Sialylation of CD55 by ST3GAL1 Facilitates Immune Evasion in Cancer

Glycosylation is a common form of cotranslational and posttranslational modification. It is a remarkable that over 80% of surface membrane and secreted proteins are N- and/or O-glycosylated. Glycosylation plays an important role in biological function, such as protein folding, protein degradation, protein–protein interaction, and signal transduction (1–4). Alteration of glycosylation is associated with an advanced cancer stage, progression, and metastasis (5). It mainly affects the terminal portion of the carbohydrate moiety of glycoproteins and glycolipids, leading to the expression of tumor-associated carbohydrate antigens. Many tumor-associated carbohydrate antigens contain sialylated structures, and increased sialylation of N-linked or O-linked oligosaccharides on cell surface glycoproteins is frequently observed in carcinoma cells (6, 7). Abnormal sialylation patterns have been correlated with tumor growth, escape from apoptosis, metastasis formation, and resistance to therapy (8). For instance, overexpression of sialyl-Lewis X or sialyl-Tn (Neu5Acα2-6GalNAcα-O-Ser/Thr) is usually associated with poor prognosis and decreased overall survival of patients with breast cancer (9–11). Such altered cell surface glycosylation often occurs as a consequence of abnormal expression of glycosyltransferases. The human genome encodes 20 different sialyltransferases involved in the biosynthesis of sialylated glycoproteins and glycolipids (12). ST3GAL1 is a glycosyltransferase that catalyzes the transfer of sialic acid from CMP-sialic acid (Cytidine-5′-monophospho-N-acetylneuraminic acid) to Galβ1-3GalNAc [galactosyl β(1,3)-N-acetylgalactosamine] substrates. In primary breast carcinomas, ST3GAL1 expression is higher in tumor than adjacent normal tissue (13). ST3GAL1 promotes cell migration, invasion, and TGFβ1-induced epithelial–mesenchymal transition in ovarian (14) and breast cancers (15). Overexpression of ST3GAL1 promotes tumor progression in a transgenic mouse mammary tumor model (16). Clinically, high expression of ST3GAL1 is associated with poor clinical outcome in patients with breast cancer (17) and glioblastoma (18). We previously showed that ST3GAL1-mediated sialylation of GFRA1 [glial cell line–derived neurotrophic factor (GDNF) receptor alpha 1; ref. 17] and vasorin (15) contributes to tumor progression in breast cancer. In this study, we identified CD55 to be another target protein and elucidated the impact of its sialylation by ST3GAL1 on immune evasion.

CD55 is also known as the decay-accelerating factor (DAF), one of the complement regulatory proteins that protect cells against complement attack. The complement system plays important roles in protection from pathogenic microorganisms, clearance of immune complexes, and complement-dependent cytotoxicity (CDC) of anticancer mAbs (19). The complement pathways can activate the membrane attack complex (MAC), comprised of C5b, C6, C7, C8, and C9 (20). Cell lysis is directly inflicted by insertion of the MAC into the target cell membrane. CD55 binds complement complexes to accelerate decay of C3 and C5 in order to prevent complement activation, thereby preventing damage to host cells (21). Thus, CD55 protects human tumor cells from complement-mediated cytotoxicity. On the other hand, loss of CD55 function with blocking anti-CD55 can enhance the C3 deposition and complement-mediated lysis of tumor cells (22). Similarly, downregulation of CD55 by siRNA sensitizes breast cancer (23) and uterine serous carcinoma (24) to herceptin-induced CDC and antibody-dependent cellular cytotoxicity (ADCC). These findings suggest that CD55 might function as a novel immune checkpoint molecule. However, it is unclear whether sialylation of CD55 contributes to its function, but it is noteworthy that neuraminidase treatment of human carcinoma cell lines enhances their sensitivity to CDC (25).

In this study, we characterized O-linked sialylation changes in CD55 upon ST3GAL1 silencing in MDA-MB-231 cells and investigated the relationship of such changes in modulating CDC and ADCC. We demonstrated that ST3GAL1-knockdown enhanced C3 deposition and susceptibility to CDC and ADCC associated with reduced O-linked sialylation of CD55. We also discerned a mechanism by which altered protein O-linked sialylation regulated CDC and ADCC in tumor immunity. These results suggest the possibility to maximize the efficacy of antibody-based immunotherapy by incorporating inhibitors of ST3GAL1 to block the immune regulator CD55, thereby overcoming the resistance of cancer cells to CDC and ADCC.

MDA-MB-231 and MCF-7 breast cancer cell lines were obtained from the American Type Culture Collection in 2008 and 2015, respectively. MDA-MB-231 cell line was authenticated by Bioresource Collection and Research Center (Taiwan) in 2018. The cell lines were maintained in DMEM medium, supplemented with 10% heat-inactivated FCS, streptomycin (100 μg/mL), and penicillin (100 μg/mL; Invitrogen). For the experimental uses, they were propagated in our laboratory for fewer than ten passages after thawing. Both cell lines were routinely tested for mycoplasma using PCR validation (Venor GeM Classic Cat. No. 11-1250, Minerva Biolabs) and found to be negative.

ST3GAL1-specific siRNA oligonucleotides [ST3GAL1: 5′-UCACUCUGAUCUUUGCAGGAACCGG-3′ (sense) and 3′-AGUGAGACUAGAAACGUCCUUGGCC-5′ (antisense)] and scramble siRNA duplexes (Cat. 12935400, Invitrogen), used as negative controls, were purchased from Invitrogen. Briefly, 2 × 10⁵ MDA-MB-231 or 1.5 × 10⁵ MCF-7 breast cancer cells were cultured in 6-well plates and transfected with ST3GAL1 siRNA duplexes at 10 nmol/L in conjunction with Lipofectamine RNAiMAX (5 μL; Invitrogen) following the manufacturer's instructions. Cells treated with siRNA for 96 hours were used in CDC and ADCC assays as described below. Control pLAS.Void (pVoid) and ST3GAL1 short-hairpin RNA (shRNA; shST3GAL1; TRCN0000231843) plasmids were purchased from RNAi core (Academia Sinica). Note that 2 × 10⁵ MDA-MB-231 or 1.5 × 10⁵ MCF-7 cells were cultured in 6-well plates and transfected with pVoid and shST3GAL1 plasmids at 2 μg in conjunction with Lipofectamine 2000 (4 μL; Invitrogen) following the manufacturer's instructions. MDA-MB-231 and MCF-7 cells stably transfected with shST3GAL1 or pVoid plasmid were selected by exposure to 2 μg/mL puromycin (17) and used in CDC experiments as described below.

A standard Europium assay was used to measure CDC and ADCC (26, 27). In brief, 2 × 10⁶ MDA-MB-231 or 1.5 × 10⁶ MCF-7 cells were treated with 10 nmol/L scramble or ST3Gal1 siRNA for 4 days. After 4 days, scramble or ST3GAL1 siRNA–treated cells (1 × 10⁶ cells/mL) were incubated for 5 minutes at 37°C with 10 to 20 mmol/L of the fluorescence-enhancing ligand BATDA [bis(acetoxymethyl)2,2′:6′,2′-terpyridine-6,6′-dicarboxylic acid; DELFIA EuTDA cytotoxicity kit, PerkinElmer, AD0116]. Intracellular esterases generate the membrane-impermeable TDA (2,2′:6′,2″-terpyridine-6,6″-dicarboxylic acid) from BATDA. Labeled tumor cells were washed and adjusted subsequently to 2 × 10⁶ cells/mL in serum-free DMEM. Note that 1 × 10⁴ BATDA-labeled cells were transferred to each well of a round-bottomed 96-well plate (Greiner) and mixed subsequently with complement-activating anti-HLA (Cat. No. 14-9983-82, eBioscience), anti-SSEA4 (monoclonal mouse IgG3 MC-813-70, R&D Systems), or anti–Globo H (VK-9, mouse IgG monoclonal antibody, AL Yu lab) at 0, 0.2, and 1 μg/mL for 30 minutes at 37°C. Normal rabbit serum obtained from The Laboratory Animal Center of Chang Gung University was added as a source of complement at final concentrations of 2.5%, 5%, and 10%, and the mixtures were incubated for an additional 120 minutes at 37°C. To determine the spontaneous release of TDA, 1 × 10⁴ BATDA-labeled cells were incubated with DMEM medium containing 1% FCS, and for evaluation of maximal release, complete cell lysis was induced by treatment with triton-x100 (Sigma, 5% v/v final concentration) for 120 minutes at 37°C. Following incubation, plates were centrifuged (500 × g, 5 minutes), and 10 μL supernatant from each well was transferred to a flat-bottomed 96-well plate (Greiner). Finally, 100 μL Europium solution (DELFIA EuTDA cytotoxicity kit, PerkinElmer, AD0116) was added to each well. After incubation at room temperature for 15 minutes, TDA released from lysed target cells was chelated with Eu3⁺. The fluorescence of the formed EuTDA chelate was measured using a Victor X3 Multimode Plate Reader (PerkinElmer). The percentage of specific release was calculated as 100% (experimental release - spontaneous release/maximal release - spontaneous release). All tests were performed in triplicate.

This study was approved by the Institutional Review Board of Human Subjects Research Ethics Committees of Academia Sinica (Taipei, Taiwan) and Chunghua Christian Hospital. In brief, 1 × 10⁶ cells/mL of scramble or ST3GAL1 siRNA–treated MDA-MB-231 cells were incubated for 5 minutes at 37°C with 10 to 20 mmol/L of the fluorescence-enhancing ligand BATDA. Labeled tumor cells were washed and adjusted subsequently to 2 × 10⁶ cells/mL DMEM containing 1% FCS. Note that 1 × 104 BATDA-labeled cells were transferred to each well of a round-bottomed 96-well plate (Greiner) and mixed subsequently with cytotoxicity-activating anti-SSEA4 (4 μg/mL) for 30 minutes at 37°C.

Peripheral blood mononuclear cells (PBMC) were isolated from blood of six healthy donors by density gradient centrifugation on Ficoll-Paque (Amersham Biosciences), washed twice with PBS, and added to the target cells (1 × 10⁵ cells/well) and incubated for an additional 120 minutes at 37°C. To determine spontaneous release of TDA, 1 × 10⁴ BATDA-labeled cells were incubated with DMEM with 1% FCS. For evaluation of maximal release, complete cell lysis was induced by triton-x100 (5% v/v final concentration). Following incubation, plates were centrifuged (500 × g, 5 minutes) and 10 μL supernatant from each well was transferred to a flat-bottomed 96-well plate (Greiner). Finally, 100 μL Europium solution was added to each well, and fluorescence of the EuTDA chelates was measured in a time-resolved fluorometer. The percentage of specific release was calculated as 100% (experimental release - spontaneous release/maximal release – spontaneous release). All tests were performed in triplicate.

RNAs were extracted from target cells with the RNeasy Mini kit (QIAGEN), according to the manufacturer's instructions, and 1 μg total RNA was reverse-transcribed into cDNA using Omiscript RT (QIAGEN). Real-time quantitative RT-PCR was performed using a TaqMan PCR detector Prism7000 (PerkinElmer Applied Biosystems) with TaqMan probes: Hs00161688_m1 (for ST3GAL1) and Hs02758991_g1 (for GAPDH, for internal control). A constant amount of cDNA was amplified using TaqMan Universal Master Mix II kit (Applied Biosystems). Data were calculated by the comparative ΔΔC_T method with the GAPDH gene as the endogenous control. Three replicates were conducted for each sample.

Flow cytometry was used for the detection of proteins expressed on the plasma membrane of siRNA- or shRNA-transfected cells. Cells were detached from plates using 0.05% trypsin-EDTA (Gibco), centrifuged at 300 × g for 5 minutes, and resuspended in FACS buffer (PBS with 1% BSA and 0.05% sodium azide; 2 × 10⁵ cells/tube). Samples were incubated with the indicated primary antibodies: anti-human HLA-ABC (Cat. No.14-9983-82, eBioscience), anti-SSEA4 (MC-813-70, R&D Systems), anti–Globo H (VK-9, AL Yu lab), anti-CD55 (NaM16-403, Santa Cruz Biotechnology), anti-CD46 (MEM-258, Abnova), anti‐CD59 (MEM-43, eBioscience), or mouse IgG1 isotype control (MAb002, R&D System) for 30 minutes on ice in the dark. Afterward, cells were washed and stained with FITC-labeled goat anti-mouse IgG (Cat. No. 1030-02, SouthernBiotech) for 30 minutes on ice in the dark. Surface proteins were detected by the EC800 Flow Cytometry Analyzer (Sony Biotechnology), and data were analyzed using FlowJo (Tree Star).

MDA-MB-231 and MCF-7 cells (4 × 10⁵ per incubation) treated with scramble or ST3GAL1 siRNA were incubated with the different concentration of mouse anti-Human HLA-ABC (Cat. No.14-9983-82, eBioscience) or anti–Globo H (VK-9, mouse IgG monoclonal antibody AL Yu lab) for 30 minutes at 37°C as indicated in the text. Cells were first washed with PBS (400 g, 3 minutes, 4°C), followed by washing with veronal buffer (Lonza BioWhittaker), and then resuspended in 20% human serum, collected from healthy donors and diluted in veronal buffer. After incubation for 15 minutes at 37°C, cells were washed twice with FACS buffer and resuspended in 100 μL FACS buffer. Cell-bound C3b was detected by incubation with rabbit polyclonal anti-C3 complex conjugated to FITC (dilution 1:100, NBP2-3318, Novus Biologicals) for 30 minutes on ice in the dark. After washing the cells twice, the cell-bound C3 complex was detected using the EC800 Flow Cytometry Analyzer, and data were analyzed using FlowJo. Background was set using control cells that consisted of cells incubated with anti-C3 complex but not exposed to human serum, and cells exposed to human serum but not incubated with sensitizing antibody.

One hundred microgram of total proteins from MDA-MB-231 cells transfected with scramble or ST3GAL1 siRNA were added to 10× Glycoprotein Denaturing Buffer (New England BioLabs) and denatured by heating the reaction at 100°C for 10 minutes. The denatured proteins were suspended in 10x G7 reaction buffer (New England BioLabs) with 0.25% Nonidet P-40 (NP-40, New England BioLabs) in a final volume of 40 μL containing the indicated glycosidases as follows: 2.5 μL of neuraminidase (P0720, New England BioLabs), 1 μL of O-glycosidase (P0733, New England BioLabs), and 1 μL of PNGase F (P0704, New England BioLabs) and incubated at 37°C for 8 hours. Proteins were separated by NuPAGE, followed by Western blotting, as described below.

For lectin pulldown assay, 1 mg MDA-MB-231 cell lysates transfected with scramble or ST3GAL1 siRNA in RIPA buffer were incubated with 30 μL PNA-agarose beads (5 mg/mL PNA, Vector Laboratories) at 4°C overnight. The agarose beads were washed 4 times with RIPA buffer and then eluted with 0.2 mol/L galactose (Sigma) in PBS. The eluate was subjected to Western blotting, as indicated below. For N- and O-glycosylation profiles of CD55, influenza hemagglutinin (HA)-CD55 expression vectors were constructed by cloning full-length CD55 cDNAs into pSin-EF2-Puromycin vector (Addgene plasmid 16579) with the influenza HA epitope at the N-terminal end (Supplementary Fig. S1). Note that 2 × 10⁵ MDA-MB-231 cells were cultured in 6-well plates and transfected with pSin-EF2-HA-CD55-Puromycin plasmid at 2 μg in conjunction with Lipofectamine 2000 (4 μL; Invitrogen) following the manufacturer's instructions. MDA-MB-231 cells stably expressing HA-CD55 were selected by exposure to 2 μg/mL puromycin.

Note that 1 × 10⁶ MDA-MB-231 cells overexpressing HA-CD55 were treated with 10 nmol/L scramble or ST3GAL1 siRNA for 96 hours. The cells were then lysed in RIPA buffer. Four milligram of cell lysate was incubated with 100 μL anti-HA agarose beads (2 mg/mL anti-HA, Sigma-Aldrich) at 4°C overnight. The agarose beads were washed 4 times with RIPA buffer and then eluted with 1× SDS sample buffer (Invitrogen). One fifth of the eluate was subjected to immunoblot analysis, as described below. The remaining eluate was subjected to SDS-PAGE, followed by Coomassie blue staining. The HA-CD55 proteins were excised from SDS-PAGE and frozen at −20°C until further processing for LC-MS/MS.

Scramble or ST3GAL1 siRNA–transfected MDA-MB-231 cells were washed with PBS and lysed in RIPA buffer containing 1× protease inhibitor cocktail (04693132001 Roche) and 1x phosphatase inhibitor cocktail (P5726, Sigma-Aldrich). Note that 20 μg protein extracts were separated on 4% to 12% NuPAGE (Invitrogen) and transferred to PVDF membranes (Millipore). The membranes were incubated with primary antibodies: anti-CD55 (NaM16-403, Santa Cruz Biotechnology) at 1:1,000, anti-actin (sc1615-R, Santa Cruz Biotechnology) at 1:2,000, anti-CD46 (MEM-258, Abnova) at 1:1,000, anti‐CD59 at 1:2,000 (MEM-43, eBioscience) at 4°C overnight, followed by Alkaline Phosphatase AffiniPure Donkey Anti-Mouse IgG (H+L; Cat. No. 715-055-150, Jackson Immunoresearch Laboratories) or Alkaline Phosphatase AffiniPure Donkey Anti-Rabbit IgG (H+L; Cat. No. 715-055-150, Jackson Immunoresearch Laboratories) at room temperature for 1 hour. The membrane was then scanned by a Typhoon9400 Variable Mode Imager (GE Healthcare Life Sciences) to detect the fluorescent signals released from catalyzed ECF substrate (GE Healthcare). The results of Western blots were quantified by ImageQuant 5.2 software (GE Healthcare).

Briefly, the excised gel bands were reduced with 10 mmol/L dithiothreitol at 37°C for 1 hour, alkylated with 50 mmol/L iodoacetamide in 25 mmol/L ammonium bicarbonate buffer for 1 hour in the dark at room temperature, and then washed twice with 25 mmol/L ammonium bicarbonate in 50% acetonitrile for 15 minutes before subjected to overnight in-gel digestion with sequencing-grade trypsin (Promega) at an enzyme-to-substrate ratio of 1:50 at 37°C. The peptide mixtures were extracted sequentially with 0.1% formic acid and 0.1% formic acid in 50% acetonitrile. The collected peptides were dried using a Speed-Vac and resuspended in 0.1% formic acid for analysis.

Online nanoLC-MS survey scan and automated data-dependent acquisition of collision-induced dissociation (CID) MS/MS under the precursor ion discovery mode were performed on a Synapt G2 HDMS mass spectrometer (Waters) fitted with a nanoAcquity UPLC system (Waters). Peptide mixtures were loaded onto a 75 μm ID, 25 cm length C18 BEH column (Waters) packed with 1.7 μm particles with a pore size of 130 Å, and were separated in 60 minutes using a gradient of 5% to 55% solvent B (acetonitrile with 0.1% formic acid) at a flow rate of 300 nL/min. Solvent A was 0.1% formic acid in water. The mass spectrometer was operated under the full software control of MassLynx 4.1 (Wateres). Glycan-specific oxonium ion fragments, m/z 204.084 for N-acetylhexosamine (HexNAc) and m/z 366.139 for hexose and N-acetylhexosamine (HexHexNAc) detected at the high-collision energy scans, were used to trigger MS/MS acquisition on the three most intense parent ions detected at the corresponding low-energy survey scan. MS profiles and MS/MS data of CD55 glycopeptides were manually identified and annotated.

The O-glycans from CD55 were released by reductive elimination with 1 mol/L NaBH4 in 0.1 N NaOH at 45°C for 16 hours. The reaction was quenched by pure acetic acid on ice, and released O-glycans were desalted by passing through a Dowex 50 × 8 column (Bio-Rad) in 5% acetic acid, after which borates were removed by coevaporation with 10% acetic acid in methanol. O-glycans were permethylated using the NaOH/dimethyl sulfoxide slurry method (28) extracted into chloroform, washed repeatedly with water, and dried down. NanoLC-MS/MS analysis of the permethylated O-glycan was carried out on a nanoLC system fitted with a 50 μm × 4 cm homemade polystyrene-divinylbenzene (PS-DVB) monolithic trap column and a 20 μm × 4 cm homemade PS-DVB–grafted open tubular analytical column coupled to an Orbitrap Elite hybrid mass spectrometer controlled by Xcalibur software (Thermo Scientific). Permethylated O-glycans were dissolved in 25% acetonitrile and separated at a constant flow rate of 150 nL/min, with a linear gradient of 0% to 40% acetonitrile containing 1 mmol/L sodium acetate over the course of 30 minutes, then increased to 80% acetonitrile in 5 minutes and held isocratically for another 10 minutes. The eluent was interfaced to the nanospray source based on the liquid junction configuration consisting of an uncoated emitter and a high-voltage platinum electrode. For each data-dependent acquisition cycle, the full-scan MS spectrum (m/z 300–2000) was acquired in the Orbitrap at 60,000 resolution (at m/z 400) with automatic gain control (AGC) target value of 1 × 10⁶. Data-dependent MS/MS acquisitions were performed by the CID experiments of the top-20 most intense ions, with the intensity threshold of 300 counts. The AGC target value and normalized collision energy applied for CID experiments were set as 30,000, 38%.

Graphs and statistical tests were carried out with Prism 6 (GraphPad Software). Statistical significance between two groups was tested using Student t tests and two-way ANOVA followed by Bonferroni multiple comparisons test. Results are expressed as mean ± SD. *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

Previously, we showed that the ratio of nonsialylated Galβ1-3GalNAc- serine/threonine core 1 O-glycan, a ligand for plant lectin peanut agglutinin (PNA), is increased after ST3GAL1 silencing (15, 17). To identify target proteins of ST3GAL1, MDA-MB-231 cells were transfected with scramble or ST3GAL1 siRNA. The expression ST3GAL1 mRNA was analyzed by qPCR and normalized to GAPDH. ST3GAL1 siRNA significantly reduced ST3GAL1 expression in MDA-MB-231 cells (Fig. 1A). To examine the protein targets by ST3GAL1 in breast cancer cells, total cell lysates from scramble or ST3GAL1 siRNA–treated MDA-MB-231 cells were prepared for lectin blot analysis after immunoprecipitated with the PNA agarose. The PNA-binding signals were increased after ST3GAL1 silencing (Fig. 1B), and the PNA pulldown proteins were identified by LTQ-FTD mass spectrometry. CD55 was identified as one of the PNA-binding proteins. The PNA pulldown showed that the signal of CD55 was increased after ST3GAL1 silencing. The molecular weight of CD55 was slightly lower after ST3GAL1 silencing (Fig. 1C), but the CD55 protein expression remained the same (Fig. 1D). Only the protein expression of CD55, but not CD46 or CD59, in PNA pulldown assays (Fig. 1C) was increased after ST3GAL1 silencing. These findings suggest that CD55 may be a target protein of ST3GAL1.

There are approximately 24 O-glycosylation sites on CD55, based on NetOGlyc 3.1 prediction (29, 30). To show that CD55 was heavily O-linked sialylated, we analyzed the glycosylation pattern of CD55 upon treatment with different deglycosylation enzymes. MDA-MB-231 cells transfected with scramble or ST3GAL1 siRNA were lysed with RIPA buffer, treated with indicated deglycosylation enzymes, and subjected to Western blotting. The calculated molecular weight of CD55 is about 41 kDa. In control cells, glycosylated CD55 migrated at a molecular weight of about 70 kDa, whereas it migrated faster after ST3GAL1 silencing (Fig. 1C). After removal of N-glycans by PNGase F treatment, CD55 shifted to lower molecular weight in control cells but even lower in ST3GAL1-silenced cells. The inability of PNGase F treatment to eliminate the difference in CD55 molecular weight between ST3GAL1-silenced and control cells suggested that ST3GAL1 mediated sialylation of O-linked glycans on CD55.

We next removed the sialic residues with neuraminidase, which catalyzes the hydrolysis of α-2,3–, α-2,6–, and α-2,8–linked sialic acid residues from glycoproteins and oligosaccharides. Treatment with neuraminidase shifted CD55 to lower but equal molecular weight in both ST3GAL1-silenced and control cells (Fig. 2). Double digestion with PNGase F and neuraminidase further reduced the molecular weight of CD55, but to comparable levels in both ST3GAL1-silenced cells and controls. These findings suggested that sialylation of CD55 was O-linked. To confirm this, cells were treated with O-glycosidase, which catalyzes the removal of core 1 and core 3 O-linked disaccharides from glycoproteins. There was a similar degree of downward shift of CD55, but CD55 remained lower in ST3GAL1-silenced cells than in control cells. Upon triple digestion with PNGase F, neuraminidase, and O-glycosidase, CD55 appeared to shift to the same molecular weight (∼50 KD) in control and ST3GAL1-silenced cells. These findings confirmed that CD55 is an O-linked sialylated glycoprotein catalyzed by ST3GAL1.

To verify that CD55 indeed carried O-linked sialylation, the glycan profiles of CD55 were analyzed. A direct LC-MS/MS analysis of the tryptic peptides from CD55 revealed that the major N-glycans of CD55 expressed in control MDA-MB-231 cells comprised complex type bi- and triantennary structures bearing terminal sialic acids, with and without core fucosylation. These were all detected as glycoforms of the expected tryptic peptide GSQWSDIEEFCNR carrying the previously reported single N-linked glycosylation site (31), the relative amount of which did not change significantly upon ST3GAL1 silencing (Fig. 3A). The same analysis failed to pinpoint the specific O-glycopeptides corresponding to any of the possible 24 predicted O-linked glycosylation sites located at serine/threonine-rich stalk region. Instead, LC-MS/MS analysis of the released, permethylated O-glycans showed that the major structures were based on core 1, with and without sialic acid, and core 2 with up to two sialic acids. The relative amount of the disialylated core 2 structure at m/z 864.426 (2+) was significantly reduced concomitant with an increase of the monosialylated core 2 structure at m/z 683.839 (2+) upon ST3GAL1 silencing (Fig. 3B and C). The LC-MS data were consistent with ST3GAL1 contributing mainly to 2,3-sialylation of O-glycans and not N-glycans of CD55. Impairment of 2,3-sialylation on the O-glycans led to a shift in abundance to nonsialylated core 1 and monosialylated core 2, at the expenses of disialylated core 2 structure. On the other hand, the relative amount of the monosialylated core 1 O-glycan structure (m/z 895.462) was found to remain similar between the control and ST3GAL1-silenced cells, probably due to compensating 2,6-sialylation. These findings indicated that ST3GAL1 mediated α2,3-sialylated O-glycans of CD55.

CD55 is known to regulate the complement system by accelerating the disassembly of C3bBb complex, thereby interrupting the complement cascade and preventing CDC (32). We next explored the possibility that ST3GAL1-mediated O-linked sialylation might affect the function of CD55. We compared the sensitivity of scramble or ST3GAL1 siRNA–transfected MDA-MB-231 cells with CDC in the presence of anti-HLA or anti-SSEA4 with rabbit complement. Cell lysis increased in a dose-dependent manner for both antibody and complement (Fig. 4A). In the presence of 10% rabbit complement, ST3GAL1 silencing enhanced cell death (from 29.6% ± 0.9% to 51.1% ± 2.2% at 0.2 μg/mL anti-HLA, P < 0.001; from 40.9% ± 5.9% to 57.4% ± 4.2% at 1 μg/mL anti-HLA, P = 0.033). In the presence of 5% rabbit complement, ST3GAL1 silencing also enhanced cell death (from 19.3% ± 3.3% to 40.6% ± 1.5% at 0.2 μg/mL anti-HLA, P < 0.001; from 27.2% ± 0.1% to 49.3% ± 2.3% at 1 μg/mL anti-HLA, P < 0.001; Fig. 4A). Anti–SSEA4-mediated CDC was also assessed because MDA-MB-231 cells have high SSEA4 expression. In the presence of 1 μg/mL anti-SSEA4, ST3GAL1 silencing enhanced cell lysis (from 12.7% ± 2.2% to 48.4% ± 4.5% at 5% rabbit complement, P < 0.001; from 22.1% ± 3.0% to 63.4% ± 10.4% at 10% rabbit complement, P = 0.003; Fig. 4B). Similarly, increased sensitivity to CDC was also observed in MDA-MB-231 cells stably transfected by ST3GAL1 shRNA plasmid in the presence of anti-HLA and anti-SSEA4 (Fig. 4C and D). To demonstrate that the impact of ST3GAL1 knockdown on sensitivity to CDC was not restricted to MDA-MB-231 cells, we performed similar studies on MCF-7 cells using anti-HLA or anti–Globo H to induce CDC. Enhanced cell lysis was also observed in ST3GAL1 siRNA–transfected MCF-7 cells (Fig. 4E and F), as well as in MCF-7 stable clones transfected with ST3GAL1 shRNA (Fig. 4G and H). In the presence of 1 μg/mL anti-HLA, ST3GAL1 silencing enhanced cell death (from 17.3% ± 2.1% in control MCF-7 cells to 27.2% ± 1.6% at 5% rabbit complement, P = 0.003; from 44.3% ± 1.5% to 49.0% ± 3.0% at 10% rabbit complement, P = 0.073; Fig. 4E). In the presence of 1 μg/mL anti–Globo H, ST3GAL1 silencing enhanced cell death (from 30.3% ± 0.9% in control MCF-7 cells to 53.4% ± 4.9% at 5% rabbit complement, P = 0.001; from 45.5% ± 2.0% to 52.0% ± 2.6% at 10% rabbit complement, P = 0.026; Fig. 4F). Thus, ST3GAL1 silencing rendered MDA-MB-231 and MCF-7 cells to become significantly more susceptible to CDC. To examine the possibility that the enhanced sensitivity to CDC upon ST3GAL1 silencing might result from increased expression of target antigens or decreased expression of membrane complement regulatory proteins (mCRP), ST3GAL1-silenced MDA-MB-231 and MCF-7 cells were analyzed by flow cytometry for their expression of mCRPs and target antigens. No changes in the expression levels of mCRPs (CD46, CD55, and CD59) and surface target antigens (HLA, SSEA4, and Globo H) were observed upon ST3GAL1 silencing (Supplementary Figs. S2–S5).

As shown above, protection from complement attack was reduced in ST3GAL1-silenced MDA-MB-231 and MCF-7 cells and might be due to expressing desialylated CD55. To further assess the decay-accelerating ability of surface desialylated CD55, we measured the deposition of C3 fragments on the cell surface. Cells were incubated under nonlytic conditions with human serum as the complement source, and cell surface–deposited C3 was detected by flow cytometry (Fig. 5A). Increased C3 deposition was observed with increasing concentration of anti-HLA (Fig. 5B). C3 deposition was significantly greater on the surface of ST3GAL1-silenced cells than control cells (geometric median fluorescence of 72.1 ± 4.8 × 10³ vs. 89.5 ± 5.5 × 10³ at 10 μg/mL, P < 0.01; 118.0 ± 6.4 × 10³ vs. 164.8 ± 0.8 × 10³ at 25 μg/mL, P < 0.01; for scramble- vs. siRNA-treated, respectively). We also assessed the deposition of C3 fragments on the ST3GAL1-silenced MCF-7 cells. Likewise, increased C3 deposition was observed with increasing concentration of anti–Globo H. C3 deposition was also significantly greater on the surface of ST3GAL1-silenced cells than control cells (geometric median fluorescence of 10.5 ± 0.01 × 10³ vs. 19.2 ± 0.4 × 10³ at 5 μg/mL, P < 0.01; 14.4 ± 0.3 × 10³ vs. 25.1 ± 0.6 × 10³ at 25 μg/mL, P < 0.01; for scramble- vs. siRNA-treated, respectively). In the presence of complement without anti–Globo H, there was a minimal basal deposition of C3 on cells, which was higher in ST3GAL1-silenced cells (geometric median fluorescence: scramble, 5.0 ± 0.1 × 10³; siRNA, 6.9 ± 0.1 × 10³; P < 0.01; Fig. 5C). These findings indicated that the decay-accelerating activity of surface-expressed CD55 was reduced in ST3GAL1-silenced cells, resulting in their increased susceptibility to CDC.

It has been shown that downregulation of CD55 in uterine serous carcinoma cells consistently and significantly increases their sensitivity to ADCC by PBMCs (24). To investigate whether desialylated CD55 in ST3GAL1-silenced cells enhanced their susceptibility to PBMC-mediated ADCC, MDA-MB-231 cells were transfected with scramble or ST3GAL1 siRNA. Using human PBMCs as effectors, anti-SSEA4, and siRNA-treated MDA-MB-231 cells as targets, an ADCC assay was performed. ST3GAL1-silenced cells were significantly more sensitive to ADCC than scramble siRNA–treated cells (Fig. 6), with a mean percent lysis of 24.2% ± 1.7% and 41.5% ± 2.4% for scramble- and siRNA-treated, respectively, at an E:T ratio of 10:1 (P < 0.05). IgG3 isotype control and anti-SSEA4 alone had negligible cell lysis in the absence of immune cells.

Posttranslational modification is critical for modulating the function of proteins. Although glycosylation is the most common posttranslational modification, it remains the least explored due to the inherent complexity and technical difficulties. Our group previously identified GFRA1 and vasorin as two target proteins of ST3GAL1, overexpression of which was shown to have an adverse impact on the clinical outcome of patients with breast cancer (15, 17). In this study, we provided evidence for CD55 to be another target protein of ST3GAL1 by demonstrating significant changes in the O-linked glycan profiles of CD55 upon ST3GAL1 silencing in breast cancer cells, with a shift from disialylated core 2 structure to nonsialylated core 1 and monosialylated core 2. Although the expression of total CD55 protein was not affected by ST3GAL1 silencing, there was a significant increase in C3 deposition on the cell surface, along with enhanced sensitivities of these cells to CDC and ADCC. Previous studies have demonstrated that removal of surfaces sialic acid from human bladder, prostate, breast, and ovarian carcinoma cells with sialidase increases sensitivity to complement-mediated lysis (25, 33). In these studies, α-(2,3,6,8,9) neuraminidases were used, which cleave all terminal sialic acid residues with α-2,3-, 2,6-, 2,8-, and α-2,9 linkages from complex carbohydrates and glycoproteins. On the other hand, our studies focused on ST3GAL1 sialyltransferase, which mediates the transfer of sialic acid residues to a Gal residue of terminal Galβ-1-3GalNAc oligosaccharide to form Neu5Acα2-3Galβ1-3GalNAc trisaccharide sequence. Following ST3GAL1 silencing, there was a small but significant increase in basal binding of C3 fragment to the cells without addition of antibody, but it did not induce significant cell lysis. It is not clear whether this increased C3 deposition represents an enhancement in C3 convertase activation or increased nonspecific adherence of the C3 complex to the ST3GAL1-silenced cells. Following ST3GAL1 silencing, binding of the anti-HLA or anti-SSEA4 to the cells was similar to control cells. However, both C3 deposition and cell lysis were significantly increased in ST3GAL1-silenced cells. The fact that enhanced lysis of ST3GAL1-silenced cells depended on the presence of both antibody and complement suggested the involvement of the classical pathway in complement activation of ST3GAL1-silenced cells.

In our study, we identified CD55 to be one of the potential target proteins of ST3GAL1. CD55 is an important complement regulatory protein, preventing cells from complement-mediated cytotoxicity. Earlier studies have demonstrated greater protein expression and O-glycosylation of CD55 in colorectal cancer than normal tissues. When the O-linked glycan was removed by enzyme digestion, the difference in the molecular weight of CD55 between cancer and normal tissues was abolished (34). Our results demonstrated only one N-linked glycosylation site on CD55 peptide at GSQWSDIEEFCNR between the SCRl and SCR2 domains, which is consistent with previous study in a CHO cell model (31). The predicted multiple O-linked glycosylation sites are enriched within the Ser/Thr-rich domain that lies between the short consensus repeat (SCR) domain and glycosyl phosphatidyl inositol (GPI) anchor. Although we have mapped the O-glycan profile on CD55, it was difficult to identify which of the 24 predicted O-linked glycosylation sites corresponded to the actual O-glycopeptides. Nevertheless, our finding of increased sensitivity of cancer cells to CDC upon ST3Gal silencing revealed the crucial role of O-linked sialylation in the function of CD55. It has been shown that deletion of the N-linked glycosylation site did not affect the function of CD55, but the Ser/Thr-rich domain of CD55 is required for complement protection (31), which is consistent with our notion that O-linked sialylation may be important. Although the loss of function of CD55 Ser/Thr-rich domain deletion could be restored by fusing amino acid residues 1–257 of DAF to the carboxyl terminus of HLA-B44 from amino acid 66 to the end of the sequence (DAF/HLA-B44 chimera), it remains to be ascertained if glycosylation of HLA-B44 contributes to the complement modulating activity of DAF/HLA-B44 chimera (31).

In addition to enhanced sensitivity to CDC, ST3GAL1 silencing sensitized MDA-MB-231 cells to ADCC. In previous studies, downregulation of CD55 by siRNA significantly increases trastuzumab-mediated killing of uterine serous carcinoma by natural killer (NK) cells (24). In another study, soluble recombinant CD55-Fc fusion proteins are shown to inhibit T-cell proliferation (35). Thus, CD55 has a negative effect on the interactions between cancer cells and NK cells and activation of T cells, implying its role as an immune checkpoint.

In addition to CD55, our previous studies demonstrated GFRA1 and vasorin to be substrates of ST3GAL1. ST3GAL1-mediated O-linked sialylation of GFRA1 enhances GDNF-induced cell growth (17). O-sialylation of vasorin reduces its binding to TGFβ1, which facilitates TGFβ1 signaling and angiogenesis (15). Here we showed that desialylation of CD55 increased C3 deposition and enhanced susceptibility to CDC and ADCC. In this context, ST3GAL1-mediated sialylation contributed to the immune evasion function of CD55. This is reminiscent of the report that N-linked glycosylation of cell surface PD-L1 promotes PD-L1/PD-1 ligation (36). Removal of N-linked glycosylation enhances anti–PD-L1-binding affinity and could guide anti–PD-1/PD-L1 therapy (37). These studies suggest targeting protein glycosylation as a potential strategy to enhance immune checkpoint therapy.

Taken together, these findings suggest that silencing ST3GAL1 may serve as a promising strategy for treatment of breast cancer and blockade of CD55-mediated immune evasion. In this era of rapid advances in antibody-based immunotherapy of cancer, it may be desirable to incorporate ST3GAL1 silencing as a strategy to block the function of CD55, which could overcome resistance of cancer cells to CDC and ADCC, thereby maximizing anticancer efficacy.

J. Yu reports grants from Ministry of Science and Technology, Taiwan, and Chang Gung Memorial Medical Foundation during the conduct of the study. A.L. Yu reports grants from Ministry of Science and Technology, Taiwan, and Chang Gung Memorial Medical Foundation during the conduct of the study. No disclosures were reported by the other authors.

W.-D. Lin: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. T.-C. Fan: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.-T. Hung: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–review and editing. H.-L. Yeo: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–review and editing. S.-H. Wang: Data curation, formal analysis, investigation, methodology. C.-W. Kuo: Data curation, formal analysis, investigation, methodology. K.-H. Khoo: Data curation, formal analysis, supervision, investigation, methodology. L.-M. Pai: Formal analysis, supervision, investigation, methodology, writing–review and editing. J. Yu: Resources, supervision, funding acquisition, writing–review and editing. A.L. Yu: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, visualization, project administration, writing–review and editing.

This work was supported by grants from the Ministry of Science and Technology and Chang Gung Medical Foundation in Taiwan to A.L. Yu (MOST 103-2321-B-182A-005, MOST 104-2321-B-182A-003, MOST 105-2321-B-182A-001, and OMRPG3C0014); J. Yu (MOST 106-3114-B-182A-001, MOST 107-2321-B-182A-005, MOST 108-2321-B-182A-004, and MOST 109-2321-B- 182A-005); and J.-T. Hung (CMRPG3G1531–CMRPG3G1533).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.