J. Biochem. 2015;157(4):197–200

doi:10.1093/jb/mvv023

Rapid communication Engineering of a 30 -sulpho-Galb14GlcNAc-specific probe by a single amino acid substitution of a fungal galectin Received January 15, 2015; accepted February 20, 2015; published online March 11, 2015

1 Institute of Traditional Chinese Medicine and Natural Products, College of Pharmacy, Jinan University, Guangzhou 510632, People’s Republic of China; 2Research Center for Stem Cell Engineering, National Institute of Advanced Industrial Science and Technology (AIST), Central 2, 1-1-1 Umezono, Tsukuba, Ibaraki 305-8568, Japan; 3Division of Functional Glycomics, Life Science Research Center, Institute of Research Promotion, Kagawa University, 1750-1 Ikenobe, Miki-cho, Kita-Gun, Kagawa 761-0793, Japan; and 4 Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology (AIST), Central 2, 1-11 Umezono, Tsukuba, Ibaraki 305-8568, Japan

*These authors contributed equally to this work. y Xinsheng Yao, Institute of Traditional Chinese Medicine and Natural Products, College of Pharmacy, Jinan University, Guangzhou 510632, People’s Republic of China. Fax: +86-20-85221559, email: [email protected] z Jun Hirabayashi, Research Center for Stem Cell Engineering, National Institute of Advanced Industrial Science and Technology (AIST), Central 2, 1-1-1 Umezono, Tsukuba, Ibaraki 305-8568, Japan. Fax: +81-29-861-3125, email: [email protected]

Among sulphated glycans, little is known about 30 -sulphation because of the lack of useful probes. In the course of molecular engineering of a fungal galectin from Agrocybe cylindracea, we found that a single substitution of Glu86 with alanine resulted in acquisition of specific binding for the 30 -sulpho-Galb1-4GlcNAc structure. Extensive glyco-technological analysis revealed that this property was obtained in a ‘loss-of-function’ manner. Though this mutant (E86A) had low total affinity, it showed substantial binding to a naturally occurring N-glycan, of which the terminal galactose is 3-sulphated. Moreover, E86A specifically bound to HeLa cells, in which galactose-3-O-sulfotransferases (Gal3ST2 or Gal3ST3) were over-expressed. Keywords: frontal affinity chromatography/galectin/ glycan microarry/glycoengineering/molecular evolution/sulphated glycan. Abbreviations: ACG, Agrocybe cylindracea galectin; Gal3ST, galactose-3-O-sulfotransferase; PAPS, 30 -phosphoadenosine-50 -phosphosulfate; ELISA, enzyme linked immunosorbent assay; HRP, horseradish peroxidase; FAC, frontal affinity chromatography; PA, pyridylamino.

ß The Authors 2015. Published by Oxford University Press on behalf of the Japanese Biochemical Society. All rights reserved

197

Downloaded from http://jb.oxfordjournals.org/ at University of California, San Diego on November 17, 2015

Dan Hu1,2,*, Hang Huang1,*, Hiroaki Tateno2, Shin-ichi Nakakita3, Takashi Sato4, Hisashi Narimatsu4, Xinsheng Yao1,y and Jun Hirabayashi2,z

Sulphated glycans play critical roles in binding to growth factors, hormones, chemokines, anti-coagulation factors, and cell adhesion molecules (1,2). Among them, 30 -sulpho-Galb1-4GlcNAc (30 -sulpho-LacNAc) forms a rare glycan motif found only in the N-glycans from thyroglobulin and Tamm-Horsfall glycoprotein (3, 4). Two galactose-3-O-sulfotransferase enzymes, Gal3ST2 and Gal3ST3, are responsible for transferring a sulphate group to the C-3 position of b-galactose in type 2 LacNAc structure (5, 6). Because the 30 -sulphoLacNAc is a preferred substrate of a1,3-fucosyltransferases, it is often fucosylated inside the cells to form 30 -sulpho-Lewis X (7, 8), a potent ligand for E- and Lselectins (2, 9). In the past, numerous studies have demonstrated that this glycoepitope plays an important role in the tumour metastasis (10—12), while the exact distribution and function of 30 -sulpho-LacNAc structure remain largely unknown. This is, at least in part, due to the lack of effective probes for its detection. Lectins, a group of carbohydrate-binding proteins, have been widely used as practical tools for cell typing, histochemical staining, and glycoprotein fractionation. Recently, the lectin microarray has been developed as an advanced method for rapid profiling of glycans (13—16). Galectins are a well-known family of lectins that are defined by an evolutionarily conserved b-galactose-binding protein structure (17, 18). In addition to the canonical binding site, individual galectins often have extended sites for modified b-galactosides, such as sialylated glycans, blood group antigens, polyN-acetyllactosamine, and sulphated glycans (19, 20). Notably, all of these specialized features are focused on the 3-OH group of galactose (20). As a tool, however, such broad specificity is disadvantageous for rigorous identification of particular glycans. In this context, recent site-directed mutagenesis studies have revealed that even a slight modification of the sugarbinding site could change their specificity rather dramatically (19, 21, 22). A fungal galectin from Agrocybe cylindracea, ACG, shows a much broader specificity than other galectins; i.e. not only for typical b-galactosides, such as Galb13/4GlcNAc (LacNAc) and Galb1-3GalNAc (T antigen), but also for 30 -substituted derivatives like those with Siaa2-3, sulpho-3 and GalNAca1-3 (23—25). These observations suggest that ACG could be a good template for engineering novel probes for particular glycans. In this context, Imamura et al. (24) recently conducted saturation mutagenesis of ACG and found that the E86D mutant has substantially lost the binding to all other glycans but retained an affinity for 30 -sialyl-LacNAc. More recently, we substituted each of five amino acid residues essential for the binding activity of ACG with Ala. Among the derived mutants (N46A, H62A, R66A, N75A and E86A), N46A showed a highly specific and even enhanced activity for GalNAca1-3Galb-containing glycans, while totally eliminating the affinity to all other canonical structures of b-galactosides (22).

D. Hu et al.

(34)

(13) (19) 50000

(6)

(10)

(33) (21)

S

to confirm that E86A has acquired a specific ability to bind to 30 -sulpho-LacNAc structure. To test whether E86A could bind endogenous glycans containing this glycoepitope, we purified the 30 -sulphoLacNAc-containing N-glycans from Tamm-Horsfall glycoprotein with reference to previous report (4; for preparation of the glycan, see Supplementary Fig. S4), and examined whether E86A really shows binding affinity to the naturally occurring 30 -sulphated glycan by frontal affinity chromatography (28). Consistent with the earlier results, E86A completely lost the binding ability to Galb-terminated (glycan No. 307, 405) or 30 sialylGalb-terminated glycans (glycan No. 505, 507, 508, 510 in Supplementary Fig. S5), while it exhibited a significant affinity to 30 -sulpho-LacNAc-containing N-glycans derived from Tamm-Horsfall glycoprotein. These results indicate that E86A is able to recognize naturally occurring 30 -sulpho-LacNAc-containing glycans. Unfortunately, the Kd value cannot be determined in this study because of the lack of an appropriate standard. However, it is estimated to be in the range 105 and 104 M assuming 30—60% of effective ligand content in a typical immobilization process (20, 28). The high specificity of the mutant E86A thus demonstrated prompted us to investigate whether this mutant could be of more practical use. To address this issue, HeLa cells were transfected with three Gal3ST enzymes, i.e. Gal3ST2, Gal3ST3 or Gal3ST4, as well as ABO transferases (ABO transferases A and B) as negative controls. The binding of E86A to these transfected cells was analysed by flow cytometry. As shown in Fig. 2, ACG bound to all these cells to a similar extent whichever transferases were transfected (data of ABO transferases A and B not shown). On the

(36)

(35) S

ACG

(38)

S

(37)

S

intensity

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Marker

0

ACG(E86A) 50000

(36)

Gal

Sulfo

GlcNAc

α-linkage

NeuAc

β-linkage

Fuc

S

6

8

4 3

2

S

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 Marker

0

Glycan NO.

Fig. 1 Analysis of the sugar-binding specificity of wild-type ACG and mutant E86A. The sugar-binding activity of ACG and E86A was analysed by glycoconjugate microarray based on previously described methods (26). ACG or E86A with an N-terminal Flag tag were expressed under the induction by 1 mM IPTG and lysed. After centrifugation, the derived supernatants were labelled by incubating with anti-Flag antibody and Cy3labelled secondary antibody. The labelled solutions were applied onto a glycoconjugate microarray and the bound fluorescent signals were detected by an evanescent-type scanner Glycostation Reader 1200 under Cy3 mode. Glycoconjugates used in the glycoconjugate microarray are listed in Supplementary Fig. S1.

198

Downloaded from http://jb.oxfordjournals.org/ at University of California, San Diego on November 17, 2015

However, this galectin had one more curious aspect: as shown in Fig. 1, E86A had lost all glycan-binding affinity except for that of 30 -sulpho-LacNAc (type II, glycan No 36 in Supplementary Fig. S1) when analysed by an evanescent-based glycoconjugate microarray (26) using the cell lysates as described previously (27). The observed affinity was highly specific, because no binding was detected towards other sulphated glycans, including 30 -sulpho-Galb1-3GlcNAc (type I, No: 34) and 30 -sulpho-Lewis A (No: 11). We also tested the specificity under different concentrations of purified E86A, and confirmed the specific binding of E86A to 30 -sulpho-LacNAc, while a background binding to other glycans at higher concentrations was observed (Supplementary Fig. S2). These results strongly suggest E86A is a good probe for 30 -sulphated type II LacNAc. To further validate the specific binding, E86A as well as three other mutants including H62A, R66A, N75A and wild-type ACG were expressed as a fusion to an Nterminal Flag tag and a C-terminal His tag in Escherichia coli, and the resultant proteins were purified by nickel affinity chromatography. The binding of each mutant and wild-type ACG were initially assessed by a sandwich binding assay using carbohydrate-conjugated PAA polymers, i.e. lactose-PAA, 30 -sialyllactose-PAA and 30 -sulpho-LacNAc-PAA, and Man-PAA as a negative control. As expected, wild-type ACG showed a prominent concentration-dependent binding to lactose-PAA, 30 -sialyllactose-PAA and 30 -sulpho-LacNAc-PAA, but no significant binding to Man-PAA (Supplementary Fig. S3). Consistent with the results of glycoconjugate microarray, E86A exhibited a substantial binding to 30 -sulpho-LacNAc-PAA, while it almost lost the binding ability to three other glycans. These results add evidence

Engineering of a 30 -sulpho-Galb1-4GlcNAc-specific probe

Cell number

60 40 20

Control ACG

HeLa (Gal3ST3)

Cell number

100 80 60 40 20 0

Cell number

60 40 20 0

0

102

103

104

105

40 20 0 102 103 104 105 Control E86A

80 60 40 20 0

0 102 103 104 105

100

Control E86A

80 60 40 20 0

0 102 103 104 105

Fluorescence intensity Fig. 2 Practical application of E86A in flow cytometry. HeLa cells were transfected with ABO transferase A, B or Gal-3-O-sulfotransferases. After 24 h, cells were harvested, washed and suspended in 1% BSA in PBS. 1  105 cells were incubated with 0.5 mg/ml FlagACG or E86A for 20 min followed by incubation with 0.5 mg/ml anti-Flag antibody and 0.5 mg/ml PE-labelled GAM antibody. The binding of lectins to the cells was analysed by FACSCanto-II cytometer.

other hand, E86A showed a more specific binding to the cells that were transfected with Gal3ST2 or Gal3ST3, but not to those transfected with Gal3ST4 or ABO transferases (data not shown). Previous studies on Gal3ST enzymes have demonstrated that both Gal3ST2 and Gal3ST3 can catalyse the 30 -sulphation of type II LacNAc (5, 6), while Gal3ST4 is responsible for 30 -sulphation of Galb1-3GalNAca (T antigen) (29). Thus, the present results provide further evidence that E86A has acquired a specific binding activity to 30 -sulpho-LacNAc (type II). As mentioned, 30 -sulpho-LacNAc is a rare sulphoglycan (3, 4). Besides, its fucosylated form, 30 -sulphoLewis X is highly expressed in various types of malignant tumour cells, serving as a potent ligand for both L- and E-selectins (2, 11). Given that 30 -sulphoLacNAc can be recognized by endogenous lectins, e.g. galectin-3 and -8, it may have unknown function(s) as a precursor of 30 -sulpho-Lewis X (30, 31). In the present study, we tailored a specific probe for detecting the 30 -sulpho-LacNAc structure from a versatile fungal galectin, ACG, which shows extremely broad specificity. Notably, this unique specificity has been obtained in a manner of ‘loss-of-function’ rather than ‘gain-of-function’, i.e. by a single amino acid substitution with Ala. Although the amino acid residue(s) involved in the sulphate recognition were not identified in the previous crystallography analysis (23), recent studies on the Nterminal domain of galectin 8 with 30 -sulpho-LacNAc have revealed importance of two amino acid residues within the S3 strand (Arg45 and Gln47) and Arg59 in the long loop between the S3 and S4 strands (30, 31).

Though these residues are not conserved in ACG, it is possible to speculate, based on the position of sulphate ion in the crystal structure of ACG with 30 -sulphoLacNAc, that the residues within the S3 strand and the S3—S4 loop are responsible for this function. In fact, R66A and H62A had completely lost their binding ability to 30 -sulpho-LacNAc, in both of which mutations locate on the S4 strand close to the supposed sulphate-binding sites, while on the other hand N75A and E86A retained this ability (Supplementary Fig. S3). Considering that Glu86 is far from the supposed sulphate binding sites, E86A mutation may have less effect on this function compared with the three other mutants. Further studies focusing the S3—S4 regions are necessary to elucidate the detailed recognition mechanism. From a practical viewpoint, enhancement of the E86A affinity to 30 -sulpho-LacNAc should be undertaken; e.g. by introducing basic amino acids (e.g. Arg, Lys) to form a salt bridge to the sulphate moiety to either the S3 strand or the S3—S4 loop region. More importantly, biological significance of the 30 -sulpho-LacNA should be elucidated by extensive approaches to this goal: they include; (i) histochemical screening of human tissues/cells using the E86A probe, (ii) identification of sulfotransferase(s) responsible for sulphation of the positive cells in (i), (iii) finding correlation between 3-O-sulphation and tumourigenesis/malignancy of these cells. In conclusion, we have succeeded in engineering a specific probe for 30 -sulpho-LacNAc, for which the biological function remains to be elucidated. This was achieved in a loss-of-function manner, an approach using a versatile lectin (ACG) as a starting material. A similar approach will be useful to engineer-specific probes for other carbohydrate epitopes to elucidate their functions.

Supplementary Data Supplementary Data are available at JB Online. Funding This work was supported in part by Grant-in-Aid for JSPS Fellows (J.H., 22-00218), grants from the National Natural Science Foundation of China (D.H., 81302673) , Natural Science Foundation of Guangdong Province (D.H., 2014A030313389), and the 111 Project (X.Y., No. B13038) from the Ministry of P. R. China. Guangdong Province Universities and Colleges Pearl River Scholar Funded Scheme (D.H., 2014). Conflict of Interest None declared.

References 1. Afratis, N., Gialeli, C., Nikitovic, D., Tsegenidis, T., Karousou, E., Theocharis, A.D., Pavao, M.S., Tzanakakis, G.N., and Karamanos, N.K. (2012) Glycosaminoglycans: key players in cancer cell biology and treatment. FEBS J 279, 1177—1197 2. Yuen, C.T., Lawson, A.M., Chai, W., Larkin, M., Stoll, M.S., Stuart, A.C., Sullivan, F.X., Ahern, T.J., and Feizi, T. (1992) Novel sulfated ligands for the cell adhesion molecule E-selectin revealed by the neoglycolipid technology among O-linked oligosaccharides on an ovarian cystadenoma glycoprotein. Biochemistry 31, 9126—9131 199

Downloaded from http://jb.oxfordjournals.org/ at University of California, San Diego on November 17, 2015

HeLa (Gal3ST4)

Control ACG

80

60

100

0 102 103 104 105

100

80

0

0 102 103 104 105

Cell number

0

Control E86A

100

Cell number

HeLa (Gal3ST2)

80

Cell number

Control ACG

100

D. Hu et al.

200

17. Barondes, S.H., Cooper, D.N., Gitt, M.A., and Leffler, H. (1994) Galectins. Structure and function of a large family of animal lectins. J. Biol. Chem. 269, 20807—20810 18. Leffler, H., Carlsson, S., Hedlund, M., Qian, Y., and Poirier, F. (2004) Introduction to galectins. Glycoconjugate J. 19, 433—440 19. Salomonsson, E., Carlsson, M.C., Osla, V., HendusAltenburger, R., Kahl-Knutson, B., Oberg, C.T., Sundin, A., Nilsson, R., Nordberg-Karlsson, E., Nilsson, U.J., Karlsson, A., Rini, J.M., and Leffler, H. (2010) Mutational tuning of galectin-3 specificity and biological function. J. Biol. Chem. 285, 35079—35091 20. Iwaki, J., Tateno, H., Nishi, N., Minamisawa, T., Nakamura-Tsuruta, S., Itakura, Y., Kominami, J., Urashima, T., Nakamura, T., and Hirabayashi, J. (2011) The Galbeta-(syn)-gauche configuration is required for galectin-recognition disaccharides. Biochim. Biophys. Acta 1810, 643—651 21. Konno, A., Yonemaru, S., Kitagawa, A., Muramoto, K., Shirai, T., and Ogawa, T. (2010) Protein engineering of conger eel galectins by tracing of molecular evolution using probable ancestral mutants. BMC Evol. Biol. 10, 43 22. Hu, D., Tateno, H., Sato, T., Narimatsu, H., and Hirabayashi, J. (2013) Tailoring GalNAc alpha 1-3Gal beta-specific lectins from a multi-specific fungal galectin: dramatic change of carbohydrate specificity by a single amino-acid substitution. Biochem. J 453, 261—270 23. Ban, M., Yoon, H.J., Demirkan, E., Utsumi, S., Mikami, B., and Yagi, F. (2005) Structural basis of a fungal galectin from Agrocybe cylindracea for recognizing sialoconjugate. J. Mol. Biol. 351, 695—706 24. Imamura, K., Takeuchi, H., Yabe, R., Tateno, H., and Hirabayashi, J. (2011) Engineering of the glycan-binding specificity of Agrocybe cylindracea galectin towards alpha(2,3)-linked sialic acid by saturation mutagenesis. J. Biochem. 150, 545—552 25. Yagi, F., Hiroyama, H., and Kodama, S. (2001) Agrocybe cylindracea lectin is a member of the galectin family. Glycoconjugate J. 18, 745—749 26. Tateno, H., Mori, A., Uchiyama, N., Yabe, R., Iwaki, J., Shikanai, T., Angata, T., Narimatsu, H., and Hirabayashi, J. (2008) Glycoconjugate microarray based on an evanescent-field fluorescence-assisted detection principle for investigation of glycan-binding proteins. Glycobiology 18, 789—798 27. Hu, D., Tateno, H., Kuno, A., Yabe, R., and Hirabayashi, J. (2012) Directed evolution of lectins with sugar-binding specificity for 6-sulpho-galactose. J. Biol. Chem. 287, 20313—20320 28. Tateno, H., Nakamura-Tsuruta, S., and Hirabayashi, J. (2007) Frontal affinity chromatography: sugar-protein interactions. Nat. Protoc. 2, 2529—2537 29. Seko, A., Hara-Kuge, S., and Yamashita, K. (2001) Molecular cloning and characterization of a novel human galactose 3-O-sulfotransferase that transfers sulfate to gal beta 1—>3galNAc residue in O-glycans. J. Biol. Chem. 276, 25697—25704 30. Ideo, H., Matsuzaka, T., Nonaka, T., Seko, A., and Yamashita, K. (2011) Galectin-8-N-domain recognition mechanism for sialylated and sulfated glycans. J. Biol. Chem. 286, 11346—11355 31. Ideo, H., Seko, A., Ohkura, T., Matta, K.L., and Yamashita, K. (2002) High-affinity binding of recombinant human galectin-4 to SO(3)(-)—>3Galbeta1—> 3GalNAc pyranoside. Glycobiology 12, 199—208

Downloaded from http://jb.oxfordjournals.org/ at University of California, San Diego on November 17, 2015

3. Spiro, R.G. and Bhoyroo, V.D. (1988) Occurrence of sulfate in the asparagine-linked complex carbohydrate units of thyroglobulin. Identification and localization of galactose 3-sulfate and N-acetylglucosamine 6-sulfate residues in the human and calf proteins. J. Biol. Chem. 263, 14351—14358 4. Hard, K., Van Zadelhoff, G., Moonen, P., Kamerling, J.P., and Vliegenthart, F.G. (1992) The Asnlinked carbohydrate chains of human Tamm-Horsfall glycoprotein of one male. Novel sulfated and novel N-acetylgalactosamine-containing N-linked carbohydrate chains. Eur. J. Biochem. 209, 895—915 5. El-Fasakhany, F.M., Uchimura, K., Kannagi, R., and Muramatsu, T. (2001) A novel human Gal-3O-sulfotransferase: molecular cloning, characterization, and its implications in biosynthesis of (SO(4)-3)Galbeta14(Fucalpha1-3)GlcNAc. J. Biol. Chem. 276, 26988—26994 6. Honke, K., Tsuda, M., Koyota, S., Wada, Y., IidaTanaka, N., Ishizuka, I., Nakayama, J., and Taniguchi, N. (2001) Molecular cloning and characterization of a human beta-Gal-30 -sulfotransferase that acts on both type 1 and type 2 (Gal beta 1-3/1-4GlcNAc-R) oligosaccharides. J. Biol. Chem. 276, 267—274 7. Chandrasekaran, E.V., Jain, R.K., and Matta, K.L. (1992) Ovarian cancer alpha 1,3-L-fucosyltransferase. Differentiation of distinct catalytic species with the unique substrate, 30 -sulpho-N-acetyllactosamine in conjunction with other synthetic acceptors. J. Biol. Chem. 267, 23806—23814 8. Degroote, S., Ducourouble, M.P., Roussel, P., and Lamblin, G. (1999) Sequential biosynthesis of sulfated and/or sialylated Lewis x determinants by transferases of the human bronchial mucosa. Glycobiology 9, 1199—1211 9. Yuen, C.T., Bezouska, K., O’Brien, J., Stoll, M., Lemoine, R., Lubineau, A., Kiso, M., Hasegawa, A., Bockovich, N.J., Nicolaou, K.C., and Feizi, T. (1994) Sulfated blood group Lewis(a). A superior oligosaccharide ligand for human E-selectin. J. Biol. Chem. 269, 1595—1598 10. Capon, C., Wieruszeski, J.M., Lemoine, J., Byrd, J.C., Leffler, H., and Kim, Y.S. (1997) Sulfated lewis X determinants as a major structural motif in glycans from LS174T-HM7 human colon carcinoma mucin. J. Biol. Chem. 272, 31957—31968 11. Shi, B.Z., Hu, P., Geng, F., He, P.J., and Wu, X.Z. (2005) Gal3ST-2 involved in tumor metastasis process by regulation of adhesion ability to selectins and expression of integrins. Biochem. Biophys. Res. Commun. 332, 934—940 12. Zhang, C.Y., Hu, P., Fu, D., Wu, W., Jia, C.Y., Zhu, X.C., and Wu, X.Z. (2010) 30 -Sulfo-Le(x) is important for regulation of integrin subunit alphaV. Biochemistry 49, 7811—7820 13. Angeloni, S., Ridet, J.L., Kusy, N., Gao, H., Crevoisier, F., Guinchard, S., Kochhar, S., Sigrist, H., and Sprenger, N. (2005) Glycoprofiling with micro-arrays of glycoconjugates and lectins. Glycobiology 15, 31—41 14. Kuno, A., Uchiyama, N., Koseki-Kuno, S., Ebe, Y., Takashima, S., Yamada, M., and Hirabayashi, J. (2005) Evanescent-field fluorescence-assisted lectin microarray: a new strategy for glycan profiling. Nat. Methods 2, 851—856 15. Pilobello, K.T., Krishnamoorthy, L., Slawek, D., and Mahal, L.K. (2005) Development of a lectin microarray for the rapid analysis of protein glycopatterns. Chembiochem 6, 985—989 16. Zheng, T., Peelen, D., and Smith, L.M. (2005) Lectin arrays for profiling cell surface carbohydrate expression. J. Am. Chem. Soc. 127, 9982—9983