Immunohistochemistry and the scoring system

Sections were deparaffinized using xylene, a series of graded ethanolic solutions, and distilled water. Tissues were then permeabilized using Triton™ X-100 (1:100, Sigma-Aldrich, St. Louis, MO) at room temperature for 10 min, followed by antigen retrieval in citrate buffer (10 mM Sodium Citrate, pH 6.0) at 37 °C for 10 min. Next, sections were blocked with Ultra V Block solution (Thermo Fisher Scientific, Waltham, MA) at room temperature for 5 min. Afterward, sections were incubated with a primary antibody at 4 °C overnight. On the following day, Primary Antibody Amplifier Quanto and AP Polymer Quanto were added sequentially at room temperature for 10 min. Signals were then visualized using the DAB Quanto Chromogen provided in the same kit. All the sections were counterstained with hematoxylin. The primary antibody used was a polyclonal rabbit anti-human MGAT5 (1:400, generated in our lab). PD-L1 expression was evaluated by VENTANA PD-L1 (SP263) assay (Roche Diagnostics, Basel, Switzerland), and TPS was evaluated by a pathologist.

To detect PHA-L binding, lectin histochemistry was performed. Briefly, following the antigen retrieval step, biotinylated PHA-L (1:400, Vector Laboratories, Newark, CA) was incubated with the sections at 4 °C overnight. The binding was captured by Streptavidin-HRP (BioGenex Laboratories, Fremont, CA) and detected using a DAB substrate (Thermo Fisher Scientific, Waltham, MA). Intensity was accessed as follows: score 3 = 66–100% of the tumor cells were intensely labeled; score 2 = 33–65% of cells were intensely labeled; score 1 = 1–32% of cells were labeled; score 0 = none of the cells were labeled. Score 0 was defined as MGAT5-negative and scores 1–3 were defined as MGAT5-positive.

Generation of an anti-MGAT5 polyclonal antibody

Partial sequence of MGAT5 was cloned from SAS cells. The primers used were 5′-GGTACCATGATTACATCTCTGTACTTACT-3′ and 5′- CTCGAGTAGGCAGTCTTTGCAGAGAG-3′. This sequence was inserted to pET30a expression vector with restriction sites of KpnI and XhoI. The recombinant protein (aa I305 – L740) was expressed in Escherichia coli (BL21/DE3 strain). Consecutive 7 doses of recombinant MGAT5, 200 μg with Freud’s complete adjuvant for the first dose and 100 μg with Freud’s incomplete adjuvant for the rest doses, were subcutaneously injected into a New Zealand rabbit. Total 50 ml serum was collected one week after the finial boost. All animal experiments were approved by the Institutional Animal Care and Use Committee of the National Taiwan University College of Medicine, Taipei, Taiwan (IACUC No. 20180161).

To purify the anti-MGAT5 antibody, a peptide containing MGAT5 sequence (H. M. Biological, Taoyuan, Taiwan) was conjugated to SulfoLink™ Coupling Resin (Thermo Fisher Scientific, Waltham, MA). The anti-serum (10 ml) was incubated with peptide conjugated agarose beads at 4 °C overnight. The anti-MGAT5 antibody was eluted using an elution buffer (0.2 M Glycine-HCl, pH2.5) and quickly neutralized using a neutralizing buffer (1 M Tris-HCl, pH8.8).

Generation of MGAT5 knockout cells

To knock out MGAT5, we used the CRISPR/Cas9 system. The small guide RNA (sgRNA) targeting MGAT5 was designed according to database predictions. The target sequence is 5′-GTGGTGGATGGGCCATACGC-3′. SAS cells were transfected with a pAll-Cas9.Ppuro plasmid (National RNAi Core Facility at Academia Sinica, Taiwan) containing Cas9 and sgRNA of MGAT5, and a pSurrogate reporter plasmid containing a sgRNA target sequence located between an in-frame EGFP cassette and an out-of-frame mCherry cassette. Viable mCherry-positive cells were sorted using a 4-laser FACSAriaIII sorter (BD Biosciences, Franklin Lakes, CA) and then cultured in 96-well plates as single cells. Single colonies with successful MGAT5 knockout were confirmed by DNA sequencing and Western blotting. Clones without MGAT5 knockout were used as the control (Mock).

Patient data and tissue samples

Total 40 patients with R/M HNSCC who failed to respond to platinum-based salvage therapy and then received either nivolumab or pembrolizumab between 2016 and 2018 were registered. Treatment response was evaluated after at least three courses of treatment and was based on response evaluation criteria in solid tumors (RECIST) criteria. Patients with stable disease, partial response, and complete remission were considered responders. All human tissues were obtained according to a protocol approved by the Institutional Review Board of National Taiwan University Hospital, Taipei, Taiwan. Written informed consents were obtained from all patients. The IRB number is 202011107RINA. The study methodologies conformed to the standards set by the Declaration of Helsinki.

Cell cultures and transfection

SAS and OEC-M1 cells were a gift from Dr. Jean-San Chia (National Taiwan University) [29]. All cell lines were authenticated by STR DNA profiling analysis. Jurkat cells were purchased from Bioresource Collection and Research Center (Hsin-Chu, Taiwan). The culture medium for SAS cells was Dulbecco′s Modified Eagle (DMEM) supplemented with 10% FBS (Thermo Fisher Scientific, Waltham, MA), and Roswell Park Memorial Institute (RPMI) 1640 supplemented with 10% FBS (Thermo Fisher Scientific, Waltham, MA) for OEC-M1 and Jurkat cells. To knock down MGAT5, cells were transfected with 20 nM siRNAs against MGAT5 (siMGAT5-1: 5′-GAAAGCGGAAGAAAGUCCUCGUUCA-3′, siMGAT5-2: 5′-GGAGACAGAAUUGUUGAGCUCAUUU-3′) using Lipofectamine RNAiMAX (Invitrogen, Waltham, MA). Non-targeting siRNA (siCtr, 5′-CAACCUCAGCCAUGUCGACUGGUUU-3′) was used as a control.

Western blot analysis and lectin pull-down assay

Proteins from cell lysates were separated on an SDS-PAGE and transferred onto a PVDF membrane, which was then blocked in 5% non-fat milk for 1 h at room temperature and incubated with a primary antibody against MGAT5 (1:1000, our lab) or PD-L1 (1:1000, Cell Signaling Technology, Danvers, MA) at 4 °C for overnight. HRP conjugated secondary antibodies (1:10000, Jackson ImmunoResearch, Philadelphia, PA) were used to detect and the protein bands.

For lectin pull-down assay, 500 μg of total proteins from cell lysates were incubated with Phaseolus vulgaris Leucoagglutinin (PHA-L) conjugated beads (Vector Laboratories, Newark, CA) for 18 h at 4 °C with constant rotation. Washed beads were boiled at 95 °C for 10 min, and PHA-L binding proteins were evaluated by Western blot analysis.

N-glycopeptide enrichment by ZIC-cHILIC StageTip and liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Intact N-glycopeptides (GPs) were enriched by ZIC-cHILIC StageTip based on methods that were previously described [30]. Briefly, ZIC-cHILIC StageTip was prepared by capping at one end with a 20 μm polypropylene frits disk (Agilent Technologies, Santa Clara, CA) enclosed in a tip-end fitting, loaded by 40 mg of ZIC-cHILIC materials resuspended in deionized water (100 μL), loaded into a 200 μL tip, and centrifuged at 3000 rpm for 2 min, then flattened the surface by adding 50 μL of deionized water and centrifuged at 6000 rpm for 2 min as a ZIC-cHILIC-StageTip for intact GPs enrichment. The protein digests were reconstituted with 50 μL of 80% ACN/0.5% TFA and loaded into the StageTips and centrifuged at 3000 rpm for 10 min; then, the flow-through was reloaded into StageTip and centrifuged at 4000 rpm for 8 min. The bound GPs were washed with 80% ACN/0.5% TFA (50 μL) twice. The GPs were stepwise eluted using 70% ACN, 65% ACN, and combined 3 elutions using 60% ACN, and 55% ACN, followed by centrifugation at 3000 rpm for 10 min, and finished at 0.5% FA three times with centrifuged at 3000 rpm for 10 min, another 4000 rpm for 8 min, and another 5000 rpm for 6 min. Finally, GPs were dried and resuspended by 20 μL of 0.1% TFA/5% ACN (v/v), desalted by reversed-phase SDB-XC StageTip, eluted by 20 μL of 80% ACN/0.1% TFA, then dried, and dissolved in 0.1% FA for LC-MS/MS. which was performed on Ultimate 3000 nanoLC system coupled with an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, San Jose, CA).

Glycoproteomic data analysis

All raw data were processed by Proteome Discoverer 2.5 (PD2.5; ThermoFisher Scientific, Waltham, MA) integrating with Byonic v4.0 (Protein Metrics, San Carlos, CA). For identification of intact GP, the raw data were queried using Byonic 4.0 with tryptic peptides with less than two cleavage sites, a precursor ion mass tolerance of 10 ppm and a fragment ion tolerance of 20 ppm for HCD spectra and 0.02 Da for EThcD spectra. Protein fasta files, including PD-L1 (Q9NZQ7, CD274) or Human (Swiss-Prot database, v2021-05-06, total 20,324 sequences) were used for protein identification. The built-in human N-glycans database (with 269 N-linked glycans from human and mammalians without NeuGc) were used for identification of glycan composition. Carbamidomethyl (C) was selected as fixed modification, deamidation (NQ) and oxidation (M) were selected as variable (common) modifications. N-glycan was selected as “rare”. Maximum of total common modification was set as 4 and rare modification was set as 1. The reversed peptide sequence identification was also considered with a protein false discovery rate (FDR) of 1%, or 20 reverse counts. The high confidence of GP sequence was also considered the two-dimensional posterior error probability (PEP2D) values with FDR < 0.01. The abundance of identified GPs from each cell was further processed by using chromatographic alignment with m/z and retention time of identified intact GPs in PD2.5 software. The LC-MS/MS data were deposited in the ProteomeXchange Consortium via the PRIDE [31] partner repository with the dataset identifier PXD045506.

The abundance ratio was calculated by the abundance of identified GP compared to sum of overall GPs abundance. The fold change of potential β1,6-branched N-GPs were compared between Mock and MGAT5 knockout SAS cells. Besides, a two-sample t-test was used to estimate the expression difference between two cell types. All tests were performed by two-tailed tests and P < 0.05 were considered significant. Both functional and physical associations of β1,6-branched N-glycoproteins were analyzed by using STRING network with high confidence (≥0.9).

Functional maps of MGAT5 protein substrates

MGAT5 protein substrates identified by the glycoproteomic analysis were used for functional map analysis. Gene Ontology (GO) over-representation analysis was performed on the differentially expressed genes by the enrichGO function in the R package clusterProfiler. Significantly enriched GO terms (P < 0.01) were constructed into an enrichment map [32] and visualized by Cytoscape.

Flow cytometry

Cells (1 × 106) were washed with PBS and then resuspended in FACS buffer. They were incubated with an anti-PD-L1 antibody (1:100, #329701, BioLegend, San Diego, CA) on ice for 30 min. After this incubation, cells were washed, Alexa Fluor 488-conjugated secondary antibodies (1:500, Jackson ImmunoResearch, West Grove, PA) were added on ice for additional 30 min.

Signals were collected on a BD LSRFortessa (BD Biosciences, Franklin Lakes, CA) with FACSDiva software (BD Biosciences, Franklin Lakes, CA), and the data were analyzed using FlowJo (Tree Star, Ashland, OR).

Ligand-binding affinity assays

One hundred micrograms of cell lysates were incubated with plates coated with an anti-PD-L1 antibody (#13684, Cell signaling technology, Danvers, MA) at 4 °C for 18 h. Subsequently, PD-1 Fc (R&D Systems, Minneapolis, MN) was added to the plates and allowed to incubate at room temperature for 2 h. After a gentle wash, an HRP-conjugated goat anti-human IgG antibody (1:10000, Jackson ImmunoResearch, Philadelphia, PA) was added and incubated at room temperature for 30 min. Following another gentle wash, bound-PD-1 Fc was detected by o-phenylenediamine dihydrochloride (OPD) substrates. An ELISA reader was used to detect O.D. at 490 nm.

CTL assay

Jurkat cells were activated with phorbol‑12 myristate‑13 acetate (PMA) (100 ng/mL) and PHA-L (5 μg/mL) for 24 h before being cocultured with cancer cells. The IncuCyte® Caspase 3/7 dye (Essen BioScience, Ann Arbor, MI) was added 24 h prior to coculturing. The effector-to-target ratio was 1:5 and the coculture time was 6 h. The fluorescent signal generated by cleaved dye was detected using fluorescence microscopy and quantified using the ImageJ software. To block PD-1, Jurkat cells were preincubated with control IgG or nivolumab (20 μg/ml, Bio X Cell, Lebanon, NH) for 30 min before coculturing.

PD-L1 purification

Lysates (15 mg) from SAS cells overexpressing HA-tagged PD-L1 were harvested and incubated with Pierce™ Anti-HA agarose (Thermo Fisher Scientific, Waltham, MA) at 4 °C for 18 h. After washing, the captured HA-PD-L1 was eluted with 0.5 mL of 0.2 M glycine (pH 2.5) and then neutralized with 50 μL 1 M Tris (pH 8.8). Eluates were desalted with an Amicon (Merck Millipore) and then run on an SDS-PAGE. The separated PD-L1 was excised and sent for LC-MS/MS analysis (Instrument Center, National Taiwan University, Taipei, Taiwan).

Statistical analysis

Differences in PHA-L binding and MGAT5 expression between non-tumor and tumor tissues were assessed using a paired Student’s t-test. Differences between intensity of fluorescent apoptotic cells were evaluated with a Student’s t-test. Correlations between MGAT5 or PD-L1 expression with treatment responses were evaluated through Fisher’s exact tests. Patients’ survival was analyzed using Kaplan–Meier analysis. P < 0.05 was considered statistically significant.