Integrated Glycoproteomic Characterization of Clear Cell Renal Cell Carcinoma

Summary

Clear cell renal cell carcinoma (ccRCC), a common form of RCC, is responsible for the high mortality rate of kidney cancer. Dysregulations of glycoproteins have been shown to associate with ccRCC. However, the molecular mechanism has not been well characterized. Here, a comprehensive glycoproteomic analysis was conducted using 103 tumors and 80 paired normal adjacent tissues. Altered glycosylation enzymes and corresponding protein glycosylation were observed, while two of the major ccRCC mutations, BAP1 and PBRM1, showed distinct glycosylation profiles. Additionally, inter-tumor heterogeneity and cross-correlation between glycosylation and phosphorylation were observed. The relation of glycoproteomic features to genomic, transcriptomic, proteomic, and phosphoproteomic changes reveals roles of glycosylation in ccRCC development with potential for therapeutic interventions. This study reports a large-scale TMT-based quantitative glycoproteomic analysis of ccRCC which can serve as a valuable resource for the community.

Introduction

Renal cell carcinoma (RCC) is a highly heterogeneous cancer of which malignant cells were formed from renal tubular epithelial cells¹. There are estimated 79,000 new cases and 13,920 deaths in 2022 with kidney and renal pelvis cancer in United States². Among the various subtypes of RCC, clear cell renal cell carcinoma (ccRCC) represents the most common form of RCC (~75%) and is responsible for the high mortality rate of kidney cancer¹. The primary treatment for localized ccRCC tumors is surgical resection because conventional radiotherapy and chemotherapy have a little effect^3,4. Although, several Food and Drug Administration (FDA)-approved agents have been applied to patients with advanced RCC targeting various intracellular pathways (e.g., mTOR), the patients’ response to these treatments is still limited^5,6. Therefore, it is essential to understand the molecular alteration associated with the development and the progression of ccRCC to derive suitable therapeutic interventions for the management of ccRCC patients. Our previous study characterized ccRCC by utilizing proteogenomic approach, which enhanced our understanding of the impact of genomic alterations of ccRCC on protein networks and signaling pathways⁷. Dysregulations of glycoproteins have been shown to associate with ccRCC development ^8,9. However, the molecular mechanism has not been well characterized.

Glycosylation is one of the most prevalent protein modifications. Dysregulation of glycoproteins have been shown to affect biological functions or disease development^10,11. Many currently FDA-approved biomarkers for cancer diagnosis and monitoring are glycoproteins, such as fucosylated alpha-fetoprotein for hepatocellular carcinoma, CA125 for ovarian cancer, and CA19–9 for pancreatic cancer¹⁰. Glycoproteins are often found on the cell surface or secreted from cells; thus, discovery of dysregulated glycoproteins/glycosylation events and their corresponding signaling pathways could potentially target the ccRCC cells for treatment. Owing to the challenges posted by the heterogeneity and complexity of glycosylation, large-scale glycoproteomic analysis has not been feasible until recently the advances in mass spectrometry (MS)-based glycoproteomic technology^12–18. Glycoproteomic studies carried out on several cancer types have revealed the functions of glycosylation in cancer development^19–23. Thus, a comprehensive characterization of ccRCC using quantitative glycoproteomic approach would allow the investigation of the roles of protein glycosylation specific to ccRCC.

In this study, we utilized TMT-based MS data to quantitatively analyze N-linked glycoproteins of 103 ccRCC tumors and 80 paired normal adjacent tissues (NATs) and quantified glycosite-specific glycosylation. Alteration in glycosylation was observed in ccRCC tumors compared to NATs with the responsible glycosylation enzymes were identified. Moreover, BAP1 and PBRM1 are two of the five major mutations of ccRCC that are mutually exclusive, where BAP1-mutant tumors are often associated with poor clinical outcome^24,25. In this particular cohort, the aggressiveness of BAP1-mutant tumors may be explained by the variations in glycosylation patterns that were found in comparison to the wildtype and PBRM1-mutant tumors. The inter-tumor heterogeneity of ccRCC was detected by subtyping the tumors into three distinct glyco-subtypes. Furthermore, different protein modifications can co-occur on the proteins that a crosstalk/cross-correlation among the protein modifications on the same or different proteins can impact the functional regulation of the proteins^26,27. Protein phosphorylation is another common protein modifications involved in many cellular processes. The association between N-linked glycosylation and phosphorylation is poorly understood. Thus, we performed an integrated analysis to predict whether crosstalk of glycosylation and phosphorylation occurred in ccRCC. We observed the association between glycan types and phosphoproteins as well as revealing distinct cross-correlation clusters among the ccRCC tumors. Overall, aforementioned findings indicate the role of protein glycosylation specific to ccRCC. Additionally, this dataset is a rich source for future studies focused on development of potential therapeutic interventions to treat the cancer.

Results

Glycoproteomic profile of ccRCC

We investigated the glycoproteomic landscape of ccRCC by analyzing 103 treatment-naïve ccRCC tumors and 80 paired-matched NATs in the current study. Figure S1A shows the experimental workflow of generating proteomic, phosphoproteomic, and glycoproteomic data for the study. The summarized clinical data can be found in Table S1, where detailed clinical information is in our previous publication⁷.

Intact glycopeptide abundances from all tissue samples, eight quality control (QC) samples, and five NCI-7 cell samples were measured in 23 TMT-10plexes. The results showed a large number of intact N-linked glycopeptides were identified in each TMT set (Figure S1B). A total of 44,181 intact glycopeptides (FDR < 1%) from 1,429 glycoproteins were identified in at least one sample across 23 TMT sets. Majority of the intact glycopeptides were from the proteins in the extracellular matrix, cell surface, lysosome, and plasma member region (Figure S1C). For the downstream analysis, intact glycopeptides quantified in greater than 50% of the samples were retained and analyzed.

We examined four glycan categories according to the monosaccharide composition of N-glycans associated with the identified glycopeptides as follows: oligomannose (High-Man), fucosylated glycans (Fucose), sialylated glycans (Sialic, including glycans with sialylation only and those with both sialylation and fucosylation), and other glycoforms (Others) that did not fit the first three categories. The percentage of identified intact N-linked glycopeptides according to glycoforms were 24.9 %, 24.3 %, 40.2 %, and 10.6% for High-Man, Fucose, Sialic, and Others, respectively. The reproducibility among inter-TMT experiments was evaluated using the QC and NCI-7. Among the QC samples, the median Spearman correlation and the median coefficient of variation (CV) was 0.95 and 18.5%, respectively (Figures 1D–E, Table S1). Similar results were observed for the NCI-7 samples with a median Spearman correlation of 0.95 and CV of 15.2% (Figures 1F–G, Table S1) indicating high reproducibility of the glycoproteomic data of ccRCC.

Figure 1.

Comparative analysis to examine alter glycosylation in ccRCC. (A) Principal component analysis between ccRCC tumors and NATs. (B) Differential analysis between tumors and NATs. Significantly altered intact glycopeptides were defined as >2-fold changes with false discovery rate (FDR)<0.01. (C) Distribution of glycan types for upregulated and downregulated intact glycopeptides. (D) Differential analysis of glycosylation enzymes on global level. Significantly altered glycosylation enzymes were defined as >1.5-fold changes with FDR<0.05. (E) Glycan types observed in differentially expressed intact glycopeptides (FDR<0.05) between high-grade and low-grade tumors. (F) Glycosylation changes in NDRN²⁴⁹SSDETFLK of GPNMB in high-grade tumors relative to low-grade tumors (FDR<0.05 as significant changes). (B, D, E, and F) The p-values were computed using two-sided Wilcoxon rank sum test and adjusted (i.e., FDR) using Benjamini-Hochberg method. See also Figure S2 and Table S2.

Altered glycosylation in ccRCC compared to NATs

We first analyzed quantitative glycoproteomic data to investigate whether ccRCC tissues showed distinct glycopeptide features in comparison to non-cancerous tissues. A clear separation between 103 tumors and 80 NATs was observed based on the abundances of intact glycopeptides by principal component analysis (Figure 1A). Further carrying out the differential analysis between ccRCCs and NATs revealed a total of 2,064 intact glycopeptides with significantly larger than 2-fold changes (FDR < 0.01), where 625 upregulated and 1,439 downregulated in tumors relative to NATs (Figure 1B, Table S2). We found a high proportion of upregulated intact glycopeptides with High-Man (44.8%) followed by Sialic (32.5%), whereas more fucosylated glycans (41.6%) were found in downregulated intact glycopeptides (Figure 1C).

The differentially expressed intact glycopeptides originated from 317 glycoproteins, where 121 and 160 glycoproteins were unique to the upregulated and downregulated glycopeptides, respectively (Figure S2A). Unique glycoproteins were subjected to the KEGG pathway enrichment analysis (Figure S2B). Pathways, such as extracellular matrix (ECM) receptor interaction, cell adhesion molecules (CAMs), and focal adhesion, were only enriched from the glycoproteins of the upregulated glycopeptides. On the other hand, glycoproteins of downregulated glycopeptides were mainly from lysosome, complement and coagulation cascades, and cholesterol metabolism pathways. In addition, distinct glycan categories were observed in the glycopeptides of the glycoproteins mapped to the aforementioned pathways (Figure S2C). For instance, in the cholesterol metabolism pathway, most of the downregulated glycopeptides were modified by fucosylated glycans, whereas high-mannose and sialylated glycans were found in the upregulated glycopeptides.

The distinct glycan profiles for differentially expressed intact glycopeptides were potentially associated with glycosylation enzymes expressed in the tumors. The altered glycosylation enzymes could provide prognostic values and/or additional therapeutic targets for treating ccRCC. Upregulation of MGAT1 and MAN1B1 and downregulation of MAN1C1 and MAN1A1 were observed in tumors relative to NATs (Figure 1D, Table S2). Similar expression patterns were found on the transcriptomic level (Figure S2D, Table S2). Decrease in MAN1C1 and MAN1A1 and elevation of MGAT1 may explain the high percentage of High-Man and complex glycans with sialic acid in tumors since MAN1C1 is responsible for trimming α-1,2-linked mannose residues from N2H8 (N=HexNAc and H=Hex) to produce N2H5 along with MAN1A1 and MAN1A2, while MGAT1 relates to the synthesis of complex glycans.

Variation in glycosylation patterns between high-grade and low-grade ccRCC tumors

In this study, we found glycopeptides with high-mannose and complex glycans with sialic acid were highly expressed in tumors compared to NATs (Figure 1C). Previous studies have shown that high-mannose or sialylated glycans are associated with cancer progression in different cancer types^28,29. It was intriguing to investigate whether glycosylation patterns varied between low-grade (G1 and G2) and high-grade (G3 and G4) ccRCC tumors. Among the 220 differentially expressed intact glycopeptides (FDR<0.05) between low-grade and high-grade tumors, High-Man were dominated in high-grade tumors, where most of the high-mannose glycans were N2H8 followed by N2H7 and N2H6 (Figure 1E, Table S2). We speculated that the elevated MAN1B1 (FDR<0.05) in high-grade relative to low-grade contributed to the High-Man patterns that were found in high-grade tumors (Table S2).

The upregulated intact glycopeptides with High-Man glycans were originated from glycoproteins with biological processes such as regulation of leukocyte activation (e.g., Transmembrane glycoprotein NMB, GPNMB), receptor-mediated endocytosis, (e.g., Scavenger receptor cysteine-rich type 1 protein M130, CD163), and integrin-mediated signaling pathway (e.g., Integrin alpha-V, ITGAV). GPNMB has been widely studied in various cancer types, which plays a role in metastasis of cancer³⁰ and a potential therapeutic target for metastatic ccRCC³¹. In this study, the global expression of GPNMB was unchanged regardless of tumor grade groups (Figure 1F, Table S2). Among the four identified glycosylation sites on GPNMB, majority of the glycosylation events showed no significant changes between high-grade and low-grade, except High-Man glycans (N2H7 and N2H8) on the glycosite, NDRN²⁴⁹SSDETFLK of GPNMB, showed significant elevation in high-grade compared to low-grade tumors (Figure 1F).

On the other hand, low-grade tumors were enriched with Sialic glycans (Figure 1E). The sialylated glycopeptides were from glycoproteins that were associated with biological processes such as cell-substrate adhesion (e.g., Cadherin-13, CDH13) and angiogenesis (e.g., Platelet endothelial cell adhesion molecule, PECAM1). In our previous work, PECAM1 was discovered as one of the gene signatures in VEGF immune desert subtype and this particular immune subtype had better survival probability⁷. We observed that sialylation was predominated on the four identified glycopeptides of PECAM1, where a significant downregulation of the glycopeptide, VN¹⁵¹CSVPEEK, attached with N5H6F1S1 (F=fucose and S=sialic acid) was observed in high-grade tumors (Figure S2E, Table S2).

Glycosylation in tumors with specific mutations

PBRM1 and BAP1 are two of the five commonly identified mutated genes in ccRCC that patients with BAP1 mutations tend to have poorer clinical outcomes than patients with PBRM1 mutations^24,25. Moreover, PBRM1 were found to be mutually exclusive with BAP1 mutations in ccRCC⁷. Therefore, we investigated the glycosylation patterns in specific mutations.

A comparative analysis was first conducted between mutant and wildtype tumors on both intact glycopeptide and global protein levels. The altered intact glycopeptides were mainly positively associated with the expression of the corresponding proteins on the global level for BAP1-mutant tumors (Figure 2A, Table S3) and PBRM1-mutant tumors (Figure 2B, Table S3) in comparison to BAP1 wildtype (BAP1-WT) and PBRM1 wildtype (PBRM1-WT) tumors, respectively. However, elevation in High-Man and Sialic glycans was observed in BAP1-mutant, whereas PBRM1-mutant tumors tended to show upregulation in Fucose and Others glycans (>1.5-fold changes with FDR < 0.05, Figure 2C, Table S3).

Figure 2.

Alteration in glycosylation in BAP1-mutant and PBRM1-mutant tumors. (A) Median log2 fold change between intact glycopeptides and corresponding glycoproteins of BAP1-mutant and BAP1-WT tumors. (B) Median log2 fold change between intact glycopeptides and corresponding glycoproteins of PBRM1-mutant and PBRM1-WT tumors. (C) Distribution of glycan types for upregulated and downregulated intact glycopeptides in BAP1-mutant and PBRM1-mutant compared to wildtypes. (D) Biological processes of BAP1-mutant-associated and PBRM1-mutant-associated glycoproteins based on the differential analysis of intact glycopeptides of BAP1-mutant vs BAP1-WT, PBRM1-mutant vs PBRM1-WT, and BAP1-mutant vs PBRM1-mutant. Significantly altered intact glycopeptides were defined as >1.5-fold changes with FDR<0.05. The p-values were computed using two-sided Wilcoxon rank sum test and adjusted (FDR) using Benjamini-Hochberg method. See also Table S3.

Examination of the intact glycopeptides with altered abundance in BAP1-mutant tumors indicated the corresponding glycoproteins were associated with immunity, inflammation, and leukocyte migration, including Angiotensin-converting enzyme 2 (ACE2), Lysosome-associated membrane glycoprotein 2 (LAMP2), and Golgi apparatus protein 1 (GLG1) (top panel of Figure 2D, Table S3). On the other hand, the glycoproteins of differentially expressed intact glycopeptides in PBRM1-mutant compared to PBRM1-WT tumors were related to metabolic process, including Agrin (AGRN), Glutamyl aminopeptidase (ENPEP), and Low-density lipoprotein receptor-related protein 2 (LRP2). By directly comparing BAP1-mutant and PBRM1-mutant tumors, we further verified that the significantly enriched glycosylation events observed in the comparison to WT tumors were undoubtedly unique to the BAP1-mutant and PBRM1-mutant tumors (bottom panel of Figure 2D, Table S3).

Subtyping ccRCC using glycoproteomics

Immune-based and global proteomic-based subtyping of ccRCC were performed in our previous study⁷. In this study, subtyping analysis was conducted using intact glycopeptides with CVs in > 25% quantile (8,111 intact glycopeptides) to examine the glycoproteomic-based inter-tumoral heterogeneity of ccRCC.

Three glyco subtypes, Glyco 1 (n=33 tumors), Glyco 2 (n=33), and Glyco 3 (n=37), were identified (Figure 3A, Table S4). Glyco 1 tumors were characterized by high grade and stage tumors and a high frequency of chromosome 14 loss, genome instability, and BAP1 mutation. The relationship between protein-based and glyco-based ccRCC subtypes was delineated according to the enrichment scores and p-value (p<0.05) from a hypergeometric test, which revealed Glyco 1 tumors were associated with Global 1 subtype (Figure 3B). Similar analysis was performed between immune-based and glyco-based subtypes, we found that Glyco 1 overlapped with two immune subtypes, CD8+ inflamed and Metabolic immune-desert (Figure 3C). On the other hand, majority of the tumors in Glyco 2 and Glyco 3 with lower grade and stage compared to Glyco 1. However, Glyco 2 overlapped with Global 2, while Glyco 3 showed a higher frequency of PBRM1 mutation as well as coincided with Global 3 and VEGF immune desert (Figures 3A–C).

Figure 3.

Inter-tumor heterogeneity of ccRCC. (A) Glycoproteomic subtyping of ccRCC. (B) Association between Glyco subtypes and global subtypes (* p-value<0.05, hypergeometric test). (C) Association between Glyco subtypes and immune subtypes (* p-value<0.05, hypergeometric test). (D) Glycan type distribution of each intact glycopeptide cluster (IPC). (E) Association among Glyco subtypes, glycan type, and IPC. (F) Examples of intact glycopeptide signatures of each glyco subtype. (G) Survival analysis of Glyco subtypes. Log-rank test p=0.015 (H) Survival analysis based on abundance of High-Man where samples with abundance above median were assigned to High group (n=51), otherwise assigned to Low group (n=52). Log-rank test p=0.013. See also Table S4.

Four intact glycopeptide clusters (IPC) were established to investigate the association between the glyco subtypes and glycan types. As shown in Figure 3D, the intact glycopeptides classified in IPC 1 mainly contain Sialic and Fucose glycans. In contrast, IPC 2 was dominated by High-Man glycans, where IPC 3 was a cluster occupied mostly by intact glycopeptides with Fucose glycans. IPC 4 was the predominant cluster of Sialic glycans. To establish the relationship between the glyco subtypes and the glycan types, we utilized the glycans associated with IPCs and averaged Z-score of the intact glycopeptides from each of the four IPCs (Figure 3E, Table S4). Intact glycopeptides in IPC 2 were highly expressed in Glyco 1 compared to the other IPCs and tumor subtypes; thus, tumors in Glyco 1 were considered as High-Man associated tumors. Moreover, an increase in the expression of intact glycopeptides from IPC 3 was observed in Glyco 2, while lower in both Glyco 1 and Glyco 3. A high frequency of intact glycopeptides was fucosylated in IPC3, therefore, Glyco 2 was regarded as Fucose-associated tumors. On the other hand, Glyco 3 showed an association with both IPC 1 and IPC 4. Since sialylated glycopeptides were dominated in IPC 1 and IPC 4, especially in IPC4, Glyco 3 was considered as Sialic glycan associated glyco subtype.

Furthermore, we conducted comparative analysis between a glyco subtype and the other subtypes to find intact glycopeptide signatures for each glyco subtype. To ensure the signatures were distinct for a particular subtype, each subtype was further compared to the NATs. Among the 8,111 intact glycopeptides, 150 intact glycopeptides with elevated expression profiles (>1.5-fold increase and FDR < 0.05) in Glyco 1 relative to Glyco 2, Glyco 3, and NATs. We also found 28 and 22 intact glycopeptides upregulated (>1.5-fold increase and FDR < 0.05) in Glyco 2 and Glyco 3, respectively (Table S4). Many of the glycopeptides were from glycoproteins related to immunity in Glyco 1 (e.g., CD63 antigen, CD63, Figure 3F). We observed glycopeptides from glycoproteins associated with urogenital system development (e.g., Laminin subunit alpha-5, LAMA5) and complement activation (e.g., Microfibril-associated glycoprotein 4, MFAP4) in Glyco 2 and Glyco 3, respectively (Figure 3F).

The probability of patient survival based on the three glyco subtypes revealed that Glyco 1 tumors were associated with the worst patient survival among the three Glyco subtypes (Figure 3G). The low survival rate of Glyco 1 subtype correlates with the features that are regarded as poor prognosticators in ccRCC, including CD8+ inflamed immune subtype, higher frequency of BAP1 mutations, and high-grade tumors. In addition, the patient outcome was evaluated based on the glycan types. We did not detect an association between survival and glycan types, except for High-Man of which high expression level of High-Man glycan type would result in poor survival probability (Figure 3H).

Cross-correlation between glycosylation and phosphorylation in ccRCC

Glycosylation and phosphorylation are two prevalent and broadly studied protein modifications, however, the cross-correlation between these two protein modifications is poorly understood. Therefore, we investigated whether there were certain interactions between N-linked glycosylation and phosphorylation that were associated with ccRCC.

It is intriguing to first explore whether glycan types affect the activation of the downstream signaling pathways of phosphoproteins in ccRCC. In this study, we classified the glycans into four major glycan types as aforementioned. The Spearman correlation was calculated between each glycan type and phosphoproteins, while only correlation > ±0.3 were selected for subsequent analysis. The associations between the glycan types and phosphoproteins could be classified into three different pathways, including metabolic pathways, RNA transport, and complement and coagulation cascades (Figure 4A, Table S5). Fucose glycans and other glycans (i.e., none High-Man and Sialic) were positively correlated with phosphoproteins in the metabolic pathways, whereas High-Man glycans were associated with phosphoproteins in the RNA transport pathway. High-Man glycans can contribute to the spread of cancer cells^28,32; thus, High-Man glycans may alter the regulation of RNA transport enhancing cancer cell proliferation. Moreover, we found that Sialic glycans were positively associated with phosphoproteins from the pathway of complement and coagulation cascades, whereas the High-Man glycans were negatively correlated with this particular pathway (Figure 4A). Studies showed that low complement activity was observed when the cell surface with abundant sialic acid, which enhanced cancer progression by allowing cancer cells to avoid complement attack^33,34. We further performed linear regression between intact glycopeptides and phosphopeptides based on the aforementioned correlated glycan types and phosphoproteins with consideration of the global expressions of corresponding glycoproteins and phosphoproteins. Potential connections among the phosphorylation and glycosylation events in metabolic pathways, RNA transport, and complement and coagulation cascades were observed (Figures 4B, S3A–B, Table S5). For example, S555 of Complement C1r subcomponent (C1R) showed associations with glycopeptides from various glycoproteins, including those involved in neutrophil mediated immunity, such as Lysosome-associated membrane glycoprotein 1 (LAMP1), Haptoglobin (HP), and Adhesion G protein-coupled receptor E5 (ADGRE5). These observations suggest that the phosphorylation of C1R-S555 may depend on the glycosylation activities of presented glycoproteins (Figure 4B).

Figure 4.

Cross-correlation between glycosylation and phosphorylation in ccRCC. (A) Correlation between glycan types and phosphoproteins. (B) Networks based on the linear models constructed using intact glycopeptides and phosphopeptides that were associated with complement and coagulation cascades. (C) Association between EGFR-N352 (N2H5) and EGFR-S1018 with p=0.037 (t-statistic). (D) Cross-Correlation (CC) clusters derived using NMF multi-omic clustering. (E) Distinct glycosylation (FLT1-N251, N4H5S1) and phosphorylation (ENO1-S373) events in CC 1 tumors relative to the other CC clusters. (F) Distinct glycosylation (CYBB-N240, N2H8) and phosphorylation (DDX3X-S102) events in CC 2 tumors relative to the other CC clusters. (G) Distinct glycosylation (TF-N630, N3H6S1) and phosphorylation (HBA2-T138) events in CC 3 tumors relative to the other CC clusters. (H) Summarized key observations for ccRCC in this study. See also Figure S3 and Table S5.

Furthermore, some proteins had been detected with both phosphorylation and glycosylation activities that these proteins were found in different biological processes, including Epidermal growth factor receptor (EGFR), PECAM1, and kidney tissue-associated proteins (e.g., LRP2). An upregulation of EGFR was observed in the majority of the ccRCC tumors⁷, which made it a potential therapeutic target. Here, a linear relationship was noted between the phosphorylation at EGFR-S1018 and glycosylation at EGFR-N352 (N2H5) (Figure 4C, Table S5). The glycosylation at EGFR-N352 is reported to be essential for EGFR to maintain its functional conformation to allow EGF binding^35,36. The phosphorylation of EGFR-S1018 is important for regulating desensitization, internalization and degradation of EGFR³⁷. The association between EGFR-S1018 and EGFR-N352 suggests a potential regulation of EGFR-S1018 activity via glycosylation of EGFR-N352.

To further delineate the tumors based on the interactions among the intact glycopeptides and phosphopeptides, a multi-omics non-negative matrix factorization (NMF)-based clustering was conducted. Three Cross-Correlation clusters (CC 1–3) were derived and overlaid with the Glyco subtypes (Figure 4D, Table S5). Intriguingly, a high percentage of CC 1 tumors (n=34) overlapped with Glyco 3, where most of the CC 2 tumors (n=35) were annotated as Glyco 1 and CC 3 tumors (n=34) were associated with Glyco 2. Each CC cluster had its own unique features composed of both intact glycopeptides and phosphopeptides, which may reflect the tumor-intrinsic biological differences among the CC clusters (Figure S3C, Table S5). Indeed, CC 1 cluster showed associations with extracellular matrix organization and receptor tyrosine kinase-related signaling pathways. On the other hand, CC 2 cluster was identified as an immune-related cluster. CC 3 cluster was mapped to molecular transport-related pathways and negatively correlated with most of the pathways associated with CC 1 and CC 2 clusters.

We further examined each CC cluster features that were differentially expressed (>1.5-fold change and FDR<0.05) compared to the other CC clusters (Table S5). In this study, an increase in FLT1-N251 with Sialic glycan (N4H5S1) was detected in CC 1 relative to the other CC clusters (Figure 4E, Table S5). The glycosylation site, N251, is in the Ig-like domain 3 of FLT1 which is responsible for its ligand binding ability^38,39. Moreover, S373 of Alpha-enolase (ENO1), downstream of FLT1 in the HIF-1 signaling pathway, was upregulated in CC1 compared to CC 2 and CC 3 (Figure 4E). Thus, we speculated that the elevation in FLT1-N251 and ENO1-S373 expression might promote tumor vascularization and alter metabolic pathways under hypoxia condition driven by HIF-1⁴⁰.

Majority of the CC 2 tumors overlapped with Glyco 1 subtype, which was correlated with CD8+ inflamed immune subtype (Figure 3C). Many significantly upregulated features of CC 2 tumors came from proteins involved in immune-related activities. We found elevated glycosylation at N240 of Cytochrome b-245 heavy chain (CYBB) with high-mannose glycan (N2H8) and phosphorylation at S102 of DEAD box protein 3, X-chromosomal (DDX3X) (Figure 4F, Table S5). CYBB is expressed by myeloid cells predominately and produces reactive oxygen species for immune response to pathogens^41,42. CYBB can promote tumor growth and cancer metastasis which may serve as an immune checkpoint target⁴¹. Phosphorylation of S102 of DDX3X mediates the activation of type I interferon signaling pathway^43,44. Studies have shown the association between DDX3X and cancer cell proliferation^45,46. Although the effects of glycosylation on CYBB-N240 and phosphorylation on DDX3X-S102 are not fully understood yet, however, we found that the global expressions of CYBB and DDX3X were unchanged in CC 2 tumors. Therefore, CYBB-N240 and/or DDX3X-S102 could be potential therapeutic targets for CC 2 tumors.

Furthermore, CC 3 was mapped to molecular transport-related biological processes involving proteins such as Serotransferrin (TF) and Hemoglobin subunit alpha (HBA2). HBA2 is responsible for oxygen transport and TF is an iron binding transport protein. Significant upregulation of glycosylation at TF-N630 (N3H6S1) and phosphorylation at HBA2-T138 were observed in CC 3 tumors compared to the remaining tumors (Figure 4G, Table S5). TF supplies irons to immature red blood cells to make hemoglobin. The N630 is a major glycosylation site for TF that sialylation can play a role in promoting tumor growth as observed in cholangiocarcinoma^47,48. Little is known about the phosphorylation of HBA2; however, we found several phosphosites on HBA2 were upregulated in CC 3 tumors relative to others in addition to T138 (Table S5). A positive correlation between phosphorylation and the function of the protein was possible in response to the oxidative stress by cancer cells^49,50.

In summary, we investigated the possibility of crosstalk between glycosylation and phosphorylation in ccRCC. Various associations were observed between glycan types and phosphoproteins as well as between intact glycopeptides and phosphopeptides. Three CC clusters were derived in addition to the Glyco subtypes. Furthermore, distinct feature profiles among ccRCC tumors were observed (Figure 4H) suggesting the potential of developing new strategies to stratify the patients for treating ccRCC.

Discussion

In this study, 44,181 intact N-linked glycopeptides (corresponding to 1,429 glycoprotein) were identified and quantified from 103 ccRCC tumors and 80 paired NATs. The differences in significantly upregulated and downregulated intact glycopeptides were observed in glycan type distribution as well as the pathways associated with the corresponding glycoproteins in ccRCC tumors compared to NATs. We found that upregulated glycopeptides mostly originated from glycoproteins involved in ECM-receptor interaction, cell adhesion molecules, or focal adhesion, which correlate well with the previous studies showing the role of N-linked glycosylation in cell adhesion and proliferation^51,52. Moreover, changes in glycan types/patterns may associate with the cancer progression⁵³. In this particular study, a high percentage of the upregulated glycopeptides contained High-Man followed by sialylated and fucosylated glycans. Increase in high-mannose glycans can impact the protein functions and enhance cancer progression^28,32. The increased level of High-Man glycans indicated incomplete/premature termination of N-glycan biosynthesis^28,54, which was correlated to the downregulation of MAN1C1 in ccRCC as observed in our data. The suppression of MAN1C1 is associated with decreased cell apoptosis in ccRCC and contributes to tumor invasion and metastasis⁵⁵. Intriguingly, we observed that intact glycopeptides with High-Man glycans were highly expressed in high-grade tumors, while elevated levels of sialylated glycopeptides were observed in low-grade tumors.

Mutations on BAP1 and PBRM1 are two of the major genomic alteration events in ccRCC, while BAP1 mutation often leads to poor clinical outcome. The attached glycan structures of differentially expressed intact glycopeptides in BAP1-mutant tumors were different from PBRM1-mutant tumors. A high percentage of upregulated glycopeptides had High-Man glycans followed by Sialic glycans in BAP1-mutant tumors, whereas the expression levels of fucosylated glycopeptides were increased in PBRM1-mutant tumors. Since high abundance of High-Man glycans can promote metastasis of cancer by enhancing the translocation and invasion of cancer cells into surrounding microenvironment^28,32, which may explain the aggressiveness of BAP1-mutant tumors compared to PBRM1-mutant tumors. Additionally, High-Man glycan can be used for survival prediction, while insignificant difference in patient survival was observed for other glycan types in this particular cohort.

Significant inter-tumor heterogeneity was detected across ccRCC tumors using glycoproteome stratifying the tumors into three glycoproteomic subtypes. Each of the Glyco subtypes had a distinct profile that most of the features (e.g., BAP1 mutation, High-Man glycans, and high-grade tumors) in Glyco 1 likely contributed to the poor patient outcome of this particular subtype through the survival analysis. By further exploring the glycosylation along with phosphorylation suggested potential crosstalk between the two protein modifications in ccRCC. The correlation between glycans and phosphoproteins revealed that sialic glycans were associated with phosphoproteins in complement and coagulation cascades. This result matched well with previous studies that lower complement activity was observed when cell surface was covered with abundant sialic acid allowing the cancer cells to escape complement attack while enhancing the progression of cancer^33,34. On the other hand, high-mannose were correlated well with RNA transport. As aforementioned, High-Man glycans can promote the spread of cancer cells; thus, High-Man glycans may play a role in altering the regulation of RNA transport by increasing cancer cell proliferation.

Furthermore, cross-correlation among glycopeptides and phosphopeptides were derived using linear regression models indicating the possibility of regulating phospho-signaling pathways through N-linked glycosylation. For instance, association between N352 and S1018 on EGFR was noted in our cohort. Glycosylation of N352 is important for stabilizing EGFR in a conformation which allows EGF binding^35,36. Although future studies will be required to confirm our speculation, we hypothesized that the glycosylation of N352 will mediate the phosphorylation of S1018, which would affect the EGFR function since S1018 is responsible for the desensitization, internalization and degradation of EGFR³⁷. By utilizing multi-omics NMF-based clustering, we were able to further differentiate the ccRCC tumors into three cross-correlation clusters (CC1–3) inferring interactions between glycosylation and phosphorylation. Variations in glycosylation patterns and phosphorylation events contributed to different biological pathways that each CC cluster was associated, further suggesting the need for developing different diagnosis strategies and/or treatment options for ccRCC patients.

This study describes glycoproteomic characterization of ccRCC using a large-scale clinical cohort with the largest number of quantified glycopeptides that has ever been reported for ccRCC. Current front-line treatment for advanced ccRCC either targets VEGF/mTOR or uses immunotherapy⁵⁶. Our results showed specific glycan types in association with cancer progression and aggressiveness along with the glycoproteomic subtyping and crosstalk of glycosylation and phosphorylation may provide new approaches to stratify ccRCC patients, with the goal of developing diagnostic and/or treatment options in future.

In summary, this study demonstrates that glycoproteome can expand our perception of the ccRCC by revealing altered biological events that were not captured in our previous proteogenomic analysis⁷. Additionally, this study provides a rich resource to the ccRCC and broader scientific communities for translational research.

Limitations of the study

The objectives of this study were to comprehensively characterize ccRCC tumors and NATs using glycoproteomic approach as well as to understand the relation of glycoproteomic features to the changes observed on genomic, transcriptomic, proteomic, and phosphoproteomic levels. Tissues collected by the Clinical Proteomic Tumor Analysis Consortium (CPTAC) program were treatment-naive and surgically resected to fulfill the above purposes. Thus, there is a limitation of extrapolation to advanced/metastatic disease treated with systemic therapy since only treatment-naïve samples available in the present cohort. Furthermore, glycoproteomic data provide additional resources to investigate biological insights of ccRCC and allow investigation of glycoproteins that are located on cell surface, membrane, and secreted from cells. However, causal effects of correlated molecular alterations cannot be determined from this study and further validation using cell lines, patient-derived xenograft (PDX) models, or clinical trials may be required.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, T. Mamie Lih (tlih1@jhmi.edu)

Materials availability

This study did not generate new, unique reagents

Data and code availability

• Clinical data and raw data files can be accessed via the Proteomic Data Commons (PDC) at: https://pdc.cancer.gov/ (PDC Study ID: PDC000127, PDC000128, PDC000471). All other data (e.g., processed data) can also be found in the PDC or in the supplementary materials.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available in the STAR Methods or upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Tissue specimens

The initial tissue cohort contained 110 RCC tumors, which was analyzed via proteogenomic in our previous work⁷. Since the current study only focused on glycoproteomic analysis of ccRCC, therefore, a total of 103 histopathologically defined ccRCC cases with an age range of 30–90 were used in the downstream data analyses. This ccRCC cohort (n=103) contained 77 males and 26 females reflecting the gender distribution of ccRCC and the previous Clinical Proteomic Tumor Analysis Consortium (CPTAC) ccRCC cohort^7,61. All human specimens were existing specimens and subjects cannot be identified. Institutional review boards at each Tissue Source Site (TSS) reviewed protocols and consent documentation, in adherence to CPTAC guidelines. The summarized clinical data can be found in Supplementary Table 1. Detailed clinical data can be accessed and downloaded from the CPTAC Data Portal⁷.

METHOD DETAILS

Experimental Design

To conduct a comprehensive glycoproteomic analysis of ccRCC, the experimental workflow of generating global proteomic, phosphoproteomic, and glycoproteomic data of ccRCC for the current study are briefly described as following (Figure S1A): (i) proteins from each cryopulverized tissue specimen were extracted and trypsin digested into peptides; (ii) peptides from each sample were labeled with TMT-10plex and then fractionated by basic reversed-phase liquid chromatography (bRPLC); (iii) a portion of the TMT labeled peptides was used for global proteomic analysis and the remaining peptides were used for phosphopeptide enrichment via immobilized metal affinity chromatography (IMAC). The flow through from the IMAC enrichment was further utilized for intact glycopeptide enrichment via mixed anion exchange (MAX) solid phase extraction column; (iv) The enriched glycopeptides were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis in data-dependent acquisition (DDA) mode; (v) since IMAC enrichment can capture both phosphopeptides and intact glycopeptides⁵⁸, therefore, we used GPQuest^58–60 to identify and quantify the glycopeptides from both IMAC phospho-enriched data and MAX glyco-enriched data. A total of 23 TMT-10plex sets were acquired and analyzed together in this study. Detailed information of materials and sample preparation (tryptic digestion, TMT labeling, and fractionation) can be found in our previous published work⁷.

Sequential enrichment of intact glycopeptides

In this study, we utilized a sequential enrichment procedure to enrich intact glycopeptides after phosphopeptide enrichment as previously described⁶². In brief, flow through and washing solution from IMAC enrichment process as described in our previous study⁷ were collected together to perform intact glycopeptide enrichment using MAX cartridge¹⁵. Peptides were diluted with 80 % ACN in 0.1 % TFA and then incubated with Fe3+-NTA beads for 30 min at room temperature. The flow through from IMAC enrichment was collected by taking the supernatant of peptides-beads mixture after centrifuging for 1 min at 1000 × g. The pellet of peptides-beads was added into the C18 stage tip and then non-phosphopeptides were washed out by washing solution (80 % ACN in 0.1 % TFA). The “flow through” and “washing solution” were dried together and stored at −8°C before MAX enrichment.

To enrich glycopeptides, MAX cartridges were sequentially conditioned with 1mL of methanol (2 times), 1mL of ACN (2 times), 1mL of 100 mM triethylammonium acetate (3 times), 1 mL of water (Fisher chemical, LC/MS grade, 5 times), and 1 mL of 95% ACN (v/v) 1% TFA (v/v) (5 times). Samples were diluted with 1 mL of 95 % ACN (v/v) 1 % TFA (v/v) and loaded onto cartridge twice. The non-glycopeptides were eluted out by 95% ACN (v/v) 1% TFA (v/v) * 3 times. Finally, glycopeptides were eluted by 50 μL of 50% ACN (v/v) 0.1% TFA (v/v), dried and stored at −80°C prior to LC-MS/MS analysis.

LC-MS/MS analysis

Intact glycopeptides were analyzed by Orbitrap Fusion Lumos Tribrid^™ (Thermo Scientific) combined with Easy nLC 1200 UPLC system (Thermo Scientific). Samples were reconstituted with 3 % ACN in 0.1 % formic acid (solvent A) and loaded onto an in-house packed column (0.75 μm I.D. × 27.5 cm length packed with ReproSil-Pur 120 C18-AQ, 1.9 μm). Loaded peptides were subjected to the gradient with 200 μL/min of flow rate as follows: 2 to 6 % B (90 % ACN 0.1 % F.A) for 1 min, 6 to 30 % B for 84 min, 30 to 60 % B for 9 min, 60 to 90 % B for 1min, isocratic 90 % B for 5 min, 90 to 50 % B for 1 min, and isocratic 50 % B for 9 min. The temperature of the column was maintained by a column heater (Phoenix-ST) at 50°C. The ion source of the mass spectrometer was set up with 1.8 kV of electrospray voltage and 305°C of ion transfer tube temperature. The precursor ion scan was acquired with 60 K resolution at 200 m/z for from 500 to 2000 m/z range and AGC value was set as 5×10⁵. Precursor ions isolated from 0.7 m/z width were fragmented by HCD with 35 % NCE and fragment ions were acquired with 50 K resolution with 1×10⁵ of AGC value for 100 ms of injection time for a total duty cycle (2 sec). The peptide charge state screening was enabled to include 2 to 6+ ions with a dynamic exclusion time of 45 sec to discriminate against previously analyzed ions between +/− 10 ppm.

QUANTIFICATION AND STATISTICAL ANALYSIS

Identification and quantification of N-linked intact glycopeptides

The DDA raw files were first converted to mzML format via the ProteoWizard 3.0 with the Peak Picking option selected for all MS levels prior to the database searching against a customized N-linked glycopeptide database containing over 30,000 known glycosite-containing peptide sequences of human species⁶³ and a glycan database containing 253 glycan compositions (GlycomeDB, http://www.glycome-db.org/) via GPQuest^58–60 for the identification of intact glycopeptides. The theoretical b-/y-ions of each targeted/decoy N-linked glycopeptide in the customized database were calculated and used as fragment ion index during database search. MS2 spectra were preprocessed in a series of procedures including spectrum de-noising, oxonium ion evaluation, and glycan composition prediction⁶⁴. For each preprocessed potential MS2 spectrum of an intact glycopeptide, the top 100 peaks were matched to the fragment ion index of candidate peptides in the database. A candidate peptide was retained if at least 6 peaks were matched. Further evaluations were performed on the spectrum and the matched candidate peptides by assessing the isotopic peaks and fragment ions (b-/y-ions and Y-type ions) to compute Morpheus score, where the candidate peptide with the highest Morpheus score was assigned to that particular MS2 spectrum. The glycan composition of the intact glycopeptide was determined by first computing the mass difference between the peptide sequence and the precursor mass followed by finding the glycan matching that mass difference in the glycan database. PSMs with summed b-/y-ion intensity less than 10% of total intensity of the spectrum were filtered out. FDR at PSM level was set at <1% to ensure the precise identification of the intact glycopeptides. Other parameters used in GPQuest as follows: 10 ppm for MS1 tolerance, 20 ppm for MS2 tolerance, maximum 1 missed cleavage and peptide length of 6 to 30 amino acids were allowed, TMT-10plex (N-terminal and lysine) and carbamidomethylation on cysteine set as fixed modifications, and oxidation on methionine set as variable modification. The intact glycopeptides were quantified using the report ions of TMT-10plex. The missing values were imputed (only for intact glycopeptides quantified in >50% of the 207 samples across the 23 TMT sets, where 11 samples were excluded from downstream data analyses for the reasons as described in our previous publication⁷) using DreamAI (https://github.com/WangLab-MSSM/DreamAI), which was the tool used in our previous study for the imputation of the phosphoproteomic data. All the downstream analyses were conducted in R (version 4.0). The processed data can be downloaded from the PDC.

Quality control

The reproducibility of 23 TMT-10plexes was evaluated using eight QC and five NCI-7 cell samples. Spearman correlation calculation along with the histograms and scatter plots were conducted using psych (version 2.0.9). The CV was calculated using goeveg (version 0.4.2) and plotted using ggplot2 (version 3.3.5).

Differential analysis

The differential analysis was carried out by calculating the median log2 fold changes between (i) tumors and NATs, (ii) low-grade and high-grade tumors, (iii) mutant tumors and NATs, (iv) mutant tumors and tumors without the particular mutation (wildtype), (v) BAP1-mutant tumors and PBRM1-mutant tumors, (vi) one Glyco subtype and the remaining Glyco subtypes, and (vii) one Cross-Correlation (CC) cluster and the remaining CC clusters on intact glycopeptide level as well as on transcriptomic, global proteomic, or phosphoproteomic level when applicable. The p-values were computed using two-sided Wilcoxon rank sum test and adjusted (false discovery rate, FDR) using Benjamini-Hochberg method when applicable. Intact glycopeptides with >2-fold changes and FDR<0.01 were considered as significantly altered in tumors relative to NATs. Glycosylation enzymes (global proteomics and transcriptomics) with >1.5-fold changes and FDR<0.05 were considered as significantly altered in tumors compared to NATs. Intact glycopeptides with FDR<0.05 were used in high-grade and low-grade tumor comparison. Intact glycopeptides and/or global proteins with >1.5-fold changes and FDR<0.05 in mutant tumors relative to NATs were used in the comparison of mutant and wildtype. For comparisons (iv), (v), (vi), and (vii), differentially expressed intact glycopeptides were defined as >1.5-fold changes and FDR<0.05.

Glycoproteomic subtyping

Intact glycopeptides with CVs in the >25% quantile were utilized for subtyping analysis. By using CancerSubtypes⁶⁵ for consensus clustering of tumor subtypes, 80% of the original sample pool was randomly subsampled without replacement and partitioned into three major clusters (Glyco subtypes 1 to 3) via hierarchical clustering, which was repeated 2000 times. The subtyping heatmap was generated using ComplexHeatmap⁶⁶ (version 2.4.3). The consensus-clustered samples were overlaid with clinical features (e.g., grade, stage) and other omics subtypes derived in our previous publication (immune subtypes and global proteomic subtypes)⁷. The intact glycopeptides were grouped into four intact glycopeptide clusters (IPCs 1 to 4) using K-means clustering. Enrichment scores between immune-based/proteomic-based subtypes in relation to the glycoproteomic-based subtypes were calculated according to the hypergeometric distribution. A hypergeometric test p-value < 0.05 was required for the consideration of overlap between different omic subtypes.

Cross-correlation analysis

The cross-correlation analysis between glycan types and phosphoproteins was based on the Spearman correlation. A correlation >±0.3 was considered as positively/negatively correlated between a glycan type and a phosphoprotein. For those correlated phosphoproteins and glycan types, the corresponding phosphopeptides and intact glycopeptides were subjected to linear regression analysis using lm function in R to examine the relation between glycosylation and phosphorylation. The resulting p-values (t-statistic) were adjusted via Benjamini-Hochberg method. Only linear regression models with FDR <0.01 were utilized to construct the networks via Cytoscape⁶⁷ (version 3.9.1). The linear regression was applied to phosphorylation and glycosylation events occurring on the same proteins as well. For the linear regression analysis, only intact glycopeptides and phosphopeptides differentially expressed between tumors and NATs were considered. To further analyze the association between glycosylation and phosphorylation, we utilized non-negative matrix factorization (NMF)-based multi-omics clustering, which was conducted similar to our previous work on pancreatic ductal adenocarcinoma¹⁹. Briefly, NMF was used to perform unsupervised clustering of tumor samples using the abundances of phosphopeptides and intact glycopeptides. Only features (i.e., phosphopeptides and intact glycopeptides) with CVs in >50% quantile were used for subsequent analysis. The feature matrix was scaled and standardized, thus all features were represented as Z-scores. Since NMF requires a non-negative input matrix, the feature matrix was further converted as follows: (i) create one data matrix with all negative numbers zeroed, (ii) create another data matrix with all positive numbers zeroed and the signs of all negative numbers removed, and (iii) concatenate both matrices resulting in a data matrix with positive values and zeros only. The resulting matrix was then subjected to NMF analysis leveraging the NMF R-package⁶⁸. To determine the optimal factorization rank k (i.e., number of clusters), a range of k from 2 to 10 was tested using default settings with 50 iterations. The optimal factorization rank k=3 (i.e., CC cluster) was selected since the product of cophenetic correlation coefficient and dispersion coefficient of the consensus matrix was the maximum compared to other tested ks. The NMF analysis was repeated using 500 iterations for the optimal factorization rank k. A list of representative features for each CC cluster was derived⁶⁹ that differential analysis was carried out using cluster-specific features by comparing one CC cluster to the remaining CC clusters as described in the Differential Analysis section. Heatmap of CC clusters was generated using ComplexHeatmap⁶⁶ (version 2.4.3), where subgroup hierarchical clustering was performed for the cluster-specific features, separately for intact glycopeptides and phosphopeptides.

Pathway enrichment and Gene Ontology (GO) analysis

All the pathway enrichment (KEGG/Reactome) and GO were performed via WebGestalt⁷⁰. The p-values were calculated using hypergeometric test and adjusted (FDR) using Benjamini-Hochberg method. For intact glycopeptides/phosphopeptides, we used the corresponding proteins to identify enriched pathways, biological processes, or cellular components.

Survival analysis

The Kaplan-Meier survival analysis was carried out and log-rank p-value was calculated using the survival (version 3.2–13) R package. For the survival analysis of Glyco subtypes, the samples were differentiated based on their subtyping assignment. For the survival analysis of glycan type (High-Man), the samples were divided into two groups based on the median abundance across 103 cases, where the samples with High-Man abundance above the median were assigned to the High group and the remaining were considered as the Low group.

Supplementary Material

Figure S1_S2_S3

Figure S1. Experimental workflow and quality control of glycoproteomic data of ccRCC. (A) Schematic of the experimental workflow used to analyze proteomic, phosphoproteomic, and glycoproteomic of ccRCC. (B) Total identified intact glycopeptides in each of the TMT-10plex set. (C) Enriched GO cellular components (FDR<0.05) of glycoproteins with intact glycopeptides detected in ccRCC. The p-values were calculated using hypergeometric test and adjusted (FDR) using Benjamini-Hochberg method. (D) Reproducibility of the glycoproteomic data of ccRCC based on QC samples. (E) Coefficient of variation of intact glycopeptides in QC samples. (F) Reproducibility of the glycoproteomic data of ccRCC based on NCI-7 samples. (G) Coefficient of variation of intact glycopeptides in NCI-7 samples. Related to Table S1.

Figure S2. Comparative analysis to examine alter glycosylation in ccRCC. (A) Glycoproteins corresponding to the upregulated and downregulated intact glycopeptides. (B) KEGG pathways enriched (FDR<0.05) using uniquely observed glycoproteins of differentially expressed intact glycopeptides. The p-values were calculated using hypergeometric test and adjusted (FDR) using Benjamini-Hochberg method. (C) Number of glycoproteins mapped and associated glycan types for each enriched KEGG pathway. (D) Differential analysis of glycosylation enzymes on transcriptomic level. Significantly altered glycosylation enzymes were defined as >1.5-fold changes and FDR<0.05. (E) Glycosylation changes in VN¹⁵¹CSVPEEK of PECAM1 in high-grade tumors relative to low-grade tumors (FDR<0.05 as significant changes). (D and E) The p-values were computed using two-sided Wilcoxon rank sum test and adjusted (FDR) using Benjamini-Hochberg method. Related to Figure 1 and Table S2.

Figure S3. Cross-correlation between glycosylation and phosphorylation in ccRCC. (A) Networks associated with metabolic pathways. (B) Networks associated with RNA transport. (C) Pathways associated with each of the CC clusters. Related to Figure 4 and Table S5.

Table S1

Supplementary Table S1. Summarized clinical data and quality control of the data. Related to Figure S1.

Table S2

Supplementary Table S2. Differential analysis of tumors vs NATs and high-grade vs low grade tumors. Related to Figures 1 and S2.

Table S3

Supplementary Table S3. BAP1-mutant tumors and PBRM1-mutant tumors. Related to Figure 2.

Table S4

Supplementary Table S4. Glycoproteomic subtyping. Related to Figure 3.

Table S5

Supplementary Table S5. Cross-correlation between glycosylation and phosphorylation. Related to Figures 4 and S3.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological Samples
Primary tumor and normal adjacent tissue samples	See Experimental Model and Subject Details	N/A
Chemicals, Peptides, and Recombinant Proteins
Aprotinin	Sigma	Catalog: A6103
Leupeptin	Roche	Catalog: 11017101001
Phenylmethylsulfonyl fluoride	Sigma	Catalog:93482
Sodium fluoride	Sigma	Catalog: S7920
Phosphatase Inhibitor Cocktail 2	Sigma	Catalog: P5726
Phosphatase Inhibitor Cocktail 3	Sigma	Catalog: P0044
Urea	Sigma	Catalog: U0631
Tris(hydroxymethyl)aminomethane	Invitrogen	Catalog: AM9855G
Ethylenediaminetetraacetic acid	Sigma	Catalog: E7889
Sodium chloride	Santa Cruz Biotechnology	Catalog: sc-295833
PUGNAc	Sigma	Catalog: A7229
Dithiothretiol	ThermoFisher Scientific	Catalog: 20291
Iodoacetamide	ThermoFisher Scientific	Catalog: A3221
Sequencing grade modified trypsin	Promega	Catalog: V511X
Lysyl endopeptidase, Mass spectrometry grade	Wako Chemicals	Catalog: 125–05061
Formic acid	Fisher Chemical	Catalog: A117–50
Acetonitrile	Fisher Chemical	Catalog: A955–4
Reversed-phase C18 SepPak	Waters	Catalog: WAT054925
4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid	Alfa Aesar	Catalog: J63218
Tandem mass tags – 10plex	ThermoFisher Scientific	Catalog: 90110
Trifluoroacetic acid	Sigma	Catalog: 302031
Ammonium Hydroxide solution	Sigma	Catalog: 338818
Hydroxylamine solution	Aldrich	Catalog: 467804
Ni-NTA agarose beads	QIAGEN	Catalog: 30410
Iron (III) chloride	Sigma	Catalog:451649
Oasis MAX Cartridge	Waters	Catalog: 186000366
Triethylammonium acetate buffer	Sigma	Catalog: 90358
Critical Commercial Assays
BCA Protein Assay Kit	ThermoFisher Scientific	Catalog: 23225
Deposited data
CPTAC ccRCC clinical data and proteomic data	This manuscript; Clark et al.7	https://pdc.cancer.gov/ (PDC000127, PDC000128, PDC000471)
Software and Algorithms
Proteowizard	Chambers et al.57	http://proteowizard.sourceforge.net/
GPQuest	Hu et al.58; Toghi Eshghi et al.59; Toghi Eshghi et al.60	https://github.com/huizhanglab-jhu/GPQuest
R v4.0	R Development Core Team	https://www.R-project.org