Introduction
Type 1 diabetes is a chronic autoimmune disease with an unknown cause, and is marked by destruction of insulin-producing pancreatic beta cells [
1]. The number of children and adolescents diagnosed with type 1 diabetes is increasing at an annual rate of approximately 3% [
2]. Although measurement of islet autoantibodies can expose the disease years before clinical diagnosis, continuous monitoring is expensive, is difficult in young children and is not adequately sensitive or specific in adults [
3,
4]. As early identification of type 1 diabetes can minimise morbidity and facilitate prevention, development of risk assessment tools is an important task, and many trials, initiatives and networks have been established with this aim [
5]
. A recently established genetic risk score showed good discriminative values [
6], but identification of additional biomarkers that could contribute to the risk assessment would be of great value.
N-glycosylation of plasma proteins is a strictly regulated and very complex enzymatic process by which different oligosaccharides are added to protein backbones [
7], modulating protein function in many instances [
8]. Glycosylation must not be confused with glycation, which is a non-enzymatic reaction, such as described for HbA
1c [
9]. Detailed genetic studies have identified genes that are important in regulating the type of glycans added, and also showed that residues flanking the
N-glycosylation motif affect the added glycan type [
10]. The human plasma
N-glycome is remarkably stable within an individual under physiological conditions [
11], and yet is very sensitive to various pathological processes, thus enabling consideration of
N-glycans as diagnostic and prognostic markers [
12‐
14]. Diabetes classification may be difficult as it is dependent on conditions at the time of diagnosis; for example, some individuals diagnosed with type 2 diabetes have islet autoantibodies [
15]. We showed previously that it is possible to differentiate between diabetes types and even identify individuals at an increased risk of developing type 2 diabetes in the future based on their
N-glycan profiles [
12,
13,
16,
17].
N-glycosylation profiling may also prove to be an asset in comparison with antibody testing due to cost reduction, which may be further reduced upon determination of diagnostically significant
N-glycan structures.
During the process of eukaryotic protein
N-glycosylation, a block of 14 sugars is transferred cotranslationally to specific asparagine residues in the endoplasmic reticulum [
18], and afterwards modified in the Golgi complex. This results in numerous modifications, such as branching, fucosylation, sialylation, etc. [
19]. Under physiological conditions, approximately 3% of glucose is used in the hexosamine biosynthesis pathway [
20], in which the donor molecule for the process of
N-glycosylation, uridine diphosphate-
N-acetylglucosamine (UDP-GlcNAc), is synthesised [
18]. The degree of glycan branching that defines glycan complexity (biantennary, triantennary and tetraantennary glycans) depends on the UDP-GlcNAc availability [
19]. Elevated glucose flux through the hexosamine biosynthesis pathway results in elevated levels of UDP-GlcNAc and potentially highly branched glycans [
19,
21]. We previously reported that complex highly branched serum
N-glycans were increased in adult type 1 diabetes patients with unregulated blood glucose [
22].
In addition, plasma levels of mannose-binding lectin, which activates one of the complement pathways upon binding to specific sugar residues [
23], were increased in a type 1 diabetes population [
24]. Aberrant
N-glycosylation of T cell proteins was implicated in the type 1 diabetes onset [
14]. Genome-wide association studies identified the glycosyltransferase gene encoding fucosyltransferase 2 as one of the susceptibility genes for type 1 diabetes [
25].
Therefore, we undertook the current study to identify plasma N-glycans characteristic of the early phase of type 1 diabetes by comparing children with type 1 diabetes and their healthy siblings, to compare these results with the N-glycan profile previously described in adult type 1 diabetes patients with unregulated blood glucose, and to describe age- and sex-dependent changes in the plasma N-glycome in children and adolescents with type 1 diabetes. As far as we are aware, this is the first study of plasma N-glycosylation changes at the onset of type 1 diabetes. We hypothesise that plasma N-glycosylation may differ between children newly diagnosed with type 1 diabetes and their healthy siblings, and thus could contribute to type 1 diabetes risk assessment.
Discussion
In the first part of this study, plasma and IgG N-glycans were analysed in a study population comprising 1917 children and adolescents (age 0.6–19.1 years) who were newly diagnosed with type 1 diabetes. A follow-up study compared the results for 188 of these participants with those for 244 unaffected siblings. As far as we are aware, these are the largest studies of plasma N-glycosylation changes in children to date, and the first to be performed at the onset of type 1 diabetes. We found significant differences in both plasma and IgG N-glycans between children with type 1 diabetes and their healthy siblings, and developed a glycan-based type 1 diabetes discriminative model. We also found significant N-glycan changes with age at diagnosis of type 1 diabetes and between sexes in diabetic patients.
Our previous study of plasma and IgG
N-glycosylation in adult type 1 diabetes patients showed that higher HbA
1c was associated with a shift toward more complex triantennary and tetraantennary plasma
N-glycans [
22]. In the current study, we observed a different change at the onset of type 1 diabetes, toward simpler
N-glycans, i.e. very simple plasma and IgG glycans with terminal mannoses and GlcNAcs, some biantennary plasma
N-glycans, and
N-glycans with bisecting GlcNAc. These changes specific to individuals with recent onset of diabetes suggest that they may have value in prediction of type 1 diabetes risk, and may not be merely the reflection of differences in glycaemic control.
A significant increase in the proportion of glycans that terminate with mannoses and glycans with bisecting GlcNAc was observed in children with type 1 diabetes in comparison with their healthy siblings. An increase in the mannose-binding lectin that binds to terminal mannoses and GlcNAcs has been reported in populations of individuals with type 1 diabetes [
24,
39]. In rheumatoid arthritis, multiple presentation of IgG
N-glycans with terminal GlcNAcs was shown to activate the mannose-binding lectin-complement cascade in the affected joints [
40], suggesting that these glycan changes may contribute to the chronic joint inflammation. In the present study, both plasma and IgG FA2B glycan (which terminates in three GlcNAcs) increased in the type 1 diabetes group. Also, the gene encoding the main protein of the complement activation pathway, complement C3 protein, has been associated with an increased risk of type 1 diabetes development among HLA-DR4/4 carriers [
41], further demonstrating an important role of the complement system in type 1 diabetes. IgGs carrying bisecting GlcNAc have been implicated in increased antibody-dependent cellular cytotoxicity [
42], an important process during virus elimination, and it has been suggested that one of the autoimmunity triggers in type 1 diabetes may be virus-derived [
43].
In addition, we observed an increase in IgG disialylation and a decrease in asialylation proportion among children with type 1 diabetes. Studies have shown that sialylated IgGs are anti-inflammatory mediators [
44]. We speculate that differences in sialylation between the studied groups may reflect the ongoing inflammation process at the onset of this disease.
A decrease in monogalactosylation proportion was also observed in children with type 1 diabetes relative to their healthy siblings. This has also been reported previously in another autoimmune disease, systemic lupus erythematosus [
45]. However, a decrease in the proportion of FA2[3]G1 and FA2[6]G1 monogalactosylated glycans was associated with poorer glycaemic control in adult type 1 diabetes patients [
22], and thus its real individual value in risk assessment should be evaluated after correcting for glycaemic differences.
Significant
N-glycosylation differences between the sexes were mainly observed upon onset of puberty, which is in line with our previous study of 170 children and adolescents [
46]. However, derived traits representing plasma and IgG bisecting structures, plasma high-mannose structures and IgG core fucose were significantly different between the sexes even before puberty. It is reasonable to examine sex differences for disease-associated glycans as there may be a hormonal component associated with type 1 diabetes given the higher prevalence of this disease in male participants after the onset of puberty [
47]. However, the proportions of some of the disease-associated risk-increasing glycans were higher and those of the disease-associated risk-decreasing glycans were lower in the same sex. Therefore, it is difficult to speculate which glycans reflect the different risk rates between the sexes. A previous study of 130 children and adolescents [
48] reported similar overall changes of IgG glycosylation with age.
ZnT8R autoantibodies (ZnT8RA) were associated with the digalactosylated IgG
N-glycan with bisecting GlcNAc (GP13), and an increase in the number of different autoantibodies, which is thought to be a better predictor of progression to type 1 diabetes than levels of any individual antibody [
3], was associated with some of the highly branched plasma
N-glycans. These results indicate that some variations in glycans reflect type 1 diabetes-specific autoimmunity. ZnT8A were detected in 81% of children who progressed to type 1 diabetes [
49], making these autoantibodies very important in predicting diabetes. However, the proportion of GP13 glycan was not significantly different between affected and unaffected siblings. In the present study, we have not distinguished between the fraction of IgG antibodies that react to type 1 diabetes antigens and the non-autoreactive IgG fraction. Studying glycosylation changes of type 1 diabetes antigen-specific IgGs would be important in future studies.
As the plasma samples analysed in the current study were collected within 3 months of type 1 diabetes onset, the study design does not allow us to conclude whether the observed
N-glycosylation changes are causative or reflective of disease status. We also acknowledge that, when studying total plasma protein
N-glycome, both variation in protein glycosylation and changes in protein concentration could affect the observed plasma glycome changes. However,
N-glycosylation appears to be strongly associated with type 1 diabetes, and this association demands further research. We were not able to perform a validation study as it is very hard to obtain samples from patients at the onset of this disease; thus, evaluation of the utility of glycan-based predictive model remains an important future step. Also, we defined puberty based on age categories, whereas clinical assessment would be a more precise measure. We were not able to standardise the glycan data against medication intake as data regarding the treatment of study participants were not available. Nevertheless, our previous study demonstrated that insulin intake has a low effect on a limited number of glycans [
50]. Some of the unaffected siblings were lost to follow-up if they were subsequently diagnosed above the age of 18 years, and therefore their type 1 diabetes status is less certain than for those individuals who were followed for an extended time. However, as the incidence of type 1 diabetes after puberty decreases markedly with increasing age [
47], it is less likely that the older individuals followed for a shorter period developed the disease. One of the major strengths of our study is that the study population comprises children at the onset of type 1 diabetes together with their unaffected siblings, without other comorbidities as seen in the adult population, allowing glycan changes related to type 1 diabetes to be investigated more precisely.
Earlier intervention to prevent type 1 diabetes would be aided by identifying individuals who are at higher risk, hence the long-running search for novel biomarkers of this disease. The main contribution of our research study was the identification of
N-glycosylation changes around the time of diabetes onset. This allowed us to develop a glycan-based predictive model that may be of clinical utility (AUC >0.9). This model outperforms a previously described genetic risk score [
51], which, when combined with additional clinical data, yielded an AUC of 0.79, and compares favourably with the recently improved version of this score [
6]. Incorporation of genetic and clinical data into the differential
N-glycosylation model could further optimise prediction. An important future step to evaluate the utility of the glycan-based predictive model would be to study whether the
N-glycosylation profile identified in children at type 1 diabetes onset also exists in individuals who have islet autoimmunity before the development of clinical diabetes. Interestingly, two unaffected siblings who had very low monogalactosylation proportions (below Q1) at the time of plasma collection were later diagnosed with diabetes. This glycan trait was significantly lower in children with type 1 diabetes in comparison with their healthy siblings.
In summary, the current study demonstrated significant changes in plasma N-glycosylation accompanying the onset of type 1 diabetes, and enabled us to develop a predictive model incorporating glycan data. Our large cohort also made it possible to confirm age- and sex-dependent N-glycosylation changes in children and adolescents studied to date on a smaller number of participants, and to reveal some new differences. An increase in the number of different type 1 diabetes-associated autoantibodies, which is a better predictor of progression to diabetes than any individual antibody, was associated with specific changes within the plasma N-glycome. These results favour further research into N-glycosylation changes and their impact on type 1 diabetes.
Acknowledgements
The authors are grateful for and would like to acknowledge the invaluable contributions of all the participants, research nurses, local investigators, administrative teams and other clinical staff. Some of the data were presented as an abstract and/or poster presentation at the 55th EASD Annual Meeting in 2019, in Glycoconjugate Journal preceding the GLYCO 25 International Symposium on Glycoconjugates in 2019, at the Infodan doktorskog studija Farmaceutsko-biokemijske znanosti in 2019, at the 29th Joint Glycobiology meeting in 2018, 2nd GlycoCom in 2018, 1st Human Glycome Project Meeting in 2018 and the 28th Joint Glycobiology Meeting in 2017.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.