Aberrations in the glycosylation of receptor tyrosine kinases: A focus on lung adenocarcinoma

EGFR

EGFR, encoded by ERBB1, is a critical RTK involved in the regulation of cell growth and survival. EGFR mutations are found in a significant proportion of NSCLC cases, with frequencies varying by region and population. EGFR mutations are more prevalent in females, Asian populations, and non-smokers, while smoking history inversely correlates with mutation frequency.^[28-30] EGFR is structurally composed of three key domains: An extracellular domain (ECD) for ligand binding, a transmembrane domain for membrane anchoring, and an intracellular cytoplasmic domain containing catalytic tyrosine kinase activity.^[31] In its inactive state, EGFR primarily exists as an auto-inhibited monomer. However, upon ligand binding, EGFR changes to a conformation that enables it to form either homodimers or heterodimers with HER-2 or other receptors. Receptor activation is possible upon the establishment of an asymmetric dimer configuration of the kinase domains, wherein one domain allosterically potentiates the activity of the other.^[32] The EGFR ECD, which has a complex, heart-shaped structure, comprises 4 subdomains. Domain I (L1), with a leucine-rich repeat structure, is involved in ligand binding, as is Domain III (L2), which also has a leucine-rich repeat structure. Domain II (CR1) and Domain IV (CR2), which are cysteine-rich domains, contribute to the dimerization of the receptor. The entire ECD adopts a tethered conformation in the absence of a ligand, preventing spontaneous activation by keeping L1 and L2 in a compact state.^[31,32] The intracellular domain of EGFR comprises a juxtamembrane region involved in tyrosine kinase domain regulation and function. It is a bilobal structure encompassing the ATP binding site (in the N-terminal lobe), the activation loop, and a substrate binding site (both in the C-terminal lobe). A C-terminal tail, rich in tyrosine residues, undergoes autophosphorylation and serves as a platform for interaction with downstream signaling proteins.^[31,32] Oncogenic EGFR mutants induced by amplification of activating mutations exhibit functional characteristics similar to those of the ligand-activated wild-type receptor. Tumors with EGFR mutations often present with specific imaging features such as ground-glass opacity and histological patterns such as lepidic growth. These features can help predict genetic changes and guide treatment decisions.^[33,34] The most prominent EGFR lesions involve the L858R point mutation in exon 21 and deletions within exon 19, often grouped as classical mutations that comprise 85-90% of EGFR kinase domain mutations.^[35] In general, EGFR mutations and extensive rearrangements observed across solid cancer entities frequently cause dysregulation of receptor endocytosis, which, in turn, enhances the receptor’s signaling potential.^[36]

EGFR is a heavily glycosylated protein and carbohydrate chains constitute approximately 25% of the 180 kDa mass of the receptor. This post-translational modification plays a fundamental role in various receptor processes, such as ligand-independent activation, growth factor binding affinity, and dimerization. Moreover, significant alterations in glycosylation are observed in various cancers, including NSCLC.^[31] N-glycosylation is crucial for the proper folding, stability, and function of EGFR. Disruption of this process can lead to altered receptor activity and localization, which may enhance cancer cell proliferation and survival [Figure 2]. The absence of N-glycosylation can lead to increased dimerization of EGFR, even in the absence of ligands, which promotes spontaneous activation of the receptor. This can result in the activation of downstream signaling pathways that drive cancer cell proliferation and survival. EGFR-targeted therapies such as erlotinib can sensitize cancer cells, suggesting that targeting N-glycosylation could be a strategy to overcome drug resistance in EGFR-mutant cancers. Disruption of N-glycosylation in cancer cells can lead to uncontrolled receptor activation by promoting spontaneous oligomerization, ultimately contributing to increased proliferation, survival, and drug resistance.^[37,38] It may also impair receptor maturation by altering subcellular localization.^[39]

Close

Figure 2:

Epidermal growth factor receptor (EGFR) signaling upon disrupted N-glycosylation. EGFR dimerization and downstream signaling occur upon proper glycosylation of EGFR and ligand binding. This leads to EGFR pathway activation through Rat sarcoma (Ras), V-Raf Murine Sarcoma Viral Oncogene Homolog B (B-RAF), mitogen-activated extracellular signal-regulated kinase 1/2 (MEK1/2), and extracellular signal-related kinase 1/2 (ERK1/2) activation resulting in cell proliferation, growth, and cell survival. Disruption of EGFR N-glycosylation enables ligand-independent dimerization and activation. P: Phospho; GTP: Guanosine triphosphate. Created in BioRender. Meder, L. (2025) https://BioRender.com/8rqpbva.

Export to PPT

The ECD of EGFR is heavily modified post-translationally by the addition of sugar, which influences receptor activity.^[40] GlyGen, a web-based platform that integrates glyco-conjugate-related information from publicly available databases, shows 17 sites for N-glycosylation and 6 sites for O-linked glycans, but not all have been verified to be related to cancer.^[41]

Glycosylation of the EGFR ECD occurs at Asn residues that fit the NXS/T consensus motif, including an atypical site at Asn32, which is located within a unique Asn-Asn-Cys sequence.^[40,42,43] This atypical site has been shown to undergo both N-glycosylation and O-fucosylation. Experimental and computational analyses have demonstrated that glycosylation is essential for maintaining both the structural integrity and stability of the protein.^[40] Through recent molecular dynamic simulations, the study of a complete monomeric glycosylated EGFR structure situated in a membrane bilayer has highlighted the critical role of N-glycosylation in enabling the EGFR ECD to adequately interact with the membrane, a prerequisite for maintaining its proper structural conformation.^[42] Specific glycosylation sites, such as N361 in Domain III and N579 in Domain IV, are critical for EGFR dimerization. Mutations at N361A increase dimerization and alter ligand sensitivity, impacting cell viability and response to inhibitors.^[44] Glycosylation at N579 affects the receptor’s conformation, influencing its dimerization state. The N579Q mutation, which prevents glycosylation, results in a higher fraction of receptors in a high-affinity state and alters phosphorylation patterns.^[45] In total, 4 glycosylation sites are located in Domain III, where the monoclonal antibody cetuximab (an anti-EGFR targeted therapy) binds to its target epitope.^[46] In addition, inhibition of N-glycosylation using OST inhibitors can overcome resistance to EGFR TKIs by disrupting EGFR signaling and reducing tumor cell viability. For instance, NGI-1, which is an OST inhibitor, was able to restore sensitivity to osimertinib, a first-generation TKI, by inducing G1 arrest and apoptosis in resistant NSCLC cells.^[47] To date, O-glycosylation in EGFR signaling has not been studied in lung cancer so far. However, C1GALT1, which is responsible for elongation of GalNAc-type O-glycosylation, and N-acetylgalactosaminyltransferase 2 (GALNT2), which regulates the early steps of mucin O-glycosylation, have been shown to enhance oncogenic EGFR signaling in brain cancers and oral squamous cell carcinomas. Both glycosyltransferases trigger EGFR terminal O-glycosylation and promote EGFR phosphorylation and downstream signaling.^[48,49]

In general, since EGFR-mutated cancer cells frequently develop resistance to EGFR blockade, attenuating the glycosylation of EGFR has emerged as a promising therapeutic approach for NSCLC. It may counteract dimerization and activation of overexpressed EGFR or can enable mAb binding. Therefore, exploring the inhibition of glycosylation as a complementary approach holds potential as a strategy to enhance therapeutic outcomes and possibly address drug resistance.

HER-2

HER-2 is an extensively studied member of the EGFR family of receptors in cancer research and is encoded by the ERBB2 gene. It is a transmembrane RTK that, when overexpressed or amplified, contributes to aggressive tumor behavior and poor prognosis.^[50] HER2 mutations are identified in approximately 2% of NSCLC, with a notable prevalence in lung adenocarcinomas. These mutations are often found in specific subgroups such as females, non-smokers, and those with adenocarcinoma histology.^[51]

HER-2 is a type I transmembrane growth factor RTK and exhibits significant structural similarity to other EGFR family receptors.^[52] This class of receptors comprises an extracellular N-terminal domain that serves as the ligand-binding site, a transmembrane domain, and an intracellular region that includes a juxtamembrane domain, a kinase catalytic domain, and a carboxy-terminal domain. In contrast to other receptors, HER-2 lacks confirmed ligands for binding; its activation occurs through heterodimerization with other EGFR family receptors bound to their respective ligands or through homodimerization.

Amplification of ERBB2 and overexpression of HER-2 are well-established characteristics in breast, gastric, and lung cancers, serving as predictive markers for therapeutic approaches aimed at HER-2.^[53] The most frequently observed and extensively studied alterations of HER-2/ERBB2 in tumor cells include protein overexpression as well as gene mutations and amplifications.^[52-55] ERBB2 mutations are independent of amplification, indicating that they represent a separate biological entity and a potential therapeutic target.^[55,56] ERBB2 overexpression or gene mutations result in an elevated presence of HER-2 on the cell surface and enhanced activation. This activation drives the uncontrolled proliferation of tumor cells.^[50,53]

The percentage of NSCLC patients with ERBB2 mutations is approximately 1-4%, while ERBB2 gene amplification occurs in 2-5% of cases and protein overexpression in 2-30% of cases. This overexpression can be induced by cytotoxic drugs through downregulation of microRNAs miR-125a and miR-125b which normally suppress HER-2 expression. The majority of patients harboring ERBB2 mutations have a history of never smoking and present with adenocarcinoma histology.^[57] The most frequent ERBB2 mutations are in-frame insertions within exon 20 (ex20ins), notably the A775_ G776insYVMA insertion/duplication. These mutations are typically found to be mutually exclusive with EGFR mutations and ALK rearrangements.^[58]

The most common treatments for cancer patients with ERBB2 mutations and/or HER-2 overexpression involve a range of targeted therapies, including pan-HER-2 inhibitors (such as afatinib and neratinib) and targeted mAbs and antibody-drug conjugates such as trastuzumab and trastuzumab emtansine (T-DM1), respectively.^[59] Trastuzumab, often used in combination with chemotherapy, has been utilized in treating HER-2-positive NSCLC. However, its efficacy as a monotherapy is limited, with a progression-free survival of about 5.1 months.^[57,60] T-DM1 has shown superior activity against HER-2-expressing tumors compared to unconjugated trastuzumab, with improved efficacy and safety.^[61]

There are studies suggesting that post-translational modifications of the HER-2 protein, rather than its overexpression, have a pivotal role in regulating endocytosis and determining the effectiveness of anti-HER-2 therapies in lung cancer cell lines and patient-derived xenograft models.^[60] A recent study that utilized mass spectrometry to create a comprehensive glycosylation profile of HER-2 showed site-specific modifications and their impact on receptor stability and activity. HER-2 has 8 identified N-glycosylation sites in its ectodomain that can interfere with trastuzumab binding to its target epitope.^[62] GlyGen references 6 sites for N-glycosylation: Asn124, Asn187, Asn259, Asn530, Asn571, and Asn629 and two sites for O-glycosylation: Thr127 and Ser998.^[41] The effectiveness of trastuzumab in treating HER-2-positive gastric cancer is significantly challenged by molecular resistance, which is influenced by extensive N-glycosylation of the HER-2 receptor. Through mass spectrometry analysis, it was found that galactose α-2,6-sialyltransferase 1 (ST6Gal1) modifies specific N-glycosylation sites within the trastuzumab-binding domain of HER-2, affecting receptor stability and therapeutic response. Inhibiting ST6Gal1 expression altered glycosylation patterns, increased HER-2 stability, and enhanced sensitivity to trastuzumab by reducing HER-2 and EGFR activation.^[63] Another way HER-2 targeting was used in the context of glycosylation, was coupling trastuzumab with a sialidase. The targeted sialidase activity toward breast cancer cells using antibody-enzyme conjugates led to desialylation of the cancer cell surface accompanied by increased NK cell killing by antibody-dependent, cell-mediated cytotoxicity.^[64]

Taken together, these findings suggest a potential new area of investigation for HER-2 and other RTK modifications in tumor cells, focusing on the role of surface receptor glycosylation across various cancer types. Altered glycosylation patterns of HER-2, including modifications in the trastuzumab-binding domain, can significantly influence receptor stability and activity as well as impact the efficacy of targeted therapies. In addition, directing de-glycosylating enzymes to RTKs might indicate a novel approach to improve immunotherapeutic approaches in cancer treatment in the future, including lung cancer treatment.

c-MET

c-MET mutations occur in approximately 3-5% of NSCLC cases, with a higher prevalence in adenocarcinoma and a notable enrichment in the sarcomatoid subtype. In addition, de novo c-MET amplifications are observed in 1-5% of NSCLC, primarily affecting adenocarcinoma.^[65] c-MET is an RTK, and its active form is a disulfide-linked heterodimer, which includes an extracellular α-subunit and a single-pass transmembrane β-subunit. The extracellular portion of c-MET features a large N-terminal semaphorin (SEMA) domain, an integrin PSI domain, along with transcription factor IPT domains. The intracellular portion of c-MET consists of three main segments: a juxtamembrane region, a tyrosine kinase domain, and a C-terminal docking site.^[66] c-MET activation occurs when it forms a homodimer in response to binding with a ligand. This binding involves high-affinity sites located within the IPT3 and IPT4 domains, as well as low-affinity sites within the SEMA domain.^[67] Hepatocyte growth factor (HGF) is the only known true endogenous ligand for c-MET.^[68] Human c-MET contains several reported glycosylation sites: 10 N-linked and three O-linked sites, according to GlyGen.^[41]

Some findings suggest that the N-glycans of the SEMA domain positively regulate HGF signaling, whereas the N-glycans of the other regions negatively regulate signaling.^[69] Specific N-glycans on c-MET regulate its proteolytic processing and HGF-induced signaling in a site-specific manner. Deletion of these N-glycans affects c-MET’s status and function. c-MET is initially synthesized as a partially glycosylated single-chain precursor (pro-MET) in the ER. Pro-MET undergoes disulfide bond formation, post-translational glycosylation, and endoproteolytic cleavage to produce the mature heterodimeric form.^[70] During receptor homodimerization and autophosphorylation, the activation of several downstream signaling pathways occurs, including the phosphoinositide 3-kinase/protein kinase B (PI3K/AKT) and RAS/extracellular signal-regulated kinase (RAS/ERK) pathways. These signaling pathways drive various cellular processes, such as cell proliferation, migration, differentiation, and morphogenesis.^[69]

Dysregulation of c-MET, resulting in aberrant signaling, can occur through various mechanisms such as gene amplification, overexpression, activating mutations, increased autocrine or paracrine ligand stimulation, HGF overexpression, formation of abnormal autocrine signaling, and interaction with other active cell-surface receptors.^[71] Elevated c-MET expression has been observed in multiple cancers, including lung, breast, ovarian, kidney, colon, thyroid, liver, and gastric carcinomas.^[68] Approximately 3-5% of NSCLC patients have c-MET dysregulation, including splice site mutations in METex14 that have been reported in about 4% of NSCLC patients.^[68,71,72] Recent advancements have led to the development and approval of several c-MET TKIs, such as capmatinib and tepotinib.^[73-75] Drug resistance is an unavoidable challenge in targeted therapies. The presence of MET amplification can lead to short-lived and heterogeneous responses.^[76] MET amplification is a well-established mechanism of acquired resistance to EGFR TKIs in NSCLC patients. It often occurs alongside secondary EGFR mutations, contributing to the complexity of resistance.^[77]

Since both transcriptional and post-translational mechanisms control the expression and activity of c-MET, NSCLC treatment may benefit from a combination of c-MET TKIs with drugs that influence c-MET maturation in cancer cells. C1GALT1 controls the formation of mucintype O-glycans, including O-glycosylation of c-MET, thereby enhancing HGF-induced dimerization and activation.^[78] GALNT2, an enzyme involved in the first step of mucin-type O-glycosylation, inhibits malignant characteristics in gastric adenocarcinoma by modulating the activity of the c-MET receptor. Knocking down GALNT2 significantly increased c-MET phosphorylation, especially in response to HGF stimulation, and reduced lectin binding to c-MET O-glycans, indicating that GALNT2 plays a role in modifying c-MET O-glycans in gastric cancer cells.^[79]

Taken together, specific N-glycans in the SEMA domain enhance HGF signaling, while those outside of this domain suppress it. O-glycosylation, particularly mediated by C1GALT1 and GALNT2, influences c-MET dimerization, activation, and downstream signaling pathways. These modifications suggest that targeting glycosylation could become a novel therapeutic option for NSCLC. Combining c-MET inhibitors with drugs targeting c-MET maturation, such as glycosylation modulators, could enhance treatment efficacy and might improve outcomes for NSCLC patients with c-MET-driven tumors.

RET

RET rearrangements occur in approximately 1.2-2.2% of NSCLC cases, with a slightly higher prevalence in adenocarcinomas.^[80,81] RET is distinguished from other RTKs by its extracellular region. This domain is characterized by 4 cadherin-like structures and a segment of 120 amino acids containing 16 cysteine residues.^[82] It is activated by glial cell line-derived neurotrophic factor (GDNF) ligands binding to GDNF family receptor alphas, leading to RET dimerization and activation. This activation triggers downstream signaling pathways such as PI3K/AKT, mitogen-activated protein kinase, and phospholipase C gamma (PLCγ), promoting cell proliferation, growth, and survival.^[82,83] Overexpression and rearrangements of RET have been identified as potential oncogenic drivers, leading to increased interest in targeted therapies. RET rearrangement is observed in approximately 1-2% of lung adenocarcinomas, with reports indicating that activating EGFR mutations can also co-exist.^[84,85] RET is also known for its fusions, such as kinesin family member 5B (KIF5B)-RET, which are found in about 1.7% of lung adenocarcinomas and often associated with younger, non-smoking patients and poorly differentiated tumors.^[86] Targeted therapies, such as RET TKIs selpercatinib and pralsetinib, show antitumor activity in lung tumors with high RET expression. Vandetanib, a multikinase inhibitor, may overcome resistance to EGFR-TKIs in RET-overexpressing lung adenocarcinomas, suggesting its potential use in combination therapies.^[87]

At present, there is a lack of studies specifically addressing the glycosylation of RET in cancer and the potential functional consequences. However, GlyGen identifies 4 reported N-glycosylation sites of RET: Asn343, Asn554, Asn763, and Asn975.^[41]

ALK and ROS1

ALK rearrangements occur in approximately 3.8-6.3% of NSCLC cases, with a higher prevalence in adenocarcinomas. ALK-positive NSCLC is more common in younger patients, women, and never-smokers. The mean age difference between ALK-positive and -negative patients is significant, with ALK-positive patients being younger by about 7 years.^[88] ALK is a receptor transmembrane protein tyrosine kinase and is involved in various signaling pathways that regulate cell growth, survival, and differentiation by its activation through binding of the extracellular ligands ALKAL1 and ALKAL2.^[89,90] Like many RTKs, ALK is structured with three distinct regions: an extra-cellular domain that binds to ligands, a single-transmembrane segment that anchors the protein, and an intracellular cytoplasmic tyrosine kinase domain critical in the development of diseases.^[91] ROS1 is also an RTK that has been implicated in various cancers, but its exact function in healthy cells remains unknown. ROS1 and ALK share significant structural similarities, particularly in their kinase domains, which have 49% amino acid sequence homology. This substantial overlap contributes to their functional similarities.^[92,93]

ALK fusions, generated by chromosomal rearrangements, can signal continuously without the normal regulatory controls. The first alteration of the ALK gene identified in human cancers was the nucleophosmin-1 (NPM1)-ALK fusion gene.^[89] In the presence of NPM1-ALK fusions, a number of interconnected pathways are activated, such as RAS/ERK, janus kinase 3 - signal transducer and activator of transcription 3, PLCγ, and PI3K/AKT, which results in cellular proliferation, survival, and phenotypic changes.^[91] Echinoderm microtubule-associated protein-like 4 (EML4)-ALK is another well-studied fusion gene that is considered to be one of the key pathogenic drivers identified in NSCLC and the most common variant in NSCLC.^[94] Beyond EML4, less common yet important ALK fusion partners in NSCLC include KIF5B, a motor protein for intracellular transport, TFG, involved in membrane trafficking, KLC1, another motor protein, and STRN, a scaffolding protein, thus demonstrating a diversity of involved proteins.^[89] ALK fusions occur in approximately 3-7% of lung cancers.^[95] EML4-ALK fusion variants arise from variable breakpoints during chromosomal rearrangements, leading to fusion proteins of differing sizes; the ALK gene frequently breaks at exon 20 and less commonly at exon 19, while the EML4 breakpoint is more variable, thus generating various fusion proteins with Variants 1 and 3 as the most common.^[96]

ROS1 rearrangements are found in a small subset of NSCLCs, with prevalence rates ranging from 0.5% to 2.6% in various studies.^[97] ROS1 also forms fusions, which are significant oncogenic drivers in various cancers, including NSCLC, glioblastoma, and others. The fusion proteins resulting from these genetic rearrangements lead to constitutive activation of the ROS1 kinase, promoting cancer cell proliferation and survival.^[98] There are at least 26 known genes that can fuse with ROS1, including cluster of differentiation 74, syndecan 4, and solute carrier family 34 member 2, among others. ROS1 fusions in NSCLC are more common in younger, non-smoking patients and these fusions are mutually exclusive with other common mutations like EGFR and Kirsten rat sarcoma viral oncogene homolog.^[98]

ROS1 and ALK have a high degree of similarity at the ATP binding site, with about 77% identity. This similarity is crucial because it allows certain inhibitors, such as crizotinib, to target both kinases effectively.^[93] Both ROS1 and ALK can develop resistance to inhibitors through similar mechanisms, such as mutations in the kinase domain. For example, ROS1 S1986Y/F and ALK C1156Y mutations are homologous and exhibit similar resistance patterns to certain TKIs. Kinase domain mutations are a major mechanism of acquired resistance in both ROS1- and ALK-positive NSCLC.^[99]

Inhibition of N-glycosylation impairs ALK phosphorylation and disrupts downstream pro-survival signaling pathways such as AKT, ERK1/2, and STAT3 in neuroblastoma cell lines. This suggests that glycosylation is essential for ALK’s role in promoting cell survival and proliferation.^[100] However, no reported glycosylation sites in ALK have been identified thus far. GlyGen lists only 18 predicted sites with 16 N- and two O-glycosylation sites.^[41] For ROS1, GlyGen shows three reported O-glycosylation sites in close proximity, namely Ser1570, Ser1577, and Ser1581, suggesting a functional relevance.^[41] Collectively, glycosylation in the context of ALK or ROS1 aberrations remains largely unexplored. Given the structural similarities between ALK and ROS1, it is plausible that glycosylation findings in one of these RTKs can be extrapolated to the other.

SUMMARY

Glycosylation plays a very important role in the regulation of RTKs relevant to solid cancers, particularly in lung cancer, and these post-translational modifications influence protein stability and function as well as therapeutic targeting. This review highlights the significance of aberrant glycosylation patterns in the key RTKs: EGFR, HER-2, ALK, ROS1, RET, and c-MET, and their potential impact on cancer treatment strategies.

N- and O-glycosylation in key domains of RTKs serve as crucial protein modifications that affect receptor activity, ligand binding, and downstream signaling pathways. EGFR, HER-2, ALK, and c-MET have been studied to some extent, and insights from their glycosylation patterns have already led to potential therapeutic interventions. For example, targeting specific glycan structures has shown promise in enhancing the efficacy of EGFR and HER2 inhibitors, possibly helping to alleviate drug resistance and improving patient responses. Furthermore, ALK glycosylation affects receptor activation and contributes to resistance mechanisms, emphasizing the need for further exploration of the role of glycosylation in fusion-driven lung cancers. On the other hand, ROS1 and RET remain relatively unexplored in terms of functional glycosylation in cancer. Their structural and functional similarities to more well-characterized RTKs provide a strong rationale for investigations to narrow our gap in understanding glycan-mediated regulatory mechanisms of RTKs in cancer.

Despite significant progress, challenges remain in translating glycosylation-based insights into clinical applications. The complexity of glycosylation patterns and their dynamic nature necessitate advanced analytical tools and comprehensive studies to identify specific glycan structures that could serve as biomarkers or therapeutic targets. In addition, understanding the interplay between glycosylation and genomic alterations is critical to developing personalized treatment strategies. Future research should focus on the integration of large-scale glycoproteomics into routine cancer diagnostics to identify additional glycosylation sites relevant to lung cancer and acquired therapy resistance. The development of glycan-targeting therapeutics and monoclonal antibodies targeting aberrantly glycosylated proteins will be critical in combined therapy regimens. In addition, fine-tuning of antibody-enzyme conjugates leading to tumor cell-specific targeted de-glycosylation activity will be necessary for enhanced personalized treatment approaches. By advancing our understanding of the role of glycosylation in lung cancer, we can lay the foundation for more precise and effective diagnostic and therapeutic approaches and ultimately improve patient outcomes.

AVAILABILITY OF DATA AND MATERIALS

Glyco-conjugate-related data have been accessed through GlyGen at https://www.glygen.org/, an international data source integrating and harmonizing publicly available data.