The impacts of long-term carbon disulfide exposure on glucose homeostasis and type 2 diabetes: a multifaceted gene-environment-lifestyle interaction study of Chinese adults

Graphical abstract

Highlights

• •Long-term CS₂ exposure impaired glucose homeostasis and increased T2D risk.
• •Gene-CS₂-lifestyle interactions on glucose dyshomeostasis and T2D were identified.
• •High genetic risk aggravated CS₂-related glucose dyshomeostasis and T2D risk.
• •Healthy lifestyle alleviated CS₂-related glucose dyshomeostasis and T2D risk.
• •Reduce CS₂ and improve lifestyle help to prevent T2D especially for those at high genetic risk.

Abstract

Introduction

The long-term impacts of exposure to carbon disulfide (CS₂), a highly concerning air toxicant listed by the Clean Air Act, and its interactions with genetic susceptibility and lifestyle on glucose homeostasis and type 2 diabetes (T2D) in the general population remain unclear and require urgent clarification.

Objectives

To investigate the interactions of CS₂ exposure, genetic susceptibility, and lifestyle on glucose homeostasis and T2D.

Methods

In this prospective study, urinary CS₂ metabolite (2-Thiothiazolidine-4-carboxylic acid, TTCA) and fasting plasma glucose (FPG) and insulin (FPI) for 5294 observations from 2523 participants were repeatedly measured to examine the cross-sectional and longitudinal associations of CS₂ exposure with glucose homeostasis and T2D by performing generalized linear mixed models or COX models. Polygenic risk score (PRS) and healthy lifestyle index (HLI) were constructed to evaluate their cross-sectional and longitudinal interactions with TTCA and to assess gene-CS₂-lifestyle interactions.

Results

TTCA was cross-sectionally and longitudinally related to glucose dyshomeostasis and T2D risk elevation. Longitudinally, compared to subjects with persistently low TTCA, those with persistently high TTCA had 0.350 (95 % CI: 0.060–0.640) mmol/L, 0.197 (0.097–0.297) ln-unit, 0.236 (0.120–0.351) ln-unit, and 56.3 % (HR: 1.563; 95 % CI: 1.018–2.401) risk increments in FPG, FPI, HOMA-IR, and T2D incidence, respectively. Cross-sectional and longitudinal interactions of TTCA with PRS and HLI were uncovered (P for interaction < 0.05). Subjects with persistently low TTCA, healthy lifestyle, and low PRS longitudinally manifested the greatest reductions in FPG (β: −0.609; 95 % CI: −1.108 to −0.108), FPI (−0.357; −0.479 to −0.236), HOMA-IR (−0.358; −0.498 to −0.219), and incident T2D risk (HR: 0.228; 95 % CI: 0.091–0.574).

Conclusion

Long-term CS₂ exposure was related to glucose dyshomeostasis and increased T2D risk, which might be exacerbated by high genetic susceptibility while mitigated by healthy lifestyle, highlighting the significance of reducing CS₂ exposure and improving lifestyle in preventing glucose dyshomeostasis and T2D, particularly among individuals with high genetic risk.

Introduction

Carbon disulfide (CS₂) is a highly volatile organic chemical that commonly persists in the environment from multiple natural phenomena like volcanic eruptions and forest fires, various chemical sectors like rubber, viscose fiber, rayon, and cellophane manufacturing, and miscellaneous human activities like cigarette smoking, fumigating, and contact with daily products, including pesticide insecticides, clothing stain removers, wool degreasers, etc. [[1], [2], [3], [4]]. Consequently, the general population is widespreadly exposed to CS₂ via inhalation, ingestion, and dermal absorption in daily life [[5], [6], [7]]. Accumulating evidence has demonstrated the severe health effects of long-term or high-level exposure to CS₂ on various physiological systems, with notable impacts on the cardiometabolic, respiratory, and nervous systems [[6], [7], [8], [9]]. The Clean Air Act has listed CS₂ as an air toxicant that may cause grievous health risk and need to be strictly controlled by the U.S. Environmental Protection Agency (EPA) to protect public health [10]. Regulatory measures including occupational CS₂ exposure limits, such as the Occupational Safety and Health Administration (OSHA) Permissible Exposure Limit (PEL) set at 20 ppm for 8-h time-weighted average (TWA) CS₂ exposure [11], alongside ambient air quality monitoring and standards (EPA reference CS₂ concentration: 0.7 mg/m³) [12] have been implemented to minimize CS₂ exposure risks.

The diabetic risk of CS₂ has been highlighted by mounting studies in workers with occupational high-level CS₂ exposure [[13], [14], [15], [16], [17], [18], [19], [20], [21], [22]], while evidence for daily-life low-level exposure to CS₂ in the general population is extremely scarce. Only our previous cross-sectional study [23] of community-based adult population reported the associations of daily-life low-level CS₂ exposure with elevated fasting plasma glucose (FPG) level and the risk of type 2 diabetes (T2D), which is a major public health issue and a global pandemic characterized by glucose dyshomeostasis (e.g., FPG elevation, impaired fasting glucose [IFG], and insulin resistance [IR]) and contributed by environmental factors, lifestyle, and genetic susceptibility as well as their interactions [[24], [25], [26], [27], [28], [29]]. Nevertheless, the effects, especially the longitudinal effects supporting causality, of CS₂ exposure on glucose homeostasis and T2D, as well as the interaction effects of CS₂ exposure with lifestyle and genetic susceptibility on glucose homeostasis and T2D are unclear. These knowledge gaps warrant exigent filling to control and prevent CS₂ exposure and resultant glucose homeostasis impairment and T2D, restrain the global T2D pandemic, and protect public health.

Therefore, we conducted a general adult population-based prospective cohort study, where we repeatedly measured 2-Thiothiazolidine-4-carboxylic acid (TTCA), which is a metabolite of CS₂ in urine and a reliable biomarker of CS₂ exposure [23,30,31], at baseline and two follow-ups to clarify the effects of daily-life chronic CS₂ exposure on glucose homeostasis and the risk of T2D. We also developed polygenic risk score (PRS) and healthy lifestyle index (HLI) to illustrate the gene-environment-lifestyle interaction effects on glucose homeostasis and T2D.

Methods

Study population

This prospective epidemiological study was derived from the Wuhan-Zhuhai cohort set up in 2011–2012 and followed up triennially [32]. Shortly, the study included 4812 Chinese adults (18–80 years) who had resided in Wuhan or Zhuhai for a minimum of 5 years. Through professional physical measurements, face-to-face standardized questionnaires, and voluntary provisions of fasting blood and morning urine specimens on the same day for all participants, we collected multitudinous sample information and data including covariates, as described in detail in the Supplementary Material. After excluding participants without adequate specimens for quantifying urinary TTCA or creatinine (Cr) (N = 1109) or genotyping (N = 1470) and those with nephritis or essential data missing (N = 131), 2102 participants were retained at baseline, of which 1649 and 1543 participated in the first (2014–2015) and second (2017–2018) follow-up visits, respectively. Altogether, 5294 observations were enrolled in the cross-sectional analyses of repeated measures design. After excluding participants who attended only once across all visits, we included 2771 observations from 2059 participants in the longitudinal analyses of repeated measures design, where incidence risks of binary outcomes were estimated after excluding prevalent cases at baseline. Fig. S1 provides the detailed flowchart of screening out the study population. Subjects signed written informed consent form before each survey. The Ethics and Human Subject Committee of Tongji Medical College, Huazhong University of Science and Technology approved this study protocol on February 10, 2011 (No. 2011-17).

Urinary TTCA detection

At baseline and follow-ups, each subject’s TTCA concentration in urine was determined by an ultra-high performance liquid chromatography (Agilent 1290 Infinity II, Agilent Technologies Inc., Santa Clara, CA, U.S.)-tandem mass spectrometry (Sciex API 6500 Triple Quad, Applied Biosystems, Foster City, CA, U.S.) [23,33] with a limit of detection of 1.0 ng/mL. Accordingly, urinary TTCA concentration was adjusted for urinary Cr level [59] and represented as µg/mmol Cr. Detailed description of the laboratory protocol with quality assurance and quality control is provided in the Supplementary Material.

Glucose homeostasis and T2D ascertainments

In this study, glucose homeostasis parameters included FPG, fasting plasma insulin (FPI), homeostasis model assessment (HOMA) of IR (HOMA-IR), HOMA of β-cell function (HOMA-B), IR, and IFG. FPG (mmol/L) was measured using a fully automatic biochemical analyzer RX Daytona (Randox Laboratories, Crumlin, U.K.) and FPI (mU/L) was measured using a fully automatic electrochemical luminescence analyzer Roche Cobas e601 (Roche Diagnostics, Basel, Switzerland). T2D was identified as using antidiabetic medicine, a diagnosis of T2D by physician, or FPG ≥ 7.0 mmol/L according to the American Diabetes Association [34]. IFG was diagnosed with an FPG level of 5.6–6.9 mmol/L while without a diagnosis of T2D. HOMA-IR was calculated as (FPG × FPI) ÷ 22.5, with a cutoff value of >2.6 identified as IR; HOMA-B was calculated as (FPI × 20) ÷ (FPG − 3.5) [35].

Genotyping and calculation of PRS

The Whole Blood DNA Extraction Kit (BioTeke, Beijing, China) was used to extract genomic DNA, which was subsequently loaded into the Infinium OmniZhongHua-8 v1.3 BeadChip (Illumina, CA, U.S.) to obtain individual genotypes [36]. More detailed procedure and quality control of genotyping have been described in our previous study [37].

PLINK 1.90 was applied for the final PRS calculation using the count of risk alleles and SNP weights with the method of clumping and thresholding. The PRS was determined as:

$${\mathit{PRS}}_i={\sum }_{k=1}^n{\beta }_k{X}_{\mathit{ik}}$$

where ${\beta }_k$ is the effect size of the $k$th SNP on glucose homeostasis parameters or T2D, $X_{\mathit{ik}}$ is the count of risk alleles ($X$ = 0, 1, or 2) for the $k$th SNP of the $i$th subject, and n is the count of SNPs associated with glucose homeostasis parameters or T2D. T2D-PRS and FPG-PRS were constructed based on 91 SNPs associated with T2D and 158 SNPs associated with FPG in BioBank Japan (BBJ) (https://jenger.riken.jp/en/result) (Tables S1–S2), respectively. FPI-PRS was constructed based on 82 SNPs associated with FPI in the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC) (https://magicinvestigators.org/downloads/), a large-scale meta-analysis of quantitative diabetes-related traits in subjects without diabetes [38] (Table S3). For HOMA-IR, IR, and IFG, given the lack of eligible SNPs for their PRSs construction, the constructed PRSs showing the most significant associations with the corresponding traits were vicariously used for further analyses in accordance with previous recommendations [38,39]. Finally, FPI-PRS was also identified as both HOMA-IR-specific PRS and IR-specific PRS, and FPG-PRS was also identified as IFG-specific PRS.

Construction of HLI

HLI was constructed based on five lifestyle factors, namely cigarette smoking, drinking, body mass index (BMI; kg/m²), physical exercise, and diet, which were highlighted in recent guidelines for the prevention of T2D and combined with data collected in this cohort [24,40,41]. In summary, each factor was assigned a score of 0, 1, or 2, with higher score indicating a healthier behavior: cigarette smoking or drinking status (current = 0, former = 1, never = 2), physical exercise (physical inactivity = 0, moderate physical exercise = 1, vigorous physical exercise = 2), BMI (≥28 = 0, 24–28 = 1, <24 = 2), and diet (insufficient dietary intake [meet 1–2 dietary criteria] = 0, moderate dietary intake [meet 3–4 dietary criteria] = 1, adequate dietary intake [meet 5–6 dietary criteria] = 2). The criteria for dietary intake were based on the Chinese Food Guide Pagoda 2022 (Chinese Nutrition Society 2022) and the specific criteria included: (1) Fruits and vegetables: ≥3 servings/day; (2) Roughage: ≥1 servings/day; (3) Meat: 1–3 servings/week; (4) Aquatic products: 2–7 servings/week; (5) Eggs and milk: ≥1 servings/day; and (6) Pickled products: ≤1 serving/week. More details on scoring are presented in Table S4 and the characteristics of lifestyle factors are presented in Table S5. The HLI (range: 0–10 scores) was then categorized into a dichotomous (healthy/unfavorable: ≥6/<6) or quartered (0–5/6/7/8–10) variable based on the median or quartile of HLI, respectively.

Data analyses

Given the right skewed distribution, data of TTCA, FPI, HOMA-IR, and HOMA-B were natural log (ln)-transformed [60]. PRS was z-score standardized. To analyze the baseline characteristics according to TTCA quartiles, variance analyses were conducted for continuous variables while Cochrane-Armitage trend tests were performed for categorical variables.

In the repeated-measure cross-sectional analyses, generalized linear mixed models (GLMMs) were exploited to estimate the βs (95 % confidence intervals [CIs]) of continuous glucose homeostasis parameters and odds ratios (ORs) (95 % CIs) of prevalent IR, IFG, and T2D associated with both quartered and continuous TTCA, with individuals modeled as random effect [[42], [43], [44]]. In the repeated-measure longitudinal analyses, participants who participated in at least two visits were categorized into three groups: persistently low group (participants with TTCA level < P50 consistently across all study periods; as the reference group), persistently high group (participants with TTCA level ≥ P50 consistently across all study periods), and inconsistent group (those neither belong to the group of persistently low nor high) [6]. Subsequently, GLMMs were employed to assess changes in continuous glucose homeostasis parameters for the persistently high group and the inconsistent group in comparison to the persistently low group, with baseline glucose homeostasis parameters additionally adjusted as confounders; and COX regression models were employed to assess the incident risks of IR, IFG, and T2D.

To examine the robustness of the associations of TTCA with glucose homeostasis and T2D, we performed sensitivity analyses by (1) excluding participants with T2D at baseline in glucose homeostasis analyses and further excluding participants with IR in IFG prevalence analysis (to minimize the possibility of reverse causality); (2) excluding participants with IFG, T2D, or IR at baseline and those with T2D or IR at follow-up in IFG incidence analysis; (3) additionally adjusting for the dietary factors (to control for the confounding effect of diet); (4) additionally adjusting for the batch of urinary TTCA measurement; and (5) performing inverse probability weighting (IPW) analysis to account for data missing and avoid potential selection bias [62].

We also examined the relationships of PRS and HLI (as quartered and continuous variables) with glucose homeostasis parameters and T2D using GLMMs or COX regression models in cross-sectional and longitudinal analyses. Thereafter, cross-classification of subjects was conducted for multifaceted joint/interaction effect estimation. Subjects were cross-classified into 4 or 8 groups based on TTCA level [low (<P50)/high (≥P50)], PRS level [low (<P50)/high (≥P50)], and HLI [healthy/unfavorable] to estimate the cross-sectional joint effects of TTCA with PRS (4 groups; those with low TTCA and low PRS were set as the reference group), TTCA with HLI (4 groups; those with high TTCA and unfavorable lifestyle were set as the reference group), HLI with PRS (4 groups; those with unfavorable lifestyle and high PRS were set as the reference group), and TTCA with both PRS and HLI (8 groups; those with high TTCA, high PRS, and unfavorable lifestyle were set as the reference group). Meanwhile, subjects were cross-classified into 4, 6, or 12 groups according to TTCA status (persistently low/inconsistent/persistently high), PRS level [low (<P50)/high (≥P50)], and HLI (healthy/unfavorable) to estimate the longitudinal joint effects of TTCA with PRS (6 groups; those with persistently low TTCA and low PRS were set as the reference group), TTCA with HLI (6 groups; those with persistently high TTCA and unfavorable lifestyle were set as the reference group), HLI with PRS (4 groups; those with unfavorable lifestyle and high PRS were set as the reference group), and TTCA with both PRS and HLI (12 groups; those with persistently high TTCA, high PRS, and unfavorable lifestyle were set as the reference group). Interaction effect was assessed by incorporating an interaction term [61] of TTCA × PRS, TTCA × HLI, PRS × HLI, or TTCA × PRS × HLI into the GLMMs or COX regression models.

According to literature review, data pre-analysis, and statistical consideration [[45], [63]], age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use, history of hypertension, history of hyperlipidemia, city, and the first 10 principal components of ancestry were generally adjusted in all regression models, where BMI, smoking status, drinking status, and physical exercise as determinants of HLI were not adjusted when HLI was involved, antidiabetic drug use as a diagnostic criterion of T2D was not adjusted when T2D was involved, and the first 10 principal components of ancestry was adjusted only when PRS was involved.

Schoenfeld residuals were used to test the proportional hazard assumption and no violation was found. All statistical analyses were implemented using R 4.4.1. The statistical significance was defined as P < 0.05 (two-sided).

Results

Study population characteristics

Table 1 summarizes the characteristics of 5294 observations (69.08 % females) with a mean age of 55.62 years. Among them, 1754 T2D-free participants at baseline who also completed follow-up visits were included in the analysis. During the 6-year follow-up period (mean follow-up time: 4.97 years; 8718 person-years), 114 incident T2D cases were identified. As urinary TTCA concentrations increased across quartiles, the levels of age, BMI, FPG, FPI, and HOMA-IR, as well as the proportions of males and observations with unfavorable lifestyle, IR, IFG, and T2D were elevated, whereas the proportions of never smokers/drinkers and the HOMA-B level were decreased. No statistically significant trend across TTCA quartiles was observed for the level of PRS. Baseline characteristics did not differ between cross-sectional and longitudinal participants (Table S6). The median TTCA concentrations were 0.76 μg/mmol Cr (8.90 ng/mL) at baseline, 0.86 μg/mmol Cr (9.44 ng/mL) at the first follow-up, and 1.33 μg/mmol Cr (13.34 ng/mL) at the second follow-up (Table S7).

Table 1.

Characteristics of study participants.

Characteristics	T2D-free participantsa (N = 1754)	Baseline participants (N = 2102)	All observations (N = 5294)	Quartiles of urinary TTCA, μg/mmol Cr				Ptrendb
				Q1 (N = 1324)	Q2 (N = 1323)	Q3 (N = 1323)	Q4 (N = 1324)
				≤0.33	(0.33, 0.93]	(0.93, 3.01]	>3.01
Age, years, mean ± SE	52.89 ± 0.28	53.50 ± 0.25	55.62 ± 0.34	55.76 ± 0.58	55.66 ± 0.56	55.06 ± 0.56	56.24 ± 0.57	0.010
Female, n (%)	1233 (70.30)	1459 (69.44)	3657 (69.08)	999 (75.45)	902 (68.18)	881 (66.59)	875 (66.09)	0.010
BMI, kg/m2, mean ± SE	24.19 ± 0.08	24.25 ± 0.07	24.67 ± 0.07	24.67 ± 0.08	24.64 ± 0.08	24.70 ± 0.08	24.71 ± 0.08	0.024
Smoking status, n (%)								<0.001
Current	252 (14.37)	305 (14.51)	782 (14.77)	143 (10.80)	211 (15.95)	216 (16.33)	212 (16.01)
Former	93 (5.30)	114 (5.42)	357 (6.74)	68 (5.14)	79 (5.97)	99 (7.48)	111 (8.38)
Never	1409 (80.33)	1683 (80.07)	4155 (78.49)	1113 (84.06)	1033 (78.08)	1008 (76.19)	1001 (75.60)
Drinking status, n (%)								<0.001
Current	242 (13.80)	282 (13.42)	836 (15.79)	165 (12.46)	207 (15.65)	247 (18.67)	217 (16.39)
Former	46 (2.62)	57 (2.71)	280 (5.29)	42 (3.17)	80 (6.05)	73 (5.52)	85 (6.42)
Never	1466 (83.58)	1763 (83.87)	4178 (78.92)	1117 (84.37)	1036 (78.31)	1003 (75.81)	1022 (77.19)
Vigorous or moderate physical exercise, n (%)	884 (50.40)	1095 (52.09)	2899 (54.76)	738 (55.74)	716 (54.12)	721 (54.50)	724 (54.68)	0.974
History of hypertension, n (%)	704 (40.14)	898 (42.74)	2461 (46.49)	561 (42.37)	621 (46.94)	628 (47.47)	651 (49.17)	0.004
History of hyperlipidemia, n (%)	997 (56.85)	1240 (59.02)	3181 (60.09)	816 (61.63)	778 (58.81)	802 (60.62)	785 (59.29)	0.435
Family history of diabetes, n (%)	94 (5.36)	127 (6.04)	283 (5.35)	72 (5.44)	82 (6.20)	71 (5.37)	58 (4.38)	0.225
Antidiabetic drug use, n (%)	−	111 (5.28)	362 (6.84)	70 (5.29)	92 (6.95)	93 (7.03)	107 (8.08)	0.157
FPG-PRS, 10^(−5), mean ± SE	1.91 ± 1.05	1.88 ± 1.09	1.29 ± 0.01	1.30 ± 0.02	1.29 ± 0.02	1.29 ± 0.02	1.27 ± 0.02	0.074
FPI-PRS, 10^(−4), mean ± SE	2.88 ± 0.11	2.87 ± 0.10	2.83 ± 0.06	2.75 ± 0.05	2.77 ± 0.05	2.80 ± 0.06	3.04 ± 0.07	0.551
T2D-PRS, 10^(−3), mean ± SE	3.22 ± 0.07	3.13 ± 0.06	3.17 ± 0.04	3.01 ± 0.05	2.98 ± 0.04	3.11 ± 0.05	3.24 ± 0.05	0.959
Unfavorable lifestyle (HLI < 6), n (%)	813 (46.35)	971 (46.22)	2169 (40.97)	468 (35.35)	549 (41.50)	560 (42.33)	592 (44.71)	<0.001
FPG, mmol/L, mean ± SE	4.60 ± 0.02	4.89 ± 0.03	5.20 ± 0.04	5.13 ± 0.07	5.20 ± 0.06	5.17 ± 0.06	5.33 ± 0.06	<0.001
FPI, mU/L, mean ± SE	9.69 ± 0.20	10.92 ± 0.33	11.58 ± 0.23	10.78 ± 0.43	11.24 ± 0.42	11.40 ± 0.42	13.19 ± 0.42	0.006
HOMA-IR, mean ± SE	2.07 ± 0.05	2.65 ± 0.12	2.86 ± 0.08	2.58 ± 0.15	2.76 ± 0.15	2.78 ± 0.15	3.43 ± 0.15	<0.001
HOMA-B, mean ± SE	121.23 ± 48.26	107.69 ± 43.43	134.60 ± 19.10	165.77 ± 39.03	134.00 ± 38.21	115.75 ± 38.31	124.08 ± 58.60	0.007
IR, n (%)	381 (21.72)	542 (25.80)	1760 (33.25)	380 (28.70)	440 (33.26)	445 (33.64)	495 (37.39)	<0.001
IFG, n (%) c	165 (9.41)	188 (9.85)	623 (13.30)	132 (11.27)	141 (12.04)	168 (14.35)	182 (15.54)	0.005
T2D, n (%)	0	194 (9.23)	610 (11.52)	141 (10.65)	144 (10.88)	146 (11.04)	179 (13.52)	0.042

Abbreviations: Q, quartile; TTCA, 2-Thiothiazolidine-4-carboxylic acid; BMI, body mass index; PRS, polygenic risk score; HLI, healthy lifestyle index; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; HOMA-B, homeostasis model assessment of β-cell function; IFG, impaired fasting glucose; IR, insulin resistance; T2D, type 2 diabetes; SE, standard error; IQR, interquartile range; Cr, creatinine.

Data were presented as least square means ± SEs for continuous variables and n (%) for categorical variables.

^aT2D-free participants were those without T2D at baseline and not lost to follow-up.

^bP_trend was estimated by Cochran-Armitage trend test for categorical variables and variance analysis for continuous variables.

^cParticipants with T2D were excluded for IFG analysis.

Relationships between TTCA and glucose homeostasis and T2D

For per ln-unit increment in TTCA level, there were 0.052 (95 % CI: 0.013–0.091) mmol/L, 0.021 (0.011–0.031) ln-unit, 0.031 (0.019–0.042) ln-unit, 16.5 % (OR: 1.165; 95 % CI: 1.083–1.253) risk, 9.1 % (1.091; 1.017–1.170) risk, and 6.1 % (1.061; 1.034–1.088) risk increases in FPG, FPI, HOMA-IR, IR prevalence, IFG prevalence, and T2D prevalence, respectively (Table 2).

Table 2.

Cross-sectional associations of urinary TTCA with glucose homeostasis and T2D (N_observation = 5294).

Variable	Estimated effect (95 % CI) by continuous urinary TTCA	Estimated effect (95 % CI) by quartiles of urinary TTCA				Ptrenda
		Q1	Q2	Q3	Q4
Consecutive outcomes
β of FPG	0.052 (0.013, 0.091) *	0 (Reference)	0.048 (−0.118, 0.214)	0.075 (−0.090, 0.241)	0.180 (0.013, 0.347) *	0.048 *
β of FPI	0.021 (0.011, 0.031) *	0 (Reference)	0.027 (−0.016, 0.069)	0.036 (−0.007, 0.078)	0.101 (0.058, 0.144) *	<0.001 *
β of HOMA-IR	0.031 (0.019, 0.042) *	0 (Reference)	0.043 (−0.006, 0.092)	0.058 (0.009, 0.107) *	0.136 (0.087, 0.186) *	<0.001 *
β of HOMA-B	−0.001 (−0.005, 0.002)	0 (Reference)	−0.001 (−0.015, 0.013)	−0.001 (−0.015, 0.013)	−0.010 (−0.024, 0.004)	0.181
Dichotomous outcomes
OR of IR	1.165 (1.083, 1.253) *	1 (Reference)	1.255 (0.996, 1.582)	1.274 (1.010, 1.607) *	1.540 (1.222, 1.941) *	<0.001 *
OR of IFG	1.091 (1.017, 1.170) *	1 (Reference)	0.983 (0.720, 1.342)	1.221 (0.900, 1.656)	1.401 (1.038, 1.892) *	0.041 *
OR of T2D	1.061 (1.034, 1.088) *	1 (Reference)	0.939 (0.740, 1.190)	1.103 (0.961, 1.266)	1.159 (1.011, 1.329) *	0.040 *

Abbreviations: Q, quartile; TTCA, 2-Thiothiazolidine-4-carboxylic acid; OR, odds ratio; CI, confidence interval; BMI, body mass index; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; HOMA-B, homeostasis model assessment of β-cell function; IR, insulin resistance; IFG, impaired fasting glucose; T2D, type 2 diabetes.

Adjusted for age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use (a diagnostic criteria of T2D that was not adjusted when T2D was an outcome), history of hypertension, history of hyperlipidemia, and city with individual included as a random effect in the generalized linear mixed models.

^*P < 0.05.

^aP_trend value across quartiles of urinary TTCA was tested by including the median of each quartile of urinary TTCA as a continuous variable in the models.

Furthermore, we found significant positive longitudinal associations between TTCA and FPG, FPI, HOMA-IR, and incident risk of T2D (Table 3). In comparison to participants with consistently low TTCA level, those with consistently high TTCA level manifested 0.350 (0.060–0.640) mmol/L, 0.197 (0.097–0.297) ln-unit, 0.236 (0.120–0.351) ln-unit, and 56.3 % (hazard ratio [HR]: 1.563; 95 % CI: 1.018–2.401) risk increasements in FPG, FPI, HOMA-IR, and incident T2D, respectively. No statistically significant association was found between TTCA and HOMA-B in cross-sectional or longitudinal analysis, therefore subsequent analyses would not consider HOMA-B.

Table 3.

Longitudinal associations of urinary TTCA with glucose homeostasis and T2D.

Variable	Estimated effect (95 % CI) by categories of urinary TTCA			Ptrendc
	Persistently low	Inconsistent	Persistently high
Consecutive outcomesa
β of FPG	0 (Reference)	0.207 (−0.074, 0.489)	0.350 (0.060, 0.640) *	0.024 *
β of FPI	0 (Reference)	0.052 (−0.036, 0.141)	0.197 (0.097, 0.297) *	0.008 *
β of HOMA-IR	0 (Reference)	0.078 (−0.024, 0.179)	0.236 (0.120, 0.351) *	0.005 *
β of HOMA-B	0 (Reference)	−0.001 (−0.002, 0.001)	−0.001 (−0.003, 0.001)	0.192
Dichotomous outcomesb
HR of IR d	1 (Reference)	1.017 (0.697, 1.483)	1.253 (0.726, 2.163)	0.949
HR of IFG e	1 (Reference)	1.076 (0.794, 1.460)	1.098 (0.764, 1.576)	0.604
HR of T2D d	1 (Reference)	1.269 (0.834, 1.906)	1.563 (1.018, 2.401) *	0.041 *

Abbreviations: TTCA, 2-Thiothiazolidine-4-carboxylic acid; HR, hazard ratio; CI, confidence interval; BMI, body mass index; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; HOMA-B, homeostasis model assessment of β-cell function; IFG, impaired fasting glucose; IR, insulin resistance; T2D, type 2 diabetes.

There were 1418, 1551, and 1754 subjects were included in the IR, IFG, and T2D models after excluding prevalent cases at baseline, and 2771 observations were included in the FPG/FPI/HOMA-IR/HOMA-B models, respectively.

Individuals were divided into low (<median TTCA) or high (≥median TTCA) TTCA group in each study period according to the median level of TTCA, and further classified into three categories of persistently low group (participants with TTCA level < median consistently across all study periods; as the reference group), persistently high group (participants with TTCA level ≥ median consistently across all study periods), and inconsistent group (those neither belong to the group of persistently low nor high).

^aAdjusted for age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use, history of hypertension, history of hyperlipidemia, city, and FPG/FPI/HOMA-IR/HOMA-B at baseline with individual included as a random effect in the generalized linear mixed models.

^bAdjusted for age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use (a diagnostic criteria of T2D that was not adjusted when T2D was an outcome), history of hypertension, history of hyperlipidemia, and city in the COX regression models.

^cP_trend value across categories of urinary TTCA was tested by including the categories of urinary TTCA as a continuous variable in the models.

^dParticipants with T2D/IR at baseline were excluded for T2D/IR incidence analysis.

^eParticipants with IFG or T2D at baseline and those with T2D at follow-up were excluded for IFG incidence analysis.

^*P < 0.05.

As shown in Tables S8–S13, the associations of TTCA with glucose homeostasis and T2D did not change apparently after excluding participants with T2D (participants with IR were additionally excluded in IFG analysis) (Tables S8–S9), conducting additional model adjustment (Tables S10–S11), and performing IPW analysis (Tables S12–S13).

Relationships of PRS and HLI with glucose homeostasis and T2D

Cross-sectionally, we discovered positive associations of PRS with FPG (β: 0.106; 95 % CI: 0.041–0.170), FPI (0.467; 0.032–0.902), HOMA-IR (0.045; 0.022–0.068), IR prevalence (OR: 1.150; 95 % CI: 1.075–1.224), IFG prevalence (1.265; 1.122–1.426), and T2D prevalence (1.313; 1.083–1.593), as well as negative associations of HLI with FPG (−0.044; −0.084 to −0.003), FPI (−0.066; −0.078 to −0.053), HOMA-IR (−0.076; −0.091 to −0.062), IR prevalence (0.724; 0.683–0.767), IFG prevalence (0.902; 0.670–1.213), and T2D prevalence (0.868; 0.806–0.936) (P and/or P_trend < 0.05; Table S14). Longitudinally, we discovered positive associations of PRS with FPG (0.179; 0.075–0.282), FPI (0.040; 0.018–0.063), HOMA-IR (0.041; 0.016–0.066), IR incidence (HR: 1.077; 95 % CI: 1.010–1.149), IFG incidence (1.141; 1.002–1.298), and T2D incidence (1.241; 1.102–1.397), as well as negative associations of HLI with FPG (−0.025; −0.089 to 0.038), FPI (−0.059; −0.074 to −0.044), HOMA-IR (−0.071; −0.087 to −0.054), IR incidence (0.868; 0.813–0.927), and T2D incidence (0.808; 0.713–0.916) (P and/or P_trend < 0.05; Table S15).

Interactions of TTCA with PRS and/or HLI on glucose homeostasis and T2D

Interaction between TTCA and PRS was generally discovered. Cross-sectionally, subjects with high levels of TTCA and PRS generally had higher levels of FPG, FPI, and HOMA-IR and higher prevalent risks of IR, IFG, and T2D (all false discovery rate (FDR)-adjusted P for interaction < 0.05; Fig. S2). Longitudinally, subjects with persistently high TTCA and high PRS manifested 0.450 (95 % CI: 0.069–0.831) mmol/L elevation in FPG, 28.8 % (HR: 1.288; 95 % CI: 1.009–1.645) risk elevation in IR incidence, and 123.7 % (2.237; 1.211–4.131) risk elevation in T2D incidence, compared with subjects with persistently low TTCA and low PRS (all P for interaction < 0.05, FDR-adjusted P for interaction < 0.10; Fig. 1).

Fig. 1.

The longitudinal joint effects between urinary TTCA and PRS on glucose homeostasis and T2D. Abbreviations: TTCA, 2-Thiothiazolidine-4-carboxylic acid; PRS, polygenic risk score; HR, hazard ratio; CI, confidence interval; BMI, body mass index; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; IR, insulin resistance; IFG, impaired fasting glucose; T2D, type 2 diabetes; FDR, false discovery rate. There were 1418, 1551, and 1754 subjects were included in the IR, IFG, and T2D models after excluding prevalent cases at baseline, and 2771 observations were included in the FPG/FPI/HOMA-IR models, respectively. FPG (A), FPI (B), and HOMA-IR (C) models (generalized linear mixed models) were adjusted for the first 10 principal components of ancestry, age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use, history of hypertension, history of hyperlipidemia, city, and FPG/FPI/HOMA-IR at baseline with individual included as a random effect. IR (D), IFG (E), and T2D (F) models (COX regression models) were adjusted for the first 10 principal components of ancestry, age, gender, BMI, smoking status, drinking status, physical exercise, family history of diabetes, antidiabetic drug use (a diagnostic criteria of T2D that was not adjusted when T2D was an outcome), history of hypertension, history of hyperlipidemia, and city.

Interaction between TTCA and HLI was generally discovered. Cross-sectionally, subjects with low level of TTCA and healthy lifestyle generally had lower levels of FPG, FPI, and HOMA-IR and lower prevalent risks of IR, IFG, and T2D (all FDR-adjusted P for interaction < 0.05; Fig. S3). Longitudinally, subjects with persistently low TTCA and healthy lifestyle manifested 0.500 (−0.866 to −0.133) mmol/L decrement in FPG, 0.178 (−0.262 to −0.094) ln-unit decrement in FPI, 0.238 (−0.334 to −0.142) ln-unit decrement in HOMA-IR, and 58.2 % (0.418; 0.219–0.801) risk decrement in T2D incidence, compared to subjects with persistently high TTCA and unfavorable lifestyle (all FDR-adjusted P for interaction < 0.05; Fig. 2).

Fig. 2.

The longitudinal joint effects between urinary TTCA and HLI on glucose homeostasis and T2D. Abbreviations: TTCA, 2-Thiothiazolidine-4-carboxylic acid; HLI, healthy lifestyle index; HR, hazard ratio; CI, confidence interval; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; IR, insulin resistance; IFG, impaired fasting glucose; T2D, type 2 diabetes; FDR, false discovery rate. There were 1418, 1551, and 1754 subjects were included in the IR, IFG, and T2D models after excluding prevalent cases at baseline, and 2771 observations were included in the FPG/FPI/HOMA-IR models, respectively. FPG (A), FPI (B), and HOMA-IR (C) models (generalized linear mixed models) were adjusted for age, gender, family history of diabetes, antidiabetic drug use, history of hypertension, history of hyperlipidemia, city, and FPG/FPI/HOMA-IR at baseline with individual included as a random effect. IR (D), IFG (E), and T2D (F) models (COX regression models) were adjusted for age, gender, family history of diabetes, antidiabetic drug use (a diagnostic criteria of T2D that was not adjusted when T2D was an outcome), history of hypertension, history of hyperlipidemia, and city.

Interaction between HLI and PRS was also generally discovered (Figs. S4–S5), and ultimately, the overall interaction between TTCA, PRS, and HLI was further uncovered. Cross-sectionally, subjects with low levels of TTCA and PRS and healthy lifestyle generally had lower levels of FPI and HOMA-IR and lower prevalent risks of IR and IFG (all FDR-adjusted P for interaction < 0.05; Fig. S6). Longitudinally, subjects with persistently low TTCA, low PRS, and healthy lifestyle manifested 0.609 (−1.108 to −0.108) mmol/L decrement in FPG, 0.357 (−0.479 to −0.236) ln-unit decrement in FPI, 0.358 (−0.489 to −0.219) ln-unit decrement in HOMA-IR, and 77.2 % (0.228; 0.091–0.574) risk decrement in T2D incidence, compared to subjects with persistently high TTCA, high PRS, and unfavorable lifestyle (all FDR-adjusted P for interaction < 0.05; Fig. 3).

Fig. 3.

The longitudinal joint effects between urinary TTCA, PRS, and HLI on glucose homeostasis and T2D. Abbreviations: TTCA, 2-Thiothiazolidine-4-carboxylic acid; PRS, polygenic risk score; HLI, healthy lifestyle index; HR, hazard ratio; CI, confidence interval; FPG, fasting plasma glucose; FPI, fasting plasma insulin; HOMA-IR, homeostasis model assessment of insulin resistance; IR, insulin resistance; IFG, impaired fasting glucose; T2D, type 2 diabetes; FDR, false discovery rate. There were 1418, 1551, and 1754 subjects were included in the IR, IFG, and T2D models after excluding prevalent cases at baseline, and 2771 observations were included in the FPG/FPI/HOMA-IR models, respectively. FPG (A), FPI (B), and HOMA-IR (C) models (generalized linear mixed models) were adjusted for the first 10 principal components of ancestry, age, gender, family history of diabetes, antidiabetic drug use, history of hypertension, history of hyperlipidemia, city, and FPG/FPI/HOMA-IR at baseline with individual included as a random effect. IR (D), IFG (E), and T2D (F) models (COX regression models) were adjusted for the first 10 principal components of ancestry, age, gender, family history of diabetes, antidiabetic drug use (a diagnostic criteria of T2D that was not adjusted when T2D was an outcome), history of hypertension, history of hyperlipidemia, and city.

Discussion

In the present prospective cohort study of general urban adult population in China, we uncovered that increased CS₂ exposure biomarker TTCA and increased PRS were associated with dysregulated glucose homeostasis and elevated T2D risk, while increased HLI was associated with improved glucose homeostasis and decreased T2D risk. Importantly, we discerned interaction effects between TTCA, PRS, and HLI, with subjects with low TTCA, low PRS, and healthy lifestyle manifested remarkable improvement in glucose homeostasis and reduction in T2D risk. These findings demonstrated that daily-life exposure of the general adult population to CS₂ was associated with dysregulated glucose homeostasis and elevated T2D risk, which were interactively aggravated by genetic susceptibility while interactively alleviated by healthy lifestyle, highlighting the public health concerns about the detrimental impacts of daily-life CS₂ exposure on glucose metabolism and T2D particularly in those with high genetic risk and underscoring the crucial protective effect of healthy lifestyle. These findings propose that reducing CS₂ exposure in daily life and adopting a healthy lifestyle especially for those with genetic susceptibility may well conduce to maintaining glucose homeostasis and reduce T2D risk, providing a novel strategy for personalized and precise T2D primary prevention.

Our discoveries provided compelling evidence that even daily-life low-level CS₂ exposure could chronically disturb glucose-insulin metabolism and increase T2D risk of the community population [[13], [14], [15], [16], [17], [18], [19], [20], [21], [22],46]. Similar results were reported in animal experiments, where Sperlingová et al. documented impaired glucose tolerance in monkeys after 20 weeks of chronical CS₂ exposure [46]. Previous studies found consistent results in occupational workers with high CS₂ exposure. Franco et al. uncovered significantly elevated blood glucose and decreased glucose tolerance indicative of latent T2D in 66 CS₂-exposed workers compared to 66 unexposed controls in Italy [13]. Similarly, a study carried out in Korea discovered significantly increased fasting glucose in 170 CS₂-poisoned workers compared to 170 controls [22]. It is notable that prior studies have predominantly concentrated on occupational population exposed to high-level CS₂ in the workplace, where urinary TTCA level was reported to be 121.04 ± 43.09 μg/mmol Cr [47], significantly exceeding the urinary TTCA level of 3.05 ± 0.88 μg/mmol Cr observed in our study participants. This stark contrast highlights the difference between occupational and daily-life CS₂ exposure, with the latter traditionally regarded as too low to warrant concern. However, our findings challenge this assumption by demonstrating that long-term CS₂ exposure of the general adults in daily life can still result in dysregulated glucose homeostasis and heighten T2D risk, providing convincing evidence to support the causal role of CS₂ exposure in impairing glucose homeostasis and elevating T2D risk.

In the present study, no significant effect of TTCA on HOMA-B, an indicator of pancreatic β-cell function, was observed, which may be attributed to several factors. First, β-cells possess compensatory mechanisms that can maintain HOMA-B stability despite early functional impairments [48], potentially masking the adverse effects of TTCA. Additionally, HOMA-B has limited sensitivity to acute or mild β-cell dysfunctions [49], which may hinder the detection of subtle functional changes. Moreover, the predominantly low-level TTCA exposure among this population might have been insufficient to induce measurable β-cell damage [50,51]. Furthermore, the relatively short follow-up period may not have been adequate to capture the long-term progression of β-cell dysfunction [50,51]. Therefore, future studies should consider increasing the sample size, extending the follow-up duration, and employing more sensitive β-cell function assessment methods to better estimate the impact of TTCA on β-cell function.

Mechanistically, although not fully understood, CS₂ exposure may interfere with glycolytic pathway, affect insulin by binding with xanthurenic acid [46], and disturb glucose metabolism by dysregulating tryptophan metabolism pathway [15]. In addition, previous in vivo studies demonstrated that dose-dependent increases in lipid peroxidation and reactive oxygen species levels from CS₂ exposure were linked to insulin resistance, impaired glucose tolerance, and β-cell dysfunction [50,51]. Animal studies conducted by Chinese researchers have demonstrated that exposure to CS₂ may interfere with embryo implantation through DNA damage and oxidative stress, as indicated by notable increases in 8-OHdG, a marker of oxidative DNA damage, and lipid peroxide levels [51]. Supporting this, occupational studies have reported that among 67 workers exposed to CS₂, serum superoxide dismutase activity, a proxy for oxidative stress, was significantly elevated compared to controls [52]. Furthermore, both exposure concentration and exposure duration were positively related with increased lipid peroxidation, measured by serum malonyl dialdehyde level [52]. Similarly, Wronska-Nofer et al. found heightened plasma TBARS, an indicator of lipid peroxidation, in viscose rayon workers exposed to CS₂ [53]. Although our previous investigation revealed that lipid peroxidation may underlying the CS₂-associated FPG elevation [23], further toxicological researches are required to determine the specific mechanisms involved, particularly in CS₂ increasing T2D risk.

Although numerous genetic variants have been uncovered to be significantly related to glucose homeostasis and T2D [39,54], single variants have limited explanatory power for heritability, it is imperative to clarify their integrated effects on glucose homeostasis and T2D by employing valid tools such as PRS. Fortunately, the PRSs specifically constructed by multiple genetic variants in our study well predicted glucose dyshomeostasis and T2D. Accordingly, the constructed PRSs were further exploited to holistically test the gene-environment interaction between CS₂ exposure and genetic susceptibility, which could also explain more heritability. For the first time, interactions between CS₂ exposure and PRSs on glucose homeostasis and T2D were discerned in our study, where subjects with high genetic susceptibility and high CS₂ exposure generally exhibited the most pronounced elevation in T2D risk and adverse effect on glucose homeostasis, indicating that subjects with high genetic risk are more likely to develop glucose dyshomeostasis and T2D after daily-life exposure to CS₂. It is plausible as interindividual variability in the adverse effects of environmental pollutants on glucose-insulin metabolism and T2D due to genetic variation amplifies susceptibility to environmental pollutants [55,56]. These discoveries are conducive to identifying high-risk (susceptible) populations and precisely preventing glucose dyshomeostasis and T2D from CS₂ exposure in daily life.

Lifestyle modifications are considered the most cost-effective means of managing glucose dyshomeostasis and T2D [57,58]. A six-year intervention study of 110,660 men and women discovered that diet and/or exercise interventions reduced the risk of developing diabetes [57]. Besides, the Diabetes Remission Clinical Trial provided two-year evidence for durable remission of T2D following diet-induced weight loss [58]. Likewise, herein we identified and quantified a protective effect of increased HLI (constructed by integrating five common lifestyle factors including diet and exercise) indicative of a healthier lifestyle on glucose dyshomeostasis and T2D through both cross-sectional and longitudinal analyses. Subsequently, the interactions between CS₂ exposure and HLI on homeostasis and T2D were detected for the first time in our current research, where subjects with a healthy lifestyle and low CS₂ exposure generally showed the most evident decrement in T2D risk and protective effect on glucose homeostasis, hinting that improvement of lifestyle may interactively alleviate the detrimental impacts of CS₂ exposure on glucose homeostasis and T2D. Altogether, reducing daily-life CS₂ exposure, adopting a healthy lifestyle, or particularly combining both strategies are highly effective for preventing glucose dyshomeostasis and T2D.

Furthermore, for the first time, we discerned the overall three-way interactions of lifestyle, CS₂ exposure, and genetic susceptibility on glucose homeostasis and T2D, and the interaction effects surpassed the sum of the individual effects of the three factors. Subjects with low CS₂ exposure, low genetic risk, and healthy lifestyle manifested the maximum reduction in risk of T2D and protective effect on glucose homeostasis. All of these discoveries contribute to the primary preventions of glucose dyshomeostasis and T2D and are of far-reaching implication in guiding precise and personalized intervention strategies to control daily-life CS₂ exposure and subsequent glucose homeostasis impairment and T2D. It is imperative to concurrently reduce CS₂ exposure and improve lifestyle especially for those with high genetic risk to maintain glucose homeostasis and reduce T2D risk.

The present research has several notable strengths. First, this is the first repeated-measure prospective cohort study allowing causal inference to assess the long-term effects of chronic CS₂ exposure in daily life on glucose homeostasis and T2D among a relatively large Chinese adult population. Second, we uncovered multifaceted gene-CS₂-lifestyle interactions on glucose homeostasis and T2D, identifying that reducing daily-life CS₂ exposure and maintaining a healthy lifestyle are feasible interventions for the control and prevention of impaired glucose homeostasis and T2D, especially in populations with high genetic risk. Natheless, several limitations should be acknowledged. First, latent measurement error may be introduced given the utilization of morning spot urine other than 24-h urine for CS₂ exposure biomarker (TTCA) measurement. Nevertheless, due to the advantages of low cost and ease of collection as well as the general representativeness of spot morning urine, it has been widespreadly accepted and adopted in numerous epidemiological researches. Second, despite many potential confounding factors were considered in our research, residual confounding is always possible in observational studies. Third, the longitudinal analyses employed baseline HLI, while it is possible that lifestyle may fluctuate over time. Nonetheless, participants who took part in the study had resided in their local communities for a minimum of 5 years and exhibited a relatively stable lifestyle (the intra-class correlation of HLI was estimated to be 0.760 over follow-up time).

Conclusions

Our repeated-measure longitudinal study in a Chinese adult population-based prospective cohort demonstrated positive associations between chronic CS₂ exposure in daily life and dysregulated glucose homeostasis and elevated T2D risk. Furthermore, the deleterious effects of CS₂ exposure on glucose homeostasis and T2D risk were intensified by genetic susceptibility while attenuated by healthy lifestyle. Our findings emphasize the necessity for precise prevention and management of glucose dyshomeostasis and T2D through the reduction of CS₂ exposure and the enhancement of lifestyle, especially in people with high genetic susceptibility.

Compliance with ethics requirement

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation by the Ethics and Human Subject Committee of Tongji Medical College, Huazhong University of Science and Technology on February 10, 2011 (ID: No. 2011-17) and with the Helsinki Declaration of 1975, as revised in 2008. Informed consent was obtained from all participants for being included in the study.

CRediT authorship contribution statement

Yueru Yang: Conceptualization, Methodology, Formal analysis, Writing – original draft, Writing – review & editing. Jiahao Song: Investigation, Methodology, Writing – review & editing. Yongfang Zhang: Investigation, Methodology, Writing – review & editing. Shuhui Wan: Investigation, Methodology, Writing – review & editing. Zhiying Huo: Investigation, Methodology, Writing – review & editing. Qing Liu: Investigation, Methodology, Writing – review & editing. Le Hong: Investigation, Methodology, Writing – review & editing. Linling Yu: Investigation, Methodology, Writing – review & editing. Wei Liu: Investigation, Methodology, Writing – review & editing. Ruyi Liang: Investigation, Methodology, Writing – review & editing. Bin Wang: Data curation, Funding acquisition, Methodology, Supervision, Writing – review & editing. Weihong Chen: Data curation, Funding acquisition, Project administration, Supervision, Writing – review & editing.

Funding

This work was supported by the Key Program of the National Natural Science Foundation of China (82241088), the National Natural Science Foundation of China (82203996), and the Hubei Provincial Natural Science Foundation of China (2025AFB591).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: