An integrative pan-cancer analysis of the molecular and biological features of glycosyltransferases

Abstract

Dear Editor, Glycosyltransferases (GTs) played important roles in cancer development and progression.1,2 Here, we conducted a pan-cancer analysis of GTs (Supporting information Table S1)3 based on the TCGA data, CCLE data, single-cell RNA sequencing datasets and our proteogenomic resource, aiming to characterize the molecular features, biological functions and clinical implications of GTs across cancer types. SEE PDF] Widespread gene expression changes of GTs in tumors were observed (Figure 2A, Supporting information Figures S4, S5A and Table S4), among them, three GTs displayed consistent expression alterations in 16 cancer types, including upregulation of ALG3, and downregulation of B3GALT2 and ST6GALNAC3 (Figure 2B). According to our proteogenomic resource of LIHC (CHCC-HBV),10 GTs that contributed to different prognosis of patients were further analyzed.

Dear Editor,

Glycosyltransferases (GTs) played important roles in cancer development and progression.^1,2 Here, we conducted a pan-cancer analysis of GTs (Supporting information Table S1)³ based on the TCGA data, CCLE data, single-cell RNA sequencing datasets and our proteogenomic resource, aiming to characterize the molecular features, biological functions and clinical implications of GTs across cancer types.

The overall mutation frequency of GTs was relatively low (0.0–11.6%). Cancer types with higher global mutation burdens exhibited higher mutation frequencies of GTs. The highest mutation frequencies were observed in UCEC (ALG13, 11.6%), SKCM (FUT9, 10.6%) and SKCM (GALNT13, 10.6%) (Figure 1A, Supporting information Figure S1A and Table S2). Survival analysis revealed that the UGGT2 mutation in COAD was linked to worse clinical outcomes, while the ALG13 mutation in UCEC was associated with better survival (Figure 1B). Furthermore, COAD patients with UGGT2 mutation showed enrichment of chronic inflammatory response, while UCEC patients with ALG13 mutation showed downregulation of response to cAMP (Figure 1C). Analysis of CCLE drug sensitivity showed colon cancer cell lines with UGGT2 mutation were resistant to EGFR inhibitors (Erlotinib and Lapatinib), and endometrial cancer cell lines with ALG13 mutation were sensitive to Panobinostat and Sorafenib (Figure 1D and Supporting information Figure S1B). It was worth noting that ALG1/2/11/14 were essential in cell survival across various cancer cell lines (Figure 1E). Widespread copy-number variations (CNVs) of GTs were found across cancer types (Figure 1F, Supporting information Figure S2 and Table S3). In addition, mutation status and CNVs of GTs in cancer cell lines of CCLE displayed a similar pattern to the TCGA pan-cancer cohort (Supporting information Figure S3).

[IMAGE OMITTED. SEE PDF]

Widespread gene expression changes of GTs in tumors were observed (Figure 2A, Supporting information Figures S4, S5A and Table S4), among them, three GTs displayed consistent expression alterations in 16 cancer types, including upregulation of ALG3, and downregulation of B3GALT2 and ST6GALNAC3 (Figure 2B). Functional analyses of these three GTs showed a strong similarity in biological functions across cancer types (Figure 2C). In addition, the expression of GTs was tightly associated with patients' prognosis (Figure 2A, Supporting information Figure S5B and Table S5). For example, decreased expression of GYS2 conveyed poor prognosis in LIHC (Figure 2D), which was consistent with previous findings that GYS2 could inhibit tumor growth via a negative feedback loop with p53.⁴ In LUAD, B3GNT3 and GALNT14 were aberrantly expressed and associated with overall survival in different LUAD cohorts (Supporting information Figure S6). The prognostic significance of GTs was also evaluated in two external cohorts of patients receiving immune checkpoint inhibitors,^5,6 and higher expression of B3GNT4 indicated worse clinical outcomes in both cohorts (Supporting information Figure S7).

[IMAGE OMITTED. SEE PDF]

The pan-cancer GT-pathway interaction (Figure 2E to G and Supporting information Table S6) and GT-protein interaction networks (Supporting information Figures S8 to S10 and Table S7) were constructed, respectively. Similar functions were enriched, such as immune response and signal transduction, suggesting the cross-talk between GTs and interacting proteins synergistically contributed to the biological alterations in cancer. In the interaction network, FBXO6 was the most common interacted protein, especially associated with KDELC2 and higher expression of KDELC2 and FBXO6 collectively contributed to the poor prognosis in LGG (Supporting information Figure S9). Further correlation analysis between GTs and tumor microenvironment (TME) was performed (Figure 3A). MFNG was significantly positively related to activated CD8⁺ T cells compared with other GTs, especially in LUAD and SKCM (Figure 3B and C), and higher expression of MFNG indeed conveyed better prognosis in LUAD and SKCM (Figure 3D). For melanoma patients treated with PD-1 blockade,⁶ elevated MFNG also indicated a satisfactory prognosis (Figure 3E). Furthermore, MFNG was found to be mainly expressed in CX3CR1⁺ cytotoxic T cells based on two single-cell RNA sequencing datasets^7,8 (Figure 3F and G). Considering previous studies,⁹ our findings of the expression of MFNG in CX3CR1⁺ cytotoxic T cells suggested that MFNG may be critically important in maintaining the function of this subset of CD8⁺ T cells. In addition to LUAD and SKCM, GTs correlating with the prognosis of patients in CD8⁺ T cell-enriched tumors were observed in other 26 types of cancers (Supporting information Table S8).

[IMAGE OMITTED. SEE PDF]

A scoring tool (GTscore) was established using the expression of GTs and this score could reflect the tumor proliferation-related activities, and predict the prognosis and treatment benefits of patients receiving immunotherapy (Supporting information Figure S11A and Table S9). In 16 cancer types, GTscore was associated with the prognosis of patients (Supporting information Figure S11B), and for immunotherapy cohort, patients with high GTscore displayed poorer prognosis and therapeutic disadvantages (Supporting information Figure S11C and D). High GTscore group showed higher levels of proliferation-related activities, such as angiogenesis, EMT and hypoxia (Supporting information Figure S12).

The proliferation subgroup was an attractive clustering part of LIHC in our previous study.¹⁰ Here, GTs were found to be significantly correlated with proliferation-related activities in LIHC (Figure 4A). Unsupervised consensus clustering based on the expression profiling of GTs could identify two clusters of LIHC patients, which showed different clinical outcomes and TME features (Figure 4B to D and Supporting information Figures S13 and S14). According to our proteogenomic resource of LIHC (CHCC-HBV),¹⁰ GTs that contributed to different prognosis of patients were further analyzed. Among them, three GTs (GALNT4, MGAT5 and UGGT2) displayed prognosis relevance at protein level (Figure 4E and F). In addition, we found that the three GTs were highly expressed in the proliferation subtype (Figure 4G), therefore, according to this observation, these three GTs were chosen for further Tissue microarray (TMA) validation and cell-based assays. The TMA comprising 154 cases showed patients with high MGAT5 or UGGT2 expression, indeed had shorter overall survival than patients with low expression (Figure 4H and I). Further analysis on the interacting proteins revealed that the expression of MGAT5 was correlated with ISLR, and the expression of UGGT2 was correlated with APP (Supporting information Figure S15). Transwell and CCK-8 assays confirmed that downregulation of GALNT4, MGAT5 or UGGT2 could inhibit the migration and proliferation of LIHC cell lines (Figure 4J and K and Supporting information Figure S16).

[IMAGE OMITTED. SEE PDF]

Further validation of the biological implications of aberrantly expressed GTs, discovery of the common substrate of GTs and deciphering the site-specific function of this substrate are necessary, and would provide vital clues for the diagnosis or treatment of cancers via targeting specific glycosylation.

ACKNOWLEDGEMENTS

We would like to thank the colleagues from our research group for their assistance.

FUNDING INFORMATION

The work was supported by Shanghai Pujiang Program (2020PJD012), National Natural Science Foundation of China (82150111, 91859105 and 81961128025) and the Science and Technology Commission of Shanghai Municipality (20JC1418900).

CONFLICT OF INTEREST

The authors declare that they have no competing interest.

References

Schjoldager KT, Narimatsu Y, Joshi HJ, Clausen H. Global view of human protein glycosylation pathways and functions. Nat Rev Mol Cell Biol. 2020; 21 (12): 729-749.

Li CW, Lim SO, Chung EM, et al. Eradication of triple-negative breast cancer cells by targeting glycosylated PD-L1. Cancer Cell. 2018; 33 (2): 187-201.e10.

Narimatsu Y, Joshi HJ, Yang Z, et al. A validated gRNA library for CRISPR/Cas9 targeting of the human glycosyltransferase genome. Glycobiology. 2018; 28 (5): 295-305.

Chen SL, Zhang CZ, Liu LL, et al. A GYS2/p53 negative feedback loop restricts tumor growth in HBV-related hepatocellular carcinoma. Cancer Res. 2019; 79 (3): 534-545.

Mariathasan S, Turley SJ, Nickles D, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature. 2018; 554 (7693): 544-548.

Liu D, Schilling B, Liu D, et al. Integrative molecular and clinical modeling of clinical outcomes to PD1 blockade in patients with metastatic melanoma. Nat Med. 2019; 25 (12): 1916-1927.

Jerby-Arnon L, Shah P, Cuoco MS, et al. A cancer cell program promotes T-cell exclusion and resistance to checkpoint blockade. Cell. 2018; 175 (4): 984-997.

Kim N, Kim HK, Lee K, et al. Single-cell RNA sequencing demonstrates the molecular and cellular reprogramming of metastatic lung adenocarcinoma. Nat Commun. 2020; 11 (1): 2285.

Song Y, Kumar V, Wei HX, Qiu J, Stanley P. Lunatic, manic, and radical fringe each promote T and B cell development. J Immunol. 2016; 196 (1): 232-243.

Gao Q, Zhu H, Dong L, et al. Integrated proteogenomic characterization of HBV-related hepatocellular carcinoma. Cell. 2019; 179 (2): 561-577.