The terminal sialic acid of stage-specific embryonic antigen-4 has a crucial role in binding to a cancer-targeting antibody

Abstract

Cancer remains a leading cause of morbidity and mortality worldwide, requiring ongoing development of targeted therapeutics such as monoclonal antibodies. Carbohydrates on embryonic cells are often highly expressed in cancer and are therefore attractive targets for antibodies. Stage-specific embryonic antigen-4 (SSEA-4) is one such glycolipid target expressed in many cancers, including breast and ovarian carcinomas. Here, we defined the structural basis for recognition of SSEA-4 by a novel monospecific chimeric antibody (ch28/11). Five X-ray structures of ch28/11 Fab complexes with the SSEA-4 glycan headgroup, determined at 1.5–2.7 Å resolutions, displayed highly similar three-dimensional structures indicating a stable binding mode. The structures also revealed that by adopting a horseshoe-shaped conformation in a deep groove, the glycan headgroup likely sits flat against the membrane to allow the antibody to interact with SSEA-4 on cancer cells. Moreover, we found that the terminal sialic acid of SSEA-4 plays a dominant role in dictating the exquisite specificity of the ch28/11 antibody. This observation was further supported by molecular dynamics simulations of the ch28/11-glycan complex, which show that SSEA-4 is stabilized by its terminal sialic acid, unlike SSEA-3, which lacks this sialic acid modification. These high-resolution views of how a glycolipid interacts with an antibody may help to advance a new class of cancer-targeting immunotherapy. Cancer remains a leading cause of morbidity and mortality worldwide, requiring ongoing development of targeted therapeutics such as monoclonal antibodies. Carbohydrates on embryonic cells are often highly expressed in cancer and are therefore attractive targets for antibodies. Stage-specific embryonic antigen-4 (SSEA-4) is one such glycolipid target expressed in many cancers, including breast and ovarian carcinomas. Here, we defined the structural basis for recognition of SSEA-4 by a novel monospecific chimeric antibody (ch28/11). Five X-ray structures of ch28/11 Fab complexes with the SSEA-4 glycan headgroup, determined at 1.5–2.7 Å resolutions, displayed highly similar three-dimensional structures indicating a stable binding mode. The structures also revealed that by adopting a horseshoe-shaped conformation in a deep groove, the glycan headgroup likely sits flat against the membrane to allow the antibody to interact with SSEA-4 on cancer cells. Moreover, we found that the terminal sialic acid of SSEA-4 plays a dominant role in dictating the exquisite specificity of the ch28/11 antibody. This observation was further supported by molecular dynamics simulations of the ch28/11-glycan complex, which show that SSEA-4 is stabilized by its terminal sialic acid, unlike SSEA-3, which lacks this sialic acid modification. These high-resolution views of how a glycolipid interacts with an antibody may help to advance a new class of cancer-targeting immunotherapy. It has long been known that glycosylation of malignant cells differs significantly from that of healthy cells (1Meezan E. Wu H.C. Black P.H. Robbins P.W. Comparative studies on the carbohydrate-containing membrane components of normal and virus-transformed mouse fibroblasts. II. Separation of glycoproteins and glycopeptides by Sephadex chromatography.Biochemistry. 1969; 8 (4307997): 2518-252410.1021/bi00834a039Crossref PubMed Scopus (187) Google Scholar). Aberrant glycosylation in cancer (associated with several glycan epitopes) has been associated with tumor progression and metastasis, and consequently, glycans are potential targets for therapeutic antibodies (2Munkley J. Elliott D.J. Hallmarks of glycosylation in cancer.Oncotarget. 2016; 7 (27007155): 35478-3548910.18632/oncotarget.8155Crossref PubMed Scopus (269) Google Scholar, 3Soliman C. Yuriev E. Ramsland P. Antibody recognition of aberrant glycosylation on the surface of cancer cells.Curr. Opin. Struct. Biol. 2017; 44 (27821276): 1-810.1016/j.sbi.2016.10.009Crossref PubMed Scopus (29) Google Scholar, 4Durrant L.G. Noble P. Spendlove I. Immunology in the clinic review series; focus on cancer: glycolipids as targets for tumour immunotherapy.Clin. Exp. Immunol. 2012; 167 (22235996): 206-21510.1111/j.1365-2249.2011.04516.xCrossref PubMed Scopus (54) Google Scholar, 5Daniotti J.L. Vilcaes A.A. Torres Demichelis V. Ruggiero F.M. Rodriguez-Walker M. Glycosylation of glycolipids in cancer: basis for development of novel therapeutic approaches.Front. Oncol. 2013; 3 (24392350): 30610.3389/fonc.2013.00306Crossref PubMed Scopus (59) Google Scholar). Specifically, glycosphingolipids (GSLs) 3The abbreviations used are: GSLglycosphingolipidBGCglucoseCDRcomplementarity-determining regionFabfragment antigen-bindingFvvariable fragmentGALβ-galactoseGLAα-galactoseMDmolecular dynamicsNGAN-acetylgalactosamineRMSFroot mean square fluctuationSIAsialic acidSLexsialyl-Lewis xSSEAstage-specific embryonic antigenST3GAL2β-galactoside α2,3-sialyltransferase 23Dthree-dimensionalPDBProtein Data BankNPINottingham Prognostic IndexBisTris2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol. are a promising class of glycan antigens for antibody targeting in cancer. Acting as structural components of cell membranes, GSLs are composed of one or more monosaccharides attached to either a sphingoid or a ceramide lipid. GSLs have important roles in immune cell function, membrane signaling, cell adhesion, apoptosis, and cell differentiation (6Zhang T. de Waard A.A. Wuhrer M. Spaapen R.M. The role of glycosphingolipids in immune cell functions.Front. Immunol. 2019; 10 (30761148): 9010.3389/fimmu.2019.00090Crossref PubMed Scopus (63) Google Scholar, 7Wang Z. Wen L. Ma X. Chen Z. Yu Y. Zhu J. Wang Y. Liu Z. Liu H. Wu D. Zhou D. Li Y. High expression of lactotriaosylceramide, a differentiation-associated glycosphingolipid, in the bone marrow of acute myeloid leukemia patients.Glycobiology. 2012; 22 (22411838): 930-93810.1093/glycob/cws061Crossref PubMed Scopus (30) Google Scholar, 8Hakomori S.I. Structure and function of glycosphingolipids and sphingolipids: recollections and future trends.Biochim. Biophys. Acta. 2008; 1780 (17976918): 325-34610.1016/j.bbagen.2007.08.015Crossref PubMed Scopus (213) Google Scholar). Furthermore, GSLs are often overexpressed in various types of human malignancies, including GD2 and GD3 in melanoma (9Nudelman E. Hakomori S. Kannagi R. Levery S. Yeh M.Y. Hellström K.E. Hellström I. Characterization of a human melanoma-associated ganglioside antigen defined by a monoclonal antibody, 4.2.J. Biol. Chem. 1982; 257 (7130176): 12752-12756Abstract Full Text PDF PubMed Google Scholar, 10Scott A.M. Lee F.T. Hopkins W. Cebon J.S. Wheatley J.M. Liu Z. Smyth F.E. Murone C. Sturrock S. MacGregor D. Hanai N. Inoue K. Yamasaki M. Brechbiel M.W. Davis I.D. et al.Specific targeting, biodistribution, and lack of immunogenicity of chimeric anti-GD3 monoclonal antibody KM871 in patients with metastatic melanoma: results of a phase I trial.J. Clin. Oncol. 2001; 19 (11579119): 3976-398710.1200/JCO.2001.19.19.3976Crossref PubMed Scopus (73) Google Scholar) and Globo-H in breast and ovarian cancers (11Chang W.W. Lee C.H. Lee P. Lin J. Hsu C.W. Hung J.T. Lin J.J. Yu J.C. Shao L.E. Yu J. Wong C.H. Yu A.L. Expression of Globo H and SSEA3 in breast cancer stem cells and the involvement of fucosyl transferases 1 and 2 in Globo H synthesis.Proc. Natl. Acad. Sci. U.S.A. 2008; 105 (18685093): 11667-1167210.1073/pnas.0804979105Crossref PubMed Scopus (126) Google Scholar). glycosphingolipid glucose complementarity-determining region fragment antigen-binding variable fragment β-galactose α-galactose molecular dynamics N-acetylgalactosamine root mean square fluctuation sialic acid sialyl-Lewis x stage-specific embryonic antigen β-galactoside α2,3-sialyltransferase 2 three-dimensional Protein Data Bank Nottingham Prognostic Index 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol. Stage-specific embryonic antigens (SSEAs) are a family of glycoconjugate antigens, consisting of SSEA-1 (also known as CD15 or Lewis X) and two related GSLs, SSEA-3 (also known as Gb5Cer) and SSEA-4 (also known as sialyl-Gb5Cer) (12Suzuki S. Ishida T. Yoshikawa K. Ueda R. Current status of immunotherapy.Jpn. J. Clin. Oncol. 2016; 46 (26819277): 191-20310.1093/jjco/hyv201Crossref PubMed Scopus (39) Google Scholar). Both SSEA-3 and SSEA-4 are known cell-surface markers of human embryonic and pluripotent stem cells. Although their biological function is relatively unclear, they are known to be overexpressed in many cancers, including breast and ovarian cancers, and may be very relevant to cancer stem cells (11Chang W.W. Lee C.H. Lee P. Lin J. Hsu C.W. Hung J.T. Lin J.J. Yu J.C. Shao L.E. Yu J. Wong C.H. Yu A.L. Expression of Globo H and SSEA3 in breast cancer stem cells and the involvement of fucosyl transferases 1 and 2 in Globo H synthesis.Proc. Natl. Acad. Sci. U.S.A. 2008; 105 (18685093): 11667-1167210.1073/pnas.0804979105Crossref PubMed Scopus (126) Google Scholar, 13Gang E.J. Bosnakovski D. Figueiredo C.A. Visser J.W. Perlingeiro R.C. SSEA-4 identifies mesenchymal stem cells from bone marrow.Blood. 2007; 109 (17062733): 1743-175110.1182/blood-2005-11-010504Crossref PubMed Scopus (454) Google Scholar, 14Noto Z. Yoshida T. Okabe M. Koike C. Fathy M. Tsuno H. Tomihara K. Arai N. Noguchi M. Nikaido T. CD44 and SSEA-4 positive cells in an oral cancer cell line HSC-4 possess cancer stem-like cell characteristics.Oral Oncol. 2013; 49 (23768762): 787-79510.1016/j.oraloncology.2013.04.012Crossref PubMed Scopus (45) Google Scholar, 15Yu C.-C. Yu C.-H. Chang Y.-C. Aberrant SSEA-4 upregulation mediates myofibroblast activity to promote pre-cancerous oral submucous fibrosis.Sci. Rep. 2016; 6 (27845370): 3700410.1038/srep37004Crossref PubMed Scopus (21) Google Scholar, 16Aloia A. Petrova E. Tomiuk S. Bissels U. Déas O. Saini M. Zickgraf F.M. Wagner S. Spaich S. Sütterlin M. Schneeweiss A. Reitberger M. Rüberg S. Gerstmayer B. Agorku D. et al.The sialyl-glycolipid stage-specific embryonic antigen 4 marks a subpopulation of chemotherapy-resistant breast cancer cells with mesenchymal features.Breast Cancer Res. 2015; 17 (26607327): 14610.1186/s13058-015-0652-6Crossref PubMed Scopus (36) Google Scholar, 17Parte S.C. Batra S.K. Kakar S.S. Characterization of stem cell and cancer stem cell populations in ovary and ovarian tumors.J. Ovarian Res. 2018; 11 (30121075): 6910.1186/s13048-018-0439-3Crossref PubMed Scopus (48) Google Scholar). SSEA-3 and SSEA-4 share a common core glycan structure (Galβ1–3GalNAcβ1–3Galα1–4Galβ1–4Glcβ1), with SSEA-4 containing a terminal sialic acid (Neu5Acα2–3Galβ1–3GalNAcβ1–3Galα1–4Galβ1–4Glcβ1), and both glycan headgroups are linked to a ceramide lipid. SSEA-4 is synthesized from an SSEA-3 precursor through the action of β-galactoside α2,3-sialyltransferase 2 (ST3GAL2). Overexpression of ST3GAL2 has been implicated in poor outcomes for various cancers, including breast and ovarian cancer, which suggests a potential role of SSEA-4 in the development or maintenance of a tumor environment (16Aloia A. Petrova E. Tomiuk S. Bissels U. Déas O. Saini M. Zickgraf F.M. Wagner S. Spaich S. Sütterlin M. Schneeweiss A. Reitberger M. Rüberg S. Gerstmayer B. Agorku D. et al.The sialyl-glycolipid stage-specific embryonic antigen 4 marks a subpopulation of chemotherapy-resistant breast cancer cells with mesenchymal features.Breast Cancer Res. 2015; 17 (26607327): 14610.1186/s13058-015-0652-6Crossref PubMed Scopus (36) Google Scholar, 18Sivasubramaniyan K. Harichandan A. Schilbach K. Mack A.F. Bedke J. Stenzl A. Kanz L. Niederfellner G. Bühring H.-J. Expression of stage-specific embryonic antigen-4 (SSEA-4) defines spontaneous loss of epithelial phenotype in human solid tumor cells.Glycobiology. 2015; 25 (25978997): 902-91710.1093/glycob/cwv032Crossref PubMed Scopus (27) Google Scholar). SSEA-4 has also been shown to be expressed on chemotherapy-resistant and cancer stem cell populations of breast and ovarian tumors (16Aloia A. Petrova E. Tomiuk S. Bissels U. Déas O. Saini M. Zickgraf F.M. Wagner S. Spaich S. Sütterlin M. Schneeweiss A. Reitberger M. Rüberg S. Gerstmayer B. Agorku D. et al.The sialyl-glycolipid stage-specific embryonic antigen 4 marks a subpopulation of chemotherapy-resistant breast cancer cells with mesenchymal features.Breast Cancer Res. 2015; 17 (26607327): 14610.1186/s13058-015-0652-6Crossref PubMed Scopus (36) Google Scholar, 17Parte S.C. Batra S.K. Kakar S.S. Characterization of stem cell and cancer stem cell populations in ovary and ovarian tumors.J. Ovarian Res. 2018; 11 (30121075): 6910.1186/s13048-018-0439-3Crossref PubMed Scopus (48) Google Scholar). As such, SSEA-4 presents an attractive target for antibody-based cancer therapeutics. Currently, the only commercially available mAbs against SSEA-3 or SSEA-4 are MC631, a rat IgM anti-SSEA-3 mAb, and MC813-70, a mouse IgG3 anti-SSEA-4 mAb. However, both antibodies demonstrate some cross-reactivity, with MC813 cross-reacting with multiple sugars, including SSEA-3 and Forssman (19Lou Y.-W. Wang P.-Y. Yeh S.-C. Chuang P.-K. Li S.-T. Wu C.-Y. Khoo K.-H. Hsiao M. Hsu T.-L. Wong C.-H. Stage-specific embryonic antigen-4 as a potential therapeutic target in glioblastoma multiforme and other cancers.Proc. Natl. Acad. Sci. U.S.A. 2014; 111 (24550271): 2482-248710.1073/pnas.1400283111Crossref PubMed Scopus (64) Google Scholar). The potential cytotoxicity of MC631 has not yet been reported, but MC813-70 has been shown to induce complement-dependent cytotoxicity of highly expressing SSEA-4 glioblastoma multiforme cell lines in vitro and suppress tumor growth in vivo (19Lou Y.-W. Wang P.-Y. Yeh S.-C. Chuang P.-K. Li S.-T. Wu C.-Y. Khoo K.-H. Hsiao M. Hsu T.-L. Wong C.-H. Stage-specific embryonic antigen-4 as a potential therapeutic target in glioblastoma multiforme and other cancers.Proc. Natl. Acad. Sci. U.S.A. 2014; 111 (24550271): 2482-248710.1073/pnas.1400283111Crossref PubMed Scopus (64) Google Scholar, 20Kannagi R. Cochran N.A. Ishigami Hakomori S. Andrews P.W. Knowles B.B. Solter D. Stage-specific embryonic antigens (SSEA-3 and -4) are epitopes of a unique globo-series ganglioside isolated from human teratocarcinoma cells.EMBO J. 1983; 2 (6141938): 2355-236110.1002/j.1460-2075.1983.tb01746.xCrossref PubMed Scopus (438) Google Scholar). However, without further engineering, these rodent mAbs are not suitable for human immunotherapy due to the human anti-rodent immune response that limits in vivo efficacy and safety. In response to the overexpression of SSEA-3 and SSEA-4 on the surface of cancer cells, a panel of anti-SSEA mAbs were developed. The lead mAb was a mouse IgG3 known as FG28/11, which has direct cytolytic activity against SSEA-4–positive cells. Following characterization, FG28/11 was chimerized as a mouse-human IgG1 mAb and renamed as ch28/11. 4J. X. Chua, M. Vankemmelbeke, R. S. McIntosh, I. Spendlove, L. G. Durrant, unpublished data. Ch28/11 is currently undergoing preclinical development against a variety of SSEA-4–positive tumor types, and its characterization will be reported elsewhere. In this study, we report the crystal structure of ch28/11 Fab bound to the SSEA-4 glycan headgroup. From three crystal forms, five independent ch28/11 Fab:SSEA-4 complexes were determined at resolutions between 1.5 and 2.7 Å. All ch28/11 Fab:SSEA-4 complexes displayed nearly identical three-dimensional (3D) structures for the glycan-antibody interface. Our findings from crystallography and molecular dynamics simulations explain the basis for ch28/11 mAb specificity for SSEA-4 and identify a critical role for the terminal sialic acid, which is not present in SSEA-3, in antibody recognition. ELISAs were used to reveal strong binding of FG28/11 to SSEA-4 and a complete absence of cross-reactivity with closely related glycans SSEA-3, Globo-H, and Forssman. All of the glycan antigens contain a GalNAcβ1–3Galα1–4Galβ1–4Glc core with varying terminal sugar residues (sialic acid in SSEA-4) (Fig. 1). For comparison, anti-SSEA-3 antibody MC631, anti-SSEA-4 antibody MC813-70, and anti-Forssman antibody M1/87 were also tested. MC631 bound well to SSEA-3 and SSEA-4 with some binding to Globo-H, and MC813 bound well to SSEA-3, SSEA-4, and Forssman. As expected, MC1/87 only bound to Forssman, and the negative control SLex-binding antibody did not interact with any of the four glycans (Fig. 1A). Therefore, FG28/11 was shown to be monospecific for SSEA-4, whereas MC631 and MC813 both displayed cross-reactivity to related glycans. The FG28/11 mAb was further assessed by glycan array, where binding was tested against 585 mammalian glycans. FG28/11 only bound to SSEA-4 with no cross-reactivity with any of the other glycans tested, whereas MC631 bound to SSEA-4, SSEA-3, and Galβ1–3GalNAcβ1–3Gal (Fig. S1). In addition, tumor cell binding was characterized by flow cytometry of a panel of cancer cell lines as well as immunostaining of tumors. FG28/11 showed strong binding to one ovarian cancer cell line (SKOV3) and weak to moderate binding to two other ovarian cancer cell lines (OVCAR5 and IGROV1) as well as two breast cancer cell lines (T47D and MCF7), but was negative for binding to colorectal cancer cell lines (COLO205 and HCT15) (Fig. 1B and Fig. S2). This was confirmed by immunostaining, where FG28/11 stained 28% (223 of 798) of breast and 14% (41 of 289) of ovarian tumors (Fig. 1C). Fab was produced by papain digestion of ch28/11 IgG and protein A purification, and the purity and uniformity were confirmed by Coomassie-stained SDS-PAGE and size-distribution analysis of dynamic light scattering (DLS) data (Fig. 2, A and B). As expected, the purified Fab was a disulfide-linked dimer of 50 kDa consisting of paired heavy (H) and light (L) chains of ∼25-kDa sizes. By DLS, the Fab sample consisted of a single population with an average hydrodynamic diameter (DH) of 6.6 nm (compared with IgG with an average DH of 13.8 nm). Thus, ch28/11 Fab was determined to be suitable for structural studies. To elucidate the structural basis for recognition of SSEA-4 by the ch28/11 mAb, crystal structures of the Fab:glycan headgroup complex were determined. Co-crystals of ch28/11 Fab and SSEA-4 hexasaccharide were obtained in three different conditions, resulting in three crystal forms, with the highest-resolution structure determined at 1.5 Å (Rwork/Rfree = 0.172/0.197) from a tetragonal P41212 crystal (Fig. 2C). The monoclinic P21 crystal structure was refined at 1.9 Å resolution (Rwork/Rfree = 0.176/0.230). Both the tetragonal and monoclinic crystals had a single ch28/11 Fab:SSEA-4 complex in the asymmetric unit. The hexagonal P62 crystal structure was refined at 2.7 Å resolution (Rwork/Rfree = 0.169/0.246) with three ch28/11 Fab:SSEA-4 complexes in the asymmetric unit (identified as complex 1, 2, and 3). X-ray diffraction data collection and crystallographic refinement statistics are presented in Table 1.Table 1Data collection and crystallographic refinement statisticsParameterch28/11 Fab:SSEA-4ch28/11 Fab:SSEA-4ch28/11 Fab:SSEA-4ch28/11 FabData collectionSpace groupP41212P21P62P21Unit cell dimensions (Å)a = 67.8, b = 67.8, c = 234.0a = 37.5, b = 69.1, c = 97.7a = 170.1, b = 170.1, c = 95.4a = 75.2, b = 69.6, c = 93.9Unit cell angles (degrees)α = 90.0, β = 90.0, γ = 90.0α = 90.0, β = 94.5, γ = 90.0α = 90.0, β = 90.0, γ = 120.0α = 90.0, β = 98.3, γ = 90.0Resolution range (Å)50–1.5 (1.6–1.5)50–1.9 (2.0–1.9)50–2.7 (2.8–2.7)50–2.5 (2.6–2.5)Number of unique reflections84,916 (13,419)39,580 (6187)41,405 (6561)33,478 (5126)Data completeness (%)99.9 (99.2)99.3 (96.8)99.7 (98.6)98.1 (93.9)Average multiplicity13.3 (13.4)3.4 (3.4)5.3 (5.2)3.4 (3.5)R-factor0.10 (0.46)0.06 (0.48)0.13 (0.69)0.06 (0.41)Rmeas0.10 (0.48)0.07 (0.57)0.14 (0.77)0.07 (0.48)Mean I/σ (I)15.8 (4.2)10.1 (1.9)10.6 (2.3)14.4 (2.6)Crystallographic refinementRwork0.1720.1760.1690.164Rfree0.1970.2300.2460.239Average B-factor from Wilson plot (Å2)19.534.452.752.1Root mean square deviation from ideal valuesBond lengths (Å)0.0180.0180.0120.014Bond angles (degrees)1.91.91.71.7Ramachandran plot values (%)Favored regions98.697.996.296.8Allowed regions1.22.13.43.0Outliers0.200.40.2Average B-factor (Å2)Protein atoms23.141.654.958.8Carbohydrate atoms26.438.361.2Water37.247.340.951.0 Open table in a new tab The electron density maps allowed fitting of the L and H polypeptide chains for each Fab structure, except for one disordered loop in the H-chain constant domain, which is distant from the antigen-binding site and often missing from Fab structures (21Fan Z.C. Shan L. Goldsteen B.Z. Guddat L.W. Thakur A. Landolfi N.F. Co M.S. Vasquez M. Queen C. Ramsland P.A. Edmundson A.B. Comparison of the three-dimensional structures of a humanized and a chimeric Fab of an anti-γ-interferon antibody.J. Mol. Recognit. 1999; 12 (10398393): 19-3210.1002/(SICI)1099-1352(199901/02)12:1%3C19::AID-JMR445%3E3.0.CO;2-YCrossref PubMed Scopus (27) Google Scholar). The missing CH1 residues were 133–137 for the two higher-resolution structures, 132–137 for complex 1 and 2, and 133–134 for complex 3 from the lowest-resolution structure (sequential numbering). In all structures, there is strong electron density corresponding to binding site residues and five of the six sugar residues of SSEA-4 (slightly weaker electron densities were associated with the terminal glucose), but the entire hexasaccharide was fitted into electron density maps (Fig. 2D and Fig. S3). The 3D structures for each of the five versions of ch28/11 Fab:SSEA-4 complexes were very similar, with SSEA-4 positioned to fill a deep groove-shaped binding site (Fig. 3). The Fab has a quaternary structure typical of most antibodies with the binding site formed by six complementarity-determining regions (CDRs), three from the L chain (identified as L1, L2, and L3), and three from the H chain (identified as H1, H2, and H3). The Fv portions (comprised of VL and VH; sequences presented in Fig. S4) were shown to be nearly identical in 3D structure, with most of the minor variations occurring in loops surrounding the binding groove (CDRs and some framework region loops). The constant regions (comprised of CL and CH1) also showed some differences (Fig. 3, A and B). A surface view of the 1.5 Å resolution structure illustrates that the center of the binding site is occupied by the H3 loop, with the SSEA-4 hexasaccharide bent around this loop in a horseshoe-like conformation (Fig. 3C). In the ch28/11 binding site, the SSEA-4 hexasaccharide is surrounded by all six CDRs (L1–3 and H1–3) and snugly fills the entire binding groove (Fig. 3 and Fig. S5). This snug fit corresponds to a stable glycan conformation where the galactose (GAL2, GLA4, and GAL5), GalNAc (NGA3), and terminal sialic acid (SIA1) residues all display very little movement between structures (Fig. 4; see figure for monosaccharide codes). In contrast, the terminal glucose (BGC6) that would normally be attached to the ceramide lipid exhibits more flexibility in its position (Fig. 4, A and B), which corresponds with it being the most solvent-exposed portion of the SSEA-4 glycan (Fig. 3). Further examining the conformations of the five SSEA-4 hexasaccharide structures in a Ramachandran-like plot showed clustering of the φ and ψ angles of the various glycosidic linkages (Fig. 4C). The same glycosidic conformations were observed for SIA1-GAL2 and GLA4-GAL5, as demonstrated by similar φ and ψ angles. GAL2-NGA3 and NGA3-GLA4 glycosidic linkages have similar ψ but different φ angles, indicating different conformations for these linkages. In line with its position in the binding site and flexibility, a distinct cluster with a larger variation of φ and ψ angles occurred for the GAL5-BGC6 glycosidic linkage (Fig. 4C). In each of the five structures, the Fab participates in numerous hydrogen-bond and van der Waals interactions with each of the six sugar residues of SSEA-4 (Fig. 5 and Fig. S6). All antibody residues involved in binding SSEA-4 are located within the L- and H-chain CDRs, except for one residue (His-33L), which forms a single hydrogen bond with the glycan ligand in each structure (see Fig. S4 for glycan contact residues and numbering schemes). A total of 791.5 Å2 of the SSEA-4 glycan headgroup is buried in the ch28/11 binding groove. Contributions to the interaction of ch28/11 mAb and glycan for the constituent monosaccharides are as follows: SIA1, 202.0 Å2; GAL2, 117.5 Å2; NGA3, 167.6 Å2; GLA4, 141.5 Å2; GAL5, 118.7 Å2; BCG6, 44.2 Å2. Within the expansive ch28/11 binding groove, contacts are made with each residue of SSEA-4. Sialic acid (SIA1) dominates the interaction with the antibody by participating in more contacts compared with any other residue in the glycan chain. It is secured in place by hydrogen bonds between the nitrogen on the N-acetyl moiety and Ser-31H, the O8 hydroxyl and Gly-33H, and from the carboxylate group to Gly-53H and Asp-54H. The GalNAc (NGA3) of SSEA-4 is involved in hydrogen bonds from the O4 and O6 hydroxyls to Tyr-93L and Ala-90L, respectively. The α-galactose (GLA4) engages in one hydrogen bond from the O4 hydroxyl to Ala-90L and two hydrogen bonds from the O6 hydroxyl to His-33L and Gly-101H. The next galactose in the chain (GAL5) is involved in three hydrogen bonds from the O2 and O3 hydroxyls to Asp-49L. In contrast, galactose at position 2 (GAL2) and glucose at position 6 (BGC6) do not engage in direct hydrogen bonding with the ch28/11 mAb. SSEA-4 is further stabilized by numerous van der Waals interactions between binding site residues and each monosaccharide in the glycan chain (Fig. 5). The ch28/11:SSEA-4 interface (at 1.5 Å resolution) has a well-defined solvent structure with 23 ordered water molecules surrounding the SSEA-4 glycan. Sialic acid (SIA1) again dominates the interactions, with eight ordered waters involved in stabilizing its conformation in the binding groove. Other monosaccharide units of SSEA-4 are involved in fewer water-mediated interactions: GAL2, 5 waters; NGA3, 4 waters; GLA4, 4 waters; GAL5, 3 waters; BCG6, 3 waters. A solvent-accessible surface representation shows SSEA-4 snugly fitted into the binding groove of the Fab, folding around the centrally located H3 loop in a horseshoe-shaped conformation, with the GalNAc hydrophobic face stacking against the Tyr-102H side chain to slot into its position in the deepest part of the binding groove (Fig. 6, A and B). Thus, the ch28/11 binding groove is totally occupied by the SSEA-4 glycan headgroup and the ordered water molecules to present an almost continuous convex surface at the end of the Fab. The absence of protruding CDR loops would permit the intact ch28/11 mAb to bind closely to the cell membrane, where the SSEA-4 hexasaccharide is directly anchored by its ceramide tail. This could be relevant for the direct cytotoxic activity of the ch28/11 antibody. To further examine the SSEA-4 binding mechanism, the crystal structure of the unliganded ch28/11 Fab was determined. X-ray diffraction data were obtained from a monoclinic P21 crystal at 2.5 Å resolution (Rwork/Rfree = 0.164/0.239; Table 1). There were two Fabs in the asymmetric unit, both with clear electron density maps in the binding site (Fig. S3). Differences in 3D structure of the binding sites are mainly due to crystal packing, where the variable domains of Fab 2 are packed against the light chain of the constant domain of Fab 1. To compare free with bound structures, the Fv regions from the unliganded Fabs were superimposed on the 1.5 Å resolution ch28/11 Fab:SSEA-4 structure. For the most part, there is little variation in binding site residues, with the exception of CDR3 of the heavy chain. In the free Fab structures, this loop appears to collapse into the binding groove, with the Tyr-102H side chain in particular shifting inward to fill part of the groove (Fig. 7 and Fig. S5). A multitude of small conformational differences occur in all H-chain CDRs around the site of interaction with sialic acid of SSEA-4. Overall, these results suggest that SSEA-4 recognition by the ch28/11 mAb represents an induced fit binding mechanism. Molecular dynamics (MD) simulations were performed to examine why the ch28/11 mAb does not cross-react with SSEA-3, which, compared with SSEA-4, is only missing the terminal sialic acid residue. Carbohydrate flexibility can be assessed by examining root mean square fluctuation (RMSF) of carbohydrate residues, which is a time-averaged measure of atom deviation from a reference position. RMSF analysis of carbohydrate residues, averaged over 1.5 μs of MD simulations, show that the central galactose and GalNAc residues (numbered 3–5) are highly stable in the binding site for both SSEA-3 and SSEA-4. Additionally, in both glycans, the terminal glucose (BCG6) was mobile in the binding site, although more so for SSEA-3. For SSEA-3, the galactose at position 2 (GAL2) displayed high RMSF values, whereas the same residue in SSEA-4 was highly stable across the MD simulations (Fig. 8, A and B). This is further evident when examining the φ versus ψ plots of each glycosidic linkage. The linkage between galactose (GAL2) and GalNAc (NGA3) is more flexible in SSEA-3, displaying a range of conformations when compared with SSEA-4. It should also be noted that the linkage between galactose (GAL5) and the terminal glucose (BGC6) is also more flexible in SSEA-3 when compared with SSEA-4 (Fig. 8, C and D). When overlaying the two main ligand conformations of SSEA-3 taken from MD simulations, there is good alignment of the three central carbohydrate residues, and the largest difference in conformation occurs with the GAL2 and BCG6 monosaccharides (Fig. 8E). Therefore, it can be concluded that the ch28/11 binding groove is more stable in the presence of SSEA-4 due to the presence of the terminal sialic acid residue absent in SSEA-