Sialylation and fucosylation of epidermal growthfactor receptor suppress its dimerization andactivation in lung cancer cellsYing-Chih Liua,b,1, Hsin-Yung Yena,b,1, Chien-Yu Chena,b, Chein-Hung Chenb, Ping-Fu Chenga,c, Yi-Hsiu Juand,Chung-Hsuan Chenb, Kay-Hooi Khooc, Chong-Jen Yud, Pan-Chyr Yangd, Tsui-Ling Hsub,2, and Chi-Huey Wongb,2

aInstitute of Biochemical Sciences, National Taiwan University, Taipei 106, Taiwan; bGenomics Research Center, Academia Sinica, Taipei 115, Taiwan;cInstitute of Biological Chemistry, Academia Sinica, Taipei 115, Taiwan; and dDepartment of Internal Medicine, College of Medicine, National TaiwanUniversity, Taipei 100, Taiwan

Contributed by Chi-Huey Wong, May 19, 2011 (sent for review February 1, 2011)

Protein glycosylation is an important posttranslational process,which regulates protein folding and functional expression. Studieshave shown that abnormal glycosylation in tumor cells affectscancer progression and malignancy. In the current study, we haveidentified sialylated proteins using an alkynyl sugar probe in twodifferent lung cancer cell lines, CL1-0 and CL1-5 with distinct inva-siveness derived from the same parental cell line. Among the iden-tified sialylated proteins, epidermal growth factor receptor (EGFR)was chosen to understand the effect of sialylation on its function.Wehave determined the differences in glycan sequences of EGFR inboth cells and observed higher sialylation and fucosylation of EGFRin CL1-5 than in CL1-0. Further study suggested that overexpressionof sialyltransferases in CL1-5 and α1,3-fucosyltransferases (FUT4 orFUT6) in CL1-5 and A549 cells would suppress EGFR dimerizationand phosphorylation upon EGF treatment, as compared to thecontrol and CL1-0 cells. Such modulating effects on EGFR dimeriza-tion were further confirmed by sialidase or fucosidase treatment.Thus, increasing sialylation and fucosylation could attenuate EGFR-mediated invasion of lung cancer cells. However, incorporation ofthe core fucose by α1,6-fucosylatransferase (FUT8) would promoteEGFR dimerization and phosphorylation.

sialic acid ∣ glycoproteomics ∣ glycan sequencing ∣ click chemistry ∣mass spectrometry

Protein glycosylation accounts for the vast majority of posttran-slational processes to affect protein folding, stability, solubi-

lity, and function. Abnormal glycosylation has been reported toassociate with malignant transformation of tumor cells. In parti-cular, the terminal position of glycans with sugars such as sialicacid and fucose is usually exhibited at a high level in tumor cells(1). These two major terminal sugars may participate in impor-tant cell–cell interactions and cell migrations in cancer metastasisand angiogenesis (2). However, little evidence has been shownto support the causative roles of sialylation and fucosylation inmalignant transformation. Because sialylation and fucosylationoccur globally, their exact roles in the growth and metastatic be-havior of tumors, especially at the molecular level, remain to beinvestigated.

Among the membrane glycoproteins, epidermal growth factorreceptor (EGFR) with tyrosine kinase activity has been the targetfor drug discovery as its activity correlates with tumor progres-sion. EGFR is overexpressed in various kinds of cancer cellsand participates in the regulation of cancer invasion, metastasis,and angiogenesis (3). It belongs to the EGFR family and trans-duces cell signaling through ligand (EGF)-induced receptordimerization (4), which then initiates its tyrosine kinase activityto direct the malignant behaviors of cancers (5). The extracellulardomains of the EGFR family members are divided into four sub-domains, with domains I and III participating in ligand binding(6) and domains II and IV for dimerization (7). In the state with-out ligand stimulation, an intramolecular interaction between

domains II and IV maintains EGFR in a tethered conformationto inhibit dimerization (8), and this conformation is openedup by EGF binding to expose the dimerization interface for re-ceptor activation (9, 10). After autophosphorylation, dimericEGFR recruits and activates various downstream signaling pro-teins to direct cell proliferation, migration, differentiation, andapoptosis (11).

Twelve N-linked glycosylation sites were reported to exist inthe extracellular region of EGFR (12). The major glycoformsof EGFR in an epidermoid carcinoma cell line, A431, have beenidentified (13). Some previous studies have indicated that theglycans could participate in regulating certain functions of EGFR(14). For example, mutagenesis to remove the Asn 420 and 579-linked glycans of EGFR and Asn 418-linked glycans of ErbB3causes ligand-independent spontaneous dimerization (15–17).Besides, overexpression of sialidase in A431 cells would enhanceEGFR activity, implying that sialylation of EGFR might inhibitits activation (18). On the other hand, core fucosylation was re-ported to aid EGFR activation: knocking down fucosyltrans-ferase (FUT) 8 attenuates EGFR phosphorylation (19), andknocking out FUT8 reduces the EGF-binding ability of EGFRand down-regulates EGFR-mediated MAPK signaling (20).Further studies have indicated that overexpressing FUT8 couldincrease the susceptibility of EGFR to specific tyrosine kinase in-hibitors, whereas knockdown of FUT8 decreases the susceptibil-ity (21). Thus, delineating the roles of glycosylation on EGFR orother proteins may provide previously undescribed insights intodisease biology and therapeutic strategy.

Lung cancer cells are known to express various sialylatedor fucosylated glycan epitopes on the surface including sLex,sLea, sTn, Ley, Globo H, polysialic acid, fucosyl GM1, GM2,etc. (1). To uncover the differences of protein glycosylationand further link to protein functions, we labeled glycosylated pro-teins with alkyne-sugar probes, followed by Cu [I]-catalyzed al-kyne-azide click chemistry (22) to identify the sialylated andfucosylated proteins from lung cancer cell lines CL1-0 andCL1-5 (23), two subpopulations that are derived from the sameparental cell line with distinct invasion capabilities. Here, weshowed that the more invasive cell line CL1-5 exhibited highersialylation and fucosylation levels and expressed more sialylatedproteins. Among these identified sialylated proteins, we chose

Author contributions: Chung-Hsuan Chen, K.-H.K., C.-J.Y., P.-C.Y., T.-L.H., and C.-H.W.designed research; Y.-C.L., H.-Y.Y., C.-Y.C., Chein-Hung Chen, P.-F.C., Y.-H.J., and T.-L.H.performed research; Y.-C.L., H.-Y.Y., C.-Y.C., Chein-Hung Chen, P.-F.C., Y.-H.J., and T.-L.H.analyzed data; and Y.-C.L., T.-L.H., and C.-H.W. wrote the paper.

The authors declare no conflict of interest.1Y.-C.L. and H.-Y.Y. contributed equally to this work.2To whom correspondence may be addressed. E-mail: tlhsu@gate.sinica.edu.tw orchwong@gate.sinica.edu.tw.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1107385108/-/DCSupplemental.

11332–11337 ∣ PNAS ∣ July 12, 2011 ∣ vol. 108 ∣ no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1107385108

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021

EGFR to further study the effect of sialylation and fucosylationon its functions, especially dimerization and phosphorylation.

ResultsIdentification of Sialyl Glycoproteins in CL1 Lung Cancer Cells. By em-ploying alkynyl sugars as probes for metabolically labeling andenriching cellular glycoproteins (22, 24), we compared the sialylglycoproteins expressed on the lung cancer cells CL1-0 and CL1-5. Fig. S1A shows the flowchart of the glycoproteomic approach.Fig. S1B shows the patterns and labeling intensity of the proteinextracts derived from these two cell lines labeling with alkynylManNAc [ManNAcyne, the precursor to CMP-alkynyl sialic acid(CMP-NeuAcyne)], and azido biotin probe. Consistently, moresialyl proteins were identified in CL1-5 cells (Fig. S1C andTable S1). Again, the merit of this labeling method was demon-strated by the results that >95% of the enriched peptides boreNXS/T sequon, and most of them were derived from membrane(72%) or secreted (13%) proteins. TheMALDI-TOFMS profilesof permethylated N glycans from CL1-0 and CL1-5 also showeddifferent sialylation levels between CL1-0 and CL1-5 cells and ahigher level of fucosylation in CL1-5 cells (Fig. S1D–F), althoughthe branching level of N glycans in CL1-0 and CL1-5 were similar(biantennary: 56.1% vs. 60.3%; triantennary: 28.5% vs. 26.7%;tetraantennary: 15.4% vs. 13.0%).

Sialylation of EGFR in CL1 Cells. Among these CL1-5 uniquelyexpressed sialyl proteins (Table S1), EGFR was chosen forfurther investigation for its important role in promoting tumorgrowth and metastasis, and the link of EGFR N-linked glycosyla-tion to its function (14, 16, 17). To verify if EGFR in CL1-5 wasoversialylated, we next examined the expression and sialylation ofEGFR in CL1-0 and CL1-5 cells. As shown in Fig. S2A, anti-EGFR antibody immunoprecipitated similar amounts of EGFRfrom very different quantities of CL1-0 and CL1-5 protein lysates(8∶1), indicating that EGFR was up-regulated in CL1-5 cells.Nevertheless, the sialylation level of EGFR as demonstrated inManNAcyne-treated cells was higher in CL1-5. To further con-firm the sialylation status, CL1-0/EGFR, an EGFR stable cloneof CL1-0 that expressed an equal level of EGFR as CL1-5 cells,and CL1-5 cells were subjected to lectin pull-down experiment bySNA (Sambucus nigra lectin, binds α2,6-linked sialic acid) andMALII (Maackia amurensis lectin II, binds α2,3-linked sialicacid). Fig. S2B showed that SNA and MALII pulled down moreEGFR from CL1-5 lysates. Accordingly, the relative percentagesof sialylated glycans in CL1-5 were also higher than in CL1-0(Figs. S2C and S1 E and F). These experiments confirmed thatthe EGFR in CL1-5 had higher sialylation than in CL1-0 cells.

Role of Sialylation on EGFR Activation. To understand how differentEGFRs might act in CL1-0 and CL1-5 cells and whether the sia-lylation status could influence EGFR activation, we next analyzedEGFR dimerization and phosphorylation upon EGF stimulationin both cells. Fig. 1A revealed that EGFR dimerization occurredin CL1-0/EGFR without EGF treatment, and EGF induced lessEGFR dimers in CL1-5 than in CL1-0/EGFR cells. To evaluatethe phosphorylation status of these two cells, cells were treatedwith EGFat various concentrations (Fig. 1B), or for different per-iods of time (Fig. 1C), and then examined for EGFR tyrosinephosphorylation. Consistent with the dimerization results, CL1-5 cells showed weaker EGFR tyrosine phosphorylation comparedto CL1-0-EGFR. Because CL1-5 showed higher sialylation thanCL1-0 and CL1-0-EGFR cells, we subsequently tested if EGF-in-duced EGFR dimerization would be intensified in CL1-5 cellsafter sialidase treatment. Accordingly, CL1-5 cells treated withα2,3/6/8-linked sialidase showed stronger EGFR dimerization(Fig. S3A) and phosphorylation (Fig. S3B), suggesting that sialy-lation suppressed ligand-induced EGFR dimerization and thedownstream signaling.

Role of Fucosylation on EGFR Activation.Our results that sialylationon EGFR could attenuate EGFR dimerization and phosphoryla-tion prompted us to inspect the effect of terminal fucosylation,another kind of glycosylation associated with inflammation andcancers, in CL1 cells. We performed quantitative-PCR to exam-ine the expression of FUT transcripts in CL1 cells. Comparing toCL1-0, CL1-5 showed a higher expression of FUT2 (2.5 folds),FUT4 (2.4 folds), FUT5 (2.4 folds), FUT6 (3.7 folds), FUT8(22.5 folds), and FUT11 (4.3 folds) (Fig. S4A), suggesting thatCL1-5 cells displayed higher α1,3- and α1,6-linked fucosylation(core fucosylation). To understand the significance of α1,3-linkedfucosylation on EGFR, we established FUT4 and FUT6 stableclones in lung adenocarcinoma A549 cells (A549-FUT4 andA549-FUT6) with Tet-off inducible protein expression system.Equal expression of EGFR in/on stable clones was confirmedby anti-EGFR immunoblotting and flow cytometric analysis(Fig. S4 B and C), and the increased fucosylation of EGFR inA549-FUT4/6 cells was detected by alkynyl fucose (Fucyne) label-ing (Fig. S4D). The results from lectin blotting confirmed that theincreased fucosylation of EGFR in these cells was mainly α1,3-linked, because AAL (Aleuria aurantia lectin, recognizes α1,2,α1,3/4, and α1,6-linked fucose), but not Lens culimaris aggluti-nin-A (recognizes α1,6-linked fucose) or Ulex europaeus agglu-tinin (recognizes α1,2-linked fucose), showed a stronger EGFR-binding signal in lectin blotting (Fig. S4E).

We next analyzed EGF-induced EGFR dimerization in A549-FUT4/6 cells. Compared with mock control, A549-FUT4 andA549-FUT6 cells showed less EGFR dimerization (Fig. 2A)and lower EGFR tyrosine phosphorylation levels (Fig. 2B). Aftertreating the cells with doxycycline to turn off FUT4 and FUT6expression, EGFR in these cells showed lower AAL binding(Fig. S5A), but increasing dimerization and tyrosine phosphory-lation levels (Fig. S5 B and C) upon EGF stimulation. These re-sults suggested that increasing α1,3-linked fucosylation on EGFRreduced receptor dimerization and activation.

Because CL1-5 expressed a very high level of FUT8 mRNAcompared to CL1-0 (Fig. S4A), we then tested if overexpressing

Fig. 1. EGF-induced EGFR dimerization and tyrosine phosphorylation.(A) Dimerization of EGFR on cell surface. (Lower) Quantitative results: first cal-culate EGFR dimer∶monomer ratio in each sample and then normalize withthe ratio of CL1-0/EGFR without EGF treatment. (B and C) Tyrosine phosphor-ylation of EGFR upon EGF stimulation at different doses (B) or for differentdurations (C). (Right) The relative intensity that was compared with phos-pho-EGFR∶EGFR ratios and normalization with the ratio of CL1-0/EGFR with-out EGF treatment. (D) Dimerization of sialidase-treated soluble form EGFR.

Liu et al. PNAS ∣ July 12, 2011 ∣ vol. 108 ∣ no. 28 ∣ 11333

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021

or knocking down FUT8 could influence the behavior of EGFR.For this purpose, we established CL1-0-FUT8, A549-FUT8stable lines (FUT8 overexpression), and CL1-5-FUT8 knock-down stable clones to examine EGF-induced EGFR dimerizationand tyrosine phosphorylation (Fig. S6). Different from what weobserved that α1,3-linked fucosylation suppressed EGF-inducedreceptor dimerization and phosphorylation, overexpressingFUT8 in CL1-0 and A549 did not influence (Fig. S6 A–F),whereas knocking down FUT8 in CL1-5 cells reduced (Fig. S6 Gand H) EGFR dimerization and phosphorylation. Our resultswere consistent with the report that FUT8 knockout cells areless sensitive to EGF treatment, and this is possibly due to thereduction of EGF-binding affinity when EGFR bears no corefucoses (20).

Site-Specific Glycoform Mapping and Glycan Sequencing of EGFR inCL1-0 and CL1-5 Cells. To profile the glycoforms of EGFR, thefull-length EGFR was overexpressed and purified for analysis.TheMALDI-TOFMS profiles (Fig. 3A) showed that EGFR fromCL1-5, compared to CL1-0, displayed a higher level of sialylatedglycan structures (highlighted by red squares). The detailed gly-cosylation patterns on EGFR were subsequently analyzed byliquid chromatography-MS/MS to reveal the glycoforms (Fig. 3Band Figs. S7 and S8). Through matching with the calculatedmasses of both tryptic peptide fragments and glycans fromConsortium for Functional Glycomics carbohydrate databases,and the appearance of fragmented glycans in MS/MS spectra,

every individual EGFR glycopeptide derived from CL1-0 andCL1-5 cells was compositionally assigned and quantified in asite-specific manner.

Ten N-linked glycosylation sites on EGFR peptides were iden-tified by Asn to Asp conversion after PNGase F treatment(Table S2). Within these 10 sites, 3 of them (Asn positions328, 337, and 599) were identified to compose mostly high man-nose-type N glycans (Man5 to Man9), 6 of them (Asn positions32, 151, 389, 420, 504, and 579) were attached mainly with com-plex-type N glycans, and Asn 544 was ligated with both high man-nose- and complex-type N glycans (Table 1 and Fig. S7). Theglycosylation analysis revealed that EGFR from CL1-0 displayedmore Man8 structure, and CL1-5 bore more bi- and triantennaryglycans attached with at least one sialic acid and one fucose re-sidue (Fig. 3B). The branching levels of complex-type N glycans

Fig. 2. EGF-induced EGFR dimerization and tyrosine phosphorylation inA549-Mock, -FUT4, and -FUT6 cells. (A) Dimerization of cell surface EGFR.(Lower) The quantitative results calculated by EGFR dimer∶monomer ratioin each sample normalized with the ratio of A549-Mock treated with20 ng∕mL EGF. (B) EGF-induced EGFR tyrosine phosphorylation. The relativeintensity indicated below was calculated as phospho-EGFR∶EGFR ratiosand normalization with the ratio of A549-Mock without EGF treatment.(C) Dimerization of fucosidase-treated soluble form EGFR.

Fig. 3. Summary of the glycoforms. (A) MALDI-TOF MS profiles of the per-methylated N glycans from EGFR immunoprecipitated from the lysates ofCL1-0 (Upper) and CL1-5 (Lower) cells. The molecular compositions of the ma-jor ½Mþ Na�þ molecular ion signals detected were assigned as annotated.High mannose-type N glycans were labeled simply as M5–M9, representingMan5–9ðGlcNAcÞ2. Monosaccharide symbols used were as follows: triangle,Fuc; square, HexNAc; circle, Hex; diamond, NeuAc. Glucose oligomer contami-nants were labeled as X. (B) The glycoforms analyzed according to glycopep-tides profiles were detected from the EGFR samples derived from CL1-0 andCL1-5 and site-specific sialic acid and fucose indexes (C) of the EGFR proteinsderived from CL1-0 and CL1-5. Types of glycans: Man, high mannose; Bi, bian-tennary; Tri, triantennary; Tetra, tetraantennary; Penta, penta-antennary;Hexa, hexa-antennary; F, fucose; S, sialic acid; N, N acetylhexosamine.Sialic acid index¼ ½ð%with one sialic acid× 1Þþ ð%with two sialic acids × 2Þþð%with three sialic acids × 3Þ þ ð%with four sialic acids × 4Þ�∕100; Fucoseindex ¼ ½ð%with one fucose × 1Þ þ ð%with two fucoses × 2Þ þ ð%with threefucoses × 3Þ�∕100.

11334 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1107385108 Liu et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021

on EGFR were similar in CL1-0 and CL1-5 cells (biantennary:82.8% vs. 82.3%; triantennary 11.8% vs. 15.4%; tetraantennary:2.7% vs. 2.0%; pentaantennary: 2.4% vs 0.3%; hexaantennary:0.4% vs. 0).

We next examined the level of sialylation and fucosylation oneach N glycosite with sialic acid and fucose indexes (Fig. 3C).Consistent with the result that EGFR in CL1-5 carried more sialicacids and fucoses, CL1-5 was identified to have more sialylatedand fucosylated glycans than CL1-0, especially at positions 389,420, and 544. Considering that CL1-0 and CL1-5 expressed muchhigher FUT8 than other FUTs (Fig. S4A), the fucose residue waspreferentially assigned to the core position in each glycan struc-ture. Detailed percentages of individual forms of glycopeptideswere shown in Fig. S8.

In Vitro Dimerization of Sialidase- and Fucosidase-Modified EGFR. Tofurther test whether sialylation and fucosylation have a role inEGFR dimerization, the FLAG-tagged soluble EGFR (sEGFR)was overexpressed in 293F cells and treated with glycosidases(α2,3/6/8/9-sialidase and α1,3/4-fucosidase). The digestion ofsEGFR with sialidase or fucosidase was confirmed by quantifyingthe remaining sialic acids attached on sialidase-treated sEGFR(Fig. S4F) or AAL lectin blotting (Fig. S4G), respectively. Ac-cording to the analysis, the sialic acids on sEGFR were removedcompletely by sialidase treatment (from approximately 6 to ap-proximately 0.2 sialic acids per sEGFRmolecule). We then testedthe in vitro dimerization of sialidase-treated sEGFR. As shown inFig. 1D, more dimers were formed in sialidase-treated sEGFRcompared to the untreated control, whereas treating with heat-inactivated sialidase did not influence sEGFR dimerization. Also,in accordance with the observations from cell-based experiments,the fucosidase-treated sEGFR yielded higher dimerization uponEGF ligation (Fig. 2C). All these results confirmed that sia-lylation and fucosylation could suppress EGFR dimerization.Together with the previous reports that Lex and sLex epitopeswere detected in EGFR of A431 cells (13, 25), our results indi-cated that Lex and sLex could affect EGFR dimerization.

Role of Sialylation and Fucosylation in EGF-Mediated Cell Invasion. Tounderstand the significance of sialylation and fucosylation onEGFR in tumor metastasis, we tested if sialidase treatment could

influence EGF-mediated CL1-5 cell invasion. CL1-5 cells werefirst confirmed for the low expression level of surface sialic acid6 h after sialidase treatment (Fig. S4H). EGF-mediated CL1-5cell invasion was then evaluated by matrigel-coated transwell as-say within 6 h. The results in Fig. 4A showed a higher invasionability of CL1-5 cells when exposed to EGF. Notably, treatingCL1-5 cells with sialidase enhanced EGF-mediated cell invasion.Nevertheless, A549-FUT4 and A549-FUT6 cells showed weakerinvasion ability than mock control cells, whereas doxycycline trea-ted A549-FUT4/6 and mock cells showed similar invasion ability(Fig. 4B). Here, our results suggested that sialylation and fuco-sylation could suppress the metastatic ability of cancer cells pos-sibly via affecting EGFR dimerization and activation.

In summary, we have identified sialylated and fucosylated pro-teins by metabolically labeling the glycoproteins of lung cancercells with ManNAcyne and Fucyne and compared the sialyland fucosyl protein profiles. Our results revealed a regulatory me-chanism of sialylation and fucosylation on EGF-mediated EGFRbehavior: Increasing sialylation and α1,3-linked fucosylationwould suppress EGFR dimerization and phosphorylation, whichcould in turn affect the metastatic ability of cancer cells (Fig. 4C).On the other hand, the core fucosylation would promote EGFRdimerization and phosphorylation.

DiscussionAberrant glycosylation has been discovered in many kinds ofcancer cells, but the exact functions of altered glycans on glyco-proteins or cellular tumorigenesis are still unclear. Many reportssuggest that structural changes in cell surface carbohydrates maypromote tumor transformation. For example, the α1,6-linkedfucosylation of complex-type glycans may be an important featureof tumor progression related to increased metastasis (26), over-expression ofN-acetylglucosaminyltransferase V can enhance theinvasion ability of glioma cells (27), and overexpression of FUT4would promote the proliferation of human epidermoid carcino-ma A431 cell by controlling the cell cycle machinery via MAPK

Table 1. Site-specific representative glycans of EGFR

Fig. 4. Sialylation and fucosylation suppress EGFR activation to modulateEGF-mediated invasion in lung cancer cells. (A) EGF-mediated invasion ofCL1-5 cells with or without sialidase treatment. (B) EGF-mediated invasionof A549-mock, A549-FUT4, and A549-FUT6 cells with or without doxycyclinetreatment. The average cell number under 100× magnifications was calcu-lated from five power fields for each sample for A and B. *, p < 0.05; **, p <0.001 by Student’s t test. (C) Sialylation and fucosylation suppress EGFRactivation.

Liu et al. PNAS ∣ July 12, 2011 ∣ vol. 108 ∣ no. 28 ∣ 11335

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021

and PI3K/AKT signaling pathway (28). Thus it is proposed thatspecific inhibitors targeting the glycosyltransferases associatedwith numerous cancers may help defend against cancers. How-ever, some other reports mention that glycosylation may havenegative effects on protein functions or cellular behavior. Forinstance, α2,6-sialylated glycans may inhibit glioma malignancybecause decreased α2,6-sialylation on glycoproteins (29, 30) oron gangliosides (31) enhances glioma formation and invasion; in-hibition of N-linked glycosylation by tunicamycin can inducefunctional E-cadherin-mediated cell–cell adhesion (32); removalof sialic acids on cell surface by sialidase enhances connexin-43gap junction functions (33). In virus infection, it has been ob-served that removal of glycans on influenza hemagglutinin en-hances binding to cell surface receptors (34). Previous studiesbased on site-directed mutagenesis of EGFR also reveal thatthe glycans at positions 420 and 579 of EGFR pose an inhibitoryeffect on EGF-independent dimerization (14, 17, 35).

In our study, we found that EGFR in CL1-5 exhibited highersialylation and fucosylation levels and resulted in lower dimeriza-tion and tyrosine phosphorylation than in CL1-0 during EGFstimulation. In addition, removal of sialic acids from EGFR bysialidase increased dimer formation of EGFR upon EGF treat-ment on cell and in vitro, and pretreating EGFR with fucosidasealso resulted in a similar dimerization enhancement in vitro.Moreover, induction of FUT4 and FUT6 overexpression in lungcancer A549 cells with Tet-off system decreased EGFR dimeriza-tion and activation, and this FUT4/6-mediated suppression wasreverted when treating the cells with doxycycline to turn offFUT4/6 overexpression (Fig. S5 and Fig. 4B). These results de-monstrated that sialylation and terminal α1,3-fucosylation couldinhibit EGFR activation in lung cancer cells, and these modifica-tions could actually affect EGFR-mediated signaling and cellularbehavior. Previous studies have reported the fractionation andidentification of EGFR glycans in A431 cells and indicated thatthe blood group A antigen, as exemplified by α-linked terminal Nacetylgalactosamine, can be highly expressed in EGFR (36, 37),whereas terminal sialylation and α1,3-fucosylation are limited inA431 cells. Thus, based on our finding, it is interesting to know ifA431 cell has enhanced EGFR dimerization and signaling. It isworthy of mentioning that the study also suggests a negative roleof blood group type A antigen/terminal N acetylgalactosaminein EGFR affinity, tyrosine kinase activation, and receptor turn-over (37).

One of the major locations that showed elevated sialylationand fucosylation of EGFR in CL1-5 was Asn position 420(Fig. 3C), the N-glycosylation site that has been shown to preventEGFR spontaneous dimerization (17). This implies that sialyla-tion and fucosylation on Asn 420 can be critical in suppressingEGFR dimerization and activation. Regarding the core fucosyla-tion, although CL1-5 expressed a much higher level of FUT8 thanCL1-0, overexpressing FUT8 in CL1-0 did not affect EGFRdimerization and phosphorylation, indicating that the core fuco-sylation could be fulfilled in the cells with even a limited amountof FUT8 expression. On the other hand, consistent with theresults reported previously, knocking down FUT8 expression inCL1-5 cells showed a slight reduction of EGFR activation, whichcould be contributed by a reduced affinity to EGF (20).

Our study provides a clear correlation between sialylation/fucosylation and EGFR dimerization/activation, but the conclu-sion is contradictory to the malignant phenotypes of CL1-5because CL1-5 cells display more sialylation/fucosylation andare still more invasive than CL1-0 (23). This is understandablebecause CL1-5 cells are selected based on the invasion ability in-stead of glycosylation patterns of the cells. This sorting systemresults in two distinct cell populations that express very differentsets of proteins, especially those for cancer progression; forexample, CL1-5 cells express more MMP-9 (23). Thus it is inap-propriate to directly link the invasive ability of CL1-5 and CL1-0

to only EGFR glycosylation in this complex situation. Neverthe-less, our results clearly showed that reducing the terminal sialyla-tion and α1,3-fucosylation on EGFR enhanced its dimerizationand activation, and according to our data, the sialic acid and fu-cose on Asn 420 glycans may be most effective to inhibit EGFRdimerization and phosphorylation.

This study raises a question that whether inhibiting sialyl-or fucosyltransferases is a practical way to counteract cancer,because EGFR, and possible other sialyl/fucosyl receptors, arewidely involved in the malignancy of many types of tumors. If sia-lylation and fucosylation suppress receptor activation, inhibitingsialylation and fucosylation may create more aggressive cancercells. Conversely, increasing sialylation or fucosylation in aggres-sive cancers may help counteract cancer progression; therefore,the enhancement of sialyl- or fucosyltransferase activity may be aunique therapeutic approach to tumors. Further attempts to ex-amine how glycoforms alter protein conformation and protein–protein interaction, the events that closely link with signalingand cellular behaviors, will provide more evidence to understandthe multifaceted roles of protein glycosylation.

Materials and MethodsEGFR Dimerization Assay. Purified FLAG-tagged sEGFR (1 μg), with or withoutglycosidase treatment, was incubated with 100 μg∕mL human EGF in 100mMNaCl∕20 mM Hepes buffer (pH 7.4) for 2 h at room temperature (RT).Cross-linker BS3 (Bis[sulfosuccinimidyl] suberate, Thermo Scientific; 0.25 mM)was added into the protein mixtures and allowed to react for 1 h at RT. Thesamples were analyzed by SDS-PAGE and immunoblotting with Peroxidase-conjugated anti-FLAG (M2) MAb (Sigma-Adrich). For analyzing the EGFR di-merization on cell surface, cells were first starved in serum-free medium over-night, followed by stimulation with EGF (0–50 ng∕mL) for 5 min. Cross-linkreaction was conducted by the addition of 1 mMBS3 and the incubation at RTfor 20 min, and terminated by adding 50 mM Tris-HCl (pH 7.4). The sampleswere analyzed by SDS-PAGE and Western blot with anti-EGFR Ab (CellSignaling).

Detection of EGFR Tyrosine Phosphorylation. Cells were starved in serum-freemedium overnight and stimulated with EGF (0–50 ng∕mL) for 5 min or with10 ng∕mL EGF for 0–10 min. The cells were washed with cold PBS immedi-ately and harvested with lysis buffer (1% NP-40, 25 mM Tris, 150 mM NaCl,3 mM KCl, pH 7.4, 1 × EDTA-free protease inhibitor cocktail from Roche) with1× phosphatase inhibitor (Roche). The EGFR in cell lysates was immunopre-cipitated with EGFR antibody (Cell Signaling) on protein G agarose (ThermoScientific). The immunoprecipitated samples were resolved SDS-PAGE andanalyzed for the phosphotyrosine signals by Western blot with antiphospho-tyrosine antibody (4G10, Millipore).

Glycosidase Treatment. For sialidase treatment on sEGFR, 1 μg sEGFR wasmixed with 28.6 mU∕μL sialidase (α2,3/6/8/9-linked, Prozyme) at 37 °C forovernight in 5 μL as instructed by the vendor. For sialidase treatment ofCL1-5 cells, 106 cells were incubated with 0.04 U∕mL sialidase (α2,3/6/8-linked,Roche) in 3 mL serum-free RPMI medium 1640 at 37 °C for 30 min. The reac-tion efficiency was further detected by sialic acid quantitation kit (Sigma-Aldrich). For cleaving fucoses on sEGFR, 1 μg sEGFR was mixed with 200 μUfucosidase (α1,3/4-fucosidase, from Calbiochem) in 10 μl of sodium phosphatebuffer (pH 5.0) at 37 °C overnight.

EGF-Mediated Invasion Assay. The 24-well transwell plates with inserts[8-μm-pore polycarbonate membrane (Corning)] were set up for invasionassay. First, the membranes of the inserts were coated with 50 μg Matrigel(BD Bioscience) in 100 μL RPMI medium 1640 (Invitrogen). Cells were resus-pended in RPMI medium 1640 at the density of 106 cells∕mL and seeded intothe inserts (100 μL∕insert) of transwell plates. The attractant human EGF(20 ng∕mL, Millipore) was diluted in RPMI medium 1640 and dispensed intothe lower chambers. Following the incubation for 6 h at 37 °C, the inserts wereremoved and the numbers of cells that migrated to the reverse side of themembrane were quantified by microscopy after nuclear staining with DAPI.

ACKNOWLEDGMENTS. We thank Professor Mien-Chie Hung (MD AndersonCancer Center, Houston, TX) for the EGFR construct, and Eric Huang,Bo-Hua Chen, and Po-Kai Chuang for technical support. The research wassupported by the National Science Council and Genomics Research Center,Academia Sinica, Taiwan.

11336 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1107385108 Liu et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021

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

2. Kannagi R, IzawaM, Koike T, Miyazaki K, Kimura N (2004) Carbohydrate-mediated celladhesion in cancer metastasis and angiogenesis. Cancer Sci 95:377–384.

3. De Luca A, et al. (2008) The role of the EGFR signaling in tumor microenvironment.J Cell Phys 214:559–567.

4. Lemmon MA (2009) Ligand-induced ErbB receptor dimerization. Exp Cell Res315:638–648.

5. Schlessinger J (2000) Cell signaling by receptor tyrosine kinases. Cell 103:211–225.6. Lax I, et al. (1989) Functional analysis of the ligand-binding site of EGF receptor

utilizing chimeric chicken human receptor molecules. EMBO J 8:421–427.7. Ogiso H, et al. (2002) Crystal structure of the complex of human epidermal growth

factor and receptor extracellular domains. Cell 110:775–787.8. Cho HS, Leahy DJ (2002) Structure of the extracellular region of HER3 reveals an

interdomain tether. Science 297:1330–1333.9. Dawson JP, et al. (2005) Epidermal growth factor receptor dimerization and activation

require ligand-induced conformational changes in the dimer interface. Mol Cell Biol25:7734–7742.

10. Ferguson KM, et al. (2003) EGF activates its receptor by removing interactions thatautoinhibit ectodomain dimerization. Mol Cell 11:507–517.

11. Citri A, Yarden Y (2006) EGF-ERBB signalling: Towards the systems level. Nat Rev MolCell Biol 7:505–516.

12. Zhen YJ, Caprioli RM, Staros JV (2003) Characterization of glycosylation sites of theepidermal growth factor receptor. Biochemistry 42:5478–5492.

13. Hanash SM, Pitteri SJ, Faca VM (2008) Mining the plasma proteome for cancer biomar-kers. Nature 452:571–579.

14. Takahashi M, et al. (2008) N-glycan of ErbB family plays a crucial role in dimer forma-tion and tumor promotion. Biochim Biophys Acta 1780:520–524.

15. Yokoe S, et al. (2007) The Asn(418)-linked N-glycan of ErbB3 plays a crucial rolein preventing spontaneous heterodimerization and tumor promotion. Cancer Res67:1935–1942.

16. Whitson KB, et al. (2005) Functional effects of glycosylation at Asn-579 of the epider-mal growth factor receptor. Biochemistry 44:14920–14931.

17. Tsuda T, Ikeda Y, Taniguchi N (2000) The Asn-420-linked sugar chain in human epider-mal growth factor receptor suppresses ligand-independent spontaneous oligomeriza-tion—Possible role of a specific chain in controllable receptor activation. J Biol Chem275:21988–21994.

18. Meuillet EJ, et al. (1999) Sialidase gene transfection enhances epidermal growth factorreceptor activity in an epidermoid carcinoma cell line, A431. Cancer Res 59:234–240.

19. Li WZ, et al. (2006) Down-regulation of trypsinogen expression is associated withgrowth retardation in alpha 1,6-fucosyltransferase-deficient mice: Attenuation ofproteinase-activated receptor 2 activity. Glycobiology 16:1007–1019.

20. Wang X, et al. (2006) Core fucosylation regulates epidermal growth factor receptor-mediated intracellular signaling. J Biol Chem 281:2572–2577.

21. Matsumoto K, et al. (2008) N-glycan fucosylation of epidermal growth factor receptormodulates receptor activity and sensitivity to epidermal growth factor receptortyrosine kinase inhibitor. Cancer Sci 99:1611–1617.

22. Hsu TL, et al. (2007) Alkynyl sugar analogs for the labeling and visualization ofglycoconjugates in cells. Proc Natl Acad Sci USA 104:2614–2619.

23. Chu YW, et al. (1997) Selection of invasive and metastatic subpopulations from ahuman lung adenocarcinoma cell line. Am J Respir Cell Mol Biol 17:353–360.

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

25. Stroop CJM, et al. (2000) Characterization of the carbohydrate chains of the secretedform of the human epidermal growth factor receptor. Glycobiology 10:901–917.

26. Dennis JW, Laferte S, Waghorne C, Breitman ML, Kerbel RS (1987) Beta-1-6 branchingof Asn-linked oligosaccharides is directly associated with metastasis. Science236:582–585.

27. Yamamoto H, et al. (2000) Beta1,6-N-acetylglucosamine-bearing N-glycans in humangliomas: Implications for a role in regulating invasivity. Cancer Res 60:134–142.

28. Yang XS, et al. (2010) Overexpression of fucosyltransferase IV promotes A431 cellproliferation through activatingMAPK and PI3K/Akt signaling pathways. J Cell Physiol225:612–619.

29. Yamamoto H, Kaneko Y, Rebbaa A, Bremer EG, Moskal JR (1997) alpha2,6-Sialyltrans-ferase gene transfection into a human glioma cell line (U373 MG) results in decreasedinvasivity. J Neurochem 68:2566–2576.

30. Yamamoto H, Oviedo A, Sweeley C, Saito T, Moskal JR (2001) alpha 2,6-Sialylation ofcell-surface N-glycans inhibits glioma formation in vivo. Cancer Res 61:6822–6829.

31. Kroes RA, et al. (2010) Overexpression of ST6GalNAcV, a ganglioside-specificalpha2,6-sialyltransferase, inhibits glioma growth in vivo. Proc Natl Acad Sci USA107:12646–12651.

32. de Freitas Junior JC, et al. (2010) Inhibition of N-linked glycosylation by tunicamycininduces E-cadherin-mediated cell-cell adhesion and inhibits cell proliferation in undif-ferentiated human colon cancer cells. Cancer Chemother Pharmacol, 10.1007/s00280-010-1477-8.

33. Li J, et al. (2010) Cell surface sialic acid inhibits Cx43 gap junction functions inconstructed Hela cancer cells involving in sialylated N-cadherin. Mol Cell Biochem344:241–251.

34. Wang CC, et al. (2009) Glycans on influenza hemagglutinin affect receptor bindingand immune response. Proc Natl Acad Sci USA 106:18137–18142.

35. Whitson KB, et al. (2005) Functional effects of glycosylation at Asn-579 of the epider-mal growth factor receptor. Biochemistry 44:14920–14931.

36. Cummings RD, Soderquist AM, Carpenter G (1985) The oligosaccharide moieties of theepidermal growth factor receptor in A-431 cells—Presence of complex-type N-linkedchains that contain terminal N-acetylgalactosamine residues. J Biol Chem260:11944–11952.

37. Defize LHK, et al. (1988) A431 cell variants lacking the blood group-A antigen displayincreased high affinity epidermal growth factor receptor number, nrotein-tyrosinekinase activity, and receptor turnover. J Cell Biol 107:939–949.

Liu et al. PNAS ∣ July 12, 2011 ∣ vol. 108 ∣ no. 28 ∣ 11337

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

10, 2

021