Patient samples

Tumor and blood samples were collected at the University Hospital Basel, and buffy coats from healthy donors were obtained from the blood bank of the University Hospital Basel, Switzerland. Sample collection and use of corresponding clinical data were approved by the local ethics committee in Basel, Switzerland (Ethikkommission Nordwestschweiz, EKNZ, Basel Stadt. Switzerland), and written informed consent was obtained from all the donors before sample collection. Healthy donors (HDs) were defined according to the inclusion criteria from the blood donation center in Basel. HD patients were 18–60 years old, had a minimal weight of 50 kilograms and were in good general health. The exclusion criteria included recent dental treatment (<14 days), tattooing, piercing or permanent makeup (<4 months), tick bites (<4 weeks), gastroscopy (<4 months), recent whole-blood donations (<3 months), surgery (<12 months), travel to certain countries with endemic diseases (up to 6 months), behavioral risk factors (sexual behavior), pregnancy and breastfeeding (1 year after giving birth). The donors were screened for infectious diseases, including HIV (AIDS), syphilis, hepatitis C and hepatitis B.

Cell lines

HeLa and B16F10 cell lines were obtained from ATCC. A549, HeLa and EL4 cells were kindly provided by the Zippelius Laboratory, and HEK293T cells were kindly provided by the Bentires Laboratory, both of which were from the Department of Biomedicine, Basel. The H1N1 viral sialidase-expressing cell lines A549-sia, HeLa-sia, and B16F10-sia as well as the EL4-GFP cell line were generated by lentiviral transduction as described below. A549-GNE KO and B16F10-GNE KO cells were generated as described before using CRISPR/CAS9 [25].

Mouse strains

The experiments were performed in accordance with the Swiss federal regulations and approved by the local ethics committee of Basel-Stadt, Switzerland (Approval 3036 and 3099). All animals were bred in-house at the Department of Biomedicine Facility (University of Basel, Switzerland) in specific pathogen-free, ventilated HEPA-filtered cages under stable housing conditions of 45–65% humidity, a temperature of 21–25 °C, and a gradual light–dark cycle with light from 7:00 am to 5 pm. Mice were provided standard food and water without restriction (license: 1007-2H).

Siglec-EloxP mice were generated in collaboration with Biocytogen Company, and LysM-Cre mice were generated as described previously [45]. To study the role of Siglec-E KO in LysM-Cre-expressing cells, Tm(Siglec-E x LysM-Cre) C57BL/6 mice were generated by crossing LysMCre mice with Siglec-EloxP mice.

Cell culture

Cell lines and primary cells were cultured at 37 °C and 5% CO2 and regularly checked for mycoplasma contamination. All of the cell lines except HEK293T cells were maintained in Dulbecco’s modified Eagle medium (DMEM, Sigma) supplemented with 10% heat-inactivated fetal bovine serum (FBS, PAA Laboratories), 1x MEM nonessential amino acid solution (Sigma) and 1% penicillin/streptomycin (Sigma).

All primary cells as well as HEK293T cells were maintained in Roswell Park Memorial Institute Medium (RPMI, Sigma) supplemented with 10% heat-inactivated fetal bovine serum (FBS, PAA Laboratories), 1x MEM nonessential amino acid solution (Sigma), 1 mM sodium pyruvate (Sigma), 0.05 mM 2-mercaptoethanol (Gibco) and 1% penicillin/streptomycin (Sigma).

Tumor digestion and splenocyte and PBMC isolation

To obtain single-cell suspensions, human and mouse tumors were mechanically dissociated and subsequently enzymatically digested using accutase (PAA Laboratories), collagenase IV (Worthington), hyaluronidase (Sigma) and DNase type IV (Sigma) for 1 h at 37 °C under constant agitation. Afterward, the samples were filtered using a 70 µM cell strainer and washed. Precision counting beads (BioLegend) were added to all mouse tumors to calculate the number of cells per gram of tumor.

PBMCs were isolated from buffy coats by density gradient centrifugation using Hisopaque-1077 (Millipore) and SepMate PBMC isolation tubes (StemCell) according to the manufacturer’s protocol, followed by red blood cell lysis using RBC lysis buffer (eBioscience) for 2 min at RT. Subsequently, the cells were washed with PBS and ready for further analysis.

For splenocyte isolation, freshly harvested murine spleens were mechanically dissociated by filtering through a 100 µM filter. After washing, the red blood cells were lysed as described above.

For murine PBMC analysis, blood was collected from the tail vein of mice on Day 14 of the experiment via tail vein puncture. After washing, the red blood cells were lysed as described above, and the samples were immediately subjected to lectin staining.

Single-cell suspensions were used immediately or frozen for later analysis in liquid nitrogen (in 90% FBS and 10% DMSO).

Tumor models

Siglec-ExLysM-Cre mice (SigEΔLysM) were injected subcutaneously into the right flank with 500,000 B16F10 melanoma, B16F10-sia, B16F10-GNE KO, EL4 lymphoma or EL4 GFP cells in phenol red-free DMEM without additives. Sex-matched Siglec-E WT (SigEWT) littermates were used as controls. Mice were between 8 and 12 weeks of age at the beginning of the experiment, and conditional knockout was confirmed by genotyping and flow cytometry.

Tumor size was measured 3 times a week using a caliper. Animals were sacrificed before reaching a tumor volume of 1500 mm3 or when they reached an exclusion criterion (ulceration, severe weight loss, severe infection or bite wounds). Tumor volume was calculated according to the following formula: tumor volume (mm3) = (d2*D)/2, where D and d are the tumor length and width in mm, respectively.

In vivo treatment

For in vivo Ly6G or Gr1 depletion, mice were injected intraperitoneally twice per week with 100 µg/mouse of anti-Ly6G depletion antibody (clone: 1A8, BioXCell) or 300 µg/mouse of anti-Gr1 depletion antibody (clone: RB6-8C5, BioXCell) in PBS. Antibody treatment began one day before tumor cell injection and was then administered twice a week up to 6 times per mouse.

For in vivo induction of TRAIL-mediated apoptosis, an anti-DR5 antibody (clone: MD5-1; BioXCell) was injected intraperitoneally twice per week. The first injection started 1 day before tumor injection, and the anti-DR5 antibody was administered up to 4 times per mouse.

For neutralization of CCL2 in vivo, mice were injected intraperitoneally 3 times a week with 200 µg/mouse of an anti-CCL2 neutralizing antibody (clone: 2H5, BioXCell) in PBS. Antibody treatment was started one day after tumor injection.

Multiparameter flow cytometry

Multiparameter flow cytometry was performed on single-cell suspensions of the cell lines, PBMCs, splenocytes or tumor homogenates. To avoid nonspecific antibody binding, rat anti-mouse FcγIII/II receptor (CD16/CD32) blocking antibodies (BD Bioscience) were administered to murine cells and Fc receptor binding inhibitor polyclonal antibodies (Invitrogen) were administered to human cells. The cells were subsequently stained with live/dead cell exclusion dye (Zombie Dyes, BioLegend). Surface staining was performed with fluorophore-conjugated antibodies (Table S1) or with lectins for 30 min at 4 °C in FACS buffer (PBS, 2% FCS, 0.5 mM EDTA). The stained samples were fixed using IC fixation buffer (eBioscience) until further analysis. For intracellular staining, the cells were fixed and permeabilized using the Foxp3/transcription factor staining buffer set (eBioscience) and 1x permeabilization buffer (eBioscience) according to the manufacturer’s instructions. All of the antibodies were titrated to obtain the optimal signal-to-noise ratio. Antibody compensation was performed by staining primary cells or using an AbC Total Antibody Compensation Bead Kit (Invitrogen).

Data were acquired by an LSR II Fortessa (BD Biosciences), a CytoFLEX (Beckman Coulter) or a Cytek Aurora (Cytek Biosciences) flow cytometer and analyzed using FlowJo 10.8 (TreeStar, Inc.). Cell sorting was performed using a BD FACSAria III or BD FACSMelody (BD Bioscience). Doublets, cell debris and dead cells were excluded before performing downstream analysis. Fluorescence-minus-one (FMO) samples were used to define the gating strategy and calculate the mean fluorescence intensity (MFI).

To assess desialylation status, cells were stained with lectins as described above. The fluorophore-coupled lectins PNA-PE (GeneTex) and SNA-FITC (VectorBiolabs) and the biotinylated lectin MALII (VectorBiolabs) were used at a final concentration of 10 µg/mL. Biotinylated lectins were detected using PE-streptavidin (BioLegend). Sialidase was stained with neuraminidase antibody (LSBio) followed by PE-conjugated anti-rabbit IgG secondary staining (Invitrogen).

Phospho-flow cytometry

Phospho-flow cytometry was used to assess the levels of phosphorylated STAT3 (pSTAT3). Frozen murine tumors were thawed, washed and cultured for 10 min with pervanadate at 37 °C in complete RPMI. Subsequently, the cells were stained for multiparameter flow cytometry as described above.

In vitro generation of human suppressive myeloid cells

To generate suppressive human myeloid cells, we used an adapted version of the protocol established by Lechner et al. [26].

A. Generation of MDSC-like cells

For in vitro MDSC induction, freshly isolated PBMCs from healthy donor buffy coats were cocultured for 7 days with different cancer cell lines (A549, A549-GNE KO, A549-sia, HeLa, or HeLa-sia) at a ratio of 1:100 in complete RPMI medium supplemented with 10 ng/mL GM-CSF (PeproTech). Cancer cells were seeded at an initial concentration of 1 × 10e4 cells/mL, and the same amount of medium supplemented with GM-CSF was added on Day 4 of the experiment. After one week, all confluent and adherent cells were collected using 0.05% trypsin-EDTA (Gibco).

B. Isolation of MDSC-like cells

For MDSC isolation, CD33 positive cells were magnetically isolated using a human CD33 positive selection kit II (StemCell) following the manufacturer’s instructions. The isolated cells were resuspended in complete fresh RPMI medium for the suppression assay.

C. Isolation of autologous CD8 T cells

Autologous CD8+ T cells were obtained from frozen PBMCs from the same donor using CD8+ microbeads from a human T-cell isolation kit (Miltenyi Biotec). To monitor cell proliferation, cells were labeled with 1.25 µM CellTrace Violet (Invitrogen) according to the manufacturer’s instructions. The washed cells were resuspended in complete RPMI and used for the suppression assay.

D. Suppression assay

Isolated MDSC-like cells and autologous CD8+ T cells were cocultured at the indicated ratios for 5 days in a U-bottom plate in complete RPMI. T cells were stimulated by the addition of 100 IU/mL IL-2 (proleukin) and anti-CD3/CD28 stimulation using loaded MACSiBead particles (Miltenyi Biotec) at a ratio of 1:1 beads to cells. Unstimulated T cells and stimulated T cells without MDSC addition were used as controls. After five days, the supernatants were frozen at −80 °C, and the cells were stained for flow cytometry.

For Siglec-9 blocking, a Siglec-9 blocking antibody (Clone 191240, R&D Systems) was added at a final concentration of 10 µg/mL. For sialidase treatment, MDSCs were pretreated and washed before being added to the wells as described below.

Human intratumoral-derived MDSC suppression assay

Fresh PBMCs and tumor homogenates were used immediately after isolation as described above. MDSCs were isolated from tumor homeogenates, and CD8 T cells were isolated from PBMCs using CD33 microbeads (Miltenyi) or CD8 microbeads (Miltenyi). The cells were cocultured at an MDSC:T-cell ratio of 1:4 in a U-bottom plate for 5 days in complete RPMI in the presence of 30 IU of IL-2 (proleukin) and human CD2/CD3/CD28 T-cell activator at a final concentration of 25 µL/mL (Immunocult, StemCell). For Siglec-9 blocking, a Siglec-9 blocking antibody (Clone 191240, R&D Systems) was added at a final concentration of 10 µg/mL. For sialidase treatment, MDSCs were pretreated and washed before being added to the wells as described below.

The supernatants were frozen at −80 °C, and the cells were stained for flow cytometry.

Murine MDSC suppression assay

Murine T cells were enriched from wild-type mouse splenocytes by negative selection using a murine pan-T-cell isolation kit (EasySep, StemCell). To monitor T-cell proliferation, isolated T cells were stained with 2.5 µM CellTrace Violet (Invitrogen) according to the manufacturer’s instructions.

Murine MDSCs were isolated from the splenocytes of tumor-bearing mice by negative selection using a murine MDSC isolation kit (EasySep, StemCell). As indicated, the obtained MDSCs were used immediately or pretreated with sialidase as described below.

Isolated MDSCs and T cells were plated at a ratio of 1:1 in a 96-well flat bottom plate and cocultured for 48 h in complete RPMI in the presence of 50 IU of IL-2 (proleukin). For T-cell stimulation, the plate was coated with anti-CD3 (clone 17A2; BioLegend) and anti-CD28 (clone 37.51; BD Biosciences) antibodies. The supernatants were frozen at −80 °C, and the cells were stained for flow cytometry.

For Siglec-E blocking, purified anti-mouse Siglec-E antibody (M1305A02; BioLegend) or rat IgG2a or κ isotype control (BioLegend) was added at a final concentration of 10 µg/mL. CCL2 blocking was performed by the addition of 50 µg/mL CCL2 (Clone 2H5, BD).

Sialidase treatment

To cleave terminal sialic acid residues, cells were treated with bacterial sialidase (Vibrio cholerae, Sigma) at a concentration of 10 µM for 20 min in PBS. Subsequently, the cells were washed with complete medium and used for downstream analysis. Additionally, viral sialidase (active H1N1; Sino Biological) and bacterial sialidase (Arthrobacter ureafaciens; Roche) were used for pretreatment, as indicated in the figure legends. If not stated otherwise, Vibrio cholerae bacterial sialidase was used.

Lentivirus production and lentiviral transduction of cell lines

To generate cell lines expressing H1N1 viral sialidase and GFP, A549, HeLa, B16F10 and EL4 cells were stably transduced with lentivirus.

For lentiviral production, 14 × 10e6 HEK293T cells were seeded 24 h before transfection in 18 mL of complete RPMI medium in a 15-cm culture dish. For the transfection mixture, 1.9 µg of pMD2. G, 3.5 µg of pCMVR8.74 and 5.4 µg of pLV transfer vector were mixed in 1.8 ml of jetOPTIMUS buffer (Polyplus). Then, 16.2 µl of jetOPTIMUS (Polyplus) was added to the prepared transfection mixture followed by 10 min of incubation. The medium was exchanged after 16 h, and the lentiviral particles were collected 24 and 48 h after medium exchange. The pooled supernatant was concentrated with 4x in-house-generated PEG-8000 solution and resuspended in PBS supplemented with 1% human serum albumin. Aliquots of the produced virus were stored at −80 °C until further use. pMD2. G and pCMVR8.74 were kindly provided by Didier Trono (Addgene plasmids #12259 and #22036).

For lentiviral transduction, 50,000 cancer cells were seeded in a 24-well plate in 500 µL of complete RPMI medium and allowed to rest overnight. The media was renewed by the addition of 100 µL of concentrated lentivirus and 8 µg/mL polybrene (Sigma). To increase transduction efficiency, spinoculation was performed, and the cells were centrifuged for 90 min at 800 × g. Afterward, the cells were incubated under standard cell culture conditions, and the transduction efficiency was frequently checked via flow cytometry staining to assess sialidase expression and PNA, MALII and SNA levels.

Cytokine and chemokine analysis

The supernatants collected from murine and human cocultures were thawed on ice, and aliquots were subsequently sent on dry ice to Eve Technologies (Canada). Cytokine and chemokine concentrations were analyzed and calculated by Eve Technology. For visualization, normalized values (z scores) of each cytokine were calculated based on the mean and standard deviation of each marker.

Bulk RNA sequencing

MDSC-like cells generated with A549 and A549-sia cancer cells from 4 different healthy donors were harvested after 7 days of coculture as described above. For purification of MDSCs, cells were stained with CD33-PE (Miltenyi) followed by CD33-positive selection using the EasySep Human PE Positive Selection Kit II (StemCell). To increase purity, the cells were further sorted for PE-positivity by an Aria III (BD Biosciences) flow cytometer. For RNA purification, the sorted cells were washed, and RNA was isolated using an RNeasy Plus Micro Kit (Qiagen), which included QIAshredder spin columns (Qiagen) for gDNA elimination.

Quality control (QC length profiling and concentration determination using RiboGreen) and library preparation (TruSeq stranded mRNA HT Kit by Illumina) were performed by the Genomics Core Facility of the University Basel. Sequencing was performed on four lanes of the Illumina NextSeq 500 instrument, resulting in 38 nt paired-end reads. The dataset was analyzed by the Bioinformatics Core Facility, Department of Biomedicine, University of Basel. cDNA reads were aligned to the ‘hg38’ genome using the Ensembl 104 gene models with the STAR tool (v2.7.10a) with default parameter values except for the following parameters: outFilterMultimapNmax=10, outSAMmultNmax=1, outSAMtype=BAM SortedByCoordinate, and outSAMunmapped=Within. At least 40 M read pairs were mapped per sample. The software R (v4.1.1) and the tool featureCounts from the Subread (v2.0.1) package from Bioconductor (v3.14) were used to count aligned reads per gene with default parameters except for -O, -M, --read2pos = 5, --primary, -s 2, -p, and -B. Further analysis steps were performed using R (v4.2.0) and multiple packages from Bioconductor (v3.15). The package edgeR (v3.38.1) was used to perform differential expression analysis. A gene was included in the analysis only if it had at least 1 count per million (CPM) in at least four samples. Gene set enrichment analysis was performed using a tool camera from the edgeR package and the Gene Ontology gene set (category C5) from MSigDB (v7.5.1). The raw sequences (FASTQ files) are available through the European Genome-Phenome Archive under accession number EGAS00001007618.

Single-cell RNA sequencing

A549 and A549-sia MDSC-like cells from 4 different healthy donors were generated and processed by fluorescence-activated cell sorting as described for bulk RNA sequencing (see above). Single-cell RNA sequencing was performed using Chromium Next GEM Single-cell 3′ with feature barcode technology from 10x Genomics according to the manufacturer’s instructions. Approximately 20,000 cells were loaded in a 10x Genomics cartridge for target recovery. Cell-barcoded 3′ gene expression libraries were sequenced on an Illumina NovaSeq 6000 system and mapped to the GRCh38 human reference genome using CellRanger v. 7.1 (10x Genomics).

For single-cell demultiplexing, we utilized patient-specific single nucleotide polymorphisms (SNPs). To genotype Bi-Allelic SNPs on single cells, Cellsnp-lite [46] was used on the output bam files and cell barcodes derived from CellRanger. The SNP list hapmap_3.3.hg38_maf.vcf of the Genome Analysis Toolkit from the Broad Institute was utilized as a SNP reference. The generated output was subsequently processed with vireo [47] to assign the identified SNPs to the 4 individuals and annotate the resulting doublets. Identified donors were matched between wells based on SNP overlap and corresponding cells were analyzed by the Seurat R package. The sequences are available through the European Genome-Phenome Archive under accession number EGAS00001007618.

Analysis of scRNA-seq data

The read count matrices were processed and analyzed in R v. 4.3.0 using Seurat v.5.0.1 [29] with default parameters for all functions, unless otherwise specified. Cell quality was investigated, and doublets and low-quality cells were removed based on the following criteria: nFeatures under 600, nCounts under 900 and a percentage of mitochondrial genes greater than 10%. Subsequently, the data were normalized to the variable features identified and scaled. After principal component analysis (PCA) reduction, the first 10 PCs were used for nearest neighbor analysis with a resolution of 0.4 for cluster identification. UMAP was also calculated for the first 10 PCs. Clusters were investigated, and those with no CD33 expression were identified as impurities and excluded from further analysis. Normalization, variable feature identification and dimensional reduction were repeated for the subset by using the first 16 PCs. A resolution of 0.2 was used, and the Leiden algorithm was used for cluster identification.

Gene set enrichment analysis was performed using the Seurat FindAllMarkers function, allowing the identification of up- and downregulated markers per cluster. The logfc.threshold was set to zero. Gene set enrichment analysis (GSEA) was performed for each cluster using the clusterprofiler package [48]. The results were ordered based on the normalized enrichment score, and only the highest enriched Gene Ontology terms were used for data interpretation.

N-glycomic profiling

For N-glycomic profiling, harvested cells were extracted with lysis buffer containing 7 M urea, 2 M thiourea, and 10 mM dithioerythreitol in 40 mM Tris buffer supplemented with 1% protease inhibitor (Roche). The cell membranes were disrupted by high-intensity focused ultrasound (HICUS) with 10 cycles of 10 s of sonication at 16× magnification and 1 min on ice in between sonication and subsequent shaking for 4 h in a cold room. The protein extracts were alkylated with 100 mM iodoacetamide in the dark for 4 h at 37 °C. Ice-cold trichloroacetic acid was added to a final concentration of 10% w/v, and the mixture was left for one hour. After centrifugation at 20,000 × g for 30 min at 4 °C, the precipitated sample pellets were washed twice with ice-cold acetone and then lyophilized. The dry protein pellets were redissolved in 50 mM ammonium bicarbonate buffer (pH 8.5), 250 units of benzonase nuclease (Sigma‒Aldrich) were added, and the mixture was incubated for 30 min at 37 °C, followed by trypsin digestion overnight. After the activity of trypsin was deactivated, the protein mixtures were further treated with PNGaseF (New England Biolab). The released glycans were removed according to previous methods [49].

For MALDI-MS analyses, the glycan samples were permethylated using the sodium hydroxide/dimethyl sulfoxide slurry method, as described by Dell et al. [50]. The samples were dissolved in 20 µL of acetonitrile. One microliter of sample mixed with 10 mg/mL 2,5-dihydroxybenzoic acid (Bruker) in 70% acetonitrile with 1 mM sodium chloride was added to a MALDI target plate and analyzed by a Bruker RapiFlexTM MALDI-TOF-TOF instrument. Permethylated high-mannose N-glycans and glycans from fetuin were used to calibrate the instrument prior to the measurement. The laser energy for each analysis was fixed, and the data were accumulated from 10 000 shots. The data were analyzed by GlycoWorkbench [51] and inspected manually. For relative quantification, the data were first deisotoped, and the peak height was used for the calculation based on the following equation:

$$\,\,\,=\frac\,\,\,\,\,\,}\,\,\,\,\,\,}\times 100 \%$$

Statistical analysis

All of the statistical analyses were performed using GraphPad Prism 9 or base functions in R v. 4.3.0. The statistical tests used and sample sizes are indicated in the figure legends.

p values > 0.05 were considered not significant, and p values < 0.05 were considered significant. Asterisks indicate the following: *p value < 0.05, **p value < 0.01, ***p value < 0.001, ****p value < 0.0001. n indicates the number of biological replicates. All of the bars within the graphs represent mean values, and the error bars represent standard errors of the mean (SEM) or standard deviation (SD) as indicated.