Association of IgG N-glycomics with prevalent and incident type 2 diabetes mellitus from the paradigm of predictive, preventive, and personalized medicine standpoint

Abstract

Objectives

Type 2 diabetes mellitus (T2DM), a major metabolic disorder, is expanding at a rapidly rising worldwide prevalence and has emerged as one of the most common chronic diseases. Suboptimal health status (SHS) is considered a reversible intermediate state between health and diagnosable disease. We hypothesized that the time frame between the onset of SHS and the clinical manifestation of T2DM is the operational area for the application of reliable risk assessment tools, such as immunoglobulin G (IgG) N-glycans. From the viewpoint of predictive, preventive, and personalized medicine (PPPM/3PM), the early detection of SHS and dynamic monitoring by glycan biomarkers could provide a window of opportunity for targeted prevention and personalized treatment of T2DM.

Methods

Case–control and nested case–control studies were performed and consisted of 138 and 308 participants, respectively. The IgG N-glycan profiles of all plasma samples were detected by an ultra-performance liquid chromatography instrument.

Results

After adjustment for confounders, 22, five, and three IgG N-glycan traits were significantly associated with T2DM in the case–control setting, baseline SHS, and baseline optimal health participants from the nested case–control setting, respectively. Adding the IgG N-glycans to the clinical trait models, the average area under the receiver operating characteristic curves (AUCs) of the combined models based on repeated 400 times fivefold cross-validation differentiating T2DM from healthy individuals were 0.807 in the case–control setting and 0.563, 0.645, and 0.604 in the pooled samples, baseline SHS, and baseline optimal health samples of nested case–control setting, respectively, which presented moderate discriminative ability and were generally better than models with either glycans or clinical features alone.

Conclusions

This study comprehensively illustrated that the observed altered IgG N-glycosylation, i.e., decreased galactosylation and fucosylation/sialylation without bisecting GlcNAc, as well as increased galactosylation and fucosylation/sialylation with bisecting GlcNAc, reflects a pro-inflammatory state of T2DM. SHS is an important window period of early intervention for individuals at risk for T2DM; glycomic biosignatures as dynamic biomarkers have the ability to identify populations at risk for T2DM early, and the combination of evidence could provide suggestive ideas and valuable insight for the PPPM of T2DM.

Supplementary information

The online version contains supplementary material available at 10.1007/s13167-022-00311-3.

Introduction

T2DM—a major metabolic disorder and global health emergency

Type 2 diabetes mellitus (T2DM), a major metabolic disorder, is expanding at a rapidly rising worldwide prevalence and has reached alarming levels, making it one of the fastest growing global health emergencies of the twenty-first century [1]. The studies reported that T2DM could cause a series of complications (including blindness, kidney failure, heart attacks, stroke) [2] and is considered a nonnegligible cause of life threatening, disability, reduced quality of life, and reduced life expectancy [1, 3]. In 2021, the International Diabetes Federation (IDF) estimates that approximately 537 million adults are living with diabetes (diagnosed or undiagnosed) worldwide, over 90% of whom suffer T2DM, and the number is projected to reach 783 million by 2045 [1]. It is worth noting that China has the highest number of patients with diabetes, accounting for 1 in 4 of all adults living with diabetes worldwide [1]. As the leading cause of disability and mortality, T2DM has become a serious global health issue, imposing a substantial economic burden on countries, health systems, diabetes patients, and their families, especially in low- and middle-income countries [4–7]. This shows that the early prevention and treatment of T2DM have become a global problem that cannot be ignored and urgently needs to be solved.

In general, the symptoms of T2DM are much less dramatic and may even be completely asymptomatic, which leads to a potentially long prediagnostic period for T2DM. Studies have reported that as many as one-third or more individuals with T2DM in the population may be undiagnosed [8], and prolonged undiagnosed T2DM could lead to a range of complications that could further result in serious health damage to the body [1]. The aforementioned phenomenon further emphasized the importance of early discrimination or diagnosis for early prediction, prevention, and personalized medicine (3PM/PPPM) of T2DM. At present, the diagnostic methods of T2DM, including fasting plasma glucose (FPG), 2-h postprandial plasma glucose test by 75 g oral glucose tolerance tests (OGTT), and/or glycosylated hemoglobin A1c (HbA1c) proposed by the World Health Organization (WHO), IDF, and American Diabetes Association [8–10], are not able to identify individuals at high risk of T2DM in the early stage before the onset of T2DM, predict the occurrence of T2DM, or monitor the dynamic changes that occur in the body during the progression of T2DM. Therefore, it is urgent to explore the potential dynamic biomarkers associated with the early pathologic stage of T2DM with high sensitivity and specificity for the early discrimination of T2DM and to facilitate the transition from reactive medical services to PPPM in the management of T2DM.

Suboptimal health status (SHS) is a time frame for PPPM in T2DM

SHS is considered a reversible intermediate state between health and a diagnosable disease and is defined as “an overall physical status between health and illness characterized by the perception of health complaints, chronic fatigue, and a constellation of physical symptoms such as the cardiovascular system, the digestive system, the immune system, and mental status; lasting for at least 3 months” [11, 12]. Recent reports have shown that SHS individuals are susceptible to physical or mental diseases, especially non-communicable diseases (NCDs), which usually present a chronic developing status by progressing from reversible SHS to irreversible severe pathological status over a couple of years [13, 14]. The studies further suggested that SHS might precede the onset of NCDs and represent the period before the onset of clinical manifestations of NCDs, involving cardiovascular disease and T2DM [14–17]. Furthermore, prediabetes, characterized by impaired glucose tolerance and impaired fasting glucose, is usually asymptomatic and an early stage in the development of T2DM, signifying that people with prediabetes are at higher risk of future development of T2DM [18]. However, the early identification of prediabetes is difficult, and as a potential tool for the identification of high-risk groups of T2DM, SHS might be widely used for the early screening of high-risk groups of T2DM due to its advantages of convenience, simplicity, acceptability, noninvasive, and cost-effectiveness. PPPM is a holistic strategy for healthcare that is carried out from the perspectives of predicting individual predisposition, providing targeted prevention, and creating personalized treatment [19, 20]. In general, T2DM is treated after onset, which is a relatively delayed response and does not fit the paradigm and goal of PPPM [13]. In other words, the early discrimination and even intervention of SHS might produce a window opportunity in the targeted prevention and personalized treatment of T2DM from the standpoint of PPPM.

Promising glycomic biosignatures for T2DM identification

N-glycosylation is considered to be one of the major glycosylation pathways and the primary mechanisms of glycans linked to proteins and is a common and essential posttranslational modification of proteins that can affect the physical, chemical, and biological properties of modified proteins with diverse glycans [21]. There are glycans attached to the majority of cell surface (i.e., membrane proteins) and secreted proteins, which can modify the functions of proteins and are involved in intra- and intercellular molecular processes in vivo, including protein folding, cell adhesion, immunoregulation, signal transduction, modulation of receptor activity, and many other disease/health-associated events [22–25]. Glycans are complex oligosaccharides composed of up to 15 monosaccharide residues and are attached to polypeptide structures on the constant fragment region of immunoglobulin G (IgG), i.e., the amide group of an asparagine side chain (Asn-297) [26, 27]. Most IgG glycans are classified as N-glycans, which can influence the effector functions of IgG, antibody half-life in the bloodstream, autoreactivity, immune complex formation, etc. [28]. Studies have indicated that alterations in IgG N-glycosylation are important for inflammatory-related physiological and pathological processes, including aging, central adiposity, hypertension, ischemic stroke, and cancers [24, 29–33]. In addition, a series of case‒control studies in multiple ethnic populations, including European, Australian, Chinese Han, and Chinese Uyghur populations, have reported changes in IgG N-glycosylation in individuals with T2DM, proving that IgG N-glycans have potential as biomarkers for the early identification of T2DM [34–37]. However, prospective evidence of altered IgG N-glycosylation in T2DM etiology is scarce.

Objective of the study

The primary objectives of this study were to investigate the association of IgG N-glycosylation with T2DM and screen potential glycomic biosignatures for distinguishing T2DM from healthy controls at an early stage based on evidence from case‒control and prospective nested case‒control studies and to evaluate the potential role of SHS in the development of T2DM in a prospective nested case‒control study. The baseline SHS and the changes in glycosylation profiles of T2DM individuals may be warning signs before the onset of T2DM. We hypothesized that the time frame between the onset of SHS and the clinical manifestation of T2DM is the operational area for the application of reliable risk assessment tools, such as IgG N-glycans. Therefore, from the perspective of PPPM, the detection of individual IgG N-glycosylation levels combined with early intervention with SHS may have important significance for the early prevention of T2DM.

Materials and methods

Study design and participants

All participants in this population-based study were recruited from a prospective cohort study. The prospective cohort was established, and baseline data collection was conducted in 2016 with yearly follow-up. All subjects were recruited from the Health Management Center of Beijing Aerospace General Hospital based on the China suboptimal health cohort study (COACS). In the present study, case–control and prospective nested case–control studies were conducted to investigate the association between IgG N-glycans and T2DM, which included 138 and 308 subjects, respectively. All individuals provided signed written informed consent, and the study was approved by the Ethics Committee of the Chinese People’s Liberation Army General Hospital, Beijing, China (No. S2016-0681–01). The implementation of the study was in accordance with the principles of the Declaration of Helsinki.

Baseline evaluation

Inclusion and exclusion criteria

The inclusion criteria of the study were individuals aged between 18 and 65 years, body mass index (BMI) ≥ 18.0 kg/m², and the resident population in Beijing. If participants had a history of somatic or psychiatric abnormalities (including mental illness, cognitive impairment, and severe physical disability), they were excluded; participants who were previously diagnosed and/or currently self-reported with hypertension, diabetes, gout, and hyperlipidemia; cardiovascular or cerebrovascular conditions (involving myocardial infarction, congestive heart failure, atrial fibrillation, atrial flutter, stroke, and transient cerebral ischemia); and any type of cancer were excluded; additionally, pregnant or lactating women were also excluded from this study.

Collection of questionnaires, blood samples, and biochemistry tests

From 2016, all subjects underwent a standardized physical examination and laboratory evaluation of T2DM risk factors, including anthropometric measurements and physical and chemical properties of blood. Additionally, the participants were required to complete a set of self-administered questionnaires (incorporating demographic questionnaire and SHS questionnaire 25 items (SHSQ-25)) with the assistance of a trained research assistant. In detail, the demographic characteristics, namely, age, sex, level of education, marital status, and household income per capita monthly, as well as lifestyle factors, including smoking, drinking, physical activity at work, and sleep duration, were collected through questionnaires. Briefly, smoking and drinking status were classified into three conditions: never (never smoking or drinking), former (former smoking or drinking), and current (currently smoking or drinking). Physical activity at work was divided into mentally demanding work and light, moderate, and heavy physically demanding work. The average sleeping hours over a 24-h period were collected as sleep duration. Moreover, anthropometric measurements involving height, weight, waist circumference, hip circumference, and systolic and diastolic blood pressures (SBP and DBP) were measured by a trained nurse or physician, and BMI and waist-to-hip ratio (WHR) were calculated from height, weight, and waist and hip circumference, respectively. More details of the measurement procedures are described in a previous report [14, 38].

Blood samples from all participants were collected at the time of baseline information acquisition, collected from the antecubital vein under fasting conditions in the morning, and stored in vacuum tubes containing/without ethylene diamine tetraacetic acid. The separated blood samples were used for routine biochemical assays and IgG N-glycan analysis. The indexes of biochemical assays included total cholesterol (TC), total triglycerides (TGs), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), FPG, HbA1c, postprandial blood glucose (PBG), alanine aminotransferase (ALT), aspartate aminotransferase (AST), creatinine (Cr), and uric acid (UA).

Determination of SHS

At the stage of baseline data collection, after screening the participants according to the inclusion and exclusion criteria, the SHSQ-25, a quick evaluation scale to distinguish SHS individuals from healthy individuals [39], was used to evaluate the SHS of the high-risk groups of T2DM during the baseline stage [11]. The multidimensionality of SHS was explained by the SHSQ-25, which involved 25 items and was divided into five domains, namely, fatigue, the cardiovascular system, the digestive system, the immune system, and mental status. All participants could be classified into two categories by SHSQ-25: an SHS score ≥ 35 represents SHS (poor health), and < 35 represents health status (optimal/ideal health) [12].

Case–control study

At the time of baseline information collection, the participants were diagnosed with T2DM based on a 2-h plasma glucose value after 75 g OGTT ≥ 11.1 mmol/L, in accordance with WHO criteria [10]. The participants with previously diagnosed T2DM or self-reported T2DM were not included in the current case‒control study. According to the inclusion and exclusion criteria of the baseline evaluation, only participants with newly identified T2DM detected by the OGTT at the time of baseline information collection were included in the case group. In fact, the participants of the case–control study were collected in 2016 based on the baseline data of the cohort, including 69 T2DM and 69 age- and sex-matched healthy individuals. In the case–control setting, the 2-h plasma glucose value after 75 g OGTT was defined as PBG.

Nested case–control study

Diagnostic criteria

During the follow-up phase, T2DM was defined as the presence of any of the following criteria on the basis of WHO criteria [10]: (1) FPG ≥ 7.0 mmol/L, (2) 2-h plasma glucose value after 75 g OGTT ≥ 11.1 mmol/L, (3) HbA1c ≥ 6.5%, or (4) physician interviews were conducted to determine whether subjects had been clinically diagnosed with T2DM since the last follow-up visit. Additionally, all the subjects were asked to check blood glucose after 2 h postprandial (i.e., PBG) during the follow-up. Specifically, the blood glucose test meal was a 100-g steamed bread meal, timed from the first bite.

Follow-up and T2DM outcomes

In the nested case–control study, T2DM outcomes were followed up yearly from 2017 to 2020. The last time point for T2DM outcome information collection was December 2020. New-onset T2DM related to medical records for all physician visits and/or hospitalizations during follow-up was obtained and reviewed by an adjudication panel consisting of the investigators. According to the inclusion and exclusion criteria, the nested case‒control study included 154 patients with new-onset T2DM and 154 age- and sex-matched healthy controls during the follow-up period from June 2017 to December 2020.

IgG N-glycomic detection and analysis

Immediately prior to the start of the IgG N-glycan detection experiment, the previously separated plasma samples were centrifuged again to remove residual lipids in the supernatant for use [40]. The experimental process of IgG N-glycan detection mainly included IgG isolation, IgG N-glycan release and labeling, and purification of labeled IgG N-glycans. In detail, first, IgG from separated plasma samples was isolated with a protein G monolithic 96-well plate (BIA Separations, Slovenia). The isolated IgG was eluted from a protein G plate using 1 mL of 0.1 M formic acid, and 1 M ammonium bicarbonate was instantly added to neutralize the eluate. Second, 30 µL of 1.33% SDS was added to the dried IgG samples, and all protein precipitates were dissolved in SDS. Then, 10 µL of 4% Igepal-CA630 was added to each sample to prevent denaturation of PNGase F caused by SDS. Third, the denaturation of IgG samples was added to N-glycosidase F (PNGase F) and incubated at 37 °C for 18 to 20 h to allow the release of N-glycans. Fourth, free N-glycans were labeled with 2-aminobenzamide (2-AB) and incubated with the labeled samples at 65 °C for 2 to 3 h. Finally, fluorescently labeled N-glycans were separated and analyzed by hydrophilic interaction liquid chromatography (HILIC) on an Acquity H-class ultra-performance liquid chromatography (UPLC) instrument (Waters, Milford, MA) producing into chromatograms for use. Previous reports described a more detailed experimental process [41, 42].

All obtained chromatograms were separated into 24 IgG N-glycan peaks (GPs, GP1-GP24) in the same manner, and the amount of each peak of IgG N-glycans was presented as a percentage of the total integrated area. In addition, 54 derived glycan traits (DGs) representing specific IgG glycosylation features, including core fucosylation, galactosylation, bisecting N-acetylglucosamine (GlcNAc), and sialylation, were calculated from the directly measured 24 GPs for analysis. The detailed description and calculation are shown in Table S1.

Statistical analysis

The assessment of statistical data (including questionnaire results, anthropometric measurements, biochemical indexes, and IgG N-glycan traits) normality was performed using the Kolmogorov–Smirnov test. Depending on whether the continuous variables followed a normal distribution, the mean ± standard deviation (SD) or median and interquartile range differences were used to describe the results, and parametric tests (Student’s t test) or nonparametric methods (Mann–Whitney U test) were used to compare the differences between the T2DM and healthy control groups. Chi-square analysis was conducted to test categorical variables, which are represented as frequencies or percentages. Correlation analyses, namely, Spearman’s rank correlation and canonical correlation analysis (CCA), were performed to investigate the relationship between IgG N-glycans and a series of T2DM-related clinical traits. The Spearman’s rank correlation results of directly measured IgG N-glycans and clinical traits are shown in heatmap (R package “psych” and “pheatmap”). In detail, CCA was conducted to determine the overall association of the two sets of variables (i.e., the directly measured glycans (x) and the clinical traits (y)) and to explore those combinations that were maximally associated with each other. After controlling for potential confounders, IgG N-glycans (24 GPs and 54 DGs) associated with T2DM were screened by multivariable logistic regression analysis and depicted by forest plots. Data from case–control and nested case–control studies were analyzed and processed using the aforementioned analysis methods.

Furthermore, in the nested case–control study, SHS was considered a stratification factor to further explore the association between T2DM and IgG N-glycans based on univariate analysis (Student’s t test or Mann–Whitney U test) and multivariable logistic regression analysis, comparing the glycan level of each stratification and establishing the SHS stratification discrimination model of T2DM. In addition, logistic regression analysis with interaction terms was applied to appraise the glycan-by-SHS interaction: T2DM ~ glycan + SHS + TC + TG_S + HDL + LDL + BMI + glycan * SHS. Specifically, logistic regression multiplicative and additive models were presented to evaluate the effect of interaction. In the logistic multiplicative model, the performance of the interaction between IgG N-glycans and SHS was evaluated by the coefficients of the interaction term of the model. In the logistic additive model, relative excess risk due to interaction (RERI), attributable proportion due to interaction (AP), and synergy index (S index) were used to assess the effects of interaction. In detail, the bootstrap method was used to calculate the point estimate and confidence interval (CI) of the RERI, AP, and S index for the interaction between two continuous variables (glycan and SHS) from the resampled samples (R package “boot”) [43].

Moreover, the internal associations among IgG N-glycans were confirmed [31], which could cause multicollinearity in the statistical models. To reduce the dimension for significant IgG N-glycans, least absolute shrinkage and selection operator (LASSO) regression was used to screen glycan variables (R package “glmnet”), and screened glycan variables were included to construct logistic regression models for discriminating T2DM and controls in the case–control and nested case–control settings. In both settings, three classification models were constructed: model 1 (clinical trait model), involving the clinical traits; model 2 (glycan model), including the screened IgG N-glycans; and model 3 (combined model), incorporating the screened IgG N-glycan variables on top of the clinical traits included in model 1. The discriminant ability of the models in delineating the T2DM from the healthy individuals was evaluated by receiver operating characteristic (ROC) curve analysis (R package “pROC”). The discriminant models were trained and evaluated using fivefold cross-validation to avoid overfitting, in which logistic regression classifiers were implemented. Next, fivefold cross-validation is repeated 400 times to protect against the influence of the random splits. The performance of the classifiers was measured by using the average value of the area under the ROC curve (AUC) of the cross-validation process to avoid biased estimation for discrimination (R package “caret”).

Data analysis and visualization were performed by SAS (version 9.4) and R programming language (version 4.1.2). To adjust the effect of the type I error, the false discovery rate (FDR) was controlled by using the Benjamini–Hochberg (BH) procedure in the Spearman rank correlation analysis and multivariable logistic regression analysis. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Results

Characteristics of participants

In total, 138 participants were included in the case–control survey, comprising 69 participants with T2DM as well as 69 age- and sex-matched healthy participants. Additionally, 154 new-onset T2DM patients during the 4-year follow-up, as well as 154 age- and sex-matched healthy individuals, were included in the nested case–control study. The demographic, anthropometric, lifestyle, and biochemical characteristics of the T2DM and healthy individuals for the case–control study and those for the nested case–control study are presented separately in Table 1.

Table 1.

Characteristics of the study participants

Variables	Case–control study					Nested case–control study
	Total (n = 138)	T2DM (n = 69)	Healthy controls (n = 69)	Statistics (χ2/Z/t)	P	Total (n = 308)	T2DM (n = 154)	Healthy controls (n = 154)	Statistics (χ2/Z/t)	P
Gender, n (%)				0.049	0.825				1.117	0.291
Male	113 (81.9)	56 (81.2)	57 (82.6)			255 (82.8)	131 (85.1)	124 (80.5)
Female	25 (18.1)	13 (18.8)	12 (17.4)			53 (17.2)	23 (14.9)	30 (19.5)
Age (years)	51 (42, 54)	51 (44, 54)	51 (41, 54)	− 0.533	0.594	47 (36, 53)	45 (36, 52)	49 (35, 54)	− 1.476	0.140
Ethnicity, n (%)				3.112	0.078				1.654	0.198
Han Chinese	135 (97.8)	69 (100.0)	66 (95.7)			298 (96.8)	151 (98.1)	147 (95.5)
Others	3 (2.2)	0 (0.0)	3 (4.3)			10 (3.2)	3 (1.9)	7 (4.5)
Levels of education, n (%)				3.665	0.160				2.695	0.260
Junior high school or below	11 (8.0)	7 (10.1)	4 (5.8)			14 (4.5)	10 (6.5)	4 (2.6)
Senior high school or technical secondary school	56 (40.6)	32 (46.4)	24 (34.8)			110 (35.7)	54 (35.1)	56 (36.4)
Bachelor degree or above	71 (51.4)	30 (43.5)	41 (59.4)			184 (59.7)	90 (58.4)	94 (61.0)
Marital status, n (%)				< 0.001	> 0.999				1.329	0.249
Spinsterhood	10 (7.2)	5 (7.2)	5 (7.2)			30 (9.7)	18 (11.7)	12 (7.8)
Married, divorced, or widowed individuals	128 (92.8)	64 (92.8)	64 (92.8)			278 (90.3)	136 (88.3)	142 (92.2)
Physical intensity at work, n (%)				0.792	0.851				4.66	0.198
Brainwork	43 (31.2)	20 (29.0)	23 (33.3)			106 (34.4)	51 (33.1)	55 (35.7)
Light physical labor	53 (38.4)	29 (42.0)	24 (34.8)			99 (32.1)	44 (28.6)	55 (35.7)
Moderate physical labor	36 (26.1)	17 (24.6)	19 (27.5)			79 (25.6)	43 (27.9)	36 (23.4)
Heavy physical labor	6 (4.3)	3 (4.3)	3 (4.3)			24 (7.8)	16 (10.4)	8 (5.2)
Household income per capita monthly, n (%)				4.226	0.121				< 0.001	> 0.999
≤ ¥3500 (503 EUR, 502 USD)	23 (16.7)	16 (23.2)	7 (10.1)			44 (14.3)	22 (14.3)	22 (14.3)
¥3501–8000 (503–1151 EUR, 502–1148 USD)	89 (64.5)	41 (59.4)	48 (69.6)			202 (65.6)	101 (65.6)	101 (65.6)
≥ ¥8001 (1151 EUR, 1148 USD)	26 (18.8)	12 (17.4)	14 (20.3)			62 (20.1)	31 (20.1)	31 (20.1)
Smoking, n (%)				0.445	0.801				4.961	0.084
No-smokers	61 (44.2)	30 (43.5)	31 (44.9)			139 (45.1)	60 (39.0)	79 (51.3)
Smokers	63 (45.7)	33 (47.8)	30 (43.5)			130 (42.2)	71 (46.1)	59 (38.3)
Former smokers	14 (10.1)	6 (8.7)	8 (11.6)			39 (12.7)	23 (14.9)	16 (10.4)
Drinking, n (%)				1.697	0.428				0.515	0.773
No-drinkers	58 (42.0)	28 (40.6)	30 (43.5)			132 (42.9)	69 (44.8)	63 (40.9)
Drinkers	73 (52.9)	39 (56.5)	34 (49.3)			165 (53.6)	80 (51.9)	85 (55.2)
Former drinkers	7 (5.1)	2 (2.9)	5 (7.2)			11 (3.6)	5 (3.2)	6 (3.9)
Sleep duration, n (%)				13.534	0.001				1.567	0.457
≤ 5 h	11 (8.0)	1 (1.4)	10 (14.5)			31 (10.1)	14 (9.1)	17 (11.0)
5–7 h	81 (58.7)	37 (53.6)	44 (63.8)			177 (57.5)	85 (55.2)	92 (59.7)
> 7 h	46 (33.3)	31 (44.9)	15 (21.7)			100 (32.5)	55 (35.7)	45 (29.2)
SBP (mmHg)	128 (121, 143)	133 (123, 146)	127 (119.5, 136)	− 1.798	0.072	128 (119, 137)	127 (119, 139)	128 (119, 136)	− 0.620	0.535
DBP (mmHg)	79 (72, 86)	82 (74, 88)	78 (72, 84)	− 1.530	0.126	77 (71, 84)	79 (71, 86)	76 (71, 83)	− 1.271	0.204
BMI (kg/m2)	26.9 (24.92, 29.44)	27.47 (25.29, 29.75)	26.54 (24.81, 28.78)	− 1.177	0.239	26.36 (24.35, 29.31)	26.41 (24.27, 29.39)	26.35 (24.38, 29.08)	− 0.089	0.929
WHR	0.93 (0.88, 0.97)	0.95 (0.90, 0.99)	0.91 (0.88, 0.95)	− 3.160	0.002	0.91 (0.88, 0.95)	0.91 (0.88, 0.95)	0.91 (0.88, 0.95)	− 0.851	0.395
TC (mmol/L)	5.18 ± 0.97	5.34 ± 0.92	5.01 ± 1.00	− 2.048	0.042	5.05 ± 0.93	5.04 ± 0.83	5.06 ± 1.03	0.189	0.850
TGs (mmol/L)	1.99 (1.27, 2.63)	2.33 (1.56, 3.13)	1.63 (1.06, 2.29)	− 3.109	0.002	1.65 (1.15, 2.29)	1.79 (1.29, 2.56)	1.56 (1.05, 2.13)	− 2.576	0.010
HDL (mmol/L)	1.18 (1.03, 1.37)	1.13 (1.03, 1.36)	1.18 (1.04, 1.38)	− 0.941	0.347	1.16 (1.03, 1.36)	1.15 (1.02, 1.30)	1.23 (1.05, 1.39)	− 2.375	0.018
LDL (mmol/L)	3.34 ± 0.81	3.45 ± 0.83	3.24 ± 0.79	− 1.573	0.118	3.31 ± 0.78	3.32 ± 0.70	3.29 ± 0.85	− 0.329	0.742
FPG (mmol/L)	6.25 (5.61, 7.95)	7.94 (7.18, 9.54)	5.65 (5.36, 5.92)	− 9.701	< 0.001	5.85 (5.45, 6.37)	6.32 (5.77, 6.66)	5.61 (5.36, 5.91)	− 8.636	< 0.001
PBG (mmol/L)	9.07 (6.79, 14.64)	14.63 (12.15, 17.18)	6.84 (5.72, 7.83)	− 9.877	< 0.001	7.32 ± 1.69	7.85 ± 1.76	6.78 ± 1.44	− 5.843	< 0.001
HbA1c (%)	5.75 (5.48, 6.80)	6.80 (6.20, 8.05)	5.50 (5.25, 5.60)	− 9.474	< 0.001	5.60 (5.40, 5.80)	5.80 (5.53, 6.00)	5.50 (5.30, 5.60)	− 7.687	< 0.001
ALT (ukat/L)	0.39 (0.29, 0.56)	0.43 (0.30, 0.70)	0.35 (0.25, 0.51)	− 2.244	0.025	0.42 (0.29, 0.70)	0.44 (0.30, 0.74)	0.41 (0.28, 0.62)	− 1.585	0.113
AST (ukat/L)	0.31 (0.28, 0.40)	0.31 (0.26, 0.44)	0.32 (0.29, 0.39)	− 0.490	0.624	0.35 (0.30, 0.45)	0.36 (0.30, 0.49)	0.34 (0.30, 0.43)	− 1.429	0.153
Cr (μmol/L)	75.50 ± 12.73	73.88 ± 11.71	77.12 ± 13.57	1.501	0.136	77.19 ± 12.75	77.08 ± 12.40	77.29 ± 13.14	0.148	0.882
UA (μmol/L)	364.25 (318.75, 431.93)	359.20 (320.45, 419.15)	380.20 (317.70, 443.20)	− 0.826	0.409	384.50 (324.82, 437.90)	389.80 (324.68, 464.32)	381.65 (325.52, 426.08)	− 1.195	0.232
SHS, n (%)									0.020	0.888
Yes	−	−	−	−	−	63 (20.5)	31 (20.1)	32 (20.8)
No	−	−	−	−	−	245 (79.5)	123 (79.9)	122 (79.2)
SHS score	−	−	−	−	−	20 (12, 33)	20 (11, 33)	21 (12, 32)	− 0.276	0.783
Fatigue score	−	−	−	−	−	10 (6, 14)	10 (7, 14)	10 (6, 14)	− 0.456	0.649
Cardiovascular system score	−	−	−	−	−	1 (0, 3)	0 (0, 3)	1 (0, 3)	− 0.201	0.841
Digestive tract score	−	−	−	−	−	1 (0, 3)	1 (0, 3)	2 (0, 3)	− 0.116	0.907
Immune system score	−	−	−	−	−	2 (1, 4)	2 (1, 4)	2 (1, 4)	− 0.178	0.859
Mental status score	−	−	−	−	−	6 (2, 10)	6 (2, 10)	6 (2, 9)	− 0.005	0.996

Data was shown as means ± SD, medians (interquartile ranges), or frequencies (percentages)

P < 0.05 was considered statistical significance

If the continuous variables were normally distributed, the t test was used to analyze the differences between the groups. Otherwise, the Mann–Whitney U test was used. χ² test analysis was conducted to test categorical variables

Transverse lines represented data unavailable

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; Cr, creatinine; DBP, diastolic blood pressure; EUR, Euro; FPG, fasting plasma glucose; HbA1c, glycosylated hemoglobin A1c; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; PBG, two-h postprandial blood glucose; SBP, systolic blood pressure; SD, standard deviation; SHS, suboptimal health status; T2DM, type 2 diabetes mellitus; TC, total cholesterol; TGs, total triglycerides; WHR, waist-to-hip ratio; UA, uric acid; ¥, Chinese Yuan; USD, United States dollar

In detail, in the case–control study, the median age of the participants was 51 years, with 81.1% being male, the majority of participants being married (92.8%) Han Chinese (97.8%), 64.5% having a family income per capita between Chinese Yuan (CNY) ¥3501–8000 (approximately 503–1151 Euro (EUR), 502–1148 United States dollar (USD)) monthly, and approximately 51.4% having a bachelor’s degree or above, but these indicators were not significantly different between T2DM and healthy controls. However, lifestyle factors and anthropometric characteristics, including sleep duration and WHR, as well as biochemical indexes, including TC, TGs, FPG, PBG, HbA1c, and ALT, were significantly higher in T2DM patients than in healthy controls (Table 1).

Furthermore, the differences in participant characteristics in the nested case–control study between T2DM and healthy controls were broadly similar to those in the case–control study. Of the 308 participants included in the nested case–control study, approximately 20.5% were SHS at baseline. The median age of the participants was 47 years, with 82.8% being male. The majority of individuals (65.6%) had a household income per capita monthly between CNY ¥3501–8000 (approximately 503–1151 EUR, 502–1148 USD), most of the participants were married (90.3%) with Han Chinese (96.8%), and approximately 59.7% of participants had a bachelor’s degree or above. The results showed that there were statistically significant differences in biochemical indexes, including TGs, HDL, FPG, PBG, and HbA1c, between the T2DM and healthy control groups (P < 0.05), whereas the differences in demographic, anthropometric, and lifestyle characteristics were not statistically significant (all P > 0.05) (Table 1). Table S2 summarizes the baseline characteristics of the study participants stratified by SHS.

Alterations in IgG N-Glycome composition in T2DM

In total, IgG N-glycome compositions (including 24 GPs and 54 DGs) were analyzed in all 138 and 308 samples included in the case–control and nested case–control settings, respectively. Out of a total of 78 IgG N-glycan traits, 59 and 73 IgG N-glycan traits were found to present abnormal distribution (P_{Kolmogorov–Smirnov} < 0.05) in the participants from the case‒control and nested case‒control settings, respectively (Table S1). Additionally, according to univariate analysis, the relative abundance of 44 glycan compositions (including 12 GPs and 32 DGs) were significantly different between T2DM and healthy controls for the case‒control study (Fig. 1a and Table S3). Furthermore, the differences in IgG N-glycan traits between T2DM and healthy controls for the nested case–control study were not statistically significant. However, after stratifying by SHS groups, IgG N-glycans with differences in compositions between T2DM and healthy controls were found in both SHS and optimal health participants. Specifically, the compositions of 5 glycan traits (GP9, GP22, DG2, DG6, and DG9) were different between T2DM and healthy controls in SHS individuals at baseline, and those of 3 glycan traits (GP7, DG12, and DG23) were different between the two groups in optimal healthy individuals at baseline (Fig. 1b and Table S3).

Fig. 1.

Comparisons of the relative abundance of IgG N-glycans between the T2DM and healthy controls. a The relative abundance of directly measured glycans in the T2DM and healthy control groups for the case–control setting. b The relative abundance of glycans in the T2DM and healthy control groups in the total participants, baseline SHS, and baseline OH participants of the nested case–control setting. Each box represents the 25th to 75th percentiles (interquartile range [IQR]). Lines inside the boxes represent the medians. GP, glycan peak; DG, derived glycan; IgG, immunoglobulin G; OH, optimal health; SHS, suboptimal health status; T2DM, type 2 diabetes mellitus

Association and complexity of IgG N-glycans with T2DM

After adjusting for potential confounders [14, 44–47], we further examined differences in glycan abundances and evaluated the association of T2DM with IgG N-glycosylation by multivariable logistic regression analysis in the case–control and nested case–control settings. Adjustment for the effect of sleep duration, WHR, TC, and TGs, comparison of T2DM against their age- and sex-matched controls showed that 22 (including 7 GPs and 15 DGs) out of 78 IgG N-glycan traits were significantly different between the groups in the case–control study (Fig. 2a and Table S4). Notably, when further controlling for the false discovery rate of 5%, we found that there were still 11 statistically significant IgG N-glycan glycans (increased relative abundance of GP1, GP20, GP22, DG7, DG9, DG18, DG46, DG51, and DG52, as well as reduced relative abundance of DG10 and DG42) between the T2DM and healthy control groups (Fig. 2b and Table S4).

Fig. 2.

The associations between IgG N-glycan traits and T2DM. AORs and 95% CIs for the association of IgG N-glycan traits with T2DM versus healthy controls adjusted for the covariates (logistic regression adjusted for sleep duration, WHR, TC, and TGs in the case–control study (a); logistic regression adjusted for TC, LDL, HDL, TGs, BMI, and/or SHS in the nested case–control study (b)). P-adjusted* < 0.05 was considered statistically significant. #Statistically significant results adjusting the aforementioned covariates and further controlling false discovery rate at 5% based on Benjamini–Hochberg method in logistic regression analysis. AOR, adjusted odds ratio; BMI, body mass index; CI, confidence interval; DG, derived glycan; GP, glycan peak; HDL, high-density lipoprotein cholesterol; IgG, immunoglobulin G; LDL, low-density lipoprotein cholesterol; TC, total cholesterol; TGs, total triglycerides; SHS, suboptimal health status; T2DM, type 2 diabetes mellitus; WHR, waist-to-hip ratio

In the nested case–control study, adjusting for the potential confounders (including TC, LDL, HDL, TGs, BMI, and SHS), the results of multivariable logistic regression showed that there were no associations between IgG N-glycan traits and T2DM. Interestingly, when the subjects were stratified according to the existence of SHS at baseline before multivariable analysis, differences in the relative abundance of IgG N-glycans were found between participants with follow-up new-onset T2DM and age- and sex-matched controls in baseline SHS and optimal health populations (Table S4). Furthermore, in SHS individuals, the relative abundance of GP22, DG2, and DG9 in participants with T2DM was significantly higher than that in healthy controls, whereas a reduced relative abundance of GP9 and DG25 was found in participants with T2DM. In the optimal healthy individuals, the levels of three glycan traits were significantly associated with T2DM, with increased relative abundance of DG12 and decreased relative abundance of GP7 and DG23 in the T2DM compared with healthy individuals (Fig. 2b). Notably, the relative abundance of GP22 and DG9 in T2DM showed an increasing trend in both case–control and nested case–control (baseline SHS individuals) settings.

IgG N-glycans and SHS interaction analysis based on the association of IgG N-glycans with T2DM in the nested case‒control study

The results of the nested case–control study also provided evidence of IgG N-glycan-by-SHS interactions. Differences in the relative abundance of glycans persisted after including the glycan-by-SHS interaction term in the logistic regression analysis between the T2DM and healthy control groups. In detail, the results showed that no additive interaction of IgG N-glycans and SHS was found on the risk of developing T2DM (all the 95% CI of RERI and AP containing 0, and the 95% CI of S involving 1) (Table S5). However, multiplicative models revealed significant interaction effects between glycan traits (GP22, DG9, and DG33) and SHS in the association between IgG N-glycans and T2DM (Table S5). Notably, the presence of a positive multiplicative interaction of GP22 (OR: 1.039, 95% CI: 1.016–1.066, P_interaction = 0.002) and DG9 (OR: 1.020, 95% CI: 1.002–1.039, P_interaction = 0.033) with SHS and a negative multiplicative interaction of DG33 (OR: 0.982, 95% CI: 0.963–0.999, P_interaction = 0.046) with SHS were observed.

Correlation between IgG N-glycans and T2DM-related clinical traits

Univariate correlation analysis through Spearman’s rank correlation method was performed to evaluate the relationship between IgG N-glycans and a series of T2DM-associated clinical traits in both case–control and nested case–control studies. The heatmap depicted that the correlation coefficients between 24 directly measured IgG N-glycan profiles and 15 T2DM-associated clinical traits (including BMI, SBP, DBP, WHR, HbA1c, FPG, TC, TGs, LDL, HDL, PBG, ALT, AST, Cr, and UA) were calculated in the case–control study and included SHS on top of the aforementioned 15 clinical traits in the nested case–control study. Notably, 18 and 19 GPs were correlated with 7 and 13 clinical traits in the case–control and nested case–control studies, respectively (Fig. 3). The more detailed correlations of 24 GPs and 54 DGs with T2DM-related clinical risk factors are presented in Table S6.

Fig. 3.

The correlation between directly measured IgG N-glycans and T2DM-related clinical traits. Univariate correlation analysis was performed to evaluate the relationship between IgG N-glycan profiles and a series of clinical traits by Spearman’s rank correlation method in the case–control (a) and nested case–control studies (b). The positive correlations are represented by red, while negative correlations are represented by blue. One, two, and three asterisks indicate that the P of the correlation coefficient is less than 0.05, 0.01, and 0.001, respectively. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; Cr, creatinine; DBP, diastolic blood pressure; FPG, fasting plasma glucose; GP, glycan peak; HbA1c, glycosylated hemoglobin A1c; HDL, high-density lipoprotein cholesterol; IgG, immunoglobulin G; LDL, low-density lipoprotein cholesterol; PBG, two-hour postprandial blood glucose; SBP, systolic blood pressure; SHS, suboptimal health status; TC, total cholesterol; TGs, total triglycerides; WHR, waist-to-hip ratio; UA, uric acid

Additionally, the results of multivariate correlation analysis, i.e., CCA, showed that only the first canonical set was statistically significant, with a canonical correlation of 0.722 (F = 1.18, P = 0.0238) in the case–control study and 0.520 (F = 1.22, P = 0.0036) in the nested case–control study. Specifically, four GPs, namely, GP24, GP18, GP19, and GP21, were significantly correlated with TGs in the first canonical set for the case–control study (Fig. 4a). The aforementioned four GPs were significantly correlated with canonical variables with loadings of − 0.590, − 0.456, − 0.443, and − 0.403, respectively, while the response variable with the highest canonical loading was − 0.422 (TGs). Furthermore, five glycans (involving GP7, GP8, GP12, GP20, and GP13) were associated with four T2DM-related clinical risk factors (including DBP, SBP, HbA1c, and WHR) in the first canonical set for the nested case‒control study, and the details corresponding to canonical loadings are presented in Fig. 4b. The canonical structure of the first pair of significant canonical variants in both case–control and nested case–control settings suggested that IgG N-glycan profiles were associated with T2DM-related clinical traits.

Fig. 4.

Canonical structures of directly measured IgG N-glycans and the clinical traits in the first canonical set for the case‒control (a) and nested case‒control settings (b). The absolute value of canonical loadings greater than 0.30 was significant. All the variables are sorted by the absolute value of their canonical loadings. The positive relationships are represented in black boxes, while negative relationships are shown in red boxes. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; Cr, creatinine; DBP, diastolic blood pressure; FPG, fasting plasma glucose; GP, glycan peak; HbA1c, glycosylated hemoglobin A1c; HDL, high-density lipoprotein cholesterol; IgG, immunoglobulin G; LDL, low-density lipoprotein cholesterol; PBG, two-hour postprandial blood glucose; SBP, systolic blood pressure; SHS, suboptimal health status; T2DM, type 2 diabetes mellitus; TC, total cholesterol; TGs, total triglycerides; WHR, waist-to-hip ratio; UA, uric acid

Discriminability of IgG N-glycome with respect to T2DM

Three types of classification models, namely, the clinical trait model (model 1), glycan model (model 2), and combined model (model 3), were fitted based on logistic regression classifiers in both case–control and nested case–control settings (Fig. 5 and Table 2). Specifically, in the case–control setting, the AUCs of model 2 and model 3 distinguishing T2DM from healthy individuals were 0.829 and 0.891, respectively, which was generally better than that of model 1 (Fig. 5a and Table 2). Significant improvement of the discriminative performance was observed in model 3 compared with model 1 (difference between AUCs (95% CI): 0.127 (0.053, 0.199)) (Table S7). In the nested case–control setting, the performance of models was presented in total participants and population stratification by SHS. The discriminative ability of the models on T2DM in the baseline SHS participants from the nested case–control study showed a similar trend as that in the case–control study, with AUCs of 0.648, 0.753, and 0.787 in model 1, model 2, and model 3, respectively (Fig. 5c). The AUCs showed significant improvement in model 3 compared to model 1, with an increment of 0.139 (95% CI: 0.004–0.274) (Table S7). The AUCs of the three models differentiating T2DM from healthy individuals were 0.627, 0.591, and 0.664 in the total participants and 0.646, 0.590, and 0.680 in the baseline optimal health status participants based on the nested–case control study, respectively (Fig. 5b, d and Table 2).

Fig. 5.

ROC curves illustrated the performance of the models in the classification of T2DM from healthy individuals. The classification models of glycans and/or clinical traits distinguishing T2DM from healthy individuals for the case–control setting (a); the classification models of glycans and/or clinical traits distinguishing T2DM from healthy individuals in the total participants (b), baseline SHS (c), and baseline optimal health participants (d) of the nested case–control setting. AUC, area under the receiver operating characteristic curve; CI, confidence interval; IgG, immunoglobulin G; T2DM, type 2 diabetes mellitus; ROC, receiver operating characteristic curve

Table 2.

The discriminative performance of models for T2DM in the case–control and the nested case–control studies

Models	Discriminative model	fivefold cross-validation	Repeated 400 times for fivefold cross-validation
	AUC (95% CI)	Average AUC	Average AUC
Case–control study
Model 1 (including WHR, TC, TGs, and sleep duration)	0.764 (0.684–0.844)	0.753	0.734
Model 2 (including GP9, GP20, GP22, DG6, DG7, DG10, DG33, DG36, and DG52)	0.829 (0.744–0.884)	0.793	0.775
Model 3 (including WHR, TC, TGs, sleep duration, GP9, GP20, GP22, DG6, DG7, DG10, DG33, DG36, and DG52)	0.891 (0.840–0.943)	0.828	0.807
Nested case–control study—total
Model 1 (including TC, TGs, HDL, LDL, BMI, WHR, and SHS)	0.627 (0.565–0.689)	0.569	0.562
Model 2 (including GP7, GP9, GP20, DG2, DG6, DG12, and DG15)	0.591 (0.528–0.655)	0.566	0.545
Model 3 (including TC, TGs, HDL, LDL, BMI, WHR, SHS, GP7, GP9, GP20, DG2, DG6, DG12, and DG15)	0.664 (0.603–0.725)	0.540	0.563
Nested case–control study—baseline SHS
Model 1 (including TC, TGs, HDL, LDL, BMI, and WHR)	0.648 (0.510–0.786)	0.616	0.615
Model 2 (including GP9, GP22, DG6, and DG9)	0.753 (0.631–0.875)	0.700	0.696
Model 3 (including TC, TGs, HDL, LDL, BMI, WHR, GP9, GP22, DG6, and DG9)	0.787 (0.673–0.902)	0.622	0.645
Nested case–control study—baseline optimal health
Model 1 (including TC, TGs, HDL, LDL, BMI, and WHR)	0.646 (0.576–0.715)	0.583	0.586
Model 2 (including GP7, DG12, and DG23)	0.590 (0.518–0.661)	0.525	0.569
Model 3 (including TC, TGs, HDL, LDL, BMI, WHR, GP7, DG12, and DG23)	0.680 (0.612–0.747)	0.591	0.604

AUC, area under the curve; BMI, body mass index; CI, confidence interval; DG, derived glycan; GP, glycan peak; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; SE, standard error; SHS, suboptimal health status; T2DM, type 2 diabetes mellitus; TC, total cholesterol; TGs, total triglycerides; WHR, waist-to-hip ratio

Model 1 and model 2 represent the model including clinical traits or glycans alone, respectively. Model 3 represents the model incorporating the clinical traits and glycans at the same time

Moreover, the results of the combined models based on fivefold cross-validation showed average AUCs of 0.828, 0.540, 0.622, and 0.591 for differentiating T2DM patients from healthy controls in the case–control and nested case–control settings (including 308 total samples, 63 baseline SHS samples, and 254 baseline optimal health samples), respectively. Similarly, the average AUCs of the combined models based on 400 repeated fivefold cross-validation differentiating T2DM from healthy individuals were 0.807 in the case–control setting and 0.563, 0.645, and 0.604 in the total samples, baseline SHS, and baseline optimal health samples of the nested case–control setting, respectively, which presented moderate discriminative ability and were generally better than models with either glycans or clinical features alone (Table 2). The average AUCs of models cross-validation discriminating T2DM from healthy controls generally decreased slightly but tended to be stable, indicating that the classification ability of the models is relatively reliable in both case‒control and nested case‒control settings. Notably, the results showed that the discriminability of model 3 (incorporating glycans and clinical traits) was generally better than that of model 1 and model 2 in both case‒control and nested case‒control settings. Although the differences between the AUCs of model 3 and the other models (model 1 and model 2) were not all statistically significant, the relatively superior classification performance of model 3 indicated that the IgG N-glycan features could enhance the discriminative ability of the models (Table S7).

Discussion

In this study, the association and discriminative potential of IgG N-glycan profiles with T2DM were investigated based on combining evidence from case–control and prospective nested case–control settings in the Chinese population. In addition, we further evaluated whether the previous SHS of T2DM patients would affect the level of IgG N-glycosylation in the nested case–control study. The IgG N-glycome composition of T2DM and age- and sex-matched healthy individuals was comprehensively analyzed based on the HILIC-UPLC measurement method. The PPPM paradigm emphasizes early risk assessment of people with high risk of diseases to achieve early prediction and prevention of diseases and to prevent the progression of diseases through personalized treatment. Furthermore, the framework of PPPM presented that the identification and development of cost-effective and reliable biomarkers, such as IgG N-glycans, are important for the prediction and prevention of diseases. This study further confirmed that altered IgG N-glycans could be detected during the progression of T2DM, which may be a promising potentiality of clinical application in the discrimination and prevention of T2DM from the perspective of PPPM.

The alteration of IgG N-glycosylation in T2DM

IgG antibodies contain complex-type biantennary N-glycan structures, and the effect of N-glycans can activate or inhibit the binding affinity of IgG toward diverse Fc receptor family members (i.e., fragment crystallizable γ receptors (FcγRs)), therefore modulating the function of IgG effectors [48]. Alteration of galactosylation, fucosylation, sialylation, and bisecting GlcNAc in IgG glycan compositions can regulate the function of the pro- and anti-inflammatory activities of IgG [49]. Studies have demonstrated that altered glycosylation of IgG is involved in many specific health conditions related to T2DM, including aging, obesity, hypertension, dyslipidemia, and metabolic syndrome [24, 29, 30, 50, 51]. Furthermore, the IgG N-glycome can reflect the intricate interplay of genetics, epigenetics, posttranslational modifications, environmental factors, etc. [52, 53], which promises attractive biomarkers for multifactorial diseases. Previous case–control studies showed that the alteration of IgG N-glycosylation was associated with T2DM in the Australian, European, and Chinese populations, suggesting that the IgG N-glycome might have potential as a biomarker for T2DM [34–37]. The current case–control and prospective nested case–control study further provides suggestive ideas and evidence for the early discrimination of T2DM using IgG N-glycome.

Only two (GP22 and DG9) IgG N-glycans that differed in the case–control setting were replicated in the nested case–control setting (baseline SHS individuals). Although the changes in GPs and DGs in patients with T2DM for the case–control study were not completely replicated in the nested case–controls, a generally consistent trend in altered IgG N-glycosylation was identified in T2DM in both settings. Specifically, the present results indicate that the individuals with increased agalactosylated glycans (GP1), sialylated/fucosylated structures with bisecting GlcNAc of IgG (GP22, DG9, DG46, DG51, and DG52), and fucosylation of digalactosylated glycan with sialic acid without bisecting GlcNAc (GP20 and DG7) and decreased galactosylation, fucosylation, and sialylation without bisecting GlcNAc of IgG (DG10 and DG42) may increase the risk for T2DM in the case–control setting. Furthermore, T2DM with baseline SHS participants from nested case–control study presented increased sialylation of fucosylated galactosylated structures with bisecting GlcNAc (GP22, DG2, and DG9), but decreased monogalactosylated core fucosylated glycan (GP9), when compared with healthy individuals; increased sialylated structures with bisecting GlcNAc (DG12) and decreased monogalactosylated glycan (GP7) were observed in T2DM with baseline optimal health status individuals, which presented basically consistent evidence with previous studies on IgG N-glycome and multiple health events [24, 30, 31, 50]. Additionally, the association of IgG N-glycans with most T2DM clinical risk factors was also identified, partly reflecting the influence of these clinical traits, which was largely in line with reports on IgG glycosylation in these clinical traits [24, 50, 54].

The relationship between changes in IgG N-glycosylation and inflammation

Although the relationship between the downregulation of IgG terminal galactose content and disease development is still under investigation, the association of potential existence seems to confirm that IgG galactosylation is one of the most prominent glycosylation alterations in many inflammatory diseases [55]. The present study showed that an increase in agalactosylated (GP1) and a decrease in monogalactosylated biantennary glycan structures (GP7 and GP9) were associated with a higher risk of T2DM. Agalactosylated IgG glycomics are able to trigger the complement system by binding to the mannose-binding lectin pathway and thus initiate an inflammatory cascade [56]. By binding to the inhibitory receptor FcγRIIB and the lectin-like receptor dectin-1, highly galactosylated IgG immune complexes suppress the inflammatory cascade and the subsequent inhibition of C5a-induced inflammation, which blocks pro-inflammatory effector functions and enhances anti-inflammatory properties [57], highlighting the complexity of the modulation of IgG effector functions by differential glycosylation. Importantly, the loss of galactose in IgG is known to act in a general pro-inflammatory humoral immune response associated with decreased immunosuppressive potential of circulating IgG, which might be the result of decreased activity of posttranslational modifications of the enzyme β4-galactosyltransferase-1 in IgG-producing plasma B cells [58]. Although the exact effect of IgG galactosylation and its influence on disease activity are not yet fully understood, decreased IgG galactosylation in T2DM patients may be a common feature of inflammatory diseases and could serve as early signals of altered glycosylation modifications during inflammation.

Most IgG N-glycans contain core fucose that impacts the recruitment of inflammatory effector cell responses [59]. IgG core fucose acts as a “safety switch,” which can reduce the affinity for the FcγRIIIA and FcγRIIIB receptors and significantly decrease antibody-dependent cell-mediated cytotoxicity (ADCC) [60]. In contrast, fucosylation in the context of bisecting GlcNac significantly enhanced ADCC. As the central branch of N-glycans, bisecting GlcNAc is biosynthesized by the glycosyltransferase GnT-III encoded by the MGAT3 gene family, which can affect the major structural diversity of N-glycans and modulate the function of target glycoproteins [61]. Our findings showed that increased fucosylation with bisecting GlcNAc (GP22, DG46, DG51, and DG52) and decreased fucosylation without bisecting GlcNAc (GP9 and DG42) could increase the risk of T2DM. Bisecting GlcNAc inhibits the modification of fucose of terminal epitopes in N-glycans, which results in the level of fucosylation presenting inconsistent trends in T2DM patients. The results further indicated that an increased relative abundance of bisecting GlcNAc and fucosylated bisected IgG could implicate a pro-inflammatory IgG effector function [61], thereby leading to the progression of healthy status or SHS toward T2DM.

Because galactosylated IgG is the substrate for sialyltransferases, the high level of IgG sialylation generally results from increased galactosylation [62]. Similar to the effect of IgG galactosylation, reduced sialylation is also typical for inflammatory conditions. Terminal sialic acid residues of IgG N-glycans can also reduce ADCC and possess anti-inflammatory properties by upregulating inhibitory FcγRIIB on effector macrophages [63, 64]. Interestingly, an inconsistent trend of altered sialylation was found in the T2DM individuals. Specifically, sialylation with bisecting GlcNAc (GP22, DG2, DG9, and DG12) showed increased relative abundance, but decreased sialylated glycan without bisecting GlcNAc (DG10) was found in T2DM participants, which was consistent with the suppression of fucosylation expression due to the addition of bisecting GlcNAc. Bisecting GlcNAc has been proven to be a general suppressor of terminal modifications (i.e., fucose, sialic acids, and others) of N-glycans, which may be a major physiological function of bisecting GlcNAc [61]. In addition, increased fucosylated and galactosylated sialylation (GP20 and DG7) was also present in T2DM participants in the case‒control setting. Notwithstanding the seemingly opposing claims, a possible explanation for this finding is that the discordant alteration of sialylation in T2DM may be the result of dynamic changes in terminal sialic acids in regulating the inflammatory state shifting from pro-inflammatory to anti-inflammatory status, which is a critical state.

Stratified analysis of the association of IgG N-glycans with T2DM based on baseline SHS subgroups

Furthermore, in the nested case–control setting, although there were no differences in IgG N-glycans between T2DM and healthy individuals in the total population, the results of stratification by baseline SHS revealed that the changes in IgG N-glycans were more pronounced in participants with new-onset T2DM with baseline SHS than T2DM with baseline healthy status. In addition, the multiplicative interaction between SHS and glycans also indicated the potential role of SHS in the development of T2DM based on the evidence of a nested case‒control study. From the paradigm of PPPM, these findings suggest that SHS is an important window period of T2DM and the early detection and management of SHS combined with IgG glycosylation for early identification, which may provide important evidence for the early prediction and prevention of T2DM. The time frame between the onset of SHS and the clinical manifestation of T2DM is the operational area for the application of reliable risk assessment tools and predictive diagnostics followed by cost-effective targeted prevention and treatments tailored to the person [13].

The ability of discriminant models to distinguish T2DM

In fact, the clinical trait models, glycan models, and combined models presented a certain discriminative ability to distinguish T2DM in the case–control and nested case–control settings. In addition, the results of the fivefold cross-validation and repetition 400 times fivefold cross-validation of models showed that the changes in the average AUCs of cross-validation compared with the point estimate of discriminative models of AUCs were not obvious, indicating that the T2DM discrimination models established in this study were relatively stable and had good credibility. It is worth noting that the models based on glycan traits or clinical traits alone have a certain discriminative ability for T2DM in both settings, but the combined models incorporating glycan traits and clinical traits could effectively explain the data for distinguishing T2DM and be superior to glycan models and clinical trait models. These results suggest that dynamic IgG N-glycan profiles might serve as biomarkers or functional effectors of the disease, potentially increasing the diagnostic accuracy for complex diseases, such as T2DM. High-throughput detection technology makes it possible to rapidly detect IgG N-glycomics in clinical laboratories. With the continuous development and optimization of high-throughput detection technology, the detection time and cost of IgG N-glycomics will be further reduced, which makes it possible to realize the clinical exploitation of IgG N-glycan biomarkers. IgG N-glycomics, a potential inflammatory biomarker, are expected to be developed into an effective and routine clinical auxiliary diagnostic and prophylactic test for T2DM patients.

Strength and limitations

The strength of this study is that IgG N-glycomes in relation to T2DM were evaluated, and altered IgG N-glycan profiles based on the combined evidence from case–control and nested case–control settings were explored; in particular, the effects of baseline SHS on IgG N-glycosylation levels in patients with new-onset T2DM were evaluated in a nested case–control setting. Furthermore, the design of nested case–control study makes use of the advantages of cohort studies, with good homogeneity and comparability of subjects, which can effectively control selection bias and recall bias and obtain higher credibility of the temporality of associations. The changes in IgG N-glycosylation in T2DM observed in this study indicate the possibility of IgG N-glycans as biomarkers for the early discrimination of T2DM.

However, these suggestions are built upon the limitations of this study, and the results should be interpreted with caution. Firstly, although the nested case–control study has some advantages over the case–control study causal inference, the evidence provided by these two designs still makes it difficult to infer a causal association between altered IgG N-glycosylation and the progression of T2DM. Secondly, the results of the nested case–control setting found fewer altered glycan profiles than those in the case–control setting, which may be because the case‒control setting usually leads to overestimation. Thirdly, the role of SHS was observed in the association between IgG N-glycans and T2DM in the nested case–control setting; however, the relatively small sample size of participants with SHS (n = 63) at baseline suggests caution in interpreting the effect of SHS in the development of T2DM. Fourth, to limit overfitting, the models were trained and evaluated by the method of internal validation, namely, fivefold cross-validation and repetition of 400 times fivefold cross-validation. External validation results should be added in future studies to further validate the stability of the model. Finally, IgG N-glycomics were detected by the HILIC-UPLC method in the present study, which focuses on the relative abundance of total IgG N-glycans. The differences in IgG Fc subclass glycan profiles associated with T2DM remain to be investigated by liquid chromatography–mass spectrometry (LC–MS), which can provide more information on the function of IgG effectors regulated by glycosylation of different IgG subclasses.

Conclusions and expert recommendations

This study comprehensively illustrated that the observed altered IgG N-glycosylation, i.e., decreased galactosylation and fucosylation/sialylation without bisecting GlcNAc, as well as increased agalactosylation and fucosylation/sialylation with bisecting GlcNAc, reflects a pro-inflammatory state of T2DM, which provides the possibility of IgG N-glycans as biomarkers for the early discrimination of T2DM by combining the evidence of case‒control and nested case‒control studies. Additionally, the results of stratification by baseline SHS revealed that altered IgG N-glycans were more pronounced in participants with new-onset T2DM with baseline SHS than in T2DM with baseline healthy status based on the nested case‒control setting. The study suggested that the combination of the assessment of SHS and characteristics of IgG N-glycan profiles could provide more information for early identification of the progression of T2DM and prediction of T2DM risk in the general population (predictive medical approaches).

SHS, as the early and reversible phase of T2DM, is an important window period of early intervention for individuals at risk for T2DM, providing new insights for the early predictive medical approach of T2DM risk in the framework of the PPPM health care system [65]. Furthermore, the assessment of SHS has the advantages of being fast, convenient, acceptable, and affordable, which is consistent with the most inclusive, equitable, and cost-effective characteristics of universal health coverage, the focus of primary health care. From the perspective of PPPM, screening of SHS is helpful for the early identification of high-risk groups of T2DM to achieve early targeted prevention of high-risk groups of T2DM. The health care providers should be aware of the importance of early management of SHS for NCDs (including T2DM) prevention and intervention and focus on the T2DM risk for individuals with SHS to provide appropriate health management advice and to contribute to health quality promotion (targeted prevention).

Glycomic biosignatures, as dynamic biomarkers reflecting the complex inflammatory process, have the ability to identify populations at risk for T2DM early or provide predictive information on the prognosis of adverse events. The combination of these macroscopic and microscopic techniques could provide suggestive ideas, evidence, and valuable insight for the PPPM of T2DM. In detail, from the viewpoint of PPPM, understanding the risk factors and identifying biomarkers of clinical progression for T2DM including any pathogenic pathways will improve the chances of developing effective therapeutics (personalized medicine). At the same time, the continuous optimization of high-throughput detection technology makes it possible to detect IgG N-glycans in clinical laboratories. Early identification of people with SHS combined with dynamic monitoring of body status by IgG N-glycans may contribute to defining high-risk groups of T2DM at an early stage, which conforms to the most active and proactive claim of primary prevention and could effectively promote the paradigm shift of T2DM management from reactive medicine to advanced methods, namely, PPPM. Future prospective cohort studies or Mendelian randomization studies further exploring the effect of glycomics (total IgG glycosylation and/or IgG subclasses Fc glycosylation) on T2DM are expected to provide a more comprehensive understanding of specific glycomic biomarkers that may serve as tools for personalized monitoring or to help prevent the progression from healthy status or SHS toward T2DM.

Supplementary information

Below is the link to the electronic supplementary material.

Abbreviations

ADCC

Antibody-dependent cell-mediated cytotoxicity

ALT

Alanine aminotransferase

AP

Attributable proportion due to interaction

AST

Aspartate aminotransferase

AUC

Area under the ROC curve

BH

Benjamini–Hochberg

BMI

Body mass index

CCA

Canonical correlation analysis

CNY

Chinese Yuan

Cr

Creatinine

CVD

Cardiovascular disease

DBP

Diastolic blood pressures

DG

Derived glycan

EUR

Euro

FcγRs

Fragment crystallizable γ receptors

FDR

False discovery rate

FPG

Fasting plasma glucose

GlcNAc

N-acetylglucosamine

GP

Glycan peak

HbA1c

Glycosylated hemoglobin A1c

HDL

High-density lipoprotein cholesterol

HILIC

Hydrophilic interaction liquid chromatography

IDF

International Diabetes Federation

IgG

Immunoglobulin G

LASSO

Least absolute shrinkage and selection operator

LC–MS

Liquid chromatography–mass spectrometry

LDL

Low-density lipoprotein cholesterol

NCDs

Non-communicable diseases

OGTT

Oral glucose tolerance tests

PBG

Postprandial blood glucose

RERI

Relative excess risk due to interaction

ROC

Receiver operating characteristic

SBP

Systolic blood pressures

SD

Standard deviation

SHS

Suboptimal health status

SHSQ-25

SHS questionnaire 25 items

S index

Synergy index

TC

Total cholesterol

TGs

Total triglycerides

T2DM

Type 2 diabetes mellitus

UA

Uric acid

UPLC

Ultra-performance liquid chromatography

USD

United States dollar

WHO

World Health Organization

WHR

Waist-to-hip ratio

2-AB

2-Aminobenzamide

3PM/PPPM

Predictive, preventive, and personalized medicine

Author contribution

QZ, WW, and YW contributed to the conception and design. XM, FW, XG, BW, and XX contributed to the acquisition and analysis of the data. XM and FW drafted the manuscript. All authors made important contributions to editing and critically revising the manuscript for important intellectual content. WW, QZ, and YW guarantee this work, have full access to all of the data, and take responsibility for the integrity of the data.

Funding

The study was supported by grants from the China-Australian Collaborative Grant (NSFC 81561128020-NHMRC APP1112767) and the National Natural Science Foundation of China (81872920). The funding organization had the role in the design and conduct of the study and the collection, management, analysis, and interpretation of the data.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Code availability

The software or software package used for the data analysis is indicated in the text, and the code used for the specific analysis can be obtained from the corresponding author.

Declarations

Ethics approval

Ethical approval was granted by the Ethics Committee of the Chinese PLA General Hospital (no. S2016-0681-01), according to the Helsinki Declaration.

Consent to participate

All subjects agreed to participate in the study and signed written informed consent.

Consent for publication

All authors have approved the final version to be published.

Conflict of interest

The authors declare no competing interests.