and D1A Abs (middle panel). FACS staining of R13A, R13W, and H29A mutagenesis did not affect KIR3DL2 R78HL82P KIR3DL2CD3ε Jurkat reporter mutant cells with DX31 and D1A reporter binding to class I HLAs (Fig. 5C, 5D). Abs (bottom panel). (B) D1A and D2A Ab titrations on 293T cells transfected with WT KIR3DL2 and the indicated mutants. (C) Effect of D0 domain (D0A- D0C and DX31), D1 domain (D1A), and D2 domain (D2A) Abs on IL-2 2 production by KIR3DL2CD3ε Jurkat reporter cells stimulated with 221B27, IL-2 production by KIR3DL2CD3ε2 Jurkat reporter cells stimulated with 221, 221B7, 221B35, and 221A3 cells. Representative data from one of three B27+ or B272 EBV-transformed B cell lines. Representative data from one independent experiments. (D) Effect of D0 domain (D0A-D0C and DX31) on of three independent experiments. Data are mean 6 SD. The Journal of Immunology 1597

I60NG62S mutagenesis inhibited KIR3DL2 recognition of We also studied the effect of L82P mutagenesis on KIR3DL2 221B27 H chains (Fig. 5C, 5D), despite the fact that the I60NG62S binding to B27 and other H chains. The L82P mutant was expressed mutant is consistently expressed at higher levels than WT KIR3DL2 at levels similar to WT, as assessed by staining with the anti- (Fig. 5A). In contrast, I60NG62S mutagenesis did not affect KIR3DL2 Ab D1A. L82P mutagenesis reduced staining with the KIR3DL2 binding to HLA-A3 and HLA-B7 as much as binding to DX31 Ab, which is consistent with the location of the DX31 epitope B27 (Fig. 5C, 5D). (Supplemental Fig. 2B). L82P mutagenesis inhibited binding to Downloaded from http://www.jimmunol.org/ by guest on September 23, 2021

FIGURE 5. Identification of amino acids involved in binding to B27 FHC dimers and other HLA class I H chains. (A and B) FACS staining of KIR3DL2CD3ε Jurkat reporter cell mutants with the indicated KIR3DL2 DX31 and D1A Abs. Two sets of mutants and WT reporter cells corresponding to the mutants in (C) were produced. The expression levels of WT and mutant constructs, as determined by staining with D1A and DX31 KIR3DL2 mAbs, are indicated by the respective geometric mean fluorescence intensities (MFIs). FACS stains of mutant KIR3DL2 reporter cells are representative of one of three independent experiments. (C) IL-2 secretion by WT and mutant KIR3DL2CD3ε reporter cells stimulated with 221 and 221B27, 221A3, or 221B7 cells. Representative data from one of three independent experiments (mean 6 SD). (D) IL-2 production by mutant KIR3DL2 reporter cells stimulated with 221B27, 221A3, and 221B7 cells and expressed as a percentage of WT KIR3DL2 reporter cell IL-2 production. Data are mean 6 SEM from three in- dependent experiments. *p , 0.05, **p , 0.01, ***p , 0.005, ****p , 0.001. 1598 ROLE OF THE D0 DOMAIN OF KIR3DL2 IN BINDING TO B27 DIMERS both B27 and other HLA class I H chains (Fig. 5C, 5D). In the D2 (Fig. 5C, 5D). Q149A and W226R mutagenesis did not have any domain, Y200A mutagenesis consistently inhibited reporter cell effect on KIR3DL2 binding to the H chains tested in this study interactions with 221B27 and 221 cells expressing other class I (Fig. 5C, 5D). HLAs (Fig. 5C, 5D). The majority of key KIR3DL2 residues identified through ex- perimental mutagenesis were persistent interface residues, and Model of KIR3DL2 binding to B27 FHC dimer many are also predicted hot spots, meaning that their mutation to We modeled B27 dimer binding to KIR3DL2 using two methods alanine carries a significant energetic penalty. To reach some that yielded similar results (Materials and Methods). Model pre- understanding of why KIR3DL2 binds strongly to B27 H chains dictions were in good agreement with observed mutagenesis compared with KIR3DL1, we also modeled KIR3DL1 B27 dimer effects. Interface residues and predicted hot spots for the model interactions. KIR3DL1 was predicted to form fewer contacts with are summarized in Supplemental Table I. the B27 dimer, suggesting a more rigid structure than KIR3DL2 We tested the robustness of the model by performing alanine (J.M. Mitchell, unpublished observations). mutagenesis on a series of KIR3DL2 residues predicted to form Residues F9, L82, and Y200 in KIR3DL2 are interface residues contacts with B27 H chains. Because the majority of the mutants in 180/180 structures and predicted hot spots. Residues R13 and tested were predicted to affect binding to one B27 H chain (chain B H29 are interface residues in all structures, whereas residues R32 in Supplemental Table I), we generated a panel of mutants pre- and R78 appear in a subset of the interfaces. dicted to alter binding to the other B27 chain (chain A): R73A in One KIR3DL2 monomer is predicted to bind via complementary the D0 domain and D142A, P143A, S144A, R145A, and Q149A asymmetric binding sites to the two B27 H chains in the B27 dimer. in the D1 domain. In addition, we made a W226R mutant in the The B27 binding face with KIR3DL2 is dominated by polar Downloaded from D2 domain that incorporates the epitope of the D2A Ab. The contacts. In this model, two C67 disulfide-bonded B27 H chains (A expression levels of these mutants compared with WT are shown and B) are arranged in reverse orientations (Fig. 6A, 6B). The in Fig. 5B. KIR3DL2 D2 domain binds to one H chain of the B27 dimer, and R73A mutagenesis reduced binding to B27 and other HLA class the D1 and D0 domains bind to complementary regions on al- I H chains (Fig. 5C, 5D). The D142A, P143A, and S144A mutants ternate H chains (Fig. 6).

increased KIR3DL2 binding to B27 H chains (Fig. 5C, 5D). The majority of contacts of B27 with KIR3DL2 are made http://www.jimmunol.org/ R145A marginally increased binding to B27 FHC. D142A muta- asymmetrically with the D1 domain, with one B27 H chain forming genesis consistently promoted binding to HLA-A3 and B7 H more contacts than the other (Fig. 6C). The D2 domain is predicted chains, whereas S144A mutation marginally increased interactions to bind to only one B27 H chain. Residues F9, R13, and H29 in the by guest on September 23, 2021

FIGURE 6. Model of KIR3DL2 binding to B27 FHC dimers. Front (A) and rear (B) views of KIR3DL2 bound to B27 dimer. B27 H chains are shown in and dark blue. The D0, D1, and D2 domains of KIR3DL2 are colored red, purple, and coral. The locations of HC10, DOC, D1A, D2A, and DX31 Ab epitopes are indicated in green. R32, implicated in KIR3DL2 ligand binding, is also shown. (C) The B27 footprint mapped to the surface of KIR3DL2, with residues colored cyan or blue to indicate the respective H chain contacts. Residues incorporated in the H chain HC10 mAb epitope are in red type. The Journal of Immunology 1599

D0 domain and Y200 in the D2 domain are predicted to contact R145. These mutations also free up some space to facilitate one B27 FHC (Fig. 7A, 7E) in a manner analogous to KIR3DL1 conformational rearrangements, but the effects are proportionally binding to HLA-B57 (PDB 3VH8). less, depending on the distance from the loop. L82 in the KIR3DL2 D0 domain is predicted to bind to K186 in Residues I60 and G62 do not make direct contact with the B27 the a3 domain of one B27 H chain (Fig. 7B). L82 makes up part of dimer. However G62 is probably important for providing flexibility the epitope for the blocking Ab DX31 and is predicted to be in the peptide loop in the D0 domain between I60 and R73 a mutagenesis hot spot in 17 of the 180 models analyzed. Additional (Supplemental Fig. 3B). In contrast, S62 in the corresponding loop residues in the D0 domain of KIR3DL2 contact alternate regions of in KIR3DL1 bound to HLA-B57 is predicted to limit the flexi- the two H chains in the B27 dimer (Fig. 7C, Supplemental Table I). bility of this region (Supplemental Fig. 3C). This added flexibility Q54 and R62 in the region of the HC10 epitope of the B27 H chain may be particularly important for accommodating the conforma- are predicted to bind residues in the D0 and D1 domains, respectively tional changes in KIR3DL2 necessary for binding to the two H (Fig. 7C). Q54 could bind to R73 and P91 in the D0 domain. R62 is chains in the B27 dimer. This offers a possible explanation for the predicted to form contacts with Q141, D142, P143, and S144 in the greater effect of I60NG62S mutagenesis on KIR3DL2 binding to D1 domain of KIR3DL2 (Figs. 6C, 7C). D1 domain binding to B27 B27 FHC dimers compared with the effect of this mutation on dimer is predicted to be strengthened by additional contacts in the binding to other HLA class I H chains (Fig. 5D). Binding of the a2 domain of the same B27 H chain and the a1domainofthe D0C Ab to this region would be expected to reduce the flexibility alternate B27 H chain (Fig. 7D, Supplemental Table I). of the KIR3DL2 D0 domain. Modeling predicts that T73 in A significant conformational change in the loop containing aa KIR3DL1 does not bind to B27 dimer (unpublished observations). residues 139–141 is present in the predicted bound structure Downloaded from compared with the predicted unbound one (Supplemental Fig. Discussion 3A). This may offer a possible explanation for the stimulatory We show that B27 FHC dimers bind more strongly to the KIR effect of D142A, P143A, and S144A mutagenesis on KIR3DL2 KIR3DL2 than do other HLA-A3, HLA-B7, HLA-B35 and HLA-G binding to B27 FHC dimer. The most dramatic effect is seen when H chains. In addition, we use a series of complementary approaches mutating D142, which is near the conformational change and to derive a model for the KIR KIR3DL2 binding to B27 FHC

probably makes space for it to mold into. The improvement in dimers. In this model, monomeric KIR3DL2 is predicted to form http://www.jimmunol.org/ binding gradually decreases for mutations in P143, S144, and multiple complementary contacts with the two B27 H chains. Monomeric KIR3DL2 binding is consistent with our observations that mutation of the cysteines in the KIR3DL2 stem does not affect B27 binding. Monomeric binding to ligand is also consistent with our biochemical evidence showing that a large fraction of KIR3DL2 is expressed at the cell surface as monomer. One B27 H chain is predicted to form contacts with KIR3DL2 D0, D1, and D2 domains, which are similar to those of KIR3DL1 bound to ligand. The other B27 H chain is predicted to form alternative complementary con- by guest on September 23, 2021 tacts with KIR3DL2 in the D0 and D1 domains. Our studies show that the D2 and D0 domains play a key role in stabilizing KIR3DL2 binding to FHC ligands. Alanine mutagenesis of F9 and Y200 significantly reduced binding. Moreover the D0 domain residues R32, R73, and L82 are directly involved in binding to the HLA class I H chains in this study, and I60 and G62 have an indirect role in stabilizing B27 dimer interactions. The aa residues R32 and R73 and I60, G62 and L82 are all nonconserved in KIR3DL1. Critically, R73 and L82 are predicted to stabilize an alternate binding face that is unique to KIR3DL2. We extend previous observations showing that KIR3DL2 binds weakly to HLA class I FHCs, in general, and binds more strongly to B27 FHC dimers. Acid treatment of APCs increased levels of HLA class I FHCs and concomitantly increased KIR3DL2 reporter cell interactions. Furthermore, these interactions are inhibited by the HLA class I H chain Ab HC10. The model also offers a possible explanation for the weak binding of KIR3DL2 to other HLA class I FHCs; our mutagenesis and Ab-inhibition studies indicated that some contact residues in KIR3DL2 are shared between B27 FHC FIGURE 7. Predicted amino acid contacts between KIR3DL2 and the dimers and other HLA class I FHCs. B27 dimer. (A and B) Predicted amino acid contacts in the two binding Of all of the HLA class FHCs that we studied, only B27 FHC registers of the KIR3DL2 D0 domain with the two B27 H chains. The R78 dimers bound strongly to KIR3DL2. KIR3DL2 reporter T cells and L82 amino acids of the DX31 Ab epitope are indicated in red in (B). consistently produced higher IL-2 in response to B27 compared C ( ) Predicted amino acid contacts in the KIR3DL2 D0 and D1 domains with stimulation with 221 B cells transfected with other class I with Q54 and R62 residues within the HLA class I H chain HC10 epitope. HLAs. Our finding of greater stimulation of KIR3DL2 reporter Amino acids that are part of the HC10 epitope are highlighted in red. (D) cells by transfected cell lines expressing B27 is also supported by Predicted amino acid contacts in the KIR3DL2 D1 domain with the B27 H + 2 chain. (E) Predicted amino acid contacts in the KIR3DL2 D2 domain with our data using B27 and B27 EBV-transformed B cell lines. We + B27. W226 in the D2A epitope is highlighted in red. Key KIR3DL2 res- observed increased stimulation of reporter cells by B27 B cells 2 idues and domains are indicated in black or purple. B27 H chain residues compared with B27 cell lines. Only KIR3DL2 interactions with and domains are indicated in blue. B27 FHC dimers were strong enough to inhibit NK killing of 1600 ROLE OF THE D0 DOMAIN OF KIR3DL2 IN BINDING TO B27 DIMERS target cells and IFN-g in the assays reported in this article. Al- We show that stronger binding of KIR3DL2 to B27 dimers though we cannot rule out the possibility of weaker functional results in increased inhibition of NK cell cytotoxicity and IFN-g interactions between KIR3DL2 and other HLA class I FHCs, as prοdyxtiοn. Lower IFN-g levels were implicated in the devel- reported by Goodridge et al. (15), our results demonstrate a much opment of spondyloarthritis and arthritis both ex vivo and in stronger functional interaction with B27 FHC dimers compared model systems (37, 38). IFN-g is critical for resolving infections with the other FHCs examined. with intracellular Gram-negative bacteria, which have been im- Transfected cell lines, EBV-immortalized B cell lines, and plicated in B27+-associated arthritis. Indeed, B27-associated an- monocytes from B27+ patients express B27 FHC dimers (22). kylosing spondylitis is characterized by low-grade intestinal Molecular modeling predicts that the stronger binding of KIR3DL2 inflammation that could result from unresolved infection (39). depends on B27 FHC dimers forming additional contacts with High levels of IFN-g inhibit production of the proarthritic cyto- KIR3DL2 compared with other class I FHCs. Consistent with this, kines IL-17 and IL-23, which have been implicated in B27+-as- we showed previously that 221 cells expressing the C67S mutant of sociated disorders (40). Thus, stronger binding of KIR3DL2 to B27, which does not form H chain dimers, bind more weakly to B27 on lymphocytes could alter the IL-17/IFN-g balance in patients KIR3DL2 (19). with B27-associated disorders. Stronger binding of KIR3DL2 to Notably, a number of point mutants that inhibited KIR3DL2 B27 also could license NK cells with greater functional potential in binding to B27 FHC dimers also inhibited KIR3DL2 binding to the B27+ individuals. other HLA class I FHCs tested. Some of the KIR3DL2 side chains In this study, we modeled KIR3DL2 binding to B27 FHC dimer. would not contact monomeric MHC in the structure suggested Our model predicts key complementary contacts of two B27 FHCs by KIR3DL1 binding ligand (PDB 3VH8). This suggests that with monomeric KIR3DL2, as well as identifies potential binding Downloaded from KIR3DL2 binds to monomeric HLA class I FHCs in two alternate sites to other monomeric HLA class I H chains. Our findings offer registers. Our model predicts critical contacts between the D1 and a possible molecular explanation for the stronger binding of B27 the D0 domains and the HLA class I region encompassed by the FHC dimers to KIR3DL2 and identify key regions in the KIR3DL2 HC10 epitope in one register. The numbers of predicted hot spots B27 interface for targeting this interaction. made with each monomer of the B27 dimer are roughly equal, suggesting this secondary interaction should generate sufficient Acknowledgments http://www.jimmunol.org/ binding strength to form a stable complex. We thank Nicolai Wagtmann (Innate Pharma) for critical reading of the Abs against the D0 domain of KIR3DL2 consistently block binding manuscript. to HLA class I FHCs and b2m-associated HLA-A3 complexes with peptide (pHLA-A3) (this study) (25). Thus, the D0 domain of Disclosures KIR3DL2 plays an important role in binding to both pHLA-A3 and The authors have no financial conflicts of interest. B27 and other HLA class I FHC ligands. Abs that directly affected binding of the KIR3DL2 D0 domain to ligands (DX31) were par- ticularly potent at inhibiting binding. In contrast, a D2 domain– References specific Ab (D2A) did not affect B27 dimer binding to KIR3DL2. 1. Brown,M.A.,K.D.Pile,L.G.Kennedy,A.Calin,C.Darke,J.Bell,B.P.Wordsworth, by guest on September 23, 2021 and F. Corne´lis. 1996. HLA class I associations of ankylosing spondylitis in the white The D2A epitope incorporates W226, which is predicted not to di- population in the United Kingdom. Ann. Rheum. Dis. 55: 268–270. rectly contact B27 H chains. The D1 domain–specific Ab D1A binds 2. Neumann-Haefelin, C. 2013. HLA-B27-mediated protection in HIV and hepatitis to residues P179 and S181 on the opposite side of the KIR3DL2 C virus infection and pathogenesis in spondyloarthritis: two sides of the same coin? Curr. Opin. Rheumatol. 25: 426–433. HLA class I–binding face. This is consistent with the lack of inhi- 3. Wang, S., G. Li, R. Ge, Z. Duan, Z. Zeng, T. Zhang, J. Gao, T. Yang, S. Liu, bition of KIR3DL2 binding to B27 FHC dimers by D1A mAb. S. Wu, et al. 2013. Association of KIR genotype with susceptibility to HLA-B27- The mAb D0C binds to residues I60 and G62, which do not positive ankylosing spondylitis. Mod. Rheumatol. 23: 538–541. 4. Shaw, J., H. Hatano, and S. Kollnberger. 2014. The biochemistry and immu- directly contact class I. G62 probably provides additional structural nology of non-canonical forms of HLA-B27. Mol. Immunol. 57: 52–58. flexibility, enabling the KIR3DL2 molecule to accommodate the 5. Carrington, M., M. P. Martin, and J. van Bergen. 2008. KIR-HLA intercourse in HIV disease. Trends Microbiol. 16: 620–627. conformational changes necessary in the D0 domain for binding to 6. Lanier, L. L. 2005. NK cell recognition. Annu. Rev. Immunol. 23: 225–274. B27 dimers. Mutation of I60 and G62 residues to the corresponding 7. Kulkarni, S., M. P. Martin, and M. Carrington. 2008. The Yin and Yang of HLA residues in KIR3DL1 inhibits KIR3DL2 binding to B27 FHC and KIR in human disease. Semin. Immunol. 20: 343–352. 8. Pende, D., R. Biassoni, C. Cantoni, S. Verdiani, M. Falco, C. di Donato, dimers, which is consistent with a critical role for this region of the L. Accame, C. Bottino, A. Moretta, and L. Moretta. 1996. The natural killer cell D0 domain for binding to ligand. receptor specific for HLA-A allotypes: a novel member of the p58/p70 family of Our model predicts that KIR3DL2 allelic variation is unlikely to inhibitory receptors that is characterized by three immunoglobulin-like domains and is expressed as a 140-kD disulphide-linked dimer. J. Exp. Med. 184: 505–518. significantly affect KIR3DL2 binding to B27 FHC dimer. None of 9. Do¨hring, C., D. Scheidegger, J. Samaridis, M. Cella, and M. Colonna. 1996. A the common allelic variants of KIR3DL2 incorporate amino acid human killer inhibitory receptor specific for HLA-A1,2. J. Immunol. 156: 3098– differences in predicted contact residues. This is in agreement with 3101. 10. Wagtmann, N., R. Biassoni, C. Cantoni, S. Verdiani, M. S. Malnati, M. Vitale, our observations that the function of KIR3DL2-expressing NK C. Bottino, L. Moretta, A. Moretta, and E. O. Long. 1995. Molecular clones of cells from different donors is consistently modulated by B27 FHC the p58 NK cell receptor reveal immunoglobulin-related molecules with diver- dimer. The fact that the KIR3DL2 binding face is formed by sity in both the extra- and intracellular domains. Immunity 2: 439–449. 11. Fauriat, C., S. Andersson, A. T. Bjo¨rklund, M. Carlsten, M. Schaffer, multiple contacts with the two B27 H chains in the B27 dimer N. K. Bjo¨rkstro¨m, B. C. Baumann, J. Michae¨lsson, H. G. Ljunggren, and suggests that KIR3DL2–B27 interactions would be relatively K. J. Malmberg. 2008. Estimation of the size of the alloreactive NK cell rep- ertoire: studies in individuals homozygous for the group A KIR haplotype. tolerant to amino acid substitutions in the different KIR3DL2 J. Immunol. 181: 6010–6019. alleles. The epitope recognized by the anti-KIR3DL2 Ab DX31 is 12. Hansasuta, P., T. Dong, H. Thananchai, M. Weekes, C. Willberg, H. Aldemir, conserved between the majority of the KIR3DL2 allelic variants. S. Rowland-Jones, and V. M. Braud. 2004. Recognition of HLA-A3 and HLA- A11 by KIR3DL2 is peptide-specific. Eur. J. Immunol. 34: 1673–1679. Two rare alleles, *047 and *054, have a histidine substitution at 13. Shaw, J., and S. Kollnberger. 2012. New perspectives on the ligands and function position 73, which is the same residue found in KIR3DL1. Given of the killer cell immunoglobulin-like receptor KIR3DL2 in health and disease. the importance of R73 in stabilizing the alternate binding face in Front. Immunol. 3: 339. 14. Kollnberger, S., L. Bird, M. Y. Sun, C. Retiere, V. M. Braud, A. McMichael, and KIR3DL2, these alleles could have altered affinity for B27 dimers P. Bowness. 2002. Cell-surface expression and immune receptor recognition of and other HLA class I ligands. HLA-B27 homodimers. Arthritis Rheum. 46: 2972–2982. The Journal of Immunology 1601

15. Goodridge, J. P., A. Burian, N. Lee, and D. E. Geraghty. 2013. HLA-F and MHC 27. De Santis, D., B. Foley, C. S. Witt, and F. T. Christiansen. 2012. The detection of class I open conformers are ligands for NK cell Ig-like receptors. J. Immunol. NK cell alloreactivity by flow cytometric CD107a assay. Methods Mol. Biol. 191: 3553–3562. 882: 477–489. 16. Vivian, J. P., R. C. Duncan, R. Berry, G. M. O’Connor, H. H. Reid, T. Beddoe, 28. Fischer, K., R. Andreesen, and A. Mackensen. 2002. An improved flow cyto- S. Gras, P. M. Saunders, M. A. Olshina, J. M. Widjaja, et al. 2011. Killer cell metric assay for the determination of cytotoxic T lymphocyte activity. immunoglobulin-like receptor 3DL1-mediated recognition of human leukocyte J. Immunol. Methods 259: 159–169. antigen B. Nature 479: 401–405. 29. McHugh, K., O. Rysnik, S. Kollnberger, J. Shaw, L. Utriainen, M. H. Al- 17. Khakoo, S. I., R. Geller, S. Shin, J. A. Jenkins, and P. Parham. 2002. The D0 Mossawi, S. Payeli, O. Marroquin, S. Milling, C. Renner, and P. Bowness. 2014. domain of KIR3D acts as a major histocompatibility complex class I binding Expression of aberrant HLA-B27 molecules is dependent on B27 dosage and peptide supply. Ann. Rheum. Dis. 73: 763–770. enhancer. J. Exp. Med. 196: 911–921. 30. Hulsmeyer,€ M., M. T. Fiorillo, F. Bettosini, R. Sorrentino, W. Saenger, 18. Fu, L., B. Hazes, and D. N. Burshtyn. 2011. The first Ig domain of KIR3DL1 A. Ziegler, and B. Uchanska-Ziegler. 2004. Dual, HLA-B27 subtype-dependent contacts MHC class I at a secondary site. J. Immunol. 187: 1816–1825. conformation of a self-peptide. J. Exp. Med. 199: 271–281. 19. Wong-Baeza, I., A. Ridley, J. Shaw, H. Hatano, O. Rysnik, K. McHugh, C. Piper, 31. Phillips, J. C., R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, S. Brackenbridge, R. Fernandes, A. Chan, et al. 2013. KIR3DL2 binds to HLA- C. Chipot, R. D. Skeel, L. Kale´, and K. Schulten. 2005. Scalable molecular B27 dimers and free H chains more strongly than other HLA class I and pro- dynamics with NAMD. J. Comput. Chem. 26: 1781–1802. motes the expansion of T cells in ankylosing spondylitis. J. Immunol. 190: 3216– 32. MacKerell, A. D., D. Bashford, M. Bellott, R. L. Dunbrack, J. D. Evanseck, 3224. M. J. Field, S. Fischer, J. Gao, H. Guo, S. Ha, et al. 1998. All-atom empirical 20. Bowness, P., A. Ridley, J. Shaw, A. T. Chan, I. Wong-Baeza, M. Fleming, potential for molecular modeling and dynamics studies of . J. Phys. F. Cummings, A. McMichael, and S. Kollnberger. 2011. Th17 cells expressing Chem. B 102: 3586–3616. KIR3DL2+ and responsive to HLA-B27 homodimers are increased in ankylosing 33. Eswar, N., D. Eramian, B. Webb, M. Y. Shen, and A. Sali. 2008. Protein structure spondylitis. J. Immunol. 186: 2672–2680. modeling with MODELLER. Methods Mol. Biol. 426: 145–159. 21. Chan, A. T., S. D. Kollnberger, L. R. Wedderburn, and P. Bowness. 2005. Ex- 34. de Vries, S. J., M. van Dijk, and A. M. Bonvin. 2010. The HADDOCK web pansion and enhanced survival of natural killer cells expressing the killer server for data-driven biomolecular docking. Nat. Protoc. 5: 883–897. immunoglobulin-like receptor KIR3DL2 in spondylarthritis. Arthritis Rheum. 35. Zhu, X., and J. C. Mitchell. 2011. KFC2: a knowledge-based hot spot prediction 52: 3586–3595. method based on interface solvation, atomic density, and plasticity features. Downloaded from 22. Payeli, S. K., S. Kollnberger, O. Marroquin Belaunzaran, M. Thiel, K. McHugh, Proteins 79: 2671–2683. J. Giles, J. Shaw, S. Kleber, A. Ridley, I. Wong-Baeza, et al. 2012. Inhibiting 36. Wainwright, S. D., P. A. Biro, and C. H. Holmes. 2000. HLA-F is a predomi- HLA-B27 homodimer-driven immune cell inflammation in spondylarthritis. nantly empty, intracellular, TAP-associated MHC class Ib protein with a re- Arthritis Rheum. 64: 3139–3149. stricted expression pattern. J. Immunol. 164: 319–328. 23. Perosa, F., G. Luccarelli, M. Prete, E. Favoino, S. Ferrone, and F. Dammacco. 37. Fert, I., N. Cagnard, S. Glatigny, F. Letourneur, S. Jacques, J. A. Smith, 2003. Beta 2-microglobulin-free HLA class I heavy chain epitope mimicry by R. A. Colbert, J. D. Taurog, G. Chiocchia, L. M. Araujo, and M. Breban. 2014. Reverse interferon signature is characteristic of antigen-presenting cells in hu- monoclonal antibody HC-10-specific peptide. J. Immunol. 171: 1918–1926.

man and rat spondyloarthritis. Arthritis Rheum. 66: 841–851. http://www.jimmunol.org/ 24. Bird, L. A., C. A. Peh, S. Kollnberger, T. Elliott, A. J. McMichael, and 38. Irmler, I. M., M. Gajda, and R. Bra¨uer. 2007. Exacerbation of antigen-induced P. Bowness. 2003. Lymphoblastoid cells express HLA-B27 homodimers both arthritis in IFN-gamma-deficient mice as a result of unrestricted IL-17 response. intracellularly and at the cell surface following endosomal recycling. Eur. J. J. Immunol. 179: 6228–6236. Immunol. 33: 748–759. 39. Ciccia, F., M. Bombardieri, A. Principato, A. Giardina, C. Tripodo, R. Porcasi, 25. Kollnberger, S., A. Chan, M. Y. Sun, L. Y. Chen, C. Wright, K. di Gleria, S. Peralta, V. Franco, E. Giardina, A. Craxi, et al. 2009. Overexpression of A. McMichael, and P. Bowness. 2007. Interaction of HLA-B27 homodimers with interleukin-23, but not interleukin-17, as an immunologic signature of subclin- KIR3DL1 and KIR3DL2, unlike HLA-B27 heterotrimers, is independent of the ical intestinal inflammation in ankylosing spondylitis. Arthritis Rheum. 60: 955– sequence of bound peptide. Eur. J. Immunol. 37: 1313–1322. 965. 26. Li, D., A. Hong, Q. Lu, G. F. Gao, B. Jin, G. R. Screaton, and X. N. Xu. 2012. A 40. Sheikh, S. Z., K. Matsuoka, T. Kobayashi, F. Li, T. Rubinas, and S. E. Plevy. novel role of CD1c in regulating CD1d-mediated NKT cell recognition by 2010. Cutting edge: IFN-gamma is a negative regulator of IL-23 in murine competitive binding to Ig-like transcript 4. Int. Immunol. 24: 729–737. macrophages and experimental colitis. J. Immunol. 184: 4069–4073. by guest on September 23, 2021