ARTICLE https://doi.org/10.1038/s41467-019-10966-8 OPEN Inactivating mutations and X-ray crystal structure of the tumor suppressor OPCML reveal cancer-associated functions James R. Birtley1,6,7, Mohammad Alomary2,7, Elisa Zanini2, Jane Antony 2, Zachary Maben1, Grant C. Weaver 1, Claudia Von Arx2, Manuela Mura2, Aline T. Marinho2, Haonan Lu 2, Eloise V.N. Morecroft2,3, Evdoxia Karali2, Naomi E. Chayen 4, Edward W. Tate 3, Mollie Jurewicz1, Lawrence J. Stern 1,8, Chiara Recchi 2,8 & Hani Gabra 2,5,8 1234567890():,; OPCML, a tumor suppressor gene, is frequently silenced epigenetically in ovarian and other cancers. Here we report, by analysis of databases of tumor sequences, the observation of OPCML somatic missense mutations from various tumor types and the impact of these mutations on OPCML function, by solving the X-ray crystal structure of this glycoprotein to 2.65 Å resolution. OPCML consists of an extended arrangement of three immunoglobulin-like domains and homodimerizes via a network of contacts between membrane-distal domains. We report the generation of a panel of OPCML variants with representative clinical mutations and demonstrate clear phenotypic effects in vitro and in vivo including changes to anchorage- independent growth, interaction with activated cognate receptor tyrosine kinases, cellular migration, invasion in vitro and tumor growth in vivo. Our results suggest that clinically occurring somatic missense mutations in OPCML have the potential to contribute to tumorigenesis in a variety of cancers. 1 Department of Pathology, University of Massachusetts Medical School, Worcester, MA 01655, USA. 2 Ovarian Cancer Action Research Centre, Department of Surgery and Cancer, Imperial College London, Du Cane Road, London W12 0NN, UK. 3 Department of Chemistry, Imperial College London, Wood Lane, London W12 0BZ, UK. 4 Computational and Systems Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London SW7 2AZ, UK. 5 Clinical Discovery Unit, Early Clinical Development, IMED Biotech Unit, AstraZeneca, Cambridge SG8 6HB, UK. 6Present address: UCB Pharma, Bath Road, Slough SL1 3WE, UK. 7These authors contributed equally: James R. Birtley, Mohammad Alomary. 8These authors jointly supervised this work: Lawrence J. Stern, Chiara Recchi, Hani Gabra. Correspondence and requests for materials should be addressed to L.J.S. (email: lawrence. [email protected]) or to C.R. (email: [email protected]) or to H.G. (email: [email protected]) NATURE COMMUNICATIONS | (2019) 10:3134 | https://doi.org/10.1038/s41467-019-10966-8 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10966-8 pioid-binding protein cell-adhesion molecule like allowed for chain tracing, which was followed by rounds of (OPCML) is a glycosylphosphatidylinositol (GPI)- automated and manual model building and refinement (see O fl anchored protein that localizes to the outer lea et of the Methods). In the final model (PDB entry 5UV6), 273 out of 286 plasma membrane and acts as a tumor suppressor (TS)1,2. residues in the mature OPCML protein were visualized (Sup- OPCML is a member of the IgLON superfamily, together with 4 plementary Figs. 2 and 3), with two molecules in the crystal- other proteins (LSAMP, NEGR1, HNT, and IgLON5), amongst lographic asymmetric unit. There are six potential N-linked which some have been hypothesized to act as tumor suppressors glycosylation sites (asparagines 44, 70, 140, 285, 293, and 306) or oncogenes3,4. However, functions of IgLON members have not and glycans were modelled at residues 70, 293, and 306. been fully characterized and their 3D structures remain unsolved. OPCML consists of three Ig-like domains termed D1, D2, and Previously, we showed that OPCML is inactivated predominantly D3, connected by short extended linkers (Fig. 1a, b, Supplemen- by somatic methylation and loss of heterozygosity in more than tary Fig. 2). The OPCML homodimer resembles an inverted V 80% of ovarian cancer patients2. When re-expressed in cancer shape (Fig. 1a), with dimerization mediated by a membrane-distal cells, OPCML inhibits proliferation in vitro and tumorigenicity D1-D1 interface stabilized by a combination of hydrophobic in vivo2 by binding and downregulating a specific subset of contacts, hydrogen bonds and salt bridges (Fig. 1c, Supplemen- receptor tyrosine kinases (RTKs)5. Although somatic promoter tary Fig. 4). This interface comprises 874 Å2 of buried surface area methylation is the predominant mechanism of silencing in and has a solvation free energy gain (ΔiG) on formation diverse tumor types1, we previously also reported a somatic calculated by PISA to be −9.5 kcal M−1. These values are larger missense mutation at P95R in an ovarian cancer patient2. Pre- than for other intermolecular interaction sites observed in the liminary data indicated that this variant was expressed in similar crystal, the next largest being crystal contacts mediated by the C- amounts as wild-type (WT) but was a loss-of-function mutant. terminal end of one molecule inserting between the B and G Here, we search for additional OPCML somatic mutations in the strands of D1 in another with 554 Å2 buried surface area and TCGA and COSMIC DNA sequencing databases from a panel of solvation free energy gain of −6.9 kcal M−1 (Supplementary nearly 30,000 patients affected by different cancers. We show that table 2 and Supplementary Fig. 5). Based on the quaternary a number of patients exhibit somatic mutations of OPCML.In structure seen in the asymmetric unit, the dimer would be order to understand how these clinical mutations might interfere anchored into the plasma membrane by two D3-linked GPI with OPCML’s tumor suppressor activity we then determine the anchors (Fig. 1a). The D3 C-terminal ends are located crystal structure of OPCML and show that these mutations fall approximately 180 Åapart, and thus the top of the D1 into several classes based on mutation type and location on the dimerization interface could extend above the plane of the protein structure. Selecting mutations representative of each class, plasma membrane by ~73 Å (Fig. 1a). we characterize their biophysical properties and demonstrate The OPCML dimer is stabilized by reciprocal salt bridges their functionality in in vitro and in vivo models. We further between Arg 65 of one monomer and Asp 80 of the other (Fig. 1c, demonstrate that particular mutations affect the oligomeric state of OPCML, its interaction with cognate RTKs and its effect on a range of phenotypes including migration, invasion, adhesion to extracellular matrix proteins and in vivo tumorigenicity. These fi fi Table 1 Data collection, phasing, and re nement statistics ndings reveal how the mutations result in functional/clinical for OPCML (MR-SAD) consequences, and these data have implications for under- standing the mechanism of action of this class of membrane- Native K2PtCl4 associated tumor suppressors. Data collection Space group P41212I4122 Cell dimensions Results a, b, c (Å) 93.6, 93.6, 262.2 98.8, 98.8, 267.1 OPCML sustains point mutations in cancer patients.Weana- α, β, γ (°) 90, 90, 90 90, 90, 90 a lyzed tumor DNA sequence data from the TCGA and COSMIC Resolution (Å) 18.5–2.65 (2.74–2.65) 48.3–3.17 (3.25–3.17) databases in order to identify possible clinical OPCML mutations. Rsym 17.1 (273.6) 18.7 (301.5) We found point mutations in the OPCML gene in 287 out of 28,132 Rpim 4.7 (73.8) 3.6 (56.6) patients and these were distributed across all cancer types (Sup- I/σ(I) 9.7 (1.2) 11.5 (1.1) CC1/2 100 (68.5) 100 (82.3) plementary Fig. 1A). Generally, the frequency of these clinical Completeness (%) 95.8 (88.8) 100.0 (100.0) mutations is uncommon, ranging from 5.5% in melanoma and Redundancy 14.4 (14.5) 28.2 (29.1) 3.3% in colon, gastric, and bladder cancer to around 0.1% in cancers Refinement Resolution (Å) 18.5–2.65 (2.74–2.65) such as breast and brain (Supplementary Fig. 1A). Nine mutations No. reflections 33374 (3021) were found to be splice site variants or stop codon-related altera- Rwork/Rfree 28.7/30.4 tions (Supplementary Table 1). The remaining 278 changes were No. atoms 4392 Protein 4216 missense mutations (Supplementary Table 1) and were mostly Ligand/ion (carbohydrate) 134 scattered evenly along the entire OPCML sequence, with the Water 42 greatest number located in domain 1 (Supplementary Fig. 1B). B factors Protein 84.6 Ligand/ion 91.0 Water 74.2 X-ray crystal structure of OPCML. To localize the clinical R.m.s deviations mutations on the three-dimensional structure of OPCML and to Bond lengths (Å) 0.006 gain insights into their role in the tumor suppressor mechanism(s) Bond angles (°) 0.99 we crystallized soluble recombinant OPCML (residues 36–316) NCS RMSD (Å) All domains 1.10 and determined its structure using a combined single-wavelength D1 (43–134) 0.24 X-ray anomalous dispersion/molecular replacement approach D2 (135–222) 0.19 (Table 1). Structure solution was complicated initially by a high D3 (223–319) 0.64 fi degree of translational pseudosymmetry, but identi cation of the aValues in parentheses are for highest-resolution shell correct space group and non-crystallographic symmetry elements 2 NATURE COMMUNICATIONS | (2019) 10:3134 | https://doi.org/10.1038/s41467-019-10966-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10966-8 ARTICLE a * * c e D1 D1 Arg 65 Thr 114 WT Leu 69 Dimer structure Asp 80 Trp 82 N N Ile 74 1 Monomer structure D2 D2 ~73Å Ile 74 D3 D3 Trp 82 Leu 69 Thr 114 Asp 80 Arg 65 Scaled intensity (log RU) 0.00 0.05 0.10 0.15 0.20 q (Å–1) C WT fit to dimer Membrane 1.5 WT fit to monomer d 1.0 Thr 94 ~180Å Val 92 Asn 93 Pro 95 0.5 0 3.1 Å 3.0 Å 0.00 0.05 0.10 0.15 0.20 b 3.2 Å Ratio of intensity (RU) q (Å–1) 2.9 Å C * Gln 97 Thr 96 WT OPCML Ser 99 f Tyr 98 C ) (arbitrary units) R ( * p 0 50 100 150 200 R (Å) Fig.