Serum Mac-2 Binding Protein Glycosylation Isomer Concentrations Are Associated With Incidence of Type 2 Diabetes

Abstract

Context

Serum Mac-2 binding protein glycosylation isomer (M2BPGi) concentrations are known to be an indicator of chronic liver injury and fibrosis.

Objective

This study aimed to investigate the association between serum M2BPGi concentrations and the development of type 2 diabetes in a Japanese community.

Methods

A total of 2143 community-dwelling Japanese individuals aged 40-79 years without diabetes at baseline were followed up for 7 years. Serum M2BPGi concentrations were divided into quintiles: Q1, ≤0.37 cutoff index (COI); Q2, 0.38-0.49 COI; Q3, 0.50-0.62 COI; Q4, 0.62-0.80 COI; and Q5, ≥0.81 COI. Cox proportional hazards models were used to estimate hazard ratios and 95% CIs for the development of type 2 diabetes.

Results

During the follow-up period, 219 individuals developed type 2 diabetes. The age- and sex-adjusted cumulative incidence of type 2 diabetes significantly increased with elevating serum M2BPGi levels (P for trend < .01). This association remained significant after adjustment for potential confounders (P for trend = .04). This significant association attenuated to a nonsignificant level after additionally adjusting for serum high-sensitivity C-reactive protein or homeostasis model assessment of insulin resistance.

Conclusion

The present study demonstrated that higher serum M2BPGi concentrations were significantly associated with higher risk of diabetes in a Japanese community. Moreover, inflammation and insulin resistance were suggested to contribute to the excess risk of diabetes in individuals with higher serum M2BPGi levels. These findings shed light on the importance of inflammation and insulin resistance when considering the pathogenesis of diabetes.

The number of individuals with type 2 diabetes is increasing worldwide, with the global diabetes prevalence estimated to become 10.2% (578 million) in 2030 and 10.9% (700 million) in 2045 (1). This is of utmost concern, because the growing burden of diabetes will increase the number of patients with chronic and acute diseases in general populations, with significant impacts on quality of life and demands on health services, and large economic costs (2). Insulin resistance is an essential component in the pathophysiology of type 2 diabetes, and the liver plays important roles in the regulation of insulin resistance (3, 4). Moreover, nonalcoholic fatty liver disease (NAFLD) has a close mutual interrelationship with type 2 diabetes (5, 6). Several previous prospective studies have reported that high levels of liver enzymes (7) and NAFLD as assessed by the fatty liver index (8) were associated with higher risk of incident diabetes. A meta-analysis has also shown that NAFLD was associated with greater risk of incident diabetes (9). In addition, hepatitis C virus (HCV) infection was reported to be associated with greater risk of incidence of diabetes (10, 11) and several studies have shown that successful treatment of HCV-infected individuals was associated with decreased incidence of type 2 diabetes (12, 13). Therefore, it would be worthwhile to investigate the influence of chronic liver diseases on the development of type 2 diabetes.

Over the last decade, immunoassay-based glycan “sugar chains” have attracted attention as a novel marker for liver fibrosis, and high concentrations of serum Mac-2 binding protein glycosylation isomer (M2BPGi) have been identified in patients with fibrosis (14). M2BPGi is secreted by hepatic stellate cells, and it functions as a messenger between these cells and Kupffer cells via Mac-2 (Galectin-3), which is secreted by hepatic stellate cells during fibrosis progression (15). Several studies have revealed a significant association between higher serum M2BPGi and the severity of liver fibrosis in chronic liver diseases, such as chronic hepatitis C, autoimmune hepatitis, and NAFLD (16). Previous hospital-based cross-sectional studies have revealed that higher serum M2BPGi levels were associated with high levels of fasting plasma glucose and HbA₁c and higher prevalence of individuals with abnormal glucose metabolism (17, 18). However, there have been no population-based prospective studies examining the association between serum M2BPGi levels and incidence of type 2 diabetes.

The aim of the present study was to investigate the association between serum M2BPGi and incidence of type 2 diabetes in a general Japanese population.

Material and Methods

Study Population

The Hisayama study is a population-based study of lifestyle-related diseases—including cardiovascular disease, diabetes, and hypertension—and their risk factors in the town of Hisayama, which is a suburb of the Fukuoka metropolitan area on Kyushu Island, Japan. An annual health examination has been repeated since 1961, and participants have been encouraged to undergo a 75-g oral glucose tolerance test (OGTT) since 1988 (19). A detailed description of this survey was published previously (20). Briefly, a total of 3028 residents aged 40-79 years (77.7% of the total population in this group) underwent the examination in 2002. Among them, 2304 individuals were eligible for follow-up by repeated annual health examinations until 2009, after excluding 28 individuals who did not consent to participate in the study, 51 with no fasting blood samples, 100 who did not complete the 75-g OGTT, 512 with diabetes mellitus at baseline, and 33 without available data of serum M2BPGi concentration. Among the 2304 eligible individuals, 2143 individuals (853 men and 1290 women) who underwent follow-up examination were finally enrolled for analysis in the present study.

Follow-up Survey and Determination of Type 2 Diabetes

The participants were followed up prospectively by an annual health examination until November 30, 2009 (the median follow-up period was 7 years; interquartile range [IQR] 5-7 years). In the baseline and follow-up periods, the study participants underwent the 75-g OGTT between 08:00 and 10:30 after an overnight fast of at least 12 hours. Plasma glucose was obtained in a fasting state and after the first 2 hours of the 75-g OGTT. Plasma glucose levels were determined by the hexokinase method. Diabetes was defined by the World Health Organization criteria in 1998 (21) as follows: fasting plasma glucose ≥7.0 mmol/L, 2-hour postload glucose ≥11.1 mmol/L, and/or the use of glucose-lowering medications (oral hypoglycemic agents or insulin). When a participant died, we reviewed all available clinical information, including the use of glucose-lowering medication. Participants were censored at the latest occasion of the health examination or at the time of death. When patients were newly administered insulin therapy during the follow-up period, they were checked to determine whether they had been diagnosed with type 1 diabetes or were positive for type 1 diabetes autoantibodies. Such individuals would have been censored, but in fact none of the participants were newly diagnosed with type 1 diabetes during the follow-up period. The participants underwent an average (±SD) of 5.0 ± 2.2 follow-up examinations.

Measurement of Mac-2 Binding Protein Glycosylation Isomer

At the baseline examination, blood samples were collected after at least 12 hours of fasting. A portion of each serum specimen was stored at −80 °C until use for the measurement of M2BPGi concentrations in 2017-2018. Serum M2BPGi was measured based on a lectin-Ab sandwich immunoassay (Catalog# M2B-1559-10, RRID:AB_2927561) using a fully automatic immunoanalyzer, HISCL-5000 (Sysmex, Kobe, Japan) (14). Glycosylated M2BP was captured by Wisteria floribunda agglutinin (WFA) immobilized on magnetic beads, and the bound product was assayed using an antihuman M2BP monoclonal antibody linked to alkaline phosphatase (ALP-aM2BP). Two types of reagent packs (an M2BP-WFA detection pack and a chemiluminescence substrate pack) were loaded in the HISCL-5000. Serum of participants (10 μL) was diluted to 60 μL with a reaction buffer solution and then mixed with WFA-coated magnetic beads solution (30 μL). After the binding reaction, an ALP-aM2BP solution (100 μL) was added to the reaction solution. The resultant conjugates were magnetically separated from the unbound components, and mixed well with a CDP-Star substrate solution (50 μL) and a stopping solution (100 μL) before reading the fluorescence. The chemiluminescence intensity was obtained within 17 minutes of the above procedure. The reaction chamber was maintained at 42 °C. The cut-off index (COI) of serum M2BPGi was calculated according to the equation reported previously (22). The measured values of M2BPGi were indexed with the obtained values using the following equation: COI = ([M2BPGi]_sample – [M2BPGi] _{negative control})/([M2BPGi] _{positive control} – [M2BPGi] _{negative control}), where [M2BPGi]_sample represents the M2BPGi count of the serum sample. Participants were categorized into 5 groups according to the quintiles of serum M2BPGi concentrations (Q1, ≤0.37 COI; Q2, 0.38-0.49 COI; Q3, 0.50-0.62 COI; Q4, 0.62-0.80 COI; and Q5, ≥0.81 COI).

Other Risk Factors

Each individual completed a self-administered questionnaire including medical history, family history of diabetes, use of antihypertensive, antidiabetic, and lipid-modifying medications, smoking habits, alcohol intake, and regular exercise. The questionnaire was checked by trained interviewers at the screening. A family history of diabetes was defined as the presence of diabetes in first-degree relatives of the individuals. Smoking habits and alcohol intake were categorized as either current use or not. Regular exercise was defined as any sports or other forms of exertion at least 3 times per week during leisure time. Body height and weight were measured in light clothing without shoes, and body mass index (BMI) was calculated (kg/m²). Blood pressure was measured 3 times in a sitting position using an automated sphygmomanometer (BP-203RV III; Omron Healthcare, Kyoto, Japan), and the mean of 3 measurements was used for the analysis. Hypertension was defined as blood pressure ≥140/90 mmHg and/or current use of antihypertensive agents. Serum levels of total cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and γ-glutamyl transpeptidase (GGT) were evaluated using the enzymic method (FUJIFILM Wako Pure Chemical Co., Osaka, Japan). The serum AST/ALT ratio was calculated by dividing serum levels of AST by ALT. Serum insulin levels were determined by a chemiluminescent enzyme immunoassay (Fujirebio Inc., Tokyo). Insulin resistance was evaluated using the homeostasis model assessment of insulin resistance (HOMA-IR) values, defined as fasting insulin (pmol/L) × fasting glucose (mmol/L)/135 (23). Serum low-density lipoprotein (LDL) cholesterol was calculated using the Friedewald formula as follows: LDL cholesterol (mmol/L) = total cholesterol (mmol/L) − HDL cholesterol (mmol/L) − 0.2 × triglycerides (mmol/L) (24). Serum high-sensitivity C-reactive protein (hs-CRP) concentrations were measured using a latex-enhanced nephelometric assay (Behring Diagnostics, Westwood, MA, USA). Hepatitis B virus surface antigen (HBsAg) was measured using a magnetic agglutination test (Shino-Test, Tokyo) and hepatitis C virus antibody (HCV Ab) was measured using a chemiluminescent enzyme immunoassay (Fujirebio Inc., Tokyo).

Statistical Analysis

We tested the trends in the means (SDs) and the frequencies of risk factors across the quintile levels of serum M2BPGi concentrations by a linear or logistic regression analysis, and the Jonckheere–Terpstra test for median values. HOMA-IR and serum concentrations of triglycerides, AST, ALT, GGT, and hs-CRP are presented as medians and IQRs in the table of the baseline characteristics of the population, and were naturally log transformed due to skewed distributions in the statistical analyses. We calculated Spearman's rank correlation coefficients between serum M2BPGi concentrations and risk factors. The incidence rate of type 2 diabetes was calculated using the person-year method after adjusting for age and sex by means of a direct method using the total examined population as the standard. We used a Cox proportional hazards model to estimate the hazard ratio (HR) and its 95% CI for the incidence of type 2 diabetes for the quintile levels of serum M2BPGi concentrations or per 1 SD increment in log-transformed serum M2BPGi concentrations (as a continuous variable). The trends in the age- and sex-adjusted incidence rates or HRs across the serum M2BPGi levels were tested by a relevant Cox model including serum M2BPGi levels assigned ordinal numbers of 1, 2, 3, 4, or 5 as categorical variables and the relevant covariates. In the multivariable analyses, we evaluated 4 different models as follows: (1) model 1, adjusted for age and sex; (2) model 2, adjusted for age, sex, family history of diabetes, hypertension, serum total cholesterol, serum HDL cholesterol, log-transformed serum triglycerides, use of lipid-modifying agents, BMI, positivity for HBsAg or HCV Ab, current smoking, current drinking, regular exercise, and number of health examinations received during follow-up; (3) model 3, adjusted for the covariates included in model 2 plus serum log-transformed hs-CRP; and (4) model 4, adjusted for the covariates included in model 2 plus log-transformed HOMA-IR. The shape of the association between serum M2BPGi concentrations and diabetic risk was also estimated by using a restricted cubic splines regression analysis with 4 knots placed at the 5th, 35th, 65^th, and 95th percentiles of serum M2BPGi concentrations (0.26, 0.46, 0.67, and 1.22 COI, respectively) (25), where the 20th percentile value (0.37 COI) was set as the reference value. The nonlinearity for the association was tested by using a likelihood ratio test comparing the relevant model with only a linear term against the model with linear and cubic spline terms. To compare the discriminatory ability for the incidence of type 2 diabetes between the models including known risk factors for diabetes with and without serum M2BPGi levels, the increase in the Harrell's c-statistics among models was evaluated and tested using a method described by Newson (26). Moreover, the increased predictive ability of the serum M2BPGi level was further estimated by the continuous net reclassification improvement (cNRI) and integrated discrimination improvement (IDI) (27). To elucidate the influence of serum M2BPGi on the risk of developing diabetes according to serum liver enzymes based on our previous study (7), we examined the combined association of serum M2BPGi and serum liver enzymes. For these analyses, levels of serum M2BPGi were divided into 2 categories by the 60th percentile value (<0.63 or ≥0.63 COI), and serum liver enzymes were divided into 2 categories by their respective median values: <18 U/L or ≥18 U/L for serum ALT and <23 U/L or ≥23 U/L for serum GGT. The interactions of serum M2BPGi levels and serum liver enzyme levels on the development of type 2 diabetes were tested by adding a multiplicative interaction term between the serum M2BPGi category and each of the liver enzyme categories to the relevant Cox model. The proportions of missing values were less than 0.12% for all the variables included in the model. A 2-sided value of P < .05 was considered to be statistically significant in all analyses. Statistical analyses were conducted using Statistical Analysis Software version 9.4 (SAS Institute, Cary, NC, USA).

Ethics Considerations

The protocol for this research project was approved by a suitably constituted Ethics Committee of the institution, namely, the Kyushu University Institutional Review Board for Clinical Research (approval no. 2021-457), and the protocol conforms to the provisions of the Declaration of Helsinki. Informed consent was obtained from all participants.

Results

The distribution of serum M2BPGi concentrations is shown in Fig. 1. The median value of serum M2BPGi concentrations was 0.55 COI (IQR 0.40-0.75 COI). When the serum M2BPGi concentrations were log transformed, the histogram showed a nearly normal distribution.

Figure 1.

Histogram of serum M2BPGi (Mac-2 binding protein glycosylation isomer) concentrations in the study population, 2002. Individuals with serum M2BPGi >2.0 COI (99th percentile point) are not included in the histogram. Abbreviations: COI, cut-off index; IQR, interquartile range; M2BPGi, Mac-2 binding protein glycosylation isomer.

Table 1 shows the baseline characteristics according to the quintile levels of serum M2BPGi concentrations. Higher serum M2BPGi levels were significantly associated with higher mean values of age, serum total cholesterol, LDL cholesterol, and BMI; higher median values of HOMA-IR, serum triglycerides, serum AST, serum ALT, and serum hs-CRP; and higher frequencies of hypertension, use of lipid-modifying agents, and positivity for HBsAg or HCV Ab. Conversely, the mean values of serum HDL cholesterol and proportions of men, current smoking, and current drinking decreased significantly with elevating serum M2BPGi levels.

Table 1.

Baseline characteristics of the study population according to serum Mac-2 binding protein glycosylation isomer levels

	Serum M2BPGi levels (COI)					P for trend
	Q1	Q2	Q3	Q4	Q5
	≤0.37	0.38-0.49	0.50-0.62	0.63-0.80	≥0.81
	(n = 429)	(n = 434)	(n = 426)	(n = 417)	(n = 437)
Age (years)	53.8 (8.7)	56.1 (9.7)	59.5 (9.7)	61.4 (10.4)	63.9 (10.1)	<.01
Male sex (%)	53.1	44.5	41.3	32.4	27.7	<.01
Family history of diabetes (%)	10.7	10.4	13.1	10.6	11.2	.81
Hypertension (%)	24.0	31.1	40.1	42	46.9	<.01
Fasting plasma glucose (mmol/L)	5.7 (0.5)	5.7 (0.5)	5.8 (0.4)	5.7 (0.5)	5.7 (0.4)	.23
HOMA-IRa	1.21 (0.86-1.69)	1.48 (1.07-2.13)	1.69 (1.11-2.37)	1.74 (1.26-2.52)	1.81 (1.29-2.72)	<.01
Serum total cholesterol (mmol/L)	5.1 (0.8)	5.3 (0.9)	5.3 (0.9)	5.4 (0.9)	5.3 (1.0)	<.01
Serum HDL cholesterol (mmol/L)	1.8 (0.4)	1.7 (0.4)	1.6 (0.4)	1.6 (0.4)	1.5 (0.4)	<.01
Serum triglycerides (mmol/L)	0.9 (0.7-1.3)	1.0 (0.8-1.5)	1.1 (0.8-1.6)	1.1 (0.8-1.6)	1.1 (0.8-1.5)	<.01
Serum LDL cholesterol (mmol/L)b	2.8 (0.8)	3.0 (0.8)	3.1 (0.8)	3.2 (0.9)	3.1 (0.8)	<.01
Use of lipid-modifying agents (%)	5.4	7.1	10.1	8.9	9.2	.03
BMI (kg/m2)	22.1 (3.1)	22.7 (3.2)	23.5 (3.2)	23.5 (3.0)	23.4 (3.1)	<.01
Serum AST (units/L)	22 (19-26)	23 (20-27)	24 (21-28)	25 (21-29)	26 (22-31)	<.01
Serum ALT (units/L)	16 (13-21)	17 (13-24)	18 (14-25)	18 (14-26)	20 (14-28)	<.01
Serum GGT (units/L)	24 (16-44)	23 (15-42)	22 (16-37)	22 (16-35)	22 (16-38)	.55
Serum AST/ALT ratio	1.44 (0.52)	1.39 (0.48)	1.37 (0.46)	1.38 (0.50)	1.40 (0.54)	.24
Serum hs-CRP (mg/L)	0.26 (0.13-0.55)	0.36 (0.19-0.76)	0.45 (0.22-0.87)	0.48 (0.27-0.94)	0.62 (0.32-1.38)	<.01
HBsAg or HCV Ab positive (%)c	5.4	4.6	5.4	6.5	15.8	<.01
Current smoking (%)	33.6	24	19.7	14.4	15.8	<.01
Current drinking (%)	62.5	53.2	44.4	33.6	31.1	<.01
Regular exercise (%)	10.0	11.8	10.3	9.4	11.4	.92
Number of health examinations received during follow-up (times)	4.7 (2.3)	5.0 (2.2)	5.1 (2.2)	5.1 (2.2)	5.0 (2.2)	.052

SI conversion factors: To convert units/L values to μkat/L, multiply serum AST, ALT, and GGT values by 0.0167. Data are presented as the mean values (SD) for continuous variables (except HOMA-IR, serum triglycerides, and serum hs-CRP), median values (interquartile range) for HOMA-IR, serum triglycerides, and serum hs-CRP, or percentages for the binary variables.

Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; COI, cut-off index; GGT, γ-glutamyl transferase; HBsAg, hepatitis B surface antigen; HCV Ab, hepatitis C antibody; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment-insulin resistance; hs-CRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein; M2BPGi, Mac-2 binding protein glycosylation isomer.

Participants with missing data of serum insulin levels (n = 23) were excluded from the analysis.

LDL cholesterol was calculated by the Friedewald equation. Participants with serum triglycerides >4.5 mmol/L (n = 24) were excluded from the analysis.

One participant with missing data for the HBsAg and HCV Ab test was excluded from the analysis.

The unadjusted and age- and sex-adjusted correlation coefficients between serum M2BPGi concentrations and other variables in the study population are shown elsewhere (Table S1 (28)). HOMA-IR, serum AST, serum ALT, serum hs-CRP, and BMI had positive correlations with serum M2BPGi concentrations. In particular, the Spearman's rank correlation coefficient between serum M2BPGi concentrations and HOMA-IR was the highest among the variables (unadjusted correlation coefficient = 0.26 [P < .01] and age- and sex-adjusted correlation coefficient = 0.30 [P < .01]). On the other hand, the serum M2BPGi concentration was inversely correlated with serum HDL cholesterol and the serum AST/ALT ratio.

During the follow-up period, 219 individuals developed type 2 diabetes. Table 2 shows the age- and sex-adjusted incidence rates and multivariable-adjusted HRs for the development of type 2 diabetes according to quintiles of serum M2BPGi levels. The incidence rates increased linearly with high serum M2BPGi levels: the rates were 10.2, 16.3, 19.3, 20.9, and 24.1 per 1000 person-years from the first to fifth quintile groups, respectively (P for trend <.01). The multivariable-adjusted HRs for the development of type 2 diabetes increased significantly with higher serum M2BPGi levels after adjusting for age, sex, family history of diabetes, hypertension, serum total cholesterol, serum HDL cholesterol, log-transformed serum triglycerides, use of lipid-modifying agents, BMI, positivity for HBsAg or HCV Ab, current smoking, current drinking, regular exercise, and number of health examinations received during follow-up (model 2, HR 1.66, 95% CI 1.003-2.74 for Q5 vs Q1, P for trend = .04). These associations were attenuated to statistically nonsignificant levels after additional adjustment for log-transformed hs-CRP and log-transformed HOMA-IR in addition to the above-mentioned confounding factors. In the restricted cubic spline regression analysis, the HR of type 2 diabetes increased linearly with greater serum M2BPGi concentrations, where the test for nonlinearity was not significant (P for nonlinearity = .28) (Fig. 2). The multivariable-adjusted HR increased by 1.52 (95% CI, 1.10-2.11) per every 1-SD increment in the log-transformed serum M2BPGi concentrations, where the SD of log-transformed serum M2BPGi concentrations was 0.48.

Table 2.

Hazard ratios for the development of type 2 diabetes according to serum Mac-2 binding protein glycosylation isomer levels

	Serum M2BPGi level (COI)					P for trend
	Q1	Q2	Q3	Q4	Q5
	≤0.37	0.38-0.49	0.50-0.62	0.63-0.80	≥0.81
Median	0.30	0.43	0.55	0.71	0.97
Number of events/participants	27/429	40/434	47/426	48/417	57/437
Age- and sex-adjusted incidence rate (103 PYs)	10.2	16.3	19.3	20.9	24.1	<.01
Hazard ratio (95% CI)
Model 1 (age and sex adjusted)	1.00 (reference)	1.48 (0.91-2.41)	1.76 (1.09-2.85)	1.91 (1.17-3.11)	2.23 (1.38-3.62)	<.01
Model 2 (multivariable adjusted)a	1.00 (reference)	1.24 (0.75-2.03)	1.23 (0.75-2.02)	1.41 (0.85-2.33)	1.66 (1.003-2.74)	.04
Model 2 + serum hs-CRP (log transformed)	1.00 (reference)	1.23 (0.75-2.02)	1.25 (0.77-2.05)	1.37 (0.83-2.26)	1.57 (0.95-2.59)	.08
Model 2 + HOMA-IR (log transformed)	1.00 (reference)	1.13 (0.69-1.86)	1.11 (0.68-1.82)	1.17 (0.70-1.95)	1.31 (0.78-2.20)	.32

Abbreviations: COI, cut-off index; HOMA-IR, homeostasis model assessment-insulin resistance; hs-CRP, high-sensitivity C-reactive protein; M2BPGi, Mac-2 binding protein glycosylation isomer; PYs, person-years.

Model 2: The risk estimates were adjusted for age, sex, family history of diabetes, hypertension, serum total cholesterol, serum high-density lipoprotein cholesterol, serum triglycerides (log transformed), use of lipid-modifying agents, body mass index, positivity for hepatitis B surface antigen or hepatitis C antibody, current smoking, current drinking, regular exercise, and number of health examinations received during follow-up.

Figure 2.

Restricted cubic spline regression analysis for the association between serum M2BPGi levels and risk of type 2 diabetes. Solid lines represent the hazard ratio, and dashed lines represent the 95% CI of the hazard ratio. Knots were placed at the 5th, 35th, 65th, and 95th percentiles (0.26, 0.46, 0.67, and 1.22 COI, respectively) of serum M2BPGi concentrations. A reference point was set at the 20th percentile (0.37 COI). Serum M2BPGi values over 2.0 COI are not present in the plots. The risk estimates were adjusted for age, sex, family history of diabetes, hypertension, serum total cholesterol, serum high-density lipoprotein cholesterol, log-transformed serum triglycerides, use of lipid-modifying agents, body mass index, positivity for HBsAg or HCV Ab, current smoking, current drinking, regular exercise, and number of health examinations received during follow-up. (The model used in this analysis was identical to Model 2 in Table 2.) Abbreviations: COI, cut-off index; HBsAg, hepatitis B surface antigen; HCV Ab, hepatitis C antibody; M2BPGi, Mac-2 binding protein glycosylation isomer.

Since some antihypertensives such as diuretics and beta-blockers have been reported to be associated with increased diabetic risk (29), we estimated the association between serum M2BPGi levels and the risk of diabetes after adjusting for systolic blood pressure taken as a continuous variable and dummy variables for use of diuretics, use of beta-blockers, and use of other classes of antihypertensive agents in the relevant Cox model, instead of variable hypertension. However, the results were substantially unchanged (Table S2 (28)). In the sensitivity analysis after excluding individuals with positivity for HBsAg or HCV Ab, the age- and sex-adjusted HRs for incidence of type 2 diabetes increased significantly with higher serum M2BPGi levels (P for trend <.01), whereas the multivariable-adjusted HRs for incidence of type 2 diabetes tended to increase with high serum M2BPGi levels, although the trend did not reach the level of statistical significance (P = .11) (Table S3 (28)).

In addition, we compared Harrell's c-statistics between the basic model including risk factors for type 2 diabetes with and without serum M2BPGi to evaluate whether serum M2BPGi improves the discrimination for type 2 diabetes. Adding the information on the serum M2BPGi level to the basic model including the afore-mentioned diabetes risk factors tended to increase Harrell's c-statistics (from 0.72 to 0.73; P = .15), though the difference did not reach the level of statistical significance. However, the changes in cNRI and IDI were significant (cNRI, 0.15, P = .04; IDI, 0.04, P = .04) (Table S4 (28)).

We also investigated the association of serum M2BPGi levels with the risk of type 2 diabetes according to serum liver enzyme levels. In these analyses, the individuals were divided into 4 groups according to serum M2BPGi levels (≥0.63 COI [median] or <0.63 COI) and serum liver enzymes, such as serum GGT levels (≥23 U/L [median] or <23 U/L) or serum ALT levels (≥18 U/L [median] or <18 U/L). The multivariable-adjusted HR of type 2 diabetes was 2.96 (95% CI 1.87-4.69) in the individuals with higher levels of both serum M2BPGi and serum GGT, and was 1.93 (95% CI 1.29-2.91) in individuals with higher levels of both serum M2BPGi and serum ALT than in individuals having lower levels of both serum M2BPGi and each serum liver enzyme (Fig. S1 (28)). Among these groups, multivariable-adjusted levels of HOMA-IR and serum hs-CRP were highest in individuals with higher levels of both serum M2BPGi levels and serum liver enzymes (Figs. S2 and S3 (28)).

Discussion

This prospective cohort study demonstrated that higher serum M2BPGi concentrations, an indicator of chronic liver injury and fibrosis, were significantly associated with greater risk of the development of type 2 diabetes after adjustment for potential confounding factors in a general Japanese population. In addition, the ability to predict future risk of type 2 diabetes was improved by adding serum M2BPGi levels to the traditional risk factors for type 2 diabetes. Moreover, serum hs-CRP and HOMA-IR, which are indicators of inflammation and insulin resistance, respectively, were positively correlated with serum M2BPGi concentrations, and the significant association between serum M2BPGi concentrations and diabetic risk attenuated to a nonsignificant level after additional adjustment for serum hs-CRP or HOMA-IR, possibly suggesting that inflammation or insulin resistance may be involved in the excess risk of diabetes in individuals with higher serum M2BPGi levels. These findings shed light on the importance of chronic liver injury and fibrosis in the pathogenesis of diabetes through inflammation and insulin resistance.

Several previous prospective studies have reported that indices of chronic liver injury, such as liver enzymes (7), or indices of NAFLD (8) were positively associated with a risk of diabetes. A recent meta-analysis revealed that individuals with NAFLD diagnosed by imaging methods have a 2-fold higher risk of developing diabetes than those without NAFLD (9). Likewise, a large population-based cohort study has shown that East Asians with HCV infection had 3-fold higher risk for incidence of diabetes (10). On the other hand, there have been no prospective studies addressing the association between serum M2BPGi concentrations and risk of diabetes. Several hospital-based cross-sectional studies have reported that higher serum M2BPGi levels were associated with higher levels of fasting plasma glucose and HbA₁c, and higher prevalence of abnormal glucose metabolism among individuals who visited a hospital for a physical checkup (17, 18). Our findings demonstrating the prospective association of serum M2BPGi concentrations with incidence of type 2 diabetes lend further support to the notion that chronic liver injury and fibrosis should be taken into account in the risk assessment of type 2 diabetes.

Several potential mechanisms could explain the positive association between serum M2BPGi and the risk of type 2 diabetes. A cross-sectional study has shown that individuals with fatty liver determined by abdominal ultrasound examination had higher levels of serum M2BPGi than individuals without fatty liver, possibly suggesting that an elevation in serum M2BPGi concentrations reflects the accumulation of fat in the liver and subsequent insulin resistance (30). In addition, a study using liver tissue collected from liver transplantation recipients has also revealed that individuals with higher inflammation in the liver exhibited higher serum M2BPGi concentrations (31), which histologically demonstrated the positive association between serum M2BPGi concentrations and liver inflammation. Our study group previously reported that higher serum CRP (32) and HOMA-IR (33) levels were associated with incidence of type 2 diabetes, and the present study showed that the association between serum M2BPGi levels and the risk of developing type 2 diabetes was attenuated to a statistically nonsignificant level after the additional adjustment for serum hs-CRP or HOMA-IR. The present study also revealed that individuals with higher levels of both serum liver enzymes (GGT and ALT) and serum M2BPGi had higher levels of serum hs-CRP and HOMA-IR and a greater risk of developing type 2 diabetes than individuals with lower levels of serum GGT, ALT, and M2BPGi. These findings suggest that inflammation and insulin resistance may be involved in the significant association between serum M2BPGi concentrations and increased risk of diabetes. Liver fibrosis is known to be caused by inflammation, and liver fibrosis induces insulin resistance (34). Furthermore, insulin resistance itself is thought to promote liver inflammation and liver fibrosis (34, 35). Hence, a close relationship exists between liver fibrosis, inflammation, and insulin resistance. Since serum M2BPGi is a biomarker of liver fibrosis, the present findings may reflect the degree of liver fibrosis. On the other hand, there are some experimental studies which support the idea that high serum M2BPGi would be a cause of inflammation and insulin resistance. These studies focused on the roles of Mac-2, which should be noted when considering the underlying mechanisms for the association of serum M2BPGi with inflammation and insulin resistance. Research using human liver tissue specimens has shown that M2BPGi induces expression of Mac-2 (15), which promotes inflammation (36, 37) and insulin resistance (38) in mice and humans. In addition, the Dallas Heart Study, a population-based cohort study, has reported that Galectin-3 (formerly called Mac-2 antigen) was associated with the prevalence and incidence of diabetes (39). Considering these results together, M2BPGi itself may be one of the factors that promote inflammation, insulin resistance, and subsequent type 2 diabetes, though further studies are needed to confirm this point.

The strengths of the present study include its longitudinal population-based design and accurate diagnoses of type 2 diabetes based on 75-g OGTT. However, some potential limitations of this study should be noted. First, the serum M2BPGi concentrations and other risk factors were based on a single measurement at baseline. In addition, we were unable to obtain information about medical treatment during the follow-up period. Given that the serum M2BPGi levels and other risk factors may have changed during follow-up, this limitation could lead to misclassification of these variables, which would weaken the association found in the present study, biasing the results toward the null hypothesis. Second, the serum levels of M2BPGi were determined using samples that had been frozen for 15 years. Such long-term storage might result in the degradation of M2BPGi. However, we believe that the influence of this limitation may have been modest, because any such degradation is likely to occur randomly among the samples. Third, we could not directly assess liver injury and fibrosis by abdominal ultrasound or biopsy. Further studies that include such data will be needed to clarify the mechanism of the association between serum M2BPGi and diabetes. Fourth, the present study showed that the frequencies of “current smoking” and “current drinking” increased with lower M2BPGi levels, which would seem somewhat counterintuitive. However, these findings may have been partly attributable to the higher frequency of men with lower M2BPGi levels than men with higher M2BPGi levels. It is also possible that alcohol drinkers or cigarette smokers with high M2BPGi levels were encouraged to stop drinking alcohol and smoking due to liver damage. Fifth, the number of incidences of type 2 diabetes was limited. This might have affected the statistical nonsignificance of adding M2BPGi levels to the traditional risk factors for type 2 diabetes when predicting future risk of the disease. Sixth, it is unclear whether high serum M2BPGi was the cause or the consequence of inflammation/insulin resistance in the present study, because serum M2BPGi, serum hs-CRP, and HOMA-IR were measured at the same time point. However, liver fibrosis is also thought to be both a cause and a consequence of inflammation and insulin resistance, as described above. Therefore, it is not easy to resolve this question. Seventh, we could not exclude individuals with overt hyperthyroidism or pregnancy due to absence of those data. However, since our study participants were community-dwelling individuals aged 40-79 years, there were likely to be no more than a few pregnant women or individuals with thyroid diseases, and thus we believe this issue would not have affected the findings. Finally, the generalizability of our findings may be limited, because these analyses were conducted in only 1 cohort of Japanese people. In terms of the probable etiologies of hepatic injury in our cohort, these may have been heterogeneous. Therefore, our findings should be validated in other large-scale cohorts of various ethnic populations.

Conclusions

In conclusion, the present study clearly demonstrated that higher serum M2BPGi concentrations were significantly associated with higher risk of type 2 diabetes in a Japanese community. Moreover, inflammation and insulin resistance may have been involved in this association. These findings highlight the importance of inflammation and insulin resistance when considering the pathogenesis of diabetes.

Abbreviations

Ab

antibody

ALT

alanine aminotransferase

AST

aspartate aminotransferase

BMI

body mass index

cNRI

continuous net reclassification improvement

COI

cutoff index

GGT

γ-glutamyl transpeptidase

HBsAg

Hepatitis B virus surface antigen

HCV

hepatitis C virus

HDL

high-density lipoprotein

HOMA-IR

homeostasis model assessment of insulin resistance

HR

hazard ratio

hs-CRP

high-sensitivity C-reactive protein

IDI

integrated discrimination improvement

LDL

low-density lipoprotein

M2BPGi

Mac-2 binding protein glycosylation isomer

NAFLD

nonalcoholic fatty liver disease

OGTT

oral glucose tolerance test

WFA

Wisteria floribunda agglutinin

Funding

This study was supported in part by the Ministry of Education, Culture, Sports, Science and Technology of Japan (JSPS KAKENHI Grant Number JP21H03200, JP19K07890, JP20K10503, JP20K11020, JP21K07522, JP21K11725, JP21K10448, JP22K07421, and JP22K17396); by the Health and Labour Sciences Research Grants of the Ministry of Health, Labour and Welfare of Japan (JPMH20FA1002); and by the Japan Agency for Medical Research and Development (JP22dk0207053). In addition, this study was sponsored by Sysmex Co. (Kobe, Hyogo, Japan). The study sponsors/funders were not involved in the design of the study; the collection, analysis, and interpretation of data; or the writing of the report. In addition, the study sponsors/funders did not impose any restrictions regarding the publication of the report.

Author Contributions

M.H. contributed to the study concept, data collection, data analysis, data interpretation and drafting of the manuscript. Y.H. contributed to the study concept, data collection, interpretation of data, and revision of the manuscript. J.H., T.H., S.S., and M.S. contributed to data collection and interpretation of data and revision of the manuscript. T.K. and H.O. contributed to interpretation of data and revision of the manuscript. T.N. was the chief investigator of the Hisayama Study and contributed to the study concept, data collection, interpretation of data and revision of the manuscript. All authors critically reviewed the manuscript and approved the final version. T.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for its integrity and accurate analysis.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.