Multi-block data integration analysis for identifying and validating targeted N-glycans as biomarkers for type II diabetes mellitus

Comprehensive understanding of spectral N-glycan data from UPLC analysis is anchored on advanced statistical methods. Integrative methods offer comprehensive means to dissect data, with the goal of transforming the data into a clinically useful information. For the first time, we have applied a powerful and advanced integrative method DIABLO, to explore N-glycan profiles interaction in real time.

Prior to applying the DIABLO method, univariate, and multivariate statistical methods (eg, student t tests and Mann Whitney U tests and chi-square test) have been used to reveal the association between T2DM and biochemical measures such as plasma glucose and lipid profiles. Surprisingly, the control group had a higher blood pressure than the cases, and this can be attributed to the medication use (glucose, lipid and lower blood pressure drugs) among the cases (Table 1). Moreover, this highlights the proportion of the population who have raised blood pressure, but they are unaware of it. WHO reports that an estimated 45% of hypertensive adults are not aware of it, although the control group in the current study cannot be said to be hypertensive. This is because hypertension is established after written measures of blood pressure above normal threshold (140 mmHg). In the present study, blood pressure was only measured once. It is not clear why the control group had a lower HDL-c but it may be attributed to genetic factors or defects in cholesterol efflux.

Medication use in T2DM can potentially affect their N-glycome. Singh et al.30 found that statin use was linked to a decrease in all fucosylated traits including diantennary and triantennary structures (A2EF, A2LF, A3EF, A3L0F). In addition, statin use was associated with an increased galactosylation in diantennary non-fucosylated (A2F0G) and in sialylated diantennary (A2SG) glycans. The study further stated that statin use negatively correlated with Alpha2,6-sialylation of triantennary (A3E) and fucosylated tetra-antennary glycans (A4FGE). Similarly, metformin correlated with a decreased fucosylation in diantennary, triantennary and tetra-antennary traits and an increase of galactosylation in diantennary glycans34.It is widely known that T2DM develops several years before clinical diagnosis. Mild symptoms such as weight loss or weight gain, fatigue, increased hunger would progressively result in persistent high plasma glucose and complications. However, because of limited sensitive, and robust biomarkers, T2DM diagnosis is often delayed. This problem appears to be solved with the advent of N-glycans. First, the GST2D score was used to predict T2DM development 6–8 years before clinical manifestation35. In another study, Cvetko et al.36 reported that individuals who were healthy at baseline but developed insulin resistance and T2DM over time, were characterised by complex and highly branched N-glycan structures. Specifically, the study identified alterations in eight N-glycans: GP10, GP16, GP18, GP19, GP20, GP26, GP32 and GP3436; with GP 32 and GP34 being the most significant in the continuum of insulin resistance and T2DM. Increasing evidence shows that T2DM patients can be distinguished from healthy individuals depending on the composition of their respective total N-glycome5,18. Thus, we explored the N-glycan traits whose expression were different in cases and controls. The present study validates that of Cvetko et al.36 and Clemens et al.35, we identified GPs 34, 32, 26, 31, 36 and 30 to be highly expressed in T2DM in the first principal axis and on the second principal axis, GPs 38, 1, 2, 25 and 20 were dominant in T2DM. Sialylated glycans (GP26, GP32, GP35 and GP36) are expressed on a1-acid glycoprotein, whereas GP18 and GP20 originates from glycoproteins a-antitrypsin. A-antitrypsin is a protease inhibitor with at least three glycosylation sites for biantennary glycans without fucosylation (site asparagine 70), bi-, tri- and tetra-antennary glycans with core and antennary fucosylation (at site asparagine 107) and site asparagine 271 is occupied by bi- and tri-antennary glycans with core- and antennary-fucosylation37. A-antitrypsin protects β-cells from apoptosis and triggers insulin secretion, hence important for preventing type I diabetes38.

Clerc et al.39, further states that triantennary (GP 30, GP 31 and GP 32) and tetraantennary (eg, GP 26, GP34, 36) glycans are expressed on kininogen-1 and histidine-rich glycoproteins. Kininogens are proteins with multiple functions including antidiuretic, antiangiogenic, antithrombotic, profibrinolytic and proinflammatory proteins. Abnormal expression of this glycoprotein is linked to diabetes40. Histidine-rich glycoproteins bind to ligands including heparan sulfate, plasminogen, are among others and regulates multiple processes such as cell adhesion, fibrinolysis, cell chemotaxis. A deficiency of this protein has been associated with thrombosis, but its role in diabetes has also been reported41. Similarly, abnormal activities of a-antitrypsin, transferrin and hemopexin are all implicated in diabetes. Of particular interest is three glycan groups (GP30, GP36 and GP38) that have been shown to have clinical relevance in maturity onset diabetes of the young (MODY)42. In fact, Juszczak et al.42, documented that GP30, GP36 and GP38 had the best discriminative power between HNF1A-MODY and early-onset type 2 diabetes. The authors explained that HNF1A is a transcription factor for the inflammatory marker C-reactive protein (CRP) and a master regulator of fucosylation; with variations in HNF1A triggering MODY. With a sensitivity of 88% and specificity of 80%, was the best amongst the three glycan groups in discriminating between individuals with damaging HNF1A alleles from those with early-onset nonautoimmune diabetes but lacked HNF1A variants. The study showed that subjects with deleterious HNF1A allele had reduced levels of these glycans than those who lacked the rare HNF1A allele42.

The findings of the current study build upon that of Keser et al.17 who also suggested that the increased branched N-glycans in T2DM can be due to dysregulation of the hexosamine biosynthesis pathway (HBP). HBP has been found to be involved in the metabolism of glucose. This pathway under normal conditions, metabolises up to 3% glucose of the total glucose in the body. However, when homeostatic mechanism is disturbed, such as in T2DM, the metabolism of glucose is heightened, producing uridine diphosphate N-acetylglucosamine (UDP-G1cNAc). UDP-G1cNAc is a substrate for glycosyltransferases that catalyses the elongation and branching of glycan chains in glycosylation. GNTs are encoded by MGAT3 [mannosyl (β-1,4-)-glycoprotein β-1,4-N-acetylglucosaminyltransferase] But specifically, GNT-I, -II, -IV and -V catalyses the biosynthesis of mono, bi, tri and tetra-antennary glycans whereas GNT extends the 1–6 arm of the glycan core with GlcNAc residue. A defective GNT glycosyltransferase in the pancreatic islets results in impaired insulin action, impaired glucose tolerance and eventually, hyperglycaemia.

Aberration of fucosylation, be it core or antennary has been implicated in our results just as stated in multiple chronic diseases43,44,45. For example, Herrera et al.46 identified core-fucosylated tetra-antennary glycan to be associated with poor breast cancer prognosis. Then Testa et al.44, showed that core-α-1,6-fucosylated diantennary glycans were associated with T2DM. Sialic acids (N-acetylneuraminic acids) are pinned to the non-reducing ends of N-glycans by way of 2,3-,2,6- linkages. When bound, they play crucial roles in the pathological conditions including cancers and viral infections, while sialic acid complex glycans have been suggested to have anti-inflammatory properties47. Removal of UDP-N-GlcNAc 2-epimerase/ N-acetylmannosamine (ManNAc) kinase, an enzyme required for the biosynthesis sialic acids, led to glomerula proteinuria in mice48. In addition, other studies have found that upregulation of β-galactoside α-2,6-sialyltransferase 1, an enzyme that catalyses terminal α2,6-sialylation, was associated with worse patient outcomes in cancer49. Other studies have also indicated that an increase in α-(2 → 3)-sialic acid correlates with tumor metastasis. For example, intravenous administration of a sialidase (enzyme that cleaves sialic acids) blocking agent caused an increase release of insulin in pancreatic islets50. It is known that hyposialylated IgG glycans stimulates endothelial FcγRIIb, which has been previously associated with insulin resistance in obese mice. In the present study, the T2DM was associated with terminal sialylation. Recently, increased sialic acids on N-glycans have been implicated in T2DM development17. The absence of sialic acids on plasma LDL-c has been shown to induce cholesterol ester accumulation in cells and hence implicated in cardiometabolic diseases. This could be a possible reason why plasma LDL-c was highly loaded in cases compared to controls51.

The main limitation of the study relates to the small sample, and which means, the results cannot be generalised. Also, there is a possibility of biological variations related to gene expression in the samples, but that was not investigated. Already a genome wide association study has identified HNFA1 α as the master regulator of fucosylation52. Moreover, Cohain et al.27 Analysis on cardiometabolic tissues revealed multiple genes that code for clinical markers including total cholesterol (DHCR7, FADS1, FADS2, MMAB, and MVK), (FLVCR1, LSS, MMAB, MVK, DHCR7, FADS1, FADS2 and VPS37D), LDL-c ( FADS1, FADS2, and LSS), HDL-c (FLVCR1, MMAB, MVK, FADS1, FADS2), and TG (VPS37D, FADS1, FADS2). Zaitseva53 also reported that most of the highly heritable N-glycan peaks such as GP1, GP2, GP4-6, GP10-11, GP16, and GP17 were core-fucosylated biantennary with reduced sialylation whereas GP 20 and GP 14 had a low heritability. We intend to use path analysis and confirmatory factor analysis to determine gene-glycan relationships.

The present study has only provided information about glycans in biological samples (glycome), without highlighting downstream changes in the transcriptome, metabolome, lipidome and proteome. Thus, combining and analysing multiomics simultaneously will provide a clearer understanding of the mechanism that underly T2DM pathogenesis.

Leave a Comment