HEG1 is a novel mucin-like membrane protein that serves as a diagnostic and therapeutic target for malignant mesothelioma

The absence of highly specific markers for malignant mesothelioma (MM) has served an obstacle for its diagnosis and development of molecular-targeting therapy against MM. Here, we show that a novel mucin-like membrane protein, sialylated protein HEG homolog 1 (HEG1), is a highly specific marker for MM. A monoclonal antibody against sialylated HEG1, SKM9-2, can detect even sarcomatoid and desmoplastic MM. The specificity and sensitivity of SKM9-2 to MM reached 99% and 92%, respectively; this antibody did not react with normal tissues. This accurate discrimination by SKM9-2 was due to the recognition of a sialylated O-linked glycan with HEG1 peptide. We also found that gene silencing of HEG1 significantly suppressed the survival and proliferation of mesothelioma cells; this result suggests that HEG1 may be a worthwhile target for function-inhibition drugs. Taken together, our results indicate that sialylated HEG1 may be useful as a diagnostic and therapeutic target for MM.

Scientific RepoRts | 7:45768 | DOI: 10.1038/srep45768 These markers often have difficulty discriminating epithelioid MM from metastatic tumors or sarcomatoid MM from some sarcomas.
A mucin-like membrane protein, which is a membrane-anchored protein modified with many glycans, can be a good cancer-related antigen 12 . Immature glycans are produced by irregular processes of carbohydrate chain synthesis in tumor cells, and are attached in clusters on mucin-like proteins [12][13][14] . Some monoclonal antibodies (mAbs) recognizing irregular glycan clusters are used clinically to measure serologic tumor markers 12,15 . Moreover, the combined recognition of a mucin-like protein and its irregular glycan attachment can detect malignant tumor cells accurately 16 . A tumor-specific mucin-like membrane protein can also become a target for antibody-utilized immunotherapy 17 or drug inhibition of cell proliferation [18][19][20][21] . If a mucin-like membrane protein with characteristic glycosylation is found on MM, it could become a specific target for accurate MM diagnosis or molecular-targeting therapy. However, changes in post-translational modifications are difficult to discover in cyclopedic analyses of gene expression. In addition, a large and heterogeneous molecular size, multiple charge states, and many heterogeneous glycans complicate detection of mucin-like proteins via general proteome analysis. Most mucin-like cancer antigens have been found by establishing tumor-specific mAbs. There are no shortcuts to find a tumor-specific mucin-like protein.
Protein HEG homolog 1 (HEG1) was first reported as the heart of glass gene regulating the concentric growth of the zebrafish heart 22 . The mouse HEG1 gene has been linked to cardiovascular organ development 23 . However, the function and structure of HEG1 has remained unclear. Here we show that sialylated HEG1, which we identified as a novel mucin-like membrane protein, is indeed a mesothelioma-related antigen, and that HEG1 expression supports the survival and proliferation of mesothelioma cells. Sialylated HEG1 may be a worthwhile target for MM diagnosis and therapy.
High specificity of SKM9-2 antigen to MM; insignificant expression of SKM9-2 antigen in non-mesothelioma tumors and non-neoplastic tissues. We investigated the expression of SKM9-2 antigen on non-mesothelioma tumors by using tissue microarrays of 24 primary tumors. SKM9-2 antigen was expressed in only 3/310 cases of unrelated tumors ( Supplementary Fig. 2). The MM specificity of SKM9-2 antigen reached 99%, which was the highest specificity that we measured among MM markers (Table 2). Although nuclear WT-1 is a good MM marker, WT-1 protein was also expressed in the cytoplasm of non-tumor cells and other tumors ( Supplementary Fig. 3). Because of the cytoplasmic staining for WT-1, it was often difficult to clearly discern nuclear staining for WT-1 ( Supplementary Fig. 3, bottom panel). Relative to WT-1, the SKM9-2 antigen exhibited better visibility.
SKM9-2 antigen was not detected in major organs (Fig. 1c). The low expression of SKM9-2 antigen by normal tissues is summarized in Supplementary  Table 1. Expression rates of marker antigens in MPM. The histologic type of MPM was classified by pathological diagnosis. Intensity and proportion of staining of MPM cells were evaluated in the entire microscopic field of the specimen. Cases were defined as positive if the proportion score was more than 0. In the immunostaining of calretinin or WT-1, staining in the nucleus, but not the cytoplasm, was considered a positive sample.
Scientific RepoRts | 7:45768 | DOI: 10.1038/srep45768 SKM9-2 often stained the capillary endothelium in near-tumor angiogenesis. This mAb was also reactive on some activated mesothelial cells, especially from a pneumothorax case. In six cases of fibrous pleurisy, spindle cells were not stained with mAb SKM9-2, whereas reactive mesothelial cells in two cases were stained with the antibody. These results suggest that SKM9-2 antigen has low expression in non-mesothelioma cells and normal tissues, except in a part of the capillary endothelium, and reactive mesothelial cells.
In summary, the mAb SKM9-2 can detect MM with 99% specificity and 92% sensitivity on histopathological specimens, which is better performance than that provided by other major MM diagnostic markers. The mAb SKM9-2 may be a good MM-specific marker for pathological diagnosis. mAb SKM9-2 recognizes a sialylated HEG1. SKM9-2 antigen was detected in lysates of several MPM cell lines as ~400 kDa bands on western blots (Fig. 2a). Flow cytometric analysis (Fig. 2b) suggested that SKM9-2 antigen was expressed on the cell surface. Neuraminidase treatment eliminated recognition of mAb SKM9-2, whereas treatment with peptide N-glycosidase F or O-glycosidase alone, which cannot digest sialylated O-linked glycans, did not affect the reactivity of mAb SKM9-2 (Fig. 2c). Proteinase K treatment also eliminated binding of mAb SKM9-2 (Fig. 2c). We reason that the mAb SKM9-2 would recognize a sialylated O-linked glycan in conjunction with peptide sequences of a large mucin-like membrane protein.
SKM9-2 antigen was purified from ACC-MESO-4, a mesothelioma cell line, by precipitation under acidic conditions and column chromatography (see Supplementary Methods). SKM9-2 antigen was fractionated as a very large molecule near the void volume on size-exclusion chromatography ( Fig. 2d and g), eluted as a broader peak in anion exchange chromatography (Fig. 2e), and purified by WGA-agarose to which sialomucin bound (Fig. 2f). These biochemical properties suggest that SKM9-2 antigen is a heavily sialylated mucin. A 400 kDa SDS-PAGE band stained with Coomassie brilliant blue (CBB) was concordant with the band detected by western blotting using mAb SKM9-2 (Fig. 2h). The band was cut from CBB-stained gel, and analyzed using nano-LC MS/MS of trypsinized peptides and a subsequent Mascot search. Results of this analysis suggest that SKM9-2 antigen is HEG1 (Fig. 2i).
To confirm that SKM9-2 antigen is sialylated HEG1, we performed a gene silencing analysis of HEG1. Suppressing the HEG1 gene with 3 different regions of siRNA (H1097, H2674, and H3671) specifically decreased HEG1 western blot signals (Fig. 2j). Lentiviral particles coding HEG1 shRNA also lightened the band detected with mAb SKM9-2, in contrast to the control lentiviral particles (copGFP). Recombinant soluble HEG1 (sHEG1) in which a His-tag was substituted for the transmembrane and cytoplasmic domains was purified from the culture supernatant of sHEG1-transfected ACC-MESO-4. The mAb SKM9-2 recognized sHEG1 at a molecular size similar to that of native HEG1 (Fig. 2j). In addition, the mAb SKM9-2 bound to a molecule in HEG1-transfected HEK293T, but not EGFP-transfected cells (Fig. 2k).
Genomic structure of human HEG1. The structure and function of full-length human HEG1 have not yet been investigated. Figure 3a shows the genomic structure of human HEG1. The human HEG1 gene (HEG1) is located between mucin-13 (MUC13) and solute carrier family 12 member 8 (SLC12A8) in chromosome 3q21.2 (GeneCards human gene database). Human HEG1 is predicted to be a type I membrane protein and to have a longer Ser/Thr rich region than the HEG1 reported for mice and zebrafish ( Fig. 3a and Supplementary Fig. 4). This Ser/Thr rich region accounts for ~70% of the mature HEG1 peptide, and is followed by three epidermal growth factor (EGF)-like domains, an unknown region, two extracellular juxtamembrane regions, a transmembrane domain, and a cytoplasmic domain (Fig. 3a). HEG1 would contain many O-linked glycans in the Ser/ Thr rich region; because the molecular size of purified HEG1 was 400 kDa, which is much larger than its predicted size of 150 kDa (Fig. 2a); HEG1 showed mucin-like characteristics during purification, as mentioned above ( Fig. 2d-g); and its various O-linked glycans were detected in a lectin microarray analysis described later (Fig. 3d-e). HEG1 does not belong to the mucin gene family because it does not contain typical tandem repeat structures. However, HEG1 has a long O-glycosylated region and EGF domains, such as membrane-associated mucins 12 . HEG1 may have a physiological function similar to that of a membrane-associated mucin.   Table 2. Expression rates of mesothelioma markers in non-MPM tumors. Intensity and proportion of staining in tumor cells were evaluated in the entire microscopic field of each specimen. Cases were defined as positive if the proportion score was more than 0. In the immunostaining of calretinin or WT-1, staining in the nucleus, but not the cytoplasm, was considered a positive sample.   Two transcript variants of HEG1 with or without exon 6 were cloned from ACC-MESO-4. In an RT-PCR analysis, mRNAs of both variants were transcribed without a distinct bias in all tested MPM cell lines (Fig. 3b,  upper panel). Thus, the two variants of HEG1 would be expressed on the cells concurrently. Slightly longer PCR products of ACC-MESO-1 or ACC-MESO-4 (Fig. 3b, upper panel), but not other cells, resulted from insertion of an additional 6 Ser residues in a characteristic poly-Ser region (the SNP present in dbSNP137) (Fig. 3a). HEG1 of ACC-MESO-4 also had two missense SNPs (Fig. 3a). Binding of mAb SKM9-2 was not affected by the presence or absence of exon 6 or these SNPs in ACC-MESO-4.

Glycosylation of HEG1 in mesothelioma cells.
Glycosylation of HEG1 was investigated using a lectin microarray 28 (Fig. 3d and Supplementary Fig. 5). Abbreviations of lectins are shown in the Methods section. HEG1 bound to AAL but not LTL, which indicates that HEG1 has sialyl Lewis X , but not Lewis X . HEG1 also bound to other Fuc binders (PSA, LCA, UEA-I). HEG1 characteristically reacted with α 2,6-linked sialic acid binders (SNA, SSA, TJA-I) and a sialylated N-acetyllactosamine (LacNAc) binder (RCA120), but not to a terminal LacNAc binder (ECA) or α 2,3-sialyl LacNAc binders (MAL, ACG). These observations suggest that α 2,6-sialyl LacNAc is attached to the nonreducing terminal of HEG1 glycan. The signals for binders of T-antigen with or without sialic acid (ABA, jacalin, PNA, ACA, and MPA) and a disialyl T-antigen binder (MAH) were weak or undetectable. Although sialyl T-antigen or disialyl T-antigen was attached to HEG1, these antigens may be present at comparatively low levels, or they may be screened by large glycans. The binding of HEG1, but not chitin binders (LEL, STL, PWN), to WGA and DSA suggests that HEG1 is a sialomucin containing a poly-LacNAc structure. The strong reaction of HEG1 to terminal GalNAc binders (TxLCI, BPL, TJA-II, and SBA), but not α -GalNAc binders (HPA, DBA, PTL-I, and GSL-I A4), indicates that some terminals of HEG1 glycan contain β -GalNAc. Furthermore, the binding of HEG1 to WFA suggests that HEG1 has a GalNAcβ 1-4GlcNAc [N, N'-diacetyllactosamine (LacdiNAc)] structure. Considering these results, the structure of O-glycan attached to HEG1 is summarized in Fig. 3e. HEG1 is attached with sialyl T-antigen, disialyl T-antigen, and many core 2 glycans with a poly-LacNAc structure. The nonreducing terminal of this poly-LacNAc structure would be attached with α 2,6-linked sialic acid, sialyl Lewis x , or LacdiNAc with or without α 2,6-linked sialic acid and/or α 1,3-linked fucose.
Although SKM9-2 antigen was not detected in major organs (Fig. 1c), HEG1 mRNA was observed in the heart and the lung ( Fig. 3c and b, lower panel). HEG1 mRNA expression in the liver, colon, kidney, prostate, and testis was relatively low (Fig. 3c). Although HEG1 protein may have low expression in some tissues other than MM, the mAb SKM9-2 can distinguish only HEG1 with glycosylation characteristic of MM.

HEG1-dependent proliferation of mesothelioma cells. Sialylated HEG1 was expressed on prolifer-
ative cells such as reactive mesothelial cells, endothelial cells in angiogenesis, and mesothelioma (Fig. 1). Since the membrane-associated mucins MUC1, MUC4, MUC13, and MUC16 support the survival and proliferation of cancer cells [18][19][20][21] , we expected that HEG1 would associate with cell proliferation. Proliferation of mesothelioma cells was suppressed by HEG1 siRNA, but not control siRNA, in a time-dependent manner (Fig. 4a). Inhibition of proliferation by HEG1 siRNA was also observed on the reduced incorporation of bromodeoxyuridine ( Supplementary Fig. 6). Cell death was partially induced by HEG1 siRNA mix 1 (H1097 plus H2674) after 48 h of treatment. Some siRNAs diminished the cell growth by ~50%, and one siRNA (H2674) strongly inhibited cell proliferation (Fig. 4b). Control siRNA and H3671 (a weak silencer of HEG1, Fig. 2j) did not significantly affect cell growth (Fig. 4b). The suppression of cell growth by H2674 was also observed on NCI-H2452, another mesothelioma cell line (Fig. 4c). In contrast, the cell growth rates of HEK293T and ACC-MESO-1, which expressed lower amounts of SKM9-2 antigen (Fig. 2k), were not affected by the siRNA (Fig. 4c). These results suggest that MM cell proliferation partly depends on HEG1 expression. The more HEG1 is expressed on the cells, the more dependently the cells may proliferate.

Discussion
Antibody against a specific glycosylation site on a membrane protein can exhibit high specificity to tumor cells 16 . This high specificity makes it possible to provide chimeric antigen receptor therapy that is more potent than tumor vaccine or antibody therapy 17 . The mAb SKM9-2 recognizes both HEG1 peptide and its sialylated O-glycosylation, and binds to MM with high specificity and sensitivity. This dual recognition of mAb SKM9-2 would be advantageous to a molecular-targeting therapy against MM using antibody. Furthermore, we found that HEG1 on mesothelioma cells contains a unique LacdiNAc structure. LacdiNAc is a rare glycosylation in human proteins, previously observed in gastric mucin and cancer antigens [29][30][31][32] . We expect that a more accurate diagnosis of MM will become possible by combining detection of SKM9-2 antigen with that of the LacdiNAc modification on HEG1.
We showed that sialylated human HEG1 was mainly expressed on the apical membrane, unlike the reported mouse HEG1 observed in the cell-cell junction 23 . Sialylated human HEG1 expression on the apical surface is consistent with this protein's heavy glycosylation and hydrophilicity. Mesothelioma cells disseminated into pleural effusion also expressed sialylated HEG1 on the cell cluster apical surface, but not at cell-cell junctions (Fig. 1b). This HEG1 expression seemed to prevent re-attachment of mesothelioma cells detached from the pleura as well as anti-adhesion effects of membrane-associated mucin 12,20,33 .
Membrane-associated mucin acts as a physical barrier and performs various physiological functions related to cell survival 12,34 . The EGF domain of MUC4 binds to ErbB2 (HER2/neu) 19 . An EGF domain of HEG1, instead of MUC4, may associate with ErbBs as a proliferation-regulating mucinous molecule on the mesothelioma cells, because MUC4 is not expressed on mesothelioma 35 . An EGF domain released from HEG1 may also bind to EGFR, analogous to binding reported for other EGFR ligands 36,37 . Although a proteolytic cleavage site characteristic of membrane-associated mucins was not found in HEG1, we speculate that the unknown region conserved Scientific RepoRts | 7:45768 | DOI: 10.1038/srep45768 among mammals (Fig. 3a and Supplementary Fig. 4) is a cleavage site digested by proteases, and that the released EGF domain of HEG1 functions as a growth factor. The cytoplasmic domain of MUC1 interacts with receptor tyrosine kinases and PI3Ks 18 . The cytoplasmic domain in HEG1 contains a predicted phosphorylation site (Ser1393) and is bound by KRIT1, an intracellular molecule 38 . HEG1 does not have a typical PxxP motif for binding protein kinase; however, HEG1 does have a P/Y/R/K rich region (1394-PYAEYPKNPR-1403), which may be bound with Src homology 3 domain 39 . We expect that the HEG1 cytoplasmic domain associates with signaling molecules for cell proliferation. Functional inhibitors of HEG1 may have anti-mesothelioma activity; a MUC1 inhibitor has been examined for anti-cancer properties 34 .
For efficacious cancer immunotherapy using antibodies, vaccines, or chimeric antigen receptors against a tumor antigen, it is desirable to target a molecule expressed specifically on the tumor. Mucin-like membrane proteins may be worthwhile targets for cancer immunotherapy due to their tumor-specific patterns of glycosylation 40,41 . In light of the association between HEG1 and mesothelioma proliferation, the mAb SKM9-2 and HEG1 may be productive diagnostic tools and therapeutic targets for MM.  Immunohistochemistry. The sections on glass slides were treated by the autostainer according to the manufacturer's instructions; calretinin, mesothelin, or a part of SKM9-2 antigen (identified as sialylated HEG1 in this study) was used with Histostainer 48 A (Nichirei Co., Tokyo, Japan); CK5/6, podoplanin, or WT-1 was used with Ventana BenchMark ULTRA (Ventana Medical Systems, Tucson, AZ). The antigen retrieval protocol was as follows: calretinin was heated for 40 min at 98 °C in Heat Processor Solution pH 9 (Nichirei Co.); mesothelin or SKM9-2 antigen was heated at 98 °C for 40 min or 121 °C for 10 min, respectively, in Target Retrieval Solution, Citrate pH 6 (Dako Japan Co, Kyoto, Japan); cytokeratin 5/6 (CK5/6) or podoplanin was heated at 95 °C for 64 min in Ultra Cell Conditioning Solution (ULTRA CC1) (Ventana Medical Systems); Wilms' tumor gene product 1 (WT-1) was heated at 95 °C for 64 min in Ultra Cell Conditioning Solution (ULTRA CC2) (Ventana Medical Systems). The primary antibodies used were as follows: calretinin, rabbit anti-calretinin polyclonal antibody (PAD:DC8) (Thermo Fisher Scientific, Rockford, IL, USA); mesothelin, mouse anti-human mesothelin mAb (5B2) (Leica Microsystems, Bannockburn, IL); CK5/6, mouse anti-cytokeratin 5,6 mAb (D5/16B4) (Nichirei Co.); podoplanin, mouse anti-podoplanin mAb (D2-40) (Ventana Medical Systems); and WT-1, mouse anti-human WT-1 mAb (6F-H2) (Ventana Medical Systems). Each antibody was used according to the respective manufacturer's instructions. SKM9-2 antigen was stained with the culture supernatant of a hybridoma clone (SKM9-2) containing mouse anti-human HEG1 mAb. Immunoreactivity was visualized using Histofine Simple Stain MAX-PO (Multi) (Nichirei Co.) with Liquid DAB+ (Dako Japan Co) [calretinin and mesothelin], the EnVision+ kits [SKM9-2 antigen], or the ultraView Universal DAB Detection Kit (Ventana Medical Systems) [CK5/6, podoplanin, and WT-1], according to the manufacturer's instructions. Finally, the sections were counterstained with hematoxylin, dehydrated, and mounted with Malinol medium (Muto Pure Chemicals, Tokyo, Japan). Immunostaining was evaluated on the basis of the intensity and proportion of staining on all tumor cells in each specimen. To evaluate the immunostaining for calretinin or WT-1, the staining in the nucleus, but not the cytoplasm, was measured. The intensity of staining was defined by scoring as follows: 2, strong staining; 1, weak staining; 0, no staining. The proportion of staining was measured in the entire microscopic field of tumor cells for each specimen and was classified by scoring as follows: 3, > 50%; 2, 50-10%; 1, 9-1%; 0, 0%. Cases were defined as positive if the proportion score was more than 0.