A Mutant Chaperone Converts a Wild-Type Protein into a Tumor-Specific Antigen

See allHide authors and affiliations

Science  13 Oct 2006:
Vol. 314, Issue 5797, pp. 304-308
DOI: 10.1126/science.1129200


Monoclonal antibodies have become important therapeutic agents against certain cancers. Many tumor-specific antigens are mutant proteins that are predominantly intracellular and thus not readily accessible to monoclonal antibodies. We found that a wild-type transmembrane protein could be transformed into a tumor-specific antigen. A somatic mutation in the chaperone gene Cosmc abolished function of a glycosyltransferase, disrupting O-glycan Core 1 synthesis and creating a tumor-specific glycopeptidic neo-epitope consisting of a monosaccharide and a specific wild-type protein sequence. This epitope induced a high-affinity, highly specific, syngeneic monoclonal antibody with antitumor activity. Such tumor-specific glycopeptidic neo-epitopes represent potential targets for monoclonal antibody therapy.

Tumor antigens expressed exclusively on the surface of cancer cells are ideal targets for monoclonal antibodies (mAb) because they are accessible, unlikely to have induced immune tolerance, and, when targeted, do not cause destruction of normal tissue (autoimmunity) (1). During tumor development, the host mounts humoral and cellular immune responses against the tumor (2). Serum from cancer patients contains high-affinity immunoglobulin G (IgG) antibodies specific for their own cancer cells but not reactive to non-neoplastic cells (3). However, the genetic origin and biochemical nature of many of these tumor-specific epitopes are incompletely understood (4). Many tumor-specific antigens result from somatic mutations that change the amino acid sequence, occur mostly in intracellular proteins, and are not readily accessible to antibodies (5). Few tumor-specific mutations have been identified in the extracellular domain of membrane proteins (6), and yet these are precisely the targets most likely to be visible to mAb.

Ag104A is a highly aggressive fibrosarcoma that arose spontaneously in an aging (female) mouse. 237 mAb (IgG2a)—a syngeneic, high-affinity mAb—shows exquisite specificity for Ag104A and does not react with a second spontaneous tumor (Ag104B) isolated from the same mouse, autologous non-neoplastic tissue (heart or lung fibroblasts, Ag104HLF), or any other tumor cell line tested (7, 8). To identify the 237 mAb epitope, the antigen was purified from Ag104A cells by 237 mAb affinity chromatography and shown by amino acid sequencing to be OTS8, a sialomucin-like transmembrane glycoprotein of 172 amino acids also known as PA2.26, Aggrus, or T1alpha (911) (Fig. 1A). OTS8 is predominantly expressed by alveolar type I cells and overexpressed in murine and human tumors of various histologic origin (11). OTS8 is thought to be involved in tumor progression and invasion, but its exact biological function is still unknown. Using a rat mAb specific for OTS8, we confirmed that Ag104A indeed expressed OTS8, as did Ag104B, Ag104HLF, and other tumor cell lines known to express OTS8 (12) (Fig. 1B). However, of all the cell lines expressing OTS8, only Ag104A-OTS8 reacted with 237 mAb. Thus, 237 mAb was specific for a form of OTS8 found only in Ag104A. Immunoblotting of embryonic tissue and adult lung tissue showed OTS8 expression but no 237 mAb antigenicity (Fig. 1C). Thus, the 237 mAb epitope was not a differentiation or oncofetal antigen but truly tumor-specific.

Fig. 1.

237 mAb showed absolute specificity to OTS8 expressed by Ag104A. (A) Amino acid sequence of OTS8; signal sequence (amino acids 1 to 24, blue), extracellular domain (ECD) (amino acids 25 to 134, red), transmembrane domain (amino acids 135 to 162, green), and the cytoplasmic region (amino acids 163 to 172, purple); OTS8-ECD contains one putative N-glycosylation site at Asn60 (♦) and 27 putative O-glycosylation sites (⚫). Peptide regions that were directly sequenced are underlined. (B) The 237 mAb recognized OTS8 expressed by Ag104A but not OTS8 expressed by other cell lines. Flow cytometric and immunoblotting of Ag104A and other cell lines for OTS8 expression and 237 mAb antigenicity. A rat IgG antibody was used as isotype control. (C) The 237 mAb antigen was not a differentiation or oncofetal antigen. OTS8 was highly expressed (45 kD) during embryogenesis (embryonic day 15) and in adult lung but did not show any reactivity to 237 mAb. Ag104A and Ag104B WCL were used as controls. (D) Confocal microscopy. OTS8-EGFP showed characteristic membrane distribution in both Ag104A and Ag104B, whereas unfused EGFP was diffusely localized. Single confocal planes (0.3 μm) are shown. Each panel is 65-μm wide. (E) 237 mAb bound to wild-type OTS8 in Ag104A. The cell lines used in (B) were retrovirally infected with pMFG OTS8-EGFP. Although the fusion protein was expressed in all cell lines as shown by anti-EGFP and anti-OTS8 immunoblotting, 237 mAb only recognized OTS8-EGFP in Ag104A WCL. OTS8 (endogenous and fusion protein) expressed in Ag104A had a molecular weight 15 to 20 kD lower than non-Ag104A OTS8.

The lack of reactivity of 237 mAb with Ag104B and Ag104HLF suggested that the antigen was not the result of a germline mutation but rather a tumor-specific mutation in Ag104A. Surprisingly, sequencing of Ag104A-OTS8 cDNA did not reveal any mutations (fig. S1). To confirm that 237 mAb did bind wild-type OTS8, we overexpressed OTS8 as a fusion protein with enhanced green fluorescent protein (EGFP). OTS8-EGFP showed appropriate localization to the membrane in both Ag104A and Ag104B (Fig. 1D). Once again, 237 mAb only recognized OTS8 when expressed in Ag104A cells (Fig. 1E), leading us to hypothesize that the unique specificity of the antigen was the result of an Ag104A-specific posttranslational modification. Moreover, the large difference in molecular weight seen between Ag104A-OTS8 and the nonantigenic OTS8 expressed in other cell lines could readily be accounted for by a difference in glycosylation. To test this idea, carbohydrate-specific oxidation with sodium periodate was carried out (13). 237 mAb immunoblotting of Ag104A OTS8-EGFP whole-cell lysate (WCL) showed positive staining on the nontreated nitrocellulose membrane, but the periodate-treated membrane had completely lost 237 mAb reactivity (Fig. 2A). Thus, carbohydrate structures of Ag104A-OTS8 were crucial for 237 mAb binding.

Fig. 2.

Altered glycosylation of OTS8 created an antigenic carbohydrate moiety at amino acid Thr77. (A) Oxidation of OTS8 carbohydrate groups with sodium periodate led to complete loss of 237 mAb immunoreactivity; the blots were reprobed with mAb to EGFP to confirm the presence of intact OTS8-EGFP protein. (B) Thr77 carried the antigenic carbohydrate moiety. Recombinant, His-tagged epitope fragments were cloned into pMFG single sequence (SS)–multiple cloning site (MCS)–(His)6–internal ribosomal entry site (IRES)–EGFP for expression in Ag104A and analyzed by anti-His and 237 mAb immunoblotting. Point mutation of Thr77 to Ser or Ala led to loss of 237 mAb antigenicity. (C) OTS8 expressed in Ag104B was a sialylated Core 1 O-glycan, but Ag104A-OTS8 was not. Immunoprecipitated (IP) OTS8-EGFP from Ag104A and Ag104B was treated with a combination of five deglycosylating enzymes and assayed by anti-EGFP and 237 mAb immunoblotting. Lane 1, N-glycanase; lane 2, O-glycanase; lane 3, O-glycanase plus sialidase A; lane 4, sialidase A; and lane 5, β(1-4) galactosidase and glucosaminidase. Ag104A-OTS8-EGFP did not change in size or 237 mAb binding. Ag104B-OTS8-EGFP decreased 20 kD but still remained 237 mAb-negative after treatment with sialidase A plus O-glycanase. (D) Ag104A-OTS8 contained primarily GalNAc. Annotated matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrum of permethylated O-glycans was obtained by beta-elimination from Ag104A-OTS8. Shown are the mass-to-charge (m/z) values of selected peaks and relative quantity percentages (mass to mass) of the O-linked glycan structures. Hex, hexose; HexNAc, N-acetylhexose; NeuNAc, N-acetylneuraminic acid (also known as sialic acid). (E) Primary O-glycan carbohydrate structures present in OTS8 expressed in Ag104A and Ag104B. The normally glycosylated OTS8 carried mainly Core 1 structures; Ag104A-OTS8 contained predominantly GalNAc.

To map the precise location of the carbohydrate epitope within the extracellular domain of OTS8 (OTS8 ECD), we developed a recombinant expression system (Fig. 2B). Regions of the OTS8 ECD were overexpressed in Ag104A as secreted His-tagged fragments and analyzed by 237 mAb immunoblotting. Fragments 25 to 75, 75 to 134, and 25 to 134 were expressed and secreted by Ag104A, but only 25 to 134 and 75 to 134 carried the antigenic carbohydrate epitope (Fig. 2B). Sequential addition of 10 amino acids to the 237 mAb-negative 25-to-75 fragment revealed that the epitope was located between amino acids 75 and 85 (Fig. 2B). This peptide, RGTKPPLEELS (14), harbors two potential O-glycosylation sites, Thr77 and Ser85. Whereas Ser85→Ala85 (S85A) and S85T point mutants remained 237 mAb-positive, T77A and T77S mutants were 237 mAb-negative (Fig. 2B); thus, the antigenic carbohydrate epitope was on Thr77. Given that 25 to 77 and 25 to 80 were 237 mAb-negative, amino acids 77 to 84 may have been needed for epitope creation and/or 237 mAb recognition (Fig. 2B).

We next compared the glycosylation profiles of nonantigenic OTS8 from Ag104B and antigenic OTS8 from Ag104A through enzymatic deglycosylation and found that nonantigenic OTS8 contained Core 1 [galactosyl-β(1-3)-N-acetylgalactosamine] with terminal sialic acid residues, whereas Ag104A-OTS8 did not (Fig. 2C). Carbohydrate analysis revealed that Ag104A-OTS8 contained primarily the monosaccharide N-acetylgalactosamine (GalNAc) (82%) (Fig. 2D and fig. S2). GalNAcα1-Ser/Thr, normally the precursor for longer O-glycans such as Core 1 (Fig. 2E), is also known as Tn antigen, an oncodevelopmental cancer–associated antigen frequently overexpressed on various cancers (15) and associated with several autoimmune diseases (16). Jurkat cells, a human T cell leukemia line known to express Tn (17), acquired 237 mAb antigenicity after being transduced to overexpress OTS8-EGFP (Fig. 3 and fig. S3). Thus, 237 mAb recognized a glycopeptidic epitope consisting of at most OTS8 amino acids 75 to 84 and GalNAc at Thr77.

Fig. 3.

The 237 mAb recognized a glycopeptidic epitope composed of GalNAc on Thr77 of OTS8. Flow cytometry of Jurkat, Jurkat EGFP, and Jurkat OTS8-EGFP for OTS8 expression and 237 mAb antigenicity. Although the parental Jurkat cells and Jurkat EGFP did not express OTS8 or the 237 mAb epitope, Jurkat OTS8-EGFP cells became positive for OTS8 expression and 237 mAb antigenicity; Ag104A was used as a control.

Because Ag104A contains GalNAc but no Core 1 structures, the normal addition of galactose to GalNAcα1-Ser/Thr was not occurring. The enzyme that catalyzes this linkage is Core 1 β1,3-galactosyltransferase (C1β3GALT) (18). In humans, C1β3GALT activity requires the coexpression of a molecular chaperone, Cosmc (Core 1 β3Gal-T–specific molecular chaperone) (19). A mouse gene (GenBank accession number NM_021550) that shows 91.5% homology to human Cosmc is X-linked (Xq23) and is predicted to encode a type II membrane protein of 316 amino acids (20). In Jurkat cells, a mutation in Cosmc leads to loss of C1β3GALT activity and ultimately to Tn overexpression (19). To examine whether defects in C1β3GALT or Cosmc could account for the abundance of Tn in Ag104A and subsequently for the creation of the 237 mAb epitope, C1β3GALT and Cosmc cDNA were sequenced. Whereas C1β3GALT cDNA did not have any mutations, Cosmc cDNA from Ag104A contained an in-frame deletion of nucleotides 509 to 587 predicted to delete 26 amino acids in the C-terminal region of the ECD (Fig. 4A). Cosmc cDNA from Ag104B and Ag104HLF was wild type, so the Ag104A-Cosmc deletion was somatic and tumor-specific. Sequencing of Ag104A genomic DNA did not show any wild-type Cosmc, suggesting a loss of heterozygosity at this locus (fig. S4). We reexpressed wild-type Cosmc in Ag104A to see if this would repair the defect in Ag104A. Notably, Ag104A cells that overexpressed wild-type Cosmc were 237 mAb-negative (Fig. 4, B and C), and OTS8 from these cells was the same size as Ag104B-OTS8 (Fig. 4C), indicating that it was now fully glycosylated. Moreover, we found that another spontaneous murine tumor, the neuroblastoma Neuro2A, also contained a Cosmc mutation resulting in Tn overexpression (fig. S5). Mutations that change amino acid sequence of a single protein create only one potential antigen, but the Cosmc mutation in Ag104A altered glycosylation globally on many cell surface proteins (fig. S6), creating many different possible glycopeptidic epitopes. Ag104A tumor-bearing mice developed Ag104A-OTS8–specific IgG (fig. S7), demonstrating the immunogenicity of aberrantly glycosylated OTS8, even in syngeneic hosts. mAb against such tumor-specific glycopeptidic epitopes can have therapeutic efficacy, as shown by the substantial anti-Ag104A activity of 237 mAb in vivo (fig. S8).

Fig. 4.

An Ag104A-specific deletion mutation in the chaperone Cosmc led to creation of the 237 mAb epitope. (A) The predicted amino acid sequences of mouse Cosmc obtained from Ag104A, Ag104B, and Ag104HLF cDNA sequencing. Ag104A contained a deletion of nucleotides 509 to 587, resulting in the loss of 26 amino acids. (B) Repair of the mutation caused loss of the 237 mAb epitope. Ag104A was retrovirally infected with pMFG WTCosmc-IRES-EGFP (IE) or the control vector pMFG IRES-EGFP. Cells were analyzed by flow cytometry using an antibody to OTS8 or 237 mAb. Ag104A cells overexpressing WTCosmc expressed OTS8 but were 237 mAb-negative. (C) OTS8 from Ag104A WTCosmc–IRES-EGFP had the same molecular weight as Ag104B-OTS8 and was not recognized by 237 mAb.

Qualitative and quantitative changes in O- and N-glycosylation are consistent features of malignancies (2123). Thus, membrane proteins with aberrant carbohydrate moieties are logical targets for active and passive immunotherapy, and much effort and progress has been made to identify tumor-associated carbohydrate antigens (22, 23) and glycopeptidic epitopes (24, 25), to elucidate the mechanisms leading to their creation (19, 2628), and to augment their antigenicity and immunogenicity (29, 30). Here, we found that a tumor-specific somatic mutation in a chaperone gene abolished the activity of a glycosyltransferase, disrupted O-glycan Core 1 synthesis, and ultimately created a glycopeptidic neo-epitope on a wild-type protein. The combination of a monosaccharide and a wild-type peptide sequence produced a tumor-specific antigen that induced the generation of a syngeneic, high-affinity mAb. Mutations in Cosmc have now been found in multiple murine (Ag104A and Neuro2A) and human tumors (LSC, Jurkat) (19, 26) and in patients with Tn syndrome (16), suggesting that this pathway may be commonly targeted in cancer cells, leading to the overexpression of Tn and subsequently to the creation of tumor-specific glycopeptidic neo-epitopes that can be targeted by mAb. The existing paradigm for tumor-specific antigen creation (from mutant gene to mutant protein to tumor antigen) must be expanded to include antigens generated by tumor-specific, posttranslational modifications of wild-type proteins.

Supporting Online Material

Materials and Methods

Figs. S1 to S8


References and Notes

View Abstract

Navigate This Article