Glycoprotein B7-H3 overexpression and aberrantglycosylation in oral cancer and immune responseJung-Tsu Chena,b,c,d, Chein-Hung Chenb, Ko-Li Kue, Michael Hsiaob, Chun-Pin Chiangc,d,e, Tsui-Ling Hsub,Min-Huey Chena,c,d,f,1, and Chi-Huey Wonga,b,1

aPh.D. Program in Translational Medicine, National Taiwan University and Academia Sinica, Taipei 100, Taiwan; bGenomic Research Center, Academia Sinica,Taipei 115, Taiwan; cGraduate Institute of Clinical Dentistry, School of Dentistry, National Taiwan University, Taipei 100, Taiwan; dDepartment of Dentistry,National Taiwan University Hospital, Taipei 100, Taiwan; eGraduate Institute of Oral Biology, School of Dentistry, National Taiwan University, Taipei 100,Taiwan; and fResearch Center for Developmental Biology and Regenerative Medicine, National Taiwan University, Taipei 100, Taiwan

Contributed by Chi-Huey Wong, August 31, 2015 (sent for review April 15, 2015; reviewed by Xi Chen and Lara K. Mahal)

The incidence and mortality rate of oral cancer continue to rise,partly due to the lack of effective early diagnosis and increasingenvironmental exposure to cancer-causing agents. To identify newmarkers for oral cancer, we used a sialylation probe to investigate theglycoproteins differentially expressed on oral cancer cells. Of theglycoproteins identified, B7 Homolog 3 (B7-H3) was significantlyoverexpressed in oral squamous cell carcinoma (OSCC), and itsoverexpression correlated with larger tumor size, advanced clinicalstage, and low survival rate in OSCC patients. In addition, knock-down of B7-H3 suppressed tumor cell proliferation, and restorationof B7-H3 expression enhanced tumor growth. It was also found thatthe N-glycans of B7-H3 from Ca9-22 oral cancer cells contain theterminal α-galactose and are more diverse with higher fucosylationand better interaction with DC-SIGN [DC-specific intercellular adhe-sion molecule-3 (ICAM-3)–grabbing nonintegrin] and Langerin on im-mune cells than that from normal cells, suggesting that the glycanson B7-H3 may also play an important role in the disease.

CD276 | proliferation | glycan sequencing | terminal α-galactose | fucose

Oral cancer is the 11th most-common cancer worldwide. Anestimated 300,373 new cases and 145,353 deaths from oral

cavity cancer (including lip cancer) occurred worldwide in 2012(1). According to the data in GLOBOCAN 2012 publishedby the World Health Organization, oral cancer incidence andmortality in men in South and Central Asia is increasing, and it hasbecome the second most common cancer in the region. Epidemi-ological studies reveal the strong association between oral cancerand the use of betel quid, alcohol, and cigarettes. Cancers of buccalmucosa, tongue, and gingiva constitute the majority of oral cancer inthe Asian population, and that is related to the exposure to thecarcinogen in these anatomical areas (2). It is estimated that ap-proximately 90–95% of oral cancers are squamous cell carcinoma(OSCC) (3), and oral cancer has been shown to progress fromhyperplasia, to mild-to-moderate dysplasia, severe dysplasia, andcarcinoma in situ (4). Surgery is the standard treatment for OSCC,but the differences in primary sites and the complex anatomy of thehead and neck give rise to intricate patterns of local invasion andregional spread. This distinction makes primary tumors difficult toeradicate once they have grown large enough to spread into adja-cent tissues. Radiotherapy is an integral part of primary or adjuvanttreatment, and chemotherapy is used as a combination therapy inadvanced OSCC (5). Despite numerous prospective trials usingvarious combination therapies to improve locoregional control,survival rates for advanced carcinogen-associated OSCC remaindismal (6). In contrast to many other cancers in which metastasis isthe primary cause of death, local recurrence is the common cause oftreatment failure and death in patients with OSCC (5). Therefore,locoregional control is a key therapeutic objective (7).Glycosylation is an important biological process that occurs

cotranslationally or posttranslationally on more than 50% ofeukaryotic proteins and can affect protein folding, stability, sol-ubility, and function. Aberrant glycosylation is often observed in

pathological conditions such as inflammation and cancer metastasis.Altered terminal fucosylation and sialylation are believed to resultfrom changes in expression and are associated with tumor malig-nancy (8). Atypical glycosylation of cell surface carbohydrates hasbeen reported to be associated with malignant transformation oforal epithelium (9), and protein-bound sugar levels were higher inplasma and tissue samples of oral cancer patients (10). A series ofstudies about dysregulation of the N-glycosylation–regulating gene,DPAGT1, drives oral cancer cell discohesion by inhibiting adhesionof E-cadherin throughWnt signaling pathway were reported (11–13).It was also found that the expression of Lewisy on EGFR promotesmigration of oral cancer cells (14).To identify new markers as diagnostic and therapeutic targets for

oral cancer, a sialylation probe was used to investigate the differ-ential expression of glycoproteins in oral cancer and normal cells.Among the glycoproteins identified, receptor B7-H3 was selectedfor further study because it was more differentially expressed incancer cells. B7-H3, also known as B7 homolog 3 or CD276 isoform 1,was discovered in 2001 (15) and is a 110-kDa, type I trans-membrane glycoprotein with four Ig-like domains that contain anearly exact tandem duplication of the IgV-IgC domain (4Ig-B7-H3).However, the potential binding partner of B7-H3 remains un-clear (16), and the functional effect of B7-H3 on T cells is con-troversial (17). B7-H3 protein was also found on various cell typesand organs. This broad expression pattern suggests more diverseimmunological and probably nonimmunological functions ofB7-H3, especially in peripheral tissues (18). Recently, B7-H3 has been

Significance

Oral squamous cell carcinoma (OSCC) is characterized by highmorbidity and mortality, and few therapeutic options. Here, weshow that the expression of B7 Homolog 3 (B7-H3) is signifi-cantly up-regulated in the tumor tissue of OSCC patients andcorrelated with increased tumor size and poor survival rate.Comparing the N-glycans of B7-H3 from Ca9-22 oral cancer cellsand Smulow–Glickman (SG) normal cells, we also found that theglycans of B7-H3 from Ca9-22 contain the terminal α-galactoseand are more diverse with higher fucosylation and better in-teraction with immune cells. These findings indicate that glycopro-tein B7-H3 is an important marker in oral cancer and itsoverexpression and aberrant glycosylation may provide a directionfor the development of diagnosis and treatment of oral cancer.

Author contributions: J.-T.C., M.H., T.-L.H., M.-H.C., and C.-H.W. designed research; J.-T.C.and K.-L.K. performed research; J.-T.C., C.-H.C., C.-P.C., and T.-L.H. analyzed data; J.-T.C.,T.-L.H., and C.-H.W. wrote the paper; and C.-P.C. contributed oral squamous cellcarcinoma specimens.

Reviewers: X.C., University of California, Davis; and L.K.M., New York University.

The authors declare no conflict of interest.1To whom correspondence may be addressed. Email: chwong@gate.sinica.edu.tw orminhueychen@ntu.edu.tw.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1516991112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1516991112 PNAS | October 20, 2015 | vol. 112 | no. 42 | 13057–13062

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

1, 2

020

found in a variety of different human cancers, including prostate(19, 20), nonsmall cell lung (NSCLC) (21), gastric (22, 23), pan-creatic (24), ovarian (25), colorectal (26), urothelial cell (27), clearcell renal cell (ccRCC) (28), and hypopharyngeal (29) cancers.We are interested in understanding the expression pattern of

B7-H3 in oral cancer and its possible underlying mechanisms. Inthis study, both the protein and the glycosylation profile of B7-H3were investigated to explore their correlation with tumor progres-sion with the expectation that the findings will provide more in-formation to better understand oral cancer and improve therapies.

ResultsProbing Glycosylation and Identification of B7-H3 in Oral Cancer CellLines. To delineate the molecular basis for aberrant glycosylationin the pathological processes of oral cancer, we investigated thedifferences in glycosylation between normal and cancer cells byusing the sialylation probe reported (30). We compared normaloral epithelial cells (Smulow–Glickman; SG), which derived fromadult human gingiva (31), and a representative oral cancer cell line(Ca9-22) by feeding cells with the peracetylated alkynyl N-ace-tylmannosamine (ManNAcyne) to label sialylated proteins. Sialy-lated proteins incorporated with an alkyne group were then reactedwith biotin-azide via Cu[I]-catalyzed click reaction and subjected toaffinity enrichment via binding to the biotin tag. The enriched gly-coproteins were treated sequentially with trypsin and PNGase F torelease the Asp-containing peptides for protein identification andglycosite mapping, respectively, through liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis (32). The resultsshowed that NPTN (neuroplastin), PRNP (major prion protein),desmoglein-2 (DSG2), integrin β-1 (ITGB1), and CD276 (B7-H3)were detected only in Ca9-22 oral cancer cells in all three batches ofsamples (SI Appendix, Table S1). Of these overexpressed sialylatedglycoproteins, B7-H3 was the most significant, although little isknown about this glycoprotein in oral cancer. We thus decided toinvestigate the relationship between B7-H3 and oral cancer. Fournormal oral epithelial cell lines and 13 OSCC cell lines from dif-ferent anatomical areas were examined. It was found that B7-H3was overexpressed in Ca9-22. However, although some cancer celllines such as HSC-3, OEC-M1, SAS, and SCC-9 have slightly higherexpression of B7-H3 than normal SG cells, the differences from thenumber of cell lines examined do not show statistical significance(SI Appendix, Fig. S1 A and B).

Correlation of B7-H3 Expression and Clinicopathological Parametersin OSCC Patients. To investigate the oncogenic potential of B7-H3in oral cancer, we collected specimens from OSCC and oralepithelial dysplasia (OED) patients and healthy donors for im-munohistochemical (IHC) staining of B7-H3. Fig. 1 shows rep-resentative IHC staining photomicrographs of normal oralmucosa, OED, and OSCC specimens. In general, normal oralepithelium was negative for B7-H3 or displayed a few B7-H3–positive cells in the basal cell layer of the epithelium (Fig. 1A).The positive B7-H3 staining on the cell surface was found pre-dominantly in the lower one-third of the epithelium in severeOED specimens (Fig. 1B). B7-H3 staining was found stronglypositive in the peripheral nonkeratinizing cells of OSCC tumornests and negative in the central keratinized area (Fig. 1 C andD). These positively stained OSCC were basal cells of the epi-thelium and were viewed as cells with rapid proliferation rates.In addition, positive membrane B7-H3 staining was also ob-served in some endothelial cells and inflammatory cells in theOSCC stromal tissue. The mean B7-H3 labeling score increasedsignificantly from normal oral mucosa (7.89 ± 14.4%) and OED(15.62 ± 8.8%) to OSCC samples (74.75 ± 47.2%, P < 0.001; SIAppendix, Table S2). The high labeling score of B7-H3 in OSCCwas significantly associated with increased tumor sizes (P = 0.0001)and advanced clinical stages (P = 0.004). No significant associationwas found between the labeling scores of B7-H3 in OSCC and

patient age, sex, location of cancer, lymph node involvement, me-tastasis, recurrence, differentiation of OSCC, alcohol consumption,betel quid chewing, or cigarette smoking (SI Appendix, Table S3).A Kaplan–Meier curve showed that the OSCC patients with theB7-H3 labeling scores ≥55% had a significantly poorer cumulativesurvival rate than those with the B7-H3 labeling scores <55% (log-rank test P = 0.005) (Fig. 1E).

Involvement of B7-H3 in Cell Proliferation and Tumor Formation.Next, we examined whether B7-H3 is involved in the pro-liferation of oral cancer cells and tumor formation in the animalstudy. By knockdown and restoration strategies, we establishedthree stable clones of B7-H3 knockdown Ca9-22 cells (shB7H3-1to -3) from two independent shRNA sequences. The knockdownefficiency in these three clones (91%, 94%, and 75%) was calcu-lated based on the mRNA level of B7-H3 (Fig. 2A). The proteinexpression level was significantly reduced in B7-H3 knockdown cells(SI Appendix, Fig. S1C). The result of in vitro cell proliferation assayshowed that B7-H3 knockdown clones had slower proliferationrates (Fig. 2B). To evaluate the effect of B7-H3 on tumor growthin vivo, the Ca9-22 shB7H3 knockdown clones were injected s.c.into the flank region of NOD-SCID mice and examined for tumorgrowth. The results showed that knocking down B7-H3 dramaticallysuppressed the growth of Ca9-22 in vivo (Fig. 2 C and D), and the

Fig. 1. Up-regulation of B7-H3 in OSCC is associated with poor prognosis.(A) Normal oral mucosa showed no B7-H3 staining in any epithelial cells.(B) Severe oral epithelial dysplasia exhibited moderate membrane B7-H3 stainingin the lower half of epithelial cells. (C) Well-differentiated OSCC showed strongmembrane B7-H3 staining in the peripheral cells of OSCC tumor nests. (D) Strongmembrane B7-H3 staining was shown at the peripheral cells of tumor nests andthe intercellular bridge. (E) Kaplan–Meier curve of overall survival in 72 patientswith OSCC stratified by the expression level of B7-H3. The duration of survivalwas measured from the beginning of the treatment to the time of death or at60 mo. The cumulative survival for patients with OSCC, with the B7-H3 labelingscores (LS) ≥ 55%, was significantly lower than that for patients with OSCC, withthe B7-H3 LS < 55% (P = 0.005, log-rank test).

13058 | www.pnas.org/cgi/doi/10.1073/pnas.1516991112 Chen et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

020

same result was also observed in orthotopic implantation (Fig. 2 Eand F). Moreover, when B7-H3 was overexpressed in controlclones to increase the expression of B7-H3 or in shB7H3 knock-down clones to restore the expression, tumor growth in vivoin both these settings was enhanced (Fig. 2 G and H). Collec-tively, our data suggest that B7-H3 acts as an oncoprotein inoral cancer.

Glycoform Analysis of B7-H3 in Ca9-22 and SG.We chose Ca9-22 oralcancer cells (gingival carcinoma) and SG normal oral epithelialcells from gingiva to further analyze the glycosylation level

of B7-H3 in different cells. Immunoblot results showed thatB7-H3 was highly expressed in Ca9-22 cells compared withnormal SG cells and other cell lines (SI Appendix, Fig. S1A).Moreover, B7-H3 protein from Ca9-22 cells showed lowermobility compared with that from SG cells. After treating withPNGase F to cleave the whole N-linked glycans, the mobilityof B7-H3 from both cell lines became the same, and the size ofprotein was ∼58 kDa as expected (SI Appendix, Fig. S2A). Thisresult indicated that N-glycosylation contributed to the differ-ences in B7-H3 mobility in protein gel, and a higher glycosylationlevel of B7-H3 from Ca9-22 cells was observed. From these data,

Fig. 2. Knockdown of B7-H3 in oral cancer cells suppressed cell proliferation in vitro and tumor growth in vivo. (A) The expression level of mRNA wasobserved in three Ca9-22 clones stably knocked down with B7-H3. Stable knockdown clones of Ca9-22/shB7H3-1 and Ca9-22/shB7H3-2 showed significantlydecreased expression of B7-H3 at the mRNA level. (B) Silencing B7-H3 in Ca9-22 slowed down their proliferation rate. Cell proliferation was observed by a liveimaging system (IncuCyte) at different time intervals ranging from 0 to 96 h (n = 8). Confluence of cells was calculated by using phase-contrast images. Dataare shown as mean ± SD (n = 8). (C and D) Knockdown of B7-H3 remarkably reduced the volumes of tumor mass. Sizes of tumors grown from s.c.-injectedshCtrl (control) and shB7H3-1 stable clones were measured by using a vernier caliper weekly. (C) Tumor growth curve. The mean ± SD (n = 3) from each groupat each time point is shown. (D) Images of tumor specimens. a shows the gross view of isolated tumors. Strong membranous staining of B7-H3 in shCtrl groups(b) and almost no staining of B7-H3 in shB7H3-1 groups (c) were observed at the end point (day 56). (E and F) Knockdown of B7-H3 also reduced the volumesof tumor mass in the organ-specific environment, but metastasis was not found. Luciferase-transduced tumor cells were injected into buccal mucosa of NOD-SCID mice, and tumor growth and metastasis were detected by using the IVIS imaging system. (E) Tumor growth curve. (F) Bioluminescence imaging of miceat the end point (day 42). The mean ± SD (n = 3) from each group at each time point is shown. (G and H) Overexpression of B7-H3 in shCtrl groups (G) orrestoration of B7-H3 in shB7H3-1 clones (H) enhanced tumor growth in mice. Data shows the mean ± SD (n = 5) from each group at each time point. P valuewas calculated by Student’s t test (*P < 0.05, **P < 0.01).

Chen et al. PNAS | October 20, 2015 | vol. 112 | no. 42 | 13059

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

1, 2

020

we conclude that both the expression and glycosylation levels ofB7-H3 were up-regulated in Ca9-22 oral cancer cell lines.To dissect the B7-H3 glycoforms involved in tumor versus normal

cells, site-specific glycoform mapping and glycan sequencing ofB7-H3 in Ca9-22 and SG cells were performed. To obtain higherprotein amounts for the glycan profiling, full-length B7-H3 wasoverexpressed in both types of cells and purified for analysis. MSprofiles showed that the glycans of B7-H3 from SG displayed ahigher level of sialylation and B7-H3 from Ca9-22 had a higher levelof fucosylation (SI Appendix, Fig. S2 B and C). The detailed gly-cosylation patterns on B7-H3 were subsequently analyzed byLC-MS/MS. Every individual B7-H3 glycopeptide derived from SGand Ca9-22 cells was compositionally assigned and quantified in asite-specific manner through matching with the calculated mass ofboth peptide fragments and glycans from the Consortium forFunctional Glycomics carbohydrate databases, and confirming withthe appearance of fragmented glycans in MS/MS spectra (Fig. 3A).

It has been verified that there are two sites of N-linked glycosyl-ation in B7-H3 (Asn-322 and 407) (33). Our result showed that thereare eight N-glycosylation sites (Asn positions 91, 104, 189, 215, 309,322, 407, and 433) in B7-H3, as determined by Asn to Asp con-version after PNGase F treatment (SI Appendix, Table S4). BecauseB7-H3 contains a nearly exact tandem duplication of the IgV-IgCdomain and there were four N-glycosylation sites located in each ofthe IgV-IgC domains, each pair of glycosylation sites, N91 and N309,N104 and N322, N189 and N407, and N215 and N433 were iden-tified through identical peptide sequence. Within four pairs ofN-glycosylation sites, one of them was identified to compose mostlyhybrid type N-glycans (N104/N322) and three of them were attachedmainly with complex N-glycan (N91/N309, N189/N407, and N215/N433) (Fig. 3B and SI Appendix, Fig. S3). The B7-H3 from SG cellscontained mostly biantennary glycans, and the most abundantstructure is BiS2F1 (42%). The B7-H3 from Ca9-22 cells contained amore diverse pattern of N-linked glycosylation, including complextypeN-glycans (BiF1: 14%, BiH1F2: 13%, BiS1F1: 11%) and hybridtype N-glycans (Man5N1F1: 6.1%) (Fig. 3 A and B).The levels of sialylation and fucosylation on each pair of

N-glycosites are shown in Fig. 3C. The B7-H3 in SG carries moresialic acid residues, whereas the B7-H3 in Ca9-22 carries morefucose residues, especially at positions N104/N322, and N189/N407(Fig. 3C). In some other OSCC cell lines, higher fucosylation inB7-H3 was also detected by AAL (Aleuria aurantia lectin, pre-ferentially binds to α1,2/3/4/6-linked fucoses) lectin blotting com-pared with normal cells (SI Appendix, Fig. S4), but the number ofcell lines used is not large enough to show statistical signifi-cance. The result of glycan analysis showed a protein mobility shiftof B7-H3 in SG cells after treatment with α2,3/6/8-neuraminidase,demonstrating that B7-H3 from SG cells had higher sialylation(SI Appendix, Fig. S2A). The higher expression level of B7H3 inCa9-22 cancer cells (SI Appendix, Table S1) is obviously due tothe higher expression of the protein instead of higher sialylation,because the sialyation level of B7-H3 is actually lower in Ca9-22than in SG cells as revealed by the glycoform analysis.Various kinds of lectins are expressed on immune cells and

interact with different types of glycans. We compared the bindingability of human immune receptors with sB7-H3 derived fromCa9-22 cells to that derived from SG cells in the presence ofcalcium and magnesium. Among a panel of 44 recombinant in-nate immunity receptor-Fc fusion proteins (34, 35), four gaveclear and consistent positive signals that were significant inB7-H3 derived from Ca9-22. They are DC-specific intercellularadhesion molecule-3 (ICAM-3)–grabbing nonintegrin (DC-SIGN),mouse Kupffer cell receptor (mKCR), Langerin, and DC-SIGN–related protein (DC-SIGNR) according to the signal intensity(SI Appendix, Fig. S5). Based on the Consortium for the Func-tional Glycomics for glycan array analysis, DC-SIGN and Langerinbind to high mannose structures, fucosyl biantennary N-glycans,and fucose (Fuc)-containing antigens, such as B antigens, Lewis, orsialyl Lewis antigens, indicating their ability to recognize variousglycans (36). This result provides further confirmation that B7-H3derived from Ca9-22 carries more diverse N-glycan structures withhigher fucosylation.

Identification of Terminal α-Galactose in B7-H3 N-Glycans. The resultsfrom glycoform analysis suggested that B7-H3 protein derived fromCa9-22 oral cancer cells had extra hexose(s) and a higher level offucosylation on its N-glycans, such as the biantennary glycan struc-ture with one hexose and two fucoses (BiH1F2). Based on thebiosynthetic pathways ofN-linked glycosylation, we propose that themain structure of BiH1F2 is composed of Manα1–6(Manα1–3)Manβ1–4GlcNAcβ1–4GlcNAcβ1, two Galβ1–4GlcNAc (N-acetyll-actosamine) units, two fucoses, and an additional terminalα-galactose (α-Gal). The fucose residues can be at the core (α1–6linked to the innermost GlcNAc residue) or terminal positions(e.g., α1–2 linked to Gal or α1–3 linked to GlcNAc residue).

Fig. 3. Glycoforms of B7-H3 derived from SG and Ca9-22 cells are determined.(A) Glycoforms of B7-H3 glycopeptides derived from SG and Ca9-22. (B) Site-specific representative glycans of B7-H3. Abbreviations of glycoforms: Bi, bian-tennary; F, fucose; Man, high mannose; H, hexose; N, N-acetylhexosamine; S,sialic acid; Tetra, tetraantennary; Tri, triantennary. (C) Comparison of site-specificsialylation and fucosylation of B7-H3 derived from SG and Ca9-22 cells. Sialic acidindex = [(%with one sialic acid × 1) + (%with two sialic acid × 2) + (%with threesialic acid × 3)]/100; Fucose index = [(% with one fucose × 1) + (% with twofucose × 2) + (% with three fucose × 3)]/100. Monosaccharide symbols used wereas follows: red triangle, Fuc; yellow circle, Gal; blue square, GlcNAc; green circle,Man; purple diamond, Neu5Ac.

13060 | www.pnas.org/cgi/doi/10.1073/pnas.1516991112 Chen et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

020

To determine the linkage of terminal Gal, we treated Ca9-22–derivedfull-length B7-H3 with α-galactosidase, an exoglycosidase that rec-ognizes terminal α-Gal. The result showed that after treating withα-galactosidase, all of the extra hexoses were cleaved (Fig. 4A). Wethen calculated the conversion rate of the extra hexose(s) cate-gorized according to biantennary, triantennary, and tetraantennaryglycan. After treatment with α-galactosidase, the percentagesof BiF(1∼3), TriF(2∼3), and TetraF(2∼3) increased from 22.4 to34.8%, 0 to 13.4%, and 0 to 1.4%, respectively, which was in accor-dance with the percentages of BiH(1∼2)F(1∼3), TriH(1∼2)F(2∼3),and TetraH(1∼2)F(2∼3) (13.4%, 6.6%, and 2.1%, respectively)before enzyme treatment (Fig. 4B). This result indicated that theextra hexose(s) were terminal α-galactoses in complex typeN-glycans in the Ca9-22 oral cancer cell-derived B7-H3. Moreover,the expression of B antigen [Galα1–3(Fucα1–2)Galβ1–4GlcNAc-R]but not α-Gal antigen (Galα1–3Galβ1–4GlcNAc-R) was observedon the cell surface of Ca9-22 (SI Appendix, Fig. S6). Our resultsuggests that the terminal α-Gal structure detected in the complextype N-glycans of B7H3 from Ca9-22 is B antigen.

DiscussionIn the current study, we report that the expression of B7-H3 issignificantly up-regulated in tumor tissues of OSCC patients andcorrelates with increased tumor size, advanced stage, and poorsurvival rate (SI Appendix, Table S3 and Fig. 1E). This finding issupported by a shRNA knockdown study with cell proliferationassay and in vivo animal studies (Fig. 2). According to thesefindings, we speculate that B7-H3 is a key factor in controllingtumor size in OSCC. This conclusion is consistent with recentstudies in which B7-H3 overexpression was observed in variouscancers, and up-regulation of B7-H3 in prostate, pancreatic, and

colorectal cancers was reported to be correlated with increasedtumor size (19, 24, 37). However, there are inconsistencies insome studies regarding the antitumor effect of B7-H3. Among thesestudies, abundant positive expression of B7-H3 in tumor tissues wasobserved in comparison with normal tissues (38, 39), but the sta-tistical analysis revealed a better prognosis with higher expression ofB7-H3 in tumor tissues. This discrepancy may be due to the lack ofstandardization of the positive B7-H3 expression levels in thesestudies. Moreover, B7-H3 is a member of the B7 family of immu-noregulatory molecules that can be induced on T cells, B cells, anddendritic cells (DCs) by a variety of inflammatory cytokines (15).We detected the expression of B7-H3 on CD14+ and CD19+ cellsbut not on CD3+ cells derived from PBMCs of healthy donors.Although the expression of B7-H3 on immune cells could poten-tially complicate the analysis, we found that the expression of B7-H3protein on intratumoral inflammatory cells in OSCC patients isscarce (2/72). Similar findings were also reported that B7-H3 pro-tein was not detected on inflammatory cells in humans with tumors.In contrast, strong consistent staining of B7-H3 was observed on thetumor vasculature and also on the tumor cells themselves (40).Aberrant glycosylation has been discovered in many kinds of

cancer cells, but little is known about this phenomenon in oralcancer. Here, we found that B7-H3 was significantly expressed inCa9-22 oral cancer cells (SI Appendix, Fig. S1A), and its glycanstructures were different compared with normal cells. Based on ourfindings, the B7-H3 derived from Ca9-22 expressed α-galactosylationand more diverse N-glycan structures with higher fucosylation (Figs.3 and 4). The terminal α-galactose (α-Gal) epitope can be found inthe glycan structures of α-Gal antigen (Galα1–3Galβ1–4GlcNAc-R)or B antigen [Galα1–3(Fucα1–2)Galβ1–4GlcNAc-R]; however, ourglycoform analysis demonstrated that the terminal α-Gal on B7-H3from Ca9-22 cells was fucosylated, suggesting the presence ofB antigen. Flow cytometric analysis of oral cancer cell lines alsorevealed the appearance of B antigen but not α-Gal antigen onCa9-22 and a subpopulation of OEC-M1 cells (SI Appendix, Fig. S6).Both Ca9-22 and OEC-M1 are gingival carcinoma and significant inthe Asian population. According to the literature, besides erythro-cytes, blood group antigen A, B, and H can be expressed in epithelialcells of many tissues including oral tissue, and their expression can bevaried among degrees of differentiation, but decreased in tumortissue of OSCC (41, 42). The loss of blood group antigens in OSCCspecimens correlates significantly with both the stage and histologicalgrade of malignancy (42). Because both Ca9-22 andOEC-M1 cancercells are derived from gingiva, a type of keratinized stratified squa-mous epithelia, the existence of B antigen could be explained (43),but the functional role of B antigen in oral cancer cells requiresfurther investigation.B7-H3–specific receptor may exert an inhibitory rather than a

stimulatory function in both T and NK cell-mediated responses.A number of recent studies have highlighted the role of thecross-talk between NK cells and dendritic cells in the earlyphases of innate immune responses, which shape the subsequentT-cell responses toward the TH1 phenotype (44). Dendritic cellsare professional antigen presenting cells and serve as a bridgelinking the innate immune system with adaptive immune re-sponse. An important family of antigen receptors on dendriticcells involved in recognition and uptake of glycan structures arethe C-type lectin receptors such as DC-SIGN and DC-SIGNR.Apparently, the tumor cells evade the immune system by tar-geting a receptor that is able to transform proinflammatory sig-nals into tolerogenic signals. DC-SIGN has therefore beenproposed to be a homeostatic receptor that can be subverted bytumors through changes in their glycan phenotype (45), but theconnection between tumor-associated B7-H3 and dendritic cellsis not known. Our data revealed that DC-SIGN and DC-SIGNRinteract with sB7-H3 derived from Ca9-22 (SI Appendix, Fig.S5A). Based on the glycoform analysis, the N-glycans of B7-H3derived from Ca9-22 are more diverse and have a higher level of

Fig. 4. Terminal α-galactose determined in complex type N-glycans of B7-H3from Ca9-22. (A) The glycoforms of B7H3 with or without α-galactosidasetreatment. To identify the presence of terminal α-galactose, we treated the full-length B7-H3 protein derived from Ca9-22 with α-galactosidase, whichcleaves terminal α1,3/4/6-linked galactose. The result showed that aftertreatment with α-galactosidase, all of the additional terminal galactose(s)(yellow circles) speculated to appear on specified complex-type N-glycanswere removed. The blue frames show the representative glycan structuresbefore α-galactosidase treatment and red frames show the representativeglycan structures after α-galactosidase treatment. (B) Comparison of theglycoforms with extra hexose(s) before and after α-galactosidase treatment.The percentages of glycoforms were calculated according to A.

Chen et al. PNAS | October 20, 2015 | vol. 112 | no. 42 | 13061

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

1, 2

020

fucosylation (Fig. 3), which may support the recognition of lectinreceptors on dendritic cells. Data revealed that binding of sB7-H3derived from Ca9-22 to dendritic cells is calcium-dependent (SIAppendix, Fig. S5B), suggesting that tumor-associated B7-H3interacts with dendritic cells through carbohydrate-lectin receptorinteraction.In summary, this study demonstrated that glycoprotein B7-H3 is

involved in the pathological process and tumor growth of OSCC,and the glycans of B7-H3 from Ca9-22 oral cancer cells containterminal α-galactoses and more diverse N-glycan structures withhigher fucosylation than that of SG normal cells. In addition, thepresence of a carbohydrate–lectin receptor interaction between thetumor-associated B7-H3 from Ca9-22 cells and immune cells wasdetected. Overall, the overexpression of B7-H3 in oral cancer tis-sues and several oral cancer cell lines, and its aberrant glycosylationin Ca9-22 oral cancer cells compared with normal SG cells found inthis study, may lead to further investigation and development ofnew diagnosis and treatment of this disease.

Materials and MethodsCell Lines and Clinical Specimens. Normal SG cells and OSCC cell line of Ca9-22(JCRB0625) were grown in DMEM (GIBCO/Invitrogen) with 10% (vol/vol) FBSand 1% of 100 units/mL penicillin and 100 μg/mL streptomycin. SG cellsbelong to normal human gingival epithelial cells. Clinical specimens usedfor IHC were obtained from National Taiwan University Hospital (NTUH, Taiwan)with informed consent and approval of institutional review board (NTUHResearch Ethics Committee).

In-Solution Tryptic Digestion, N-Glycosite Determination, Assigning Glycopeptidesand Exoglycosidases Treatment. Purified full-length B7-H3-FLAGs (5 μg) was ly-ophilized and redissolved in 5 μL of 100 mM NH4HCO3, and the same volume of2,2,2-trifluoroethanol (Sigma-Aldrich) was added to denature proteins. Then thesamples were subjected to reduction with 10 mM DTT (Sigma-Aldrich) and al-kylation with 20 mM iodoacetamide (Sigma-Aldrich). After quenching with10 mM DTT, 0.5 μg of trypsin and 1 μg of chymotrypsin were added, and themixture was incubated at 25 °C overnight, followed by heat inactivation oftrypsin and chymotrypsin at 95 °C for 10 min and dried in a SpeedVac evaporatorand redissolved with 20 μL of 0.1% formic acid for LC-MS/MS analysis. De-termination of N-linked glycosylation sites was based on the MS data of thepeptides prepared from sequential trypsin/chymotrypsin and PNGase F digestion.Assignment of glycopeptides was based on matching the measured masses withthe database, which combined predicted masses of tryptic and chymotrypticpeptides (by Protein Digest Simulator Basic) and N- and O-linked glycans (fromConsortium of Functional Glycomics), as well as the retention time of the gly-copeptides by in-house software. Each assigned glycopeptide was confirmed bythe appearance of glycan fragments in its MS/MS spectra.

To detect the presence of terminal α-galactose, full-length B7-H3 glyco-peptide samples were treated with α-galactosidase (Prozyme) at 37 °Covernight in a 10-μL reaction under the conditions suggested by the vender.

ACKNOWLEDGMENTS. We thank the Mass Spectrometry Core at the GenomicsResearch Center Academia Sinica, Taiwan, Dr. Hsin-Yung Yen for glycoproteomicanalyses, Dr. Yi-Ping Wang for pathological evaluation, Prof. Jiiang-Huei Jeng forproviding SG cells, and Dr. Huang Han-Wen andMr. Tsung-Ching Lai for technicalsupport. The research was supported by the National Science Council andGenomics Research Center, Academia Sinica, Taiwan.

1. Torre LA, et al. (2015) Global cancer statistics, 2012. CA Cancer J Clin 65(2):87–108.2. Krishna Rao SV, Mejia G, Roberts-Thomson K, Logan R (2013) Epidemiology of oral

cancer in Asia in the past decade–an update (2000-2012). Asian Pac J Cancer Prev14(10):5567–5577.

3. Chen YK, Huang HC, Lin LM, Lin CC (1999) Primary oral squamous cell carcinoma: Ananalysis of 703 cases in southern Taiwan. Oral Oncol 35(2):173–179.

4. Brennan M, et al. (2007) Management of oral epithelial dysplasia: A review. Oral SurgOral Med Oral Pathol Oral Radiol Endod 103(Suppl:S19):e11–e12.

5. Argiris A, Karamouzis MV, Raben D, Ferris RL (2008) Head and neck cancer. Lancet371(9625):1695–1709.

6. Jemal A, et al. (2008) Cancer statistics, 2008. CA Cancer J Clin 58(2):71–96.7. Argiris A, Li Y, Forastiere A (2004) Prognostic factors and long-term survivorship in

patients with recurrent or metastatic carcinoma of the head and neck. Cancer 101(10):2222–2229.

8. Dube DH, Bertozzi CR (2005) Glycans in cancer and inflammation–potential fortherapeutics and diagnostics. Nat Rev Drug Discov 4(6):477–488.

9. Dennis JW, Granovsky M, Warren CE (1999) Glycoprotein glycosylation and cancerprogression. Biochim Biophys Acta 1473(1):21–34.

10. Manoharan S, Padmanabhan M, Kolanjiappan K, Ramachandran CR, Suresh K (2004)Analysis of glycoconjugates in patients with oral squamous cell carcinoma. Clin ChimActa 339(1-2):91–96.

11. Jamal B, et al. (2012) Aberrant amplification of the crosstalk between canonical Wntsignaling and N-glycosylation gene DPAGT1 promotes oral cancer. Oral Oncol 48(6):523–529.

12. Liu G, et al. (2013) N-glycosylation induces the CTHRC1 protein and drives oral cancercell migration. J Biol Chem 288(28):20217–20227.

13. Nita-Lazar M, et al. (2009) Overexpression of DPAGT1 leads to aberrant N-glycosyla-tion of E-cadherin and cellular discohesion in oral cancer. Cancer Res 69(14):5673–5680.

14. Lin WL, Lin YS, Shi GY, Chang CF, Wu HL (2015) Lewisy promotes migration of oralcancer cells by glycosylation of epidermal growth factor receptor. PLoS One 10(3):e0120162.

15. Chapoval AI, et al. (2001) B7-H3: A costimulatory molecule for T cell activation andIFN-gamma production. Nat Immunol 2(3):269–274.

16. Leitner J, et al. (2009) B7-H3 is a potent inhibitor of human T-cell activation: No ev-idence for B7-H3 and TREML2 interaction. Eur J Immunol 39(7):1754–1764.

17. Steinberger P, et al. (2004) Molecular characterization of human 4Ig-B7-H3, a memberof the B7 family with four Ig-like domains. J Immunol 172(4):2352–2359.

18. Loos M, Hedderich DM, Friess H, Kleeff J (2010) B7-h3 and its role in antitumor im-munity. Clin Dev Immunol 2010:683875.

19. Roth TJ, et al. (2007) B7-H3 ligand expression by prostate cancer: A novel marker ofprognosis and potential target for therapy. Cancer Res 67(16):7893–7900.

20. Zang X, et al. (2007) B7-H3 and B7x are highly expressed in human prostate cancerand associated with disease spread and poor outcome. Proc Natl Acad Sci USA104(49):19458–19463.

21. Sun Y, et al. (2006) B7-H3 and B7-H4 expression in non-small-cell lung cancer. LungCancer 53(2):143–151.

22. Dai W, Shen G, Qiu J, Zhao X, Gao Q (2014) Aberrant expression of B7-H3 in gastricadenocarcinoma promotes cancer cell metastasis. Oncol Rep 32(5):2086–2092.

23. Arigami T, et al. (2011) B7-H3 expression in gastric cancer: A novel molecular bloodmarker for detecting circulating tumor cells. Cancer Sci 102(5):1019–1024.

24. Yamato I, et al. (2009) Clinical importance of B7-H3 expression in human pancreaticcancer. Br J Cancer 101(10):1709–1716.

25. Zang X, et al. (2010) Tumor associated endothelial expression of B7-H3 predicts sur-vival in ovarian carcinomas. Mod Pathol 23(8):1104–1112.

26. Sun J, et al. (2010) Clinical significance and regulation of the costimulatory moleculeB7-H3 in human colorectal carcinoma. Cancer Immunol Immunother 59(8):1163–1171.

27. Boorjian SA, et al. (2008) T-cell coregulatory molecule expression in urothelial cellcarcinoma: Clinicopathologic correlations and association with survival. Clin CancerRes 14(15):4800–4808.

28. Crispen PL, et al. (2008) Tumor cell and tumor vasculature expression of B7-H3 predictsurvival in clear cell renal cell carcinoma. Clin Cancer Res 14(16):5150–5157.

29. Katayama A, et al. (2011) Expression of B7-H3 in hypopharyngeal squamous cellcarcinoma as a predictive indicator for tumor metastasis and prognosis. Int J Oncol38(5):1219–1226.

30. Hsu TL, et al. (2007) Alkynyl sugar analogs for the labeling and visualization of gly-coconjugates in cells. Proc Natl Acad Sci USA 104(8):2614–2619.

31. Smulow JB, Glickman I (1966) An epithelial-like cell line in continuous culture fromnormal adult human gingiva. Proc Soc Exp Biol Med 121(4):1294–1296.

32. Hanson SR, et al. (2007) Tailored glycoproteomics and glycan site mapping usingsaccharide-selective bioorthogonal probes. J Am Chem Soc 129(23):7266–7267.

33. Chen R, et al. (2009) Glycoproteomics analysis of human liver tissue by combination ofmultiple enzyme digestion and hydrazide chemistry. J Proteome Res 8(2):651–661.

34. Hsu TL, et al. (2013) The surface carbohydrates of the Echinococcus granulosus larvainteract selectively with the rodent Kupffer cell receptor. Mol Biochem Parasitol192(1-2):55–59.

35. Chen PC, et al. (2014) Role of N-linked glycans in the interactions of recombinant HCVenvelope glycoproteins with cellular receptors. ACS Chem Biol 9(7):1437–1443.

36. Hsu TL, et al. (2009) Profiling carbohydrate-receptor interaction with recombinantinnate immunity receptor-Fc fusion proteins. J Biol Chem 284(50):34479–34489.

37. Arigami T, et al. (2010) B7-h3 ligand expression by primary breast cancer and asso-ciated with regional nodal metastasis. Ann Surg 252(6):1044–1051.

38. Loos M, et al. (2009) Expression of the costimulatory molecule B7-H3 is associatedwith prolonged survival in human pancreatic cancer. BMC Cancer 9:463.

39. Wu CP, et al. (2006) Relationship between co-stimulatory molecule B7-H3 expressionand gastric carcinoma histology and prognosis. World J Gastroenterol 12(3):457–459.

40. Seaman S, et al. (2007) Genes that distinguish physiological and pathological angio-genesis. Cancer Cell 11(6):539–554.

41. Gao S, et al. (2004) Genetic and epigenetic alterations of the blood group ABO genein oral squamous cell carcinoma. Int J Cancer 109(2):230–237.

42. Campi C, et al. (2007) Secretor status and ABH antigens expression in patients withoral lesions. Med Oral Patol Oral Cir Bucal 12(6):E431–E434.

43. Ravn V, Dabelsteen E (2000) Tissue distribution of histo-blood group antigens. APMIS108(1):1–28.

44. Castriconi R, et al. (2004) Identification of 4Ig-B7-H3 as a neuroblastoma-associatedmolecule that exerts a protective role from an NK cell-mediated lysis. Proc Natl AcadSci USA 101(34):12640–12645.

45. García-Vallejo JJ, van Kooyk Y (2009) Endogenous ligands for C-type lectin receptors:The true regulators of immune homeostasis. Immunol Rev 230(1):22–37.

13062 | www.pnas.org/cgi/doi/10.1073/pnas.1516991112 Chen et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

020