Overexpression of ST6GalNAcV, a -specific α2,6-, inhibits glioma growth in vivo

Roger A. Kroesa, Huan Heb, Mark R. Emmettb,c, Carol L. Nilssond, Franklin E. Leach IIIe, I. Jonathan Amstere, Alan G. Marshallb,c, and Joseph R. Moskala,1

aThe Falk Center for Molecular Therapeutics, Department of Biomedical Engineering, Northwestern University, Evanston, IL 60201; bNational High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32310; cDepartment of Chemistry and Biochemistry, Florida State University, Tallahassee, FL 32306; dPfizer Global Research and Development, San Diego, CA 92121; and eDepartment of Chemistry, University of Georgia, Athens, GA 30602

Edited* by Senitiroh Hakomori, Pacific Northwest Research Institute, Seattle, WA, and approved June 10, 2010 (received for review August 31, 2009)

Aberrant cell-surface glycosylation patterns are present on virtually (9, 10) and Nakamura et al. (11) has shown that human neuro- all tumors and have been linked to tumor progression, metastasis, blastomas, medulloblastomas, and astrocytomas shed ganglio- and invasivity. We have shown that expressing a normally quies- sides that can be detected in patient serum and cerebrospinal cent, -specific α2,6-sialyltransferase (ST6Gal1) gene in fluid. Li et al. (12) showed that shed ganglioside GD2 was gliomas inhibited invasivity in vitro and tumor formation in vivo. To markedly immunosuppressive and, as such, may facilitate tumor identify other glycogene targets with therapeutic potential, we cre- formation and progression. ated a focused 45-mer oligonucleotide microarray platform repre- We used focused 45-mer oligonucleotide microarrays coupled senting all of the cloned human glycotranscriptome and examined with quantitative RT-PCR (qRT-PCR) analyses to comprehen- fi sively query the human glyco-transcriptome (13) and found the glycogene expression pro les of 10 normal human brain speci- fi - α mens, 10 malignant gliomas, and 7 human glioma cell lines. Among signi cantly lower levels of ganglioside selective 2,6-sialyl- transferase, ST6 (α-N-acetyl-neuraminyl-2,3-β-galactosyl-1,3)-N- the many significant changes in glycogene expression observed, acetylgalactosaminide α2,6-sialyltransferase 5 (ST6GalNAcV), of particular interest was the observation that an additional α2,6- in malignant glioblastomas (92 primary tumor specimens and 7 sialyltransferase, ST6 (α-N-acetyl-neuraminyl-2,3-β-galactosyl-1,3)- N α human glioma cell lines) relative to normal brain (34 specimens). -acetylgalactosaminide 2,6-sialyltransferase 5 (ST6GalNAcV), was GD1α is the terminal structure of α-series and is expressed at very low levels in all glioma and glioma cell lines exam- synthesized from GM1b through the transfer of to its ined compared with normal brain. ST6GalNAcV catalyzes the forma- N-acetylgalactosamine residue through an α2,6-linkage with α tion of the terminal 2,6-sialic acid linkages on gangliosides. Stable GM1b α2,6-sialyltransferase ST6GalNAcV, which, in mice, is transfection of ST6GalNAcV into U373MG glioma cells produced (i)no only active toward GM1b (12). Significant down-regulation of change in α2,6-linked sialic acid-containing , (ii)in- ST6GalNAcV has been demonstrated in lung squamous cell creased expression of GM2α and GM3 gangliosides and decreased carcinoma (14). Its involvement in the inflammatory breast expression of GM1b, Gb3, and Gb4, (iii) marked inhibition of in vitro cancer phenotype (15) and in breast cancer metastases to the invasivity, (iv)modified cellular adhesion to fibronectin and laminin, brain (16) has been reported. Alterations in the adhesive prop- (v) increased adhesion-mediated protein tyrosine phosphorylation of erties of HeLa cells by modulation of ST6GalNAcV expression HSPA8, and (vi) inhibition of tumor growth in vivo. These results has also been demonstrated (17). Interestingly, there is little to strongly suggest that modulation of the synthesis of specific glioma no α2,6-sialytransferase (ST6Gal1) activity, the enzyme that cell-surface alters invasivity in a manner that may sialylates glycoprotein-associated, N-linked , in have significant therapeutic potential. malignant glioblastomas. Moreover, studies have shown that expression of this sialyltransferase in human glioma model sys- glycosyltransferase | | glioblastoma | heat shock 70 kDa tems inhibited both tumor growth and invasivity, by altering the protein 8 | glycosynapse capability of the α3β1 integrin, because of its altered terminal sialic acid composition, and reflected in changes in focal adhesion kinase activity (5, 6). berrations in cell-surface glycosylation patterns are a dis- Thus, studies were undertaken to evaluate the role of Atinguishing feature of all vertebrate tumor cells, including fi ST6GalNAcV in glioma invasivity by creating stable trans- brain tumors (1). Cell-surface glycoproteins have been identi ed fectants and evaluating them for alterations in cell-surface oli- that are associated with the invasive potential of malignant gliomas gosaccharide expression, invasive potential in vitro, intracellular (2). The overexpression of the glycosyltransferase, GnT-V in gli- protein phosphorylation, and tumorigenicity in vivo. omas, led to increased invasivity of these tumor cells in vitro, suggesting that β1,6-N-acetylglucosamine–bearing N- play Results a role in glioma invasivity (3), consistent with earlier reports Stable Transfection. We have demonstrated the relative absence showing a strong correlation between metastatic potential and the of ST6GalNAcV mRNA in clinical gliomas and glioma cell lines, β expression of tri- and tetra-antennary 1,6NAG-bearing N-glycans including the U373MG cell line (13). The coding region of the (4). And recently, direct demonstration of a role for the oligo- ST6GalNAcV gene (UnigeneID: 181819), subcloned into the saccharide component of the glioma-associated integrin, α3β1, in pcDNA3 expression vector, was used for the generation of regulating glioma invasivity has been reported (5, 6). a panel of stable ST6GalNAcV-expressing cell lines. We chose Gangliosides are sialic acid-bearing glycosphingolipids that are an important part of the cell surface expressed fl on all mammalian cells. They in uence tumor growth and pro- Author contributions: R.A.K. and J.R.M. designed research; R.A.K., H.H., M.R.E., C.L.N., and gression through modulation of adhesion, migration, and an- F.E.L. performed research; M.R.E., I.J.A., and A.G.M. contributed new reagents/analytic giogenesis, and their expression is markedly altered in a variety tools; R.A.K., H.H., M.R.E., C.L.N., F.E.L., and J.R.M. analyzed data; and R.A.K. and J.R.M. of brain tumors and brain tumor model systems. Hamasaki et al. wrote the paper. (7) have reported that GT1b ganglioside can be used as a brain The authors declare no conflict of interest. metastasis marker. Mennel and Lell (8) also corroborated earlier *This Direct Submission article had a prearranged editor. reports that simpler ganglioside patterns were associated with 1To whom correspondence should be addressed. E-mail: [email protected]. the malignancy progression, but not the proliferation patterns, of This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. astrocytomas. And a series of reports by Ladisch and coworkers 1073/pnas.0909862107/-/DCSupplemental.

12646–12651 | PNAS | July 13, 2010 | vol. 107 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.0909862107 Downloaded by guest on October 2, 2021 S25) by use of a detailed, sensitive, semiquantitative method using extraction, one-step nano-liquid chromatography (nLC) separation, and high-resolution mass analysis (18, 19). We observed significant modulation of specific gangliosides in the α-series pathway (Fig. 2). As ST6GalNAcV directs the synthesis of GD1α in mice, we initially expected that the increased cell surface SNA reactivity was due to specific increases in this rare, brain-specific ganglioside. The lipid profiling data clearly dem- onstrate that, despite no change in GD1α content in the trans- fectants, there was an ≈7-fold increase in another α2,6- sialoganglioside, GM2α, with a concomitant ≈3-fold decrease in GM1b and ≈2-fold decreases in Gb3 and Gb4. A 2-fold increase in GM3 was also observed in the transfectant clones. Sialic acid linked to GalNAc is thought to be relatively rare in most tissues. Although GM2α may constitute only a minor ganglioside com- ponent, the lack of appropriate standards does not allow mea- surement of its absolute concentration. Nonetheless, the 7-fold increase in GM2α containing the d18:1-C16:0 structure common in many tumor cells had the highest ion abundance Fig. 1. ST6GalNAcV mRNA expression in U373MG/ST6GalNAcV clones. (A) measured across all transfectant samples (Table S2 and Fig. S2). Transcript abundance, normalized to GAPDH, was calculated by qRT-PCR. Data Electron-induced dissociation (EID) experiments, performed represent mean (± SD) of triplicate analyses. (**P < 0.01; ***P < 0.001; two- on selected ions from the nLC effluent after peak-parking to tailed, unpaired Student’s t test). n/s, not significant (P > 0.05). (B–E) Immu- differentiate between globo-series Gb4 and iso-globoseries iGb4 nohistochemical analysis after incubation with 10 μg/mL FITC-SNA (Matreya) isomers, confirmed the presence of Gb4 and not iGb4 (Fig. S3). and 1% BSA. Staining of U373MG parental cells (B), pcDNA3-transfected EID fragmentation of Gb3/iGb3 was also performed, but be- control cells (C), and ST6GalNAcV-transfected cells, clones S8 and S25 (D and E). cause of the highly symmetrical nature of the residue, we were unable differentiate between Gb3 and iGb3. Because Gb4 is biosynthesized from Gb3 and not iGb3, Gb3 is the most likely four individually isolated clones for further study and quanti- isomer detected. tated ST6GalNAcV mRNA expression by qRT-PCR (Fig. 1). The morphology of these clones was more rounded and less In Vitro Invasivity. The four clones examined represented a spec- dendritic than the parental U373MG cells or the control trum of ST6GalNAcV overexpression ranging from 2- to 25-fold. pcDNA3 transfectants. The proliferation rates of the trans- The decrease in in vitro invasivity was directly related to the level fectants did not significantly differ from controls, with doubling of expression of ST6GalNAc mRNA, with clones S8 and S25, the times of ≈30 h (Table S1). highest expressers, displaying ≈20% of the invasivity of the control cells (Fig. 3). Thus, within the limits of quantitation used Expression of Cell Surface Glycoconjugates. Prominent cell surface in these analyses, it appears compelling that the greater the immunohistochemical staining of the transfectant clones with ST6GalNAcV mRNA levels, the greater the quantity of cell FITC-SNA was detectable only in the highest expressers (S8 and surface-expressed α2,6-linked and the lesser the in S25; Fig. 1). The effects of ST6GalNAcV overexpression on vitro invasivity of the transfectants. overall glycoprotein sialylation, evaluated by using SNA lectin Western blot analysis, demonstrated no alterations in terminal In Vitro Adhesion. One of the primary mechanisms to modulate α2,6-sialylation (Fig. S1). cellular invasivity is via differential adhesion to extracellular ma- We compared polar lipid changes in control U373MG cells trix (ECM). Compared with controls, a marked reduction in ad- and the two highest expressing ST6GalNAc transfectants (S8 and hesion to fibronectin (≈20–40%) and increase (≈40–60%) in MEDICAL SCIENCES

Fig. 2. α-Series ganglioside profile of ST6GalNAcV transfectant (S25) via nLC-MS analysis (Left) with gangliosides demonstrating significant overexpression in the transfectants outlined in red, and those demonstrating significantly decreased expression outlined in green. The approximate fold-difference in expression is indicated. Linear quadrupole ion trap product ion mass spectrum for precursor ions of m/z 1354.783 (Right), with fragmentation pattern consistent with the structure of GM2α.

Kroes et al. PNAS | July 13, 2010 | vol. 107 | no. 28 | 12647 Downloaded by guest on October 2, 2021 Fig. 3. In vitro invasion assay. Evaluation of the relative invasivity of the transfected subclones compared with pcDNA3-transfected controls. The data are the average ± SD (bars) values of assays performed in triplicate. (**P < 0.01; ***P < 0.001; two-tailed, unpaired Student’s t test). n/s, not significant (P > 0.05).

adhesion to laminin was observed in the two clones (S8 and S25) demonstrating comparatively high levels of ST6GalNAcV mRNA expression and cell surface α2,6-linked gangliosides (Fig. 4). No significant difference in adhesion to collagen was demonstrated.

Adhesion-Mediated Protein Tyrosine Phosphorylation. Altered ad- hesivity of U373MG cells to fibronectin by modulation of cell surface N-linked glycoconjugates is mediated by altered tyrosine phosphorylation of p125FAK (5). Regulation of laminin-mediated signaling in other systems has also been described (20). After adhesion to laminin, both the qualitative and quantitative pattern of phosphorylated proteins in each of the clones was identical to that observed in control cells (Fig. S4). After adhesion to fibro- nectin, however, there was a striking difference in the level of phosphorylation of a 75-kDa protein in all of the transfectants. The abundance of this 75-kDa phosphoprotein directly correlated with the level of ST6GalNAcV expression (Fig. 5A). The identity of this 75-kDa phosphoprotein (pp75) and the signaling pathway altered by aberrant expression in these transfectants was examined.

Identification of pp75. After Off-Gel Electrophoretic (OGE) and antiphosphotyrosine Western blot analyses, pp75 was identified in the fraction corresponding to an approximate pI range of Fig. 5. Adhesion-mediated protein tyrosine phosphorylation. (A) Cells were incubated in fibronectin-coated plates, unattached cells were removed, and phosphotyrosine-modified proteins were detected as described in Methods. Lane 1, parental U373MG cells; lane 2, pcDNA3 control cells; lane 3, clone S8; lane 4, clone S23; lane 5, clone S25; lane 6, clone S32. (B) Identification of HSPA8 by nanoLC-MS/MS. The precursor ion shown is m/z 740.88, and the resultant peaks searched against the IPI human database by using both the Mascot 2.2 search engine and the SEQUEST algorithm. Eleven unique tryptic peptides, as shown, matched the heat shock cognate 71-kDa protein. (C) Immunoprecipitation analysis of phospho-HSPA8 in the ST6GalNAcV-trans- fected cells. Lanes 1 and 2: pcDNA3-transfected control cells and ST6Gal- NAcV-transfected clone (S25) were incubated in fibronectin-coated plates. HSPA8 protein content was detected by Western blot analysis using HRP- conjugated anti-HSPA8 (Millipore) and visualization by ECL. Lanes 3 and 4: After adhesion to fibronectin, pcDNA3-transfected control cells and ST6GalNAcV-transfected clone (S25) were lysed in RIPA buffer, 200 μgof precleared lysate protein immunoprecipitated with antibody to HSPA8 (1B5; AbCam) and analyzed by using an antiphosphotyrosine antibody, as de- scribed in Methods.

analysis 5.3–5.6. After excision of the corresponding region of a sister Coomassie-stained gel, trypsin digestion and nano LC- MS/MS analyses identified two major proteins: the clathrin un- Fig. 4. In vitro adhesion assay. Relative adhesion of the two highest coating ATPase HSPA8/HSC70 (with 11 unique peptides rep- ST6GalNAcV-expressing transfectants (clones S8 and S25) on fibronectin- (A), resenting 212 of 646 amino acids spanning >99.5% of the laminin- (B), or collagen type I-coated (C) plates compared with control B pcDNA3 cells. Data are average ± SEM (bars) values of three values taken protein; Fig. 5 ) and KEAP1 (with 3 unique peptides repre- from a representative experiment. n/s, not significant; *P < 0.05; **P < 0.01; senting 66 of 624 amino acids spanning 48% of the protein). ***P < 0.001 from a Student’s t test (unpaired). Immunoprecipitation of extracts with the respective primary

12648 | www.pnas.org/cgi/doi/10.1073/pnas.0909862107 Kroes et al. Downloaded by guest on October 2, 2021 onstrated compensatory shunting of ganglioside biosynthesis toward the normally quiescent “0” pathway in murine cells but not in human cells, with differing alternative glycosphingolipid pathways used to functionally compensate for aberrant ganglio- side biosynthesis (21, 22). The reduced expression of Gb3 and Gb4 observed in the transfectants is also consistent with such potential shunting of ganglioside biosynthetic pathways toward the synthesis of GM2α. At any rate, the stable overexpression of ST6GalNAcV leads to the production of nascent α2,6-linked sialoglycoconjugates on the cell surface. It also follows that the relatively low, yet significant, cell surface SNA reactivity observed in the clones exhibiting the highest levels of ST6GalNAcV expression is likely due to the low affinity of this lectin for sialic acid that is α2,6-linked to N-acetylgalactosamine, as opposed to galactose. The modest increase in GM3 synthesis, a ganglioside pos- sessing modulatory effects on glioma growth, adhesion, and invasivity (23), is also intriguing. At high concentrations, GM3 inhibits glioma cell proliferation predominantly by inhibition of Fig. 6. Intracranial tumorigenicity. Glioma cells (1.25 × 106) were stereo- tyrosine phosphorylation of EGF, PDGF, or FGF receptors. In tactically injected into the basal ganglia of SCID mice. Eight weeks later, our studies, no changes in cellular proliferation or in tyrosine brains were harvested, and sections stained with H&E (A–C). The pcDNA3 phosphorylation of these receptors in the transfectant clones vector-transfected U373MG cells (A) formed large tumors (arrow). The were observed. High concentrations of GM3 also inhibit tumor ST6GalNAcV transfectants [clone S8 (B) and clone S25 (C)] formed signifi- cell invasion predominantly via interference with integrin re- cantly smaller tumors, with tumors primarily surrounding the needle-tract ceptor function, which would also be reflected in events down- (arrows). (D) Differences in tumor maximal cross-sectional area between stream of the receptor. The predominant integrin in U373MG groups were determined by ANOVA followed by Fisher’s PLSD post hoc test. cells is α3β1, whose primary downstream effect is tyrosine Ten mice per cell line were used in each group. Data are expressed as av- phosphorylation of p125FAK (5). We also did not observe ad- erage ± SEM (bars). (*** P < 0.0001). hesion-mediated alterations in p125FAK expression in the trans- fectants. We interpret these observations to suggest that the 2- fold increase in GM3 content in the transfectants was not suffi- followed by antiphosphotyrosine Western blot analy- cient to explain the differences in in vitro invasivity or inhibition sis demonstrated a significant increase in HSPA8 phosphoryla- fi C of in vivo tumor growth that we have demonstrated. tion after adhesion to bronectin (Fig. 5 ). No change in Keap1 In addition to their well studied roles as epitopes/cell surface phosphorylation was observed (Fig. S5). markers, gangliosides participate in multiple biological functions, including inter- and intracellular signaling functions (24). Aber- Intracranial Tumorigenicity. Intracranial tumor growth by ST6Gal- rantly expressed cell surface gangliosides have been demon- NAcV-transfected U373MG (clones S8 and S25), parental strated to directly impact intracellular signaling. Alteration of U373MG, and pcDNA3 vector-transfected control cells was glycosylation in glycolipids affects intracellular localization of examined after stereotactic implantation into SCID mice (Fig. 6 integrin, src, and caveolin into or out of glycolipid-enriched A–C ). Both parental and vector-transfected U373MG cells microdomains (25). The association of GD2 with the integrin/ formed large tumors in vivo. Tumors formed by the ST6GalNAcV- FAK macromolecular complex has also been demonstrated (26). – transfected U373MG clones were, on average, 20 25% the size of GD3-mediated inhibition of laminin 5-dependent cell motility is D those formed by vector-transfected U373MG cells (Fig. 6 ). due to disruption of specific interactions between CD9 and α3 integrin in GD3-containing microdomains (27). Ganglioside Discussion alterations in epithelial cells leading to changes in (i) adhesion to In this study, we have demonstrated that ST6GalNAcV over- specific extracellular matrix components, (ii) relative rates of expression in tumorigenic, human U373MG glioma cells leads to cellular proliferation and apoptosis, (iii) protease activation and (i) increased expression of GM2α and GM3 gangliosides, and de- function, and (iv) disruption of cell surface integrin:growth factor creased expression of GM1b, Gb3, and Gb4, (ii) no change in α2,6- receptor associations have also been described (28). Signifi- linked sialic acid-containing glycoproteins, (iii) marked inhibition cantly, GT1b-induced apoptosis in SCC12 cells by direct binding of in vitro invasivity with no effect on cellular proliferation, (iv) to α5β1 leads to decreased activity of the integrin-linked kinase/ modified cellular adhesion to fibronectin leading to significantly protein kinase B/AKT pathway (29). Although the detailed increased adhesion-mediated protein tyrosine phosphorylation of fi molecular mechanisms of speci c glycosphingolipid modulation MEDICAL SCIENCES HSPA8 that directly correlated with decreased in vitro invasivity, on glioma invasivity may differ, it is nonetheless clear that ab- and (vi) significant inhibition of in vivo tumor growth. These results errant signal transduction plays a pivotal role. strongly suggest that modulation of the synthesis of specific glioma The opposing effects of ST6GalNAcV overexpression on ad- cell-surface glycosphingolipids alters adhesion and invasivity in hesion to laminin vs. fibronectin are seemingly paradoxical. Up- a manner that may play a significant role in gliomagenesis. regulated expression of ST6GalNAcV may result in signal In mice, ST6GalNAcV encodes an enzyme catalyzing the for- transduction cascades that culminate in different membrane mation of GD1α, a relatively minor brain-specific ganglioside in protein expression profiles that, on one hand, are favorable to the “0” biosynthetic pathway (21). The lack of GD1α and the adhesion to laminin, and on the other hand, not favorable to elevated expression of GM2α in the U373MG stable transfectants adhesion to fibronectin. In support of this hypothesis, we did not described here is unexpected and may result from compen- measure any significant changes in signal transduction (measured satory mechanisms after long-term ST6GalNAcV expression. In at the level of laminin-mediated tyrosine phosphorylation) in the humans, the “a” and “b” biosynthetic pathways are far more ac- transfectants. Another possibility is that the effects on the tive than the “0” pathway, which would result in synthesis of GT1a receptors responsible for adhesion to different substrates may be from GD1a and GQ1b from GT1b. Because we did not detect any differentially susceptible to glycolipid-associated alterations in of these products in our lipidomic analysis, it may be more likely membrane microdomains in the transfectants. Altered glioma that the synthesis of GM2α results from altered substrate avail- cell adhesion to laminin has been demonstrated not to be suffi- ability. Recent studies in GM3-deficient fibroblasts have dem- cient to affect invasion (30).

Kroes et al. PNAS | July 13, 2010 | vol. 107 | no. 28 | 12649 Downloaded by guest on October 2, 2021 Unlike the effects of adhesion to fibronectin on protein tyrosine Methods phosphorylation in glioma cells expressing the glycoprotein and Stable Transfection. Cells were maintained exactly as de- FAK ST6Gal1 (5), no decrease in p125 phosphorylation was ob- scribed (5). The 1.3-kb protein coding region of the human ST6GalNAcV served in these studies. There was, however, a direct relationship gene (UnigeneID: 181819), subcloned into the pcDNA3 expression vector, between the level of overexpression of ST6GalNAcV and the level was transfected into human U373MG glioma cells, and G418-resistant of protein tyrosine phosphorylation of a unique 75-kDa (p75) transfectants screened for ST6GalNAcV expression by qRT-PCR. protein after adhesion to fibronectin. Detailed phosphoproteomic analysis demonstrated that overexpression of ST6GalNAcV leads Immunohistochemistry. Cell surface α2,6-linked sialoglycoconjugate expres- to enhanced de novo phosphorylation of HSPA8/HSC70 in re- sion was confirmed by using fluorescein isothiocyanate (FITC)-conjugated sponse to attachment to fibronectin. We hypothesize that this Sambucus nigra agglutinin (SNA) (Matreya) exactly as described (5). SNA phosphorylation is caused by the expression of GM2α ganglioside recognizes epitopes containing α2,6-linked sialic acid linked to either Gal or synthesized by this enzyme and plays a key role in regulating al- GalNAc (45). tered adhesivity. It seems reasonable to generalize that these data imply that ganglioside signal transduction is an important part of SNA Lectin Blot Analysis. Expression of cell-surface glycoproteins was evalu- the mechanisms that underlie glioma invasivity. ated by SNA lectin blot analysis, as described (5), by using the DIG Glycan Further supporting this hypothesis, it has been reported that Differentiation Kit (Roche Molecular Biochemical), exactly as per the HSPA8 is colocalized with cytoskeletal-associated proteins in the manufacturer’s recommendations. extended pseudopodial protrusions of highly invasive variants of canine kidney tumor cells (31). There is increasing evidence that nLC-MS Detection of Polar . Polar lipids were extracted from cells and phosphorylation of cell-surface-expressed heat shock proteins analyzed by nLC-MS as described (18, 19) and detailed in SI Methods. Data- (HSPs) are intimately involved in tumor cell adhesion, motility, dependent MS/MS was performed in the linear quadrupole ion trap (CID) and invasion (32, 33). The effect of specific tyrosine phosphory- during collection of the ICR time-domain data. Mean fold change was cal- lation of HSPA8 in mitotic and tumor cells (34) on cellular lo- culated by direct comparison of average ion signal magnitude in trans- calization and function (35) has been demonstrated. In human fectants vs. control cells, the combined data representing different ceramide D54MG glioma cells, siRNA-mediated knockdown of HSPA8 chain compositions. To differentiate between Gb4/iGb4 isomers, EID exper- expression is associated with decreased glioma cell migration iments were also performed as detailed in SI Methods. through vitronectin-coated membranes (36). Alteration in HSPA8 phosphorylation has been demonstrated in integrin-mediated Invasion Assay. In vitro invasivity was examined by using Biocoat Matrigel platelet adhesion to collagen (37). The authors proposed that heat Invasion Chambers, exactly as described (5). shock proteins, including HSPA8, act as signaling scaffolds regu- lating platelet adhesion and spreading (38). HSPA8 is coexpressed Cell Adhesion Assay. Cell adhesion to human fibronectin, laminin, and col- with, and anchored by, cytoskeletal β- on the cell surface of lagen type I was measured, also as described (5). HTLV-susceptible T cells and is an integral part of a complex supported by palmitoyl (16:0)-oleoyl (18:1)-phosphatidylglycerol Adhesion-Mediated Protein Tyrosine Phosphorylation. Cells were incubated on in the lipid bilayer (39). The N-terminal ATPase-containing do- fibronectin-coated plates, unattached cells were removed, and Western blot main of HSPA8 involved in uncoating of clathrin-coated vesicles analysis was performed by using a HRP-conjugated antiphosphotyrosine has been shown to directly bind to cell surface 3′-sulfogalactoli- antibody (clone 4G10; Millipore) performed as described (5). pids, sulfogalactosyl ceramide, and sulfogalactoglycerolipid (40). Because membrane localization of HSPs appears to be restricted Phosphoproteomic Analysis. After adhesion to fibronectin as above, 300 μgof to lipid rafts/glycosynapses on transformed cells (41) and is spe- cell lysate protein was fractionated on an Agilent 3100 OFFGel Electropho- cifically associated with Gb3 in colon and pancreatic tumor resis Apparatus by using 24-cm IPG strips, pI 3–10 (GE Healthcare) in urea, cells (42), perhaps the differential phospho-HSPA8-mediated thiourea, DTT, glycerol, and the corresponding GE ampholine under con- signal transduction modulated by specific membrane ganglioside ditions recommended by Agilent. The buffer in each well, containing fo- content provides the scaffold for signal transduction cascades cused proteins, was recovered for Western blot analyses to identify in glial tumors. appropriate phosphoprotein-containing fractions by using an anti- Hakomori (43, 44) has discussed models of how altered mem- phosphotyrosine antibody. Protein identification in the corresponding band, brane glycosphingolipids (GSL) can influence the tumorigenic excised from Coomassie-stained sister gels, was performed by nanoLC-MS/ phenotype, based on their ability to form clusters at the cell surface MS of tryptic fragments, as described (46) on a Thermo Instruments LTQ-FT and interact with various functional components on the cell equipped with a Dionex Ultimate 3000 2D microcapillary HPLC system. membrane. These clusters (or “glycosynapses”) interact with such functional membrane proteins, including integrins, growth factor Immunoprecipitation. After adhesion to fibronectin-coated plates, cells were receptors, tetraspanins, and nonreceptor cytoplasmic protein lysed in RIPA buffer and 200 μg of precleared lysate protein was incubated kinases (e.g., src kinases and small G proteins) that ultimately with 5 μg/mL antibody to HSPA8 (1B5; AbCam). After incubation with Protein control GSL-modulated cell adhesion, growth, and motility. It is A/G Plus Agarose (Santa Cruz Biotechnology), beads were washed, dena- through this framework that modulation of cell surface carbohy- tured, and analyzed by Western blot analysis using antiphosphotyrosine drate expression and subsequent disorganization of cell mem- antibody, as above. brane components can mediate cell signaling, leading to changes in cellular phenotype. Intracranial Tumorigenicity. Intracranial tumorigenicity of ST6GalNAcV-, From the “glycosynapse” perspective, our results can be sum- pcDNA3 vector-transfected cells, and parental U373 MG cells was examined in marized in the following way: Manipulation of the synthesis of SCID mice (C.B-17 scid/scid, 6 wk of age; Charles River Laboratories), as de- cell surface gangliosides by ST6GalNAcV gene transfer alters scribed (6) and detailed in SI Methods. All mice were maintained in the the composition of specific membrane domains. Alteration in the animal facility of Northwestern University, Evanston, IL, and approved by composition of specific membrane domains, in turn, alters ad- their Institutional Animal Care and Use Committee. hesivity to ECM components, increases the phosphorylation of HSPA8 protein, and decreases the invasive potential and in vivo ACKNOWLEDGMENTS. We thank Mary Schmidt and Dr. Jeffrey Burgdorf for tumorigenicity of glioma cells. Based on this and previous studies expert technical assistance. Proteomics and informatics services were pro- (3, 5, 6), modulating the expression of α2,6-sialic acids on either vided by the CBC-UIC Research Resources Center Proteomics and Informatics gangliosides or the N-linked oligosaccharides expressed on cell Services Facility established by a grant from The Searle Funds at the Chicago Community Trust to the Chicago Biomedical Consortium. This research was surface signaling molecules (i.e., integrins) by modulating glyco- supported by a grant from the Falk Foundation (Chicago) (to J.R.M), gene expression may have therapeutic potential for the treat- National Science Foundation Grant DMR 0654118, and a grant from the ment of malignant gliomas. State of Florida.

12650 | www.pnas.org/cgi/doi/10.1073/pnas.0909862107 Kroes et al. Downloaded by guest on October 2, 2021 1. Mikkelsen T, Bjerkvig R, Laerum O, Rosenblum ML, eds (1998) Brain Tumor Invasion 26. Aixinjueluo W, et al. (2005) Mechanisms for the apoptosis of small cell lung cancer (Wiley Liss, New York). cells induced by anti-GD2 monoclonal antibodies: roles of anoikis. J Biol Chem 280: 2. Paulus W, Baur I, Beutler AS, Reeves SA (1996) Diffuse brain invasion of glioma cells 29828–29836. – requires beta 1 integrins. Lab Iinvest 75:819 826. 27. Kawakami Y, et al. (2002) Tetraspanin CD9 is a “proteolipid,” and its interaction with 3. Yamamoto H, et al. (2000) Beta1,6-N-acetylglucosamine-bearing N-glycans in human alpha 3 integrin in microdomain is promoted by GM3 ganglioside, leading to gliomas: Implications for a role in regulating invasivity. Cancer Res 60:134–142. inhibition of laminin-5-dependent cell motility. J Biol Chem 277:34349–34358. 4. Dennis JW, Laferte S, Waghorne C, Breitman ML, Kerbel RS (1987) Beta 1-6 branching 28. Wang XQ, Sun P, Paller AS (2002) Ganglioside modulation regulates epithelial cell of Asn-linked oligosaccharides is directly associated with metastasis. Science 236: adhesion and spreading via ganglioside-specific effects on signaling. J Biol Chem 277: 582–585. – 5. Yamamoto H, Kaneko Y, Rebbaa A, Bremer EG, Moskal JR (1997) alpha2,6- 40410 40419. Sialyltransferase gene transfection into a human glioma cell line (U373 MG) results in 29. Wang XQ, Sun P, Paller AS (2001) Inhibition of integrin-linked kinase/protein kinase B/ decreased invasivity. J Neurochem 68:2566–2576. Akt signaling: Mechanism for ganglioside-induced apoptosis. J Biol Chem 276: 6. Yamamoto H, Oviedo A, Sweeley C, Saito T, Moskal JR (2001) Alpha2,6-sialylation of 44504–44511. cell-surface N-glycans inhibits glioma formation in vivo. Cancer Res 61:6822–6829. 30. Merzak A, Koochekpour S, Pilkington GJ (1995) Adhesion of human glioma cell lines 7. Hamasaki H, et al. (1999) GT1b in human metastatic brain tumors: GT1b as a brain to fibronectin, laminin, vitronectin and collagen I is modulated by gangliosides in metastasis-associated ganglioside. Biochim Biophys Acta 1437:93–99. vitro. Cell Adhes Commun 3:27–43. 8. Mennel HD, Lell B (2005) Ganglioside (GD2) expression and intermediary filaments in 31. Jia Z, et al. (2005) Tumor cell pseudopodial protrusions. Localized signaling domains – astrocytic tumors. Clin Neuropathol 24:13 18. coordinating cytoskeleton remodeling, cell adhesion, glycolysis, RNA translocation, 9. Ladisch S, et al. (1987) Shedding of GD2 ganglioside by human neuroblastoma. Int J and protein translation. J Biol Chem 280:30564–30573. Cancer 39:73–76. 32. Lee KJ, et al. (2006) Release of heat shock protein 70 (Hsp70) and the effects of 10. Ladisch S, Chang F, Li R, Cogen P, Johnson D (1997) Detection of medulloblastoma and extracellular Hsp70 on matric metalloproteinase-9 expression in human monocytic astrocytoma-associated ganglioside GD3 in cerebrospinal fluid. Cancer Lett 120:71–78. – 11. Nakamura O, Iwamori M, Matsutani M, Takakura K (1991) Ganglioside GD3 shedding U937 cells. Exp Mol Med 38:364 374. by human gliomas. Acta Neurochir (Wien) 109:34–36. 33. Tsutsumi S, Neckers L (2007) Extracellular heat shock protein 90: A role for – 12. Li R, Gage D, McKallip R, Ladisch S (1996) Structural characterization and in vivo a molecular chaperone in cell motility and cancer metastasis. Cancer Sci 98:1536 1539. fi fi immunosuppressive activity of neuroblastoma GD2. Glycoconj J 13:385–389. 34. Rush J, et al. (2005) Immunoaf nity pro ling of tyrosine phosphorylation in cancer 13. Kroes RA, Dawson G, Moskal JR (2007) Focused microarray analysis of glyco-gene cells. Nat Biotechnol 23:94–101. expression in human glioblastomas. J Neurochem 103 (Suppl 1):14–24. 35. Chu A, Matusiewicz N, Stochaj U (2001) Heat-induced nuclear accumulation of hsc70s 14. Inamura K, et al. (2005) Two subclasses of lung squamous cell carcinoma with is regulated by phosphorylation and inhibited in confluent cells. FASEB J 15: different gene expression profiles and prognosis identified by hierarchical clustering 1478–1480. and non-negative matrix factorization. Oncogene 24:7105–7113. 36. Vila-Carriles WH, Zhou ZH, Bubien JK, Fuller CM, Benos DJ (2007) Participation of the fi 15. Dressman HK, et al. (2006) Gene expression pro les of multiple breast cancer chaperone Hsc70 in the trafficking and functional expression of ASIC2 in glioma cells. – phenotypes and response to neoadjuvant chemotherapy. Clin Cancer Res 12:819 826. J Biol Chem 282:34381–34391. 16. Bos PD, et al. (2009) Genes that mediate breast cancer metastasis to the brain. Nature 37. Gear AR, Simon CG, Polanowska-Grabowska R (1997) Platelet adhesion to collagen 459:1005–1009. activates a phosphoprotein complex of heat-shock proteins and protein phosphatase 17. Jaluria P, Betenbaugh M, Konstantopoulos K, Frank B, Shiloach J (2007) Application of 1. J Neural Transm 104:1037–1047. microarrays to identify and characterize genes involved in attachment dependence in 38. Polanowska-Grabowska R, Gear AR (2000) Heat-shock proteins and platelet function. HeLa cells. Metab Eng 9:241–251. – 18. He H, et al. (2007) Method for lipidomic analysis: p53 expression modulation of Platelets 11:6 22. , ganglioside, and phospholipid composition of U87 MG glioblastoma cells. 39. Sagara Y, et al. (2001) Phosphatidylglycerol participates in syncytium formation Anal Chem 79:8423–8430. induced by HTLV type 1-bearing cells. AIDS Res Hum Retroviruses 17:125–135. 19. He H, et al. (2010) Polar lipid remodeling and increased sulfatide expression are 40. Mamelak D, et al. (2001) The aglycone of sulfogalactolipids can alter the sulfate ester associated with the glioma therapeutic candidates, wild type p53 elevation and the substitution position required for hsc70 recognition. Carbohydr Res 335:91–100. -1 inhibitor, irinotecan. Glycoconj J 27:27–38. 41. Multhoff G, et al. (1995) A stress-inducible 72-kDa heat-shock protein (HSP72) is 20. Todeschini AR, Dos Santos JN, Handa K, Hakomori SI (2007) Ganglioside GM2- expressed on the surface of human tumor cells, but not on normal cells. Int J Cancer tetraspanin CD82 complex inhibits met and its cross-talk with integrins, providing 61:272–279. a basis for control of cell motility through glycosynapse. J Biol Chem 282:8123–8133. 42. Gehrmann M, Liebisch G, Schmitz G, Anderson R, Steinem C, De Maio A, Pockley G, 21. Shevchuk NA, et al. (2007) Alteration of ganglioside synthesis by GM3 synthase Multhoff G (2008) Tumor-specific Hsp70 plasma membrane localization is enabled by knockout in murine embryonic fibroblasts. Biochim Biophys Acta 1771:1226–1234. the glycosphingolipid Gb3. PLoS One 3:e1925. 22. Liu Y, Su Y, Wiznitzer M, Epifano O, Ladisch S (2008) Ganglioside depletion and EGF 43. Hakomori S (2004) -to-carbohydrate interaction, through glycosynapse, responses of human GM3 synthase-deficient fibroblasts. Glycobiology 18:593–601. as a basis of cell recognition and membrane organization. Glycoconj J 21:125–137. 23. Fujimoto Y, et al. (2005) Ganglioside GM3 inhibits proliferation and invasion of 44. Regina Todeschini A, Hakomori SI (2008) Functional role of glycosphingolipids and glioma. J Neurooncol 71:99–106. 24. Hakomori S, Yamamura S, Handa AK (1998) Signal transduction through glyco gangliosides in control of cell adhesion, motility, and growth, through glycosynaptic – (sphingo)lipids. Introduction and recent studies on glyco(sphingo)lipid-enriched microdomains. Biochim Biophys Acta 1780:421 433. microdomains. Ann N Y Acad Sci 845:1–10. 45. Shibuya N, et al. (1987) The elderberry (Sambucus nigra L.) bark lectin recognizes the 25. Kazui A, Ono M, Handa K, Hakomori S (2000) Glycosylation affects translocation of Neu5Ac(alpha 2-6)Gal/GalNAc sequence. J Biol Chem 262:1596–1601. integrin, Src, and caveolin into or out of GEM. Biochem Biophys Res Commun 273: 46. Kinter MA, Sherman NE (2000) Protein Sequencing and Identification Using Tandem 159–163. Mass Spectrometry (Wiley Interscience, New York). MEDICAL SCIENCES

Kroes et al. PNAS | July 13, 2010 | vol. 107 | no. 28 | 12651 Downloaded by guest on October 2, 2021