Diabetes Care 1

Bin Wang,1,2 Weihong Qiu,1,2 Shijie Yang,1,2 Acrylamide Exposure and Limin Cao,1,2 Chunmei Zhu,1,2 Jixuan Ma,1,2 Wei Li,1,2 Zhuang Zhang,1,2 Tao Xu,1,2 Oxidative DNA Damage, Lipid Xing Wang,1,2 Man Cheng,1,2 Ge Mu,1,2 Dongming Wang,1,2 Yun Zhou,1,2 Peroxidation, and Fasting Plasma Jing Yuan,1,2 and Weihong Chen1,2 Alteration: Association andMediation Analyses inChinese Urban Adults PDMOOYHAT EVCSRESEARCH SERVICES /HEALTH https://doi.org/10.2337/dc19-2603

OBJECTIVE Acrylamide exposure from daily-consumed has raised global concern. We aimed to assess the exposure-response relationships of internal acrylamide exposure with oxidative DNA damage, lipid peroxidation, and fasting plasma glucose (FPG) alteration and investigate the mediating role of oxidative DNA damage and lipid peroxidation in the association of internal acrylamide exposure with FPG.

RESEARCH DESIGN AND METHODS FPG and urinary biomarkers of oxidative DNA damage (8-hydroxy-deoxyguanosine [8-OHdG]), lipid peroxidation (8-iso-prostaglandin-F2a [8-iso-PGF2a]), and acryl- exposure (N-acetyl-S-[2-carbamoylethyl]-L-cysteine [AAMA], N-acetyl-S-[2- carbamoyl-2-hydroxyethyl]-L-cysteine [GAMA]) were measured for 3,270 general adults from the Wuhan-Zhuhai cohort. The associations of urinary acrylamide 1Department of Occupational and Environmental metabolites with 8-OHdG, 8-iso-PGF2a, and FPG were assessed by linear mixed Health, School of Public Health, Tongji Medical models. The mediating roles of 8-OHdG and 8-iso-PGF2a were evaluated by College, Huazhong University of Science and mediation analysis. Technology, Wuhan, Hubei, China 2Key Laboratory of Environment and Health, RESULTS Ministry of Education & Ministry of Environmen- tal Protection, and State Key Laboratory of We found significant linear positive dose-response relationships of urinary acryl- Environmental Health (Incubating), School of amide metabolites with 8-OHdG, 8-iso-PGF2a, and FPG (except GAMA with FPG) Public Health, Tongji Medical College, Huazhong and 8-iso-PGF2a with FPG. Each 1-unit increase in log-transformed level of AAMA, University of Science and Technology, Wuhan, Hubei, China AAMA 1 GAMA (SUAAM), or 8-iso-PGF2a was associated with a 0.17, 0.15, or P P < Corresponding author: Weihong Chen, wchen@ 0.23 mmol/L increase in FPG, respectively ( and/or trend 0.05). Each 1% mails.tjmu.edu.cn S increase in AAMA, GAMA, or UAAM was associated with a 0.19%, 0.27%, or 0.22% Received 28 December 2019 and accepted 31 increase in 8-OHdG, respectively, and a 0.40%, 0.48%, or 0.44% increase in 8-iso- March 2020 PGF2a, respectively (P and P trend < 0.05). Increased 8-iso-PGF2a rather than This article contains supplementary material 8-OHdG significantly mediated 64.29% and 76.92% of the AAMA- and SUAAM- online at Figshare: https://doi.org/10.2337/ associated FPG increases, respectively. figshare.12076134. © 2020 by the American Diabetes Association. CONCLUSIONS Readers may use this article as long as the work is Exposure of the general adult population to acrylamide was associated with FPG properly cited, the use is educational and not for profit, and the work is not altered. More infor- elevation, oxidative DNA damage, and lipid peroxidation, which in turn partly mation is available at https://www.diabetesjournals mediated acrylamide-associated FPG elevation. .org/content/license. Diabetes Care Publish Ahead of Print, published online April 28, 2020 2 Acrylamide, Oxidative Damage, and FPG Diabetes Care

Acrylamide is an a,b-unsaturated car- damage and lipid peroxidation have been drugs and free of kidney disease were bonyl-containing volatile organic com- rarely investigated in epidemiological included in our final cross-sectional anal- pound with high reactivity and water studies. Urinary 8-hydroxy-deoxyguano- yses. There were no significant differ- solubility that is extensively used in sine (8-OHdG) and 8-iso-prostaglandin- ences between excluded and included gel experiments, water purification, oil F2a (8-iso-PGF2a) are separate critical participants in basic characteristics, such and sugar refining, and production of biomarkers of oxidativeDNA damage and as BMI, smoking status, drinking status, toiletries, varnishes, and plastics (1,2). lipid peroxidation that reflect the global and physical activity (all P . 0.05). The Besides, acrylamide is present in ciga- oxidative stress status of the body and research protocol was evaluated and rette mainstream smoke (3) and various play important roles in approved by the Ethics and Human Sub- commonly consumed heated and and glucose dyshomeostasis (13–15). ject Committee of Tongji Medical Col- beverages, particularly , Thus, assessment of the relationships be- lege, Huazhong University of Science and crisps, and , attracting a global tween acrylamide exposure and 8-OHdG Technology. All participants gave written concern (2,4). In addition to occupational and 8-iso-PGF2a in a general population informed consent for participation and activities, smoking and are potential would help to understand the carcino- storage and use of biospecimens. sources of acrylamide exposure for genic and hyperglycemic potentials of humans. acrylamide to humans and the potential Data and Biospecimen Collection As a proven animal and mechanism of acrylamide causing oxida- Standardized questionnaires and physi- human neurotoxicant with ubiquitous tive stress and affecting blood glucose cal examinations were carried out to presence in daily life, acrylamide has levels in humans (16). collect information on demographics, raised worldwide concern for its poten- In the current study, we measured FPG socioeconomics, lifestyles, anthropo- tial adverse health effect on the public and urinary 8-OHdG and 8-iso-PGF2a for metrics, medical care, diseases, and bio- (2). The World Health Organization has 3,270 general Chinese adults from the chemical indexes of urine and blood, called for further urgent research on the Wuhan-Zhuhai cohort. To evaluate ac- including FPG and fasting lipids. FPG health hazards of acrylamide, sounding rylamide exposure, we comprehensively was immediately measured by an enzy- an international health alarm for its measured urinary metabolites of both matic colorimetric method with a fully health risk (5). Studies in zebrafish and acrylamide (N-acetyl-S-[2-carbamoylethyl]- automated biochemical analyzer (RX rodents have shown a significant dose- L-cysteine [AAMA]) and daytona; Randox Laboratories, Crumlin, dependent increase in blood glucose (N-acetyl-S-[2-carbamoyl-2-hydroxyethyl]- U.K.). Biospecimens, including fasting levels after acrylamide exposure (6–8). L-cysteine [GAMA]) (17), a more reactive blood and morning urine, were collected High fasting plasma glucose (FPG) is an electrophilic and genotoxic epoxide bio- on the same day and separately stored important indicator for diabetes and the transformed from acrylamide by cyto- at 280°C and 220°C, respectively, for third leading risk factor for global disease chrome P450 2E1 (18). Urinary AAMA further analyses. Mean arterial pressure burden that accounts for 6.53 million and GAMA are widely accepted internal (MAP) was computed as [(diastolic pres- deaths (9). However, epidemiological biomarkers of acrylamide exposure from sure 3 2 1 systolic pressure) / 3]. study of the effect of acrylamide expo- various sources (19,20). We aimed to Participants who had smoked $1 ciga- sure on FPG is limited. A pilot study estimate the associations of urinary ac- rette/day or who had drank one or more conducted among 14 healthy volunteers rylamide metabolites with FPG, urinary times per week for .6 months were reported an FPG increase after a 4-week 8-OHdG, and 8-iso-PGF2a. We also in- defined as smokers and drinkers, respec- intake of acrylamide-rich food but a lack vestigatedthe mediating roles of 8-OHdG tively; otherwise, they were separately of statistical significance (10). Studies and 8-iso-PGF2a in the relationships identified as nonsmokers and non- with larger sample sizes are warranted between urinary acrylamide metabolites drinkers. Physical activity was defined to assess the association of acrylamide and FPG alteration to identify the po- as taking regular exercise two or more exposure with FPG alteration. tential mechanistic link. times per week and each time for $20 min Acrylamide has been listed as a prob- for .6 months. Education degree was able human carcinogen (group 2A) by the RESEARCH DESIGN AND METHODS categorized as low (#9years)andhigh International Agency for Research on Study Population (.9 years) education levels. Family in- (IARC) on the basis of the suffi- Participants were drawn from the Wuhan- come was divided into low income cient evidence for carcinogenicity in an- Zhuhai cohort, a Chinese community- (,40,000 yuan/year) and high income imals and mechanistic considerations, based prospective cohort that was ($40,000 yuan/year). while the epidemiological evidence is established in 2012 and comprised of inadequate (1). High priority has been 4,812 adults (18–80 years of age) who Urinary Acrylamide Metabolites given by the IARC to assess acrylamide had dwelled in Wuhan or Zhuhai city for Determination as a human carcinogen in future IARC at least 5 years, as described previously Determination of urinary acrylamide me- monographs (11). The unifying factor (21). Our current analyses were re- tabolites,includingAAMAandGAMA,was available for determining the toxicity stricted to 3,491 participants who had performed by an ultra-high-performance and carcinogenicity of chemicals is oxi- sufficient urine samples for measure- liquid chromatography system (Agilent dative stress, which then causes oxida- ments of acrylamide metabolites, creat- 1290 Infinity II; Agilent Technologies, tive DNA damage and lipid peroxidation inine (Cr), 8-OHdG, and 8-iso-PGF2a.Of Santa Clara, CA) coupled with electro- (12). Nevertheless, the associations of the 3,491 participants, 3,270 without spray tandem mass spectrometry (SCIEX acrylamide exposure with oxidative DNA missing data or taking hypoglycemic Triple Quad 6500; Applied Biosystems, care.diabetesjournals.org Wang and Associates 3

Foster City, CA) following a previously test for dichotomous variables and var- were 3.28 (1.96–5.36), 0.50 (0.32–0.75), published method (22) with minor mod- iance analysis for continuous variables. and 3.79 (2.32–6.11) for AAMA, GAMA, ifications. A 100-mL thawed aliquot of We conducted both continuous (each and SUAAM, respectively. As summa- urine from each participant was diluted 1-unit increase) and categorical (across rized inTable 1,themean ageof the study with mixed internal standards (AAMA- quartiles) analyses to assess the rela- population (29.88% male) was 53.02 d3 and GAMA-d3) and ammonium ac- tionships among urinary acrylamide me- years. As the SUAAM levels gradually etate to 1 mL for measurements of tabolites, biomarkers of oxidative stress increased, the concentrations of FPG, AAMA and GAMA, which were accu- (8-OHdG and 8-iso-PGF2a), and FPG by 8-iso-PGF2a, and 8-OHdG and the pro- rately and precisely quantified with R2 using linear mixed models with adjust- portions of males, smokers, drinkers, and of the standard calibration curve .0.998, ment for age, sex, BMI, smoking status, individuals with a family history of di- the recoveries of labeled internal stand- drinking status, physical activity, educa- abetes were increased, while the MAP ards in a range of 80–120%, the relative tion level, family income, family history levels and the percentages of partici- recoveries of target compounds in a of diabetes, total cholesterol, triglycer- pants with physical activity or high family range of 85–111%, and the coefficients ides, and MAP. City (Wuhan/Zhuhai) income were decreased. No trend was of variation ,10%. The limits of de- was simultaneously included as a ran- observed in age, BMI, education level, tection (LODs) were 0.6 ng/mL for dom effect in the models. We also fit- total cholesterol, and triglycerides. AAMA and 1.5 ng/mL for GAMA, and ted restricted cubic spline (RCS) models concentrationspffiffiffi below LOD were set at to graphically characterize their dose- Associations of Urinary Acrylamide LOD/ 2. Total urinary acrylamide me- response relationships. Stratified analyses Metabolites With FPG, 8-Iso-PGF2a, tabolites (SUAAM) was defined as the were conducted by major characteristics, and 8-OHdG sum of AAMA and GAMA. Valid concen- including age, sex, BMI, smoking status, As shown in Table 2, significant associ- trations of urinary acrylamide metabo- drinking status, and physical activity. ations of urinary acrylamide metabolites lites were corrected for urinary Cr and Effect modification by each stratified with increased FPG (except GAMA with finally expressed as mg/mmol Cr. characteristic was estimated by adding FPG), 8-iso-PGF2a, and 8-OHdG were an interaction term in the linear mixed found after adjusting for potential con- Urinary 8-OHdG Determination models. founders (P and/or P trend , 0.05). Each Urinary 8-OHdG was measured using We further performed mediation anal- 1-unit increase in log-transformed con- solid phase extraction followed by high- ysis to evaluate the mediating role of centration of AAMA or SUAAM was performance liquid chromatography with 8-OHdG or 8-iso-PGF2a in the associa- associated with a 0.17 mmol/L or electrochemical detection (Waters 2465; tions of urinary metabolites with FPG 0.15 mmol/L increase in FPG, respec- Waters, Milford, MA) according to a pre- using a method described by MacKinnon tively. Each 1% increase in AAMA, viously developed method (23) with mi- et al. (25). The following two linear mixed GAMA, or SUAAM was associated with nor modification. The detailed procedure models were fitted to estimate the direct a 0.40%, 0.48%, or 0.44% increase in has been described in our previously and indirect (mediated) effects: 8-iso-PGF2a, respectively, and 0.19%, published literature (24). Valid 8-OHdG 0.27% or 0.22% increase in 8-OHdG, concentration was calibrated by urinary respectively. Besides, significantly (all M 5 w0 1 wUAAMXUAAM 1 wCXC 1 v1 Cr and finally presented as mmol/mol P , 0.05) linear positive dose-response 5 z 1 z 1 z 1 z 1 v Cr. Y 0 UAAMXUAAM OSXOS CXC 2 relationships of urinary acrylamide me- tabolites with FPG (except GAMA with Urinary 8-Iso-PGF2a Determination FPG), 8-iso-PGF2a, and 8-OHdG were where Y signifies outcome (FPG), M signifies Urinary 8-iso-PGF2a was detected by a found by categorical analyses and were mediator (8-OHdG or 8-iso-PGF2a), X commercially available ELISA kit (Cayman UAAM visually shown by RCS models (Supple- signifies exposure (urinary acrylamide Chemical, Ann Arbor, MI) according to mentary Fig. 1). The positive association metabolite), X signifies confounder, the manufacturer’s instructions. The de- C of GAMA with FPG was not found to be z denotes direct effect, and w 3 tection range was 0.8–500 pg/mL, and UAAM UAAM statistically significant (Table 2 and Sup- z denotes mediated effect by the the sensitivity was 3 pg/mL. The intra- OS plementary Fig. 1). mediator. We further computed the CI of and interassay coefficients of variance Stratified analyses showed that sex the mediated effect with the PRODCLIN were ,5% and ,10%, respectively. Valid (male/female), age (,55/$55 years), program (25), and the proportion medi- 2 8-iso-PGF2a level was adjusted by uri- BMI (,24/$24 kg/m ), smoking status ated by the mediator with the equation nary Cr and finally shown as ng/mmol (smokers/nonsmokers), drinking status [w 3 z /(z 1 w 3 z ) 3 Cr. UAAM OS UAAM UAAM OS (drinkers/nondrinkers), and physical ac- 100%]. All statistical analyses were tivity (active/inactive) did not modify the performed using SAS 9.4 software (SAS Statistical Analyses significant associations of urinary acryl- Concentrations of urinary acrylamide Institute, Cary, NC). amide metabolites (except GAMA) with metabolites, 8-OHdG, and 8-iso-PGF2a FPG, and these associations were per- were log10-transformed because of their RESULTS sistent in subgroups of females, age ,55 2 right-skewed distributions (examined by Characteristics of the Study Population years, BMI ,24 kg/m , physical inactiv- Kolmogorov-Smirnov test). Basic charac- The median levels of urinary acrylamide ity, nonsmokers, and both drinkers and teristics by quartiles of SUAAM were metabolites with interquartile range nondrinkers (P and/or P trend , 0.05) analyzed by Cochran-Armitage trend (mg/mmol Cr) of the 3,270 participants (Table 3 and Supplementary Tables 1 and cyaie xdtv aae n FPG and Damage, Oxidative Acrylamide, 4

Table 1—Characteristics of study participants by quartiles of SUAAM and in all participants (N 5 3,270) Quartiles of SUAAM (mg/mmol Cr) Characteristic All participants Q1 (#2.32) Q2 (2.33–3.79) Q3 (3.80–6.11) Q4 (.6.11) P for trend Participants, n 3,270 817 818 818 817 Male sex 977 (29.88) 223 (27.29) 195 (23.84) 247 (30.20) 312 (38.19) ,0.001 Age (years) 53.02 6 12.82 53.22 6 12.61 52.50 6 13.37 52.79 6 12.5 53.57 6 12.76 0.500 BMI (kg/m2) 24.00 6 3.53 24.02 6 3.41 24.01 6 3.55 24.01 6 3.45 23.97 6 3.72 0.781 Smokers 485 (14.83) 56 (6.85) 60 (7.33) 129 (15.77) 240 (29.38) ,0.001 Drinkers 418 (12.78) 82 (10.04) 80 (9.78) 113 (13.81) 143 (17.50) ,0.001 Physical activity 1,595 (48.78) 433 (53.00) 394 (48.17) 403 (49.27) 365 (44.68) 0.002 Education levels 0.299 Low (#9 years) 1,957 (59.85) 479 (58.63) 510 (62.35) 511 (62.47) 457 (55.94) High (.9 years) 1,313 (40.15) 338 (41.37) 308 (37.65) 307 (37.53) 360 (44.06) Family income, yuan/year 0.015 Low (,40,000) 1,822 (55.72) 452 (55.32) 421 (51.47) 458 (55.99) 491 (60.10) High ($40,000) 1,448 (44.28) 365 (44.68) 397 (48.53) 360 (44.01) 326 (39.90) Family history of diabetes 209 (6.39) 40 (4.90) 56 (6.85) 49 (5.99) 64 (7.83) 0.037 MAP (mmHg) 95.57 6 13.26 96.99 6 14.12 95.10 6 12.59 95.49 6 13.18 94.70 6 13.03 0.001 Total cholesterol (mmol/L) 5.12 6 1.33 5.18 6 1.16 5.07 6 1.19 5.18 6 1.76 5.07 6 1.10 0.298 Triglycerides (mmol/L) 1.51 6 1.23 1.54 6 1.41 1.46 6 1.07 1.49 6 1.16 1.55 6 1.25 0.772 FPG (mmol/L) 4.85 6 1.57 4.71 6 1.48 4.76 6 1.54 4.92 6 1.5 5.02 6 1.74 ,0.001 8-iso-PGF2a (ng/mmol Cr) 61.46 39.20 54.93 69.00 98.20 ,1 0.00 (38.06–108.26) (25.77–62.55) (36.58–87.09) (45.46–111.46) (60.38–174.36) 8-OHdG (mmol/mol Cr) 62.12 51.92 62.86 66.13 72.21 ,0.001 (28.24–122.51) (23.05–102.88) (28.18–116.92) (31.34–126.00) (31.04–157.08) Data are n (%) or mean 6 SD or median (interquartile range). ibtsCare Diabetes care.diabetesjournals.org Wang and Associates 5

Table 2—Associations of urinary acrylamide metabolites with FPG, 8-iso-PGF2a, and 8-OHdG (N 5 3,270) Effect estimates (95% CI) by quartiles of metabolites Effect estimates by Variablecontinuous metabolites Q1 Q2 Q3 Q4 P for trend* AAMA (mg/mmol Cr) #1.96 1.97–3.28 3.29–5.36 .5.36 Estimated change (mmol/L) for FPG 0.17 (0.01, 0.33) 0 (Ref) 0.11 (20.03, 0.26) 0.17 (0.02, 0.31) 0.20 (0.05, 0.35) 0.008 Adjusted b for 8-iso-PGF2a 0.40 (0.37, 0.44) 0 (Ref) 0.14 (0.11, 0.18) 0.21 (0.18, 0.25) 0.34 (0.31, 0.37) ,0.001 Adjusted b for 8-OHdG 0.19 (0.13, 0.26) 0 (Ref) 0.10 (0.05, 0.16) 0.15 (0.09, 0.21) 0.19 (0.13, 0.25) ,0.001 GAMA (mg/mmol Cr) #0.32 0.33–0.50 0.51–0.75 .0.75 Estimated change (mmol/L) for FPG 0.03 (20.15, 0.21) 0 (Ref) 0.01 (20.14, 0.15) 0.01 (20.14, 0.15) 0.05 (20.10, 0.19) 0.581 Adjusted b for 8-iso-PGF2a 0.48 (0.44, 0.52) 0 (Ref) 0.14 (0.11, 0.17) 0.22 (0.19, 0.25) 0.33 (0.30, 0.37) ,0.001 Adjusted b for 8-OHdG 0.27 (0.20, 0.34) 0 (Ref) 0.12 (0.06, 0.18) 0.15 (0.09, 0.20) 0.18 (0.12, 0.24) ,0.001 ∑UAAM (mg/mmol Cr) #2.32 2.33–3.79 3.80–6.11 .6.11 Estimated change (mmol/L) for FPG 0.15 (20.01, 0.32) 0 (Ref) 0.06 (20.09, 0.20) 0.15 (0.01, 0.30) 0.17 (0.02, 0.32) 0.014 Adjusted b for 8-iso-PGF2a 0.44 (0.40, 0.47) 0 (Ref) 0.14 (0.11, 0.17) 0.22 (0.19, 0.25) 0.34 (0.31, 0.38) ,0.001 Adjusted b for 8-OHdG 0.22 (0.15, 0.28) 0 (Ref) 0.09 (0.04, 0.15) 0.14 (0.09, 0.20) 0.19 (0.13, 0.24) ,0.001 Adjusted for age, sex (male/female), BMI, smoking status (smokers/nonsmokers), drinking status (drinkers/nondrinkers), physical activity (active/inactive), education level (low/high), family income (low/high), family history of diabetes (yes/no), total cholesterol, triglycerides, and MAP and included city (Wuhan/Zhuhai) as a random effect in the linear mixed models. Ref, reference. *P for trend across quartiles of urinary acrylamide metabolites was tested by including the median of each quartile of urinary acrylamide metabolites as a continuous variable in the linear mixed models.

2). Besides, the significant associations of Mediation Analysis attention. Our findings highlight the SUAAM with 8-iso-PGF2a and 8-OHdG As presented in Table 4, the associations public health concern regarding the were not modified by the stratified char- of FPG with AAMA and SUAAM rather carcinogenic and hyperglycemic effects acteristics and were persistent in almost than GAMA were significantly mediated on the general population from wide- all subgroups (P and/or P trend , 0.05), by elevated 8-iso-PGF2a, which medi- spread acrylamide exposure and provide except that the relationship between ated 64.29% and 76.92% of the AAMA- vital clues for illuminating the mecha- SUAAM and 8-OHdG in the small- and SUAAM-associated FPG increases, nisms whereby acrylamide raises blood sample-sized subgroups of smokers or respectively. Unfortunately, the associ- glucose. Policies or advices from author- drinkers was not significant (Table 3). ations of urinary acrylamide metabolites ities are warranted to restrict or control Stratified analyses for the associations with FPG were not found to be mediated acrylamide exposure of the public of 8-iso-PGF2a and 8-OHdG with AAMA by 8-OHdG (data not shown). In addition, from various sources, particularly daily- and GAMA (Supplementary Tables 1 we found no interaction effects be- consumed acrylamide-containing foods, and 2) showed similar results with tween urinary acrylamide metabolites such as potato crisps (average content a SUAAM. and 8-iso-PGF2 or 8-OHdG on FPG (all 752 mg/kg), coffee (average content . P 0.05). 509 mg/kg), and toasts (average content Associations of 8-Iso-PGF2a and 446 mg/kg) (2), which are estimated to 8-OHdG With FPG CONCLUSIONS Supplementary Table 3 presents the as- In this study, we found that urinary account for one-third of the calories sociation of 8-iso-PGF2a with FPG in all acrylamide metabolites were positively consumed by U.S. and European popu- participants and stratified by major char- associated, in a dose-dependent manner, lations (26). acteristics. After adjusting for potential with FPG (except GAMA with FPG) and The hyperglycemic effect of acrylam- confounders, each 1-unit increase in log- urinary 8-iso-PGF2a and 8-OHdG in a ide has been documented in experimen- fi transformed concentration of 8-iso-PGF2a general urban adult population. Besides, tal studies in zebra sh and rodents, fi was associated with a 0.23 mmol/L in- 8-iso-PGF2a was dose-dependently as- where a signi cant dose-dependent in- crease in FPG in all participants (P , 0.05). sociated with FPG increase and signifi- crease in blood glucose was shown after We also found a significant (all P , 0.05) cantly mediated AAMA- and SUAAM- acrylamide exposure (6–8). However, linear positive dose-response relation- associated FPG elevations. Our findings epidemiological evidence was limited. ship between 8-iso-PGF2a and FPG by indicate that acrylamide exposure is as- Naruszewicz et al. (10) recruited 14 Polish categorical analyses and RCS regressions sociated with FPG elevation and in- adults to chronically ingest acrylamide- (SupplementaryFig.2).Inaddition,stratified creased oxidative DNA damage and lipid containing potato chips and found that analyses found that such association was peroxidation. Moreover, lipid peroxida- FPG was increased after 28 days of in- not modified by major characteristics and tion may be further involved in the mech- gestion, although no statistical signifi- was persistent (P and/or P trend , 0.05) anism underlying acrylamide-associated cance was detected because of the in almost all subgroups except in the FPG elevation. small study sample size. Besides, Lin small-sample-sized subgroups of males Our findings have substantial public et al. (27) cross-sectionally evaluated and smokers (Supplementary Table 3). health implications. Acrylamide expo- but failed to find a statistically signifi- We did not observe a significant relation- sure along with a resultant potential cant association between urinary AAMA ship between 8-OHdG and FPG (data health hazard is a major public health and serum glucose among 675 Taipei not shown). issue that has attracted worldwide residents aged 12–30 years. Our present cyaie xdtv aae n FPG and Damage, Oxidative Acrylamide, 6

Table 3—Stratified analysis for the associations of SUAAM with FPG, 8-iso-PGF2a, and 8-OHdG (N 5 3,270) FPG 8-iso-PGF2a 8-OHdG Estimated change P for P for Adjusted b P for P for Adjusted b P for P for Stratification characteristic (mmol/L) (95% CI)* trend† modification‡ (95% CI)* trend† modification‡ (95% CI)* trend† modification‡ Sex 0.519 0.173 0.066 Male (n 5 977) 0.25 (20.11, 0.61) 0.710 0.41 (0.34, 0.47) ,0.001 0.11 (20.01, 0.23) 0.043 Female (n 5 2,293) 0.12 (20.06, 0.31) 0.037 0.44 (0.40, 0.48) ,0.001 0.25 (0.17, 0.32) ,0.001 Age (years) 0.323 0.387 0.704 ,55 (n 5 1,702) 0.25 (0.04, 0.46) 0.048 0.46 (0.41, 0.51) ,0.001 0.19 (0.09, 0.28) ,0.001 $55 (n 5 1,568) 0.09 (20.17, 0.35) 0.112 0.42 (0.37, 0.47) ,0.001 0.24 (0.15, 0.33) ,0.001 BMI (kg/m2) 0.426 0.133 0.343 ,24 (n 5 1,722) 0.12 (20.11, 0.34) 0.022 0.46 (0.41, 0.51) ,0.001 0.24 (0.16, 0.33) ,0.001 $24 (n 5 1,548) 0.18 (20.07, 0.42) 0.201 0.41 (0.36, 0.46) ,0.001 0.19 (0.09, 0.29) ,0.001 Smoking status 0.370 0.540 0.059 Smokers (n 5 485) 0.01 (20.53, 0.54) 0.871 0.46 (0.37, 0.56) ,0.001 0.04 (20.13, 0.21) 0.588 Nonsmokers (n 5 2,785) 0.18 (0.01, 0.35) 0.009 0.43 (0.40, 0.47) ,0.001 0.24 (0.17, 0.31) ,0.001 Drinking status 0.097 0.510 0.058 Drinkers (n 5 418) 0.57 (0.04, 1.11) 0.302 0.41 (0.31, 0.51) ,0.001 0.03 (20.16, 0.21) 0.697 Nondrinkers (n 5 2,852) 0.11 (20.07, 0.28) 0.038 0.44 (0.40, 0.47) ,0.001 0.24 (0.18, 0.31) ,0.001 Physical activity 0.594 0.632 0.667 Active (n 5 1,595) 0.07 (20.19, 0.33) 0.240 0.43 (0.38, 0.48) ,0.001 0.21 (0.12, 0.30) ,0.001 Inactive (n 5 1,675) 0.24 (0.04, 0.45) 0.010 0.45 (0.40, 0.49) ,0.001 0.21 (0.12, 0.30) ,0.001 *Adjusted for age, sex (male/female), BMI, smoking status (smokers/nonsmokers), drinking status (drinkers/nondrinkers), physical activity (active/inactive), education level (low/high), family income (low/high), family history of diabetes (yes/no), total cholesterol, triglycerides, and MAP and included city (Wuhan/Zhuhai) as a random effect in the linear mixed models. †P for trend across quartiles of SUAAM in each subgroup was tested by including the median of each SUAAM quartile as a continuous variable in the linear mixed models. ‡P for modification of each characteristic was calculated by including a product term of SUAAM with each stratified characteristic in the linear mixed models. ibtsCare Diabetes care.diabetesjournals.org Wang and Associates 7

Table 4—Mediated effects by 8-iso-PGF2a on associations between urinary acrylamide metabolites and FPG (N 5 3,270) Urinary metabolite Direct effects (95% CI)* Mediated effects (95% CI)† Proportion mediated by 8-iso-PGF2a,% AAMA 0.05 (20.12, 0.21) 0.09 (0.02, 0.15) 64.29 GAMA 20.07 (20.27, 0.13) 0.10 (0.02, 0.18) d‡ SUAAM 0.03 (20.16, 0.20) 0.10 (0.03, 0.17) 76.92

AAMA, GAMA, SUAAM, FPG, and 8-iso-PGF2a were modeled as continuous variables with log10-transformations. *Adjusted for age, sex (male/female), BMI, smoking status (smokers/nonsmokers), drinking status (drinkers/nondrinkers), physical activity (active/inactive), education level (low/high), family income (low/high), family history of diabetes (yes/no), total cholesterol, triglycerides, MAP, and 8-iso-prostaglandin-F2a and included city (Wuhan/Zhuhai) as a random effect in the linear mixed models. †Tested by the PRODCLIN program; CIs not containing a 0 value denote statistical significance. ‡Proportion mediated by 8-iso-PGF2a was not calculated because of the insignificant total effect. large general adult population–based cross-sectional study conducted on 76 In- Our study has some vital strengths. study clearly demonstrated a significant dian Mauritians showed a positive re- First, we provide unequivocal epidemi- positive dose-response relationship be- lationship between 8-iso-PGF2a and FPG ological evidence for the significant as- tween AAMA and FPG. Such discrepancy (31), and similar results were also found sociations of acrylamide exposure with is probably due to the differences in age, by Brinkmann et al. (32) and Altomare oxidative DNA damage, lipid peroxida- ethnicity, genetics, lifestyles, and sample et al. (33). Besides, Rytter et al. (34) tion, and FPG elevation. Second, we size of the study population. Noteworthy, investigated 56 Swedish patients with further assessed the mediating role of the association of GAMA with FPG was diabetes and found that urinary 8-iso- oxidative DNA damage and lipid perox- not observed in our study.Unlike AAMA, a PGF2a, but not 8-OHdG, was positively idation and found for the first time that direct urinary metabolite of acrylamide, associated with blood glucose. These lipid peroxidation significantly mediated GAMA is a direct urinary metabolite of studies support our findings of a signif- acrylamide-associated FPG elevation, glycidamide biotransformed from acryl- icantrelationshipbetweenFPGand8-iso- providing an important clue for a further amide. It seems likely that acrylamide PGF2a rather than 8-OHdG. However, a mechanism of study. Third, our present rather than glycidamide may be the sub- lack of statistical relationship between study was conducted in a large and stantial contributor for FPG elevation. 8-iso-PGF2a and FPG was reported in representative Chinese general adult The capacity of acrylamide in inducing a cross-sectional study conducted in a population from central and southern oxidative damage to DNA and lipid has 77-year-old Swedish population (n 5 China, making our results better repre- been identified in in vitro and in vivo 765), where such an association was as- sentative. Fourth, urinary metabolites studies (28–30), while assessed in few sumed to be concealed by the advanced of both acrylamide and its epoxide gly- epidemiological studies. Similar to our age and age-related complications (35). cidamide (i.e., AAMA, GAMA) were ac- results, a cross-sectional study conducted Our present study provides clear evidence curately measured to comprehensively in 800 adolescents and young adults for the positive dose-response relation- estimate the individual acrylamide expo- in Taiwan found a significant positive ship between 8-iso-PGF2a and FPG in a sure levels and their associations with association between urinary AAMA and large general population after adjusting 8-iso-PGF2a, 8-OHdG, and FPG, making 8-OHdG (27), a biomarker of oxidative for potential confounders. our fi ndings more convincing. The limi- DNA damage. Unfortunately, investiga- Further mediation analysis found that tations that should be noted for our tors of that study (27) did not consider urinary 8-iso-PGF2a rather than 8-OHdG present study included a lack of external or measure GAMA, another important significantly mediated the positive asso- acrylamide exposure data (e.g., diet), urinary biomarker of acrylamide expo- ciations of AAMA and SUAAM with FPG, which should be taken into consideration sure directly metabolized from glyci- indicating the involvement of lipid per- to be collected in our future studies. damide, which is much more reactive oxidation rather than oxidative DNA Besides, the cross-sectional design limits and largely responsible for the geno- damage in the mechanisms whereby inference on temporality. Additionally, toxicity and carcinogenicity of acryl- acrylamide increases FPG. Acrylamide- we used spot morning rather than 24-h amide (18). In our study, we found that induced lipid peroxidation could cause urine samples to measure acrylamide both AAMA and GAMA were posi- adipose inflammatory changes and in- metabolite levels, which may be subject tively associated with urinary 8-OHdG sulin signaling abnormalities, which could to measurement error. However, be- and urinary 8-iso-PGF2a, a biomarker lead to insulin resistance and finally, FPG cause of the stable lifestyles and dietary of lipid peroxidation. Furthermore, the elevation (36,37). Besides, acrylamide- patterns and nonexistent occupational associations of 8-OHdG and 8-iso-PGF2a induced lipid peroxidation could raise exposure of our study population, par- with GAMA were found to be stronger FPG by causing b-cell dysfunction and ticipants’ exposure levels of acrylamide than with AAMA, suggesting a proba- apoptosis (38,39). The unobserved me- was supposed to be relatively stable. blystrongeroxidativedamagepotency diating role of 8-OHdG might be inter- In conclusion, our study results consti- of glycidamide toward DNA and lipid preted by its novel biological functions tute an unequivocal demonstration that than acrylamide. in anti-inflammation and amelioration exposure of the general population to Several epidemiological studies with of insulin resistance (40,41). More studies acrylamide is significantly associated with relatively small sample sizes have eval- are warranted to elucidate the mechanisms FPGelevation,oxidativeDNAdamage,and uated the associations of urinary 8-iso- underlying acrylamide-induced glucose lipid peroxidation, which in turn signifi- PGF2a and 8-OHdG with blood glucose. A elevation. cantly mediates acrylamide-associated 8 Acrylamide, Oxidative Damage, and FPG Diabetes Care

FPG elevation. Our study findings sup- of 84 behavioural, environmental and occupa- for the indirect effect: program PRODCLIN. Behav port the hypothesis of an environmental tional, and metabolic risks or clusters of risks Res Methods 2007;39:384–389 etiology of FPG elevation (adverse out- for 195 countries and territories, 1990-2017: 26. Mucci LA, Wilson KM. Acrylamide intake through diet and human cancer risk. J Agric come) with acrylamide exposure induc- a systematic analysis for the Global Burden of Disease Study 2017 [published corrections ap- Food Chem 2008;56:6013–6019 ing lipid peroxidation (a mechanistic pear in Lancet 2019;393:132; Lancet 2019;393: 27. Lin CY, Lee HL, Chen YC, et al. Positive association link). Further research is warranted to e44]. Lancet 2018;392:1923–1994 between urinary levels of 8-hydroxydeoxyguanosine corroborate our findings and illuminate 10. Naruszewicz M, Zapolska-Downar D, Kosmider´ and the acrylamide metabolite N-acetyl-S-(propio- A, et al. Chronic intake of potato chips in humans namide)-cysteine in adolescents and young adults. the underlying mechanisms. – increases the production of reactive oxygen radicals J Hazard Mater 2013;261:372 377 by leukocytes and increases plasma C-reactive pro- 28. Pan X, Zhu L, Lu H, Wang D, Lu Q, Yan H. tein: a pilot study. Am J Clin Nutr 2009;89:773–777 Melatonin attenuates oxidative damage induced Acknowledgments. All participants included in 11. CoglianoV,BaanR,StraifK,etal.Future priorities by acrylamide in vitro and in vivo. Oxid Med Cell our study and all members of our study team are for IARC monographs. Lancet Oncol 2008;9:708 Longev 2015;2015:703709 greatly acknowledged. 12. Wang S, Bai Y, Deng Q, et al. Polycyclic 29. Abdel-Daim MM, Abd Eldaim MA, Hassan Funding. This study was supported by the Key aromatic hydrocarbons exposure and lung func- AG. Trigonella foenum-graecum ameliorates Program of the National Natural Science Foun- tion decline among coke-oven workers: a four- acrylamide-induced toxicity in rats: roles of oxi- fl dation of China (91543207) and the Major Re- year follow-up study.Environ Res2016;150:14–22 dative stress, proin ammatory cytokines, and – search Program of the National Natural Science 13. Wu LL, Chiou CC, Chang PY, Wu JT. Urinary DNA damage. Biochem Cell Biol 2015;93:192 198 Foundation of China (91843302). 8-OHdG: a marker of oxidative stress to DNA 30. Alturfan AA, Tozan-Beceren A, Sehirli AO, Duality of Interest. No potential conflicts of and a risk factor for cancer, atherosclerosis and Demiralp E, Sener G, Omurtag GZ. Resveratrol interest relevant to this article were reported. diabetics. Clin Chim Acta 2004;339:1–9 ameliorates oxidative DNA damage and protects Author Contributions. B.W. performed the 14. Klaunig JE, Kamendulis LM. The role of against acrylamide-induced oxidative stress in – statistical analyses, interpreted the data, and oxidative stress in carcinogenesis. Annu Rev rats. Mol Biol Rep 2012;39:4589 4596 drafted the manuscript. B.W., W.Q., S.Y., L.C., Pharmacol Toxicol 2004;44:239–267 31. Gopaul NK, Manraj MD, Heb´ e´ A, et al. C.Z., J.M., W.L., Z.Z., T.X., X.W., M.C., G.M., D.W., 15. Dalle-Donne I, Rossi R, Colombo R, Giustarini Oxidative stress could precede endothelial dys- Y.Z., J.Y., and W.C. contributed to the data D, Milzani A. Biomarkers of oxidative damage in function and insulin resistance in Indian Mau- acquisition, manuscript revision, review for im- human disease. Clin Chem 2006;52:601–623 ritians with impaired glucose metabolism. – portant intellectual content, and final version 16. Tangvarasittichai S. Oxidative stress, insulin Diabetologia 2001;44:706 712 approval. B.W. and W.C. conceived and designed resistance, dyslipidemia and type 2 diabetes 32. Brinkmann C, Neumann E, Blossfeld J, et al. fl fi this study. W.C. was responsible for the design of mellitus. World J Diabetes 2015;6:456–480 In uence of glycemic status and physical tness the Wuhan-Zhuhai cohort study and obtaining 17. Boettcher MI, Bolt HM, Drexler H, Angerer J. on oxidative stress and the peroxiredoxin system funding.W.C. isthe guarantorof thiswork and, as Excretion of mercapturic acids of acrylamide and in the erythrocytes of non-insulin-dependent such, had full access to all the data in the study glycidamide in human urine after single oral type 2 diabetic men. Exp Clin Endocrinol Diabetes and takes responsibility for the integrity of the administration of deuterium-labelled acrylam- 2011;119:559–564 data and the accuracy of the data analysis. ide. Arch Toxicol 2006;80:55–61 33. Altomare E, Vendemiale G, Chicco D, 18. Besaratinia A, Pfeifer GP. Genotoxicity of Procacci V, Cirelli F. Increased lipid peroxidation References acrylamide and glycidamide. J Natl Cancer Inst in type 2 poorly controlled diabetic patients. 1. Acrylamide. IARC Monogr Eval Carcinog Risks 2004;96:1023–1029 Diabete Metab 1992;18:264–271 Hum 1994;60:389–433 19. Boettcher MI, Schettgen T, Kutting¨ B, 34. Rytter E, Vessby B, Asgard˚ R, et al. Glycaemic 2. Joint FAO/WHO Expert Committee on Food Pischetsrieder M, Angerer J. Mercapturic acids status in relation to oxidative stress and inflam- Additives & World Health Organization. Evalua- of acrylamide and glycidamide as biomarkers of mation in well-controlled type 2 diabetes sub- tion of certain food contaminants: sixty-fourth the internal exposure to acrylamide in the gen- jects. Br J Nutr 2009;101:1423–1426 report of the Joint FAO/WHO Expert Committee eral population. Mutat Res 2005;580:167–176 35. Helmersson J, Vessby B, Larsson A, Basu S. on Food Additives [Internet], 2006. Geneva, World 20. Sams C, Jones K, Warren N, et al. Towards a Association of type 2 diabetes with cyclooxygenase- Health Organization. Available from http:// biological monitoring guidance value for acryl- mediated inflammation and oxidative stress in www.who.int/iris/handle/10665/43258. Ac- amide. Toxicol Lett 2015;237:30–37 an elderly population. Circulation 2004;109: cessed 28 November 2019 21. Song Y, Hou J, Huang X, et al. The Wuhan- 1729–1734 3. SmithCJ,PerfettiTA, RumpleMA,RodgmanA, Zhuhai(WHZH)cohortstudyof environmentalair 36. Fantuzzi G. Adipose tissue, adipokines, and Doolittle DJ. “IARC group 2A ” re- particulate matter and the pathogenesis of car- inflammation. J Allergy Clin Immunol 2005;115: ported in cigarette mainstream smoke. Food diopulmonary diseases: study design, methods 911–919; quiz 920 Chem Toxicol 2000;38:371–383 and baseline characteristics of the cohort. BMC 37. Ruskovska T, Bernlohr DA. Oxidative stress 4. Mottram DS, Wedzicha BL, Dodson AT. Ac- Public Health 2014;14:994 and protein carbonylation in adipose tissue - rylamide is formed in the . 22. Alwis KU, Blount BC, Britt AS, Patel D, Ashley implications for insulin resistance and diabetes Nature 2002;419:448–449 DL. Simultaneous analysis of 28 urinary VOC mellitus. J Proteomics 2013;92:323–334 5. Kapp C. WHO urges more research into metabolites using ultra high performance liquid 38. Sasson S. Nutrient overload, lipid peroxida- acrylamide in food. Lancet 2002;360:64 chromatography coupled with electrospray ion- tion and pancreatic beta cell function. Free Radic 6. Ghorbel I, Elwej A, Chaabene M, et al. Effects of ization tandem mass spectrometry (UPLC-ESI/ Biol Med 2017;111:102–109 acrylamide graded doses on metallothioneins I and MSMS). Anal Chim Acta 2012;750:152–160 39. Maulucci G, Daniel B, Cohen O, Avrahami Y, II induction and DNA fragmentation: biochemical 23. Yuan J, Chen L, Chen D, et al. Elevated serum Sasson S. Hormetic and regulatory effects of lipid and histomorphological changes in the liver of adult polybrominated diphenyl ethers and thyroid- peroxidation mediators in pancreatic beta cells. rats. Toxicol Ind Health 2017;33:611–622 stimulating hormone associated with lympho- Mol Aspects Med 2016;49:49–77 7. Kim SM, Baek JM, Lim SM, et al. Modified cytic micronuclei in Chinese workers from an 40. Choi S, Choi HH, Lee SH, et al. Anti-inflam- lipoproteins by acrylamide showed more athero- E-waste dismantling site. Environ Sci Technol matory effects of 8-hydroxy-29-deoxyguanosine genic properties and exposure of acrylamide in- 2008;42:2195–2200 on lipopolysaccharide-induced inflammation via duces acute hyperlipidemia and fatty liver changes 24. Xiao L, Zhou Y, Ma J, et al. Oxidative DNA Rac suppression in Balb/c mice. Free Radic Biol in zebrafish. Cardiovasc Toxicol 2015;15:300–308 damage mediates the association between uri- Med 2007;43:1594–1603 8. Lee HW, Pyo S. Acrylamide induces adipocyte nary metals and prevalence of type 2 diabetes 41. Huh JY, Jung I, Piao L, Ha H, Chung MH. 8- differentiation and obesity in mice. Chem Biol mellitus in Chinese adults. Sci Total Environ 2018; Hydroxy-2-deoxyguanosine ameliorates high-fat Interact 2019;298:24–34 627:1327–1333 diet-induced insulin resistance and adipocyte 9. GBD 2017 Risk Factor Collaborators. Global, 25. MacKinnon DP, Fritz MS, Williams J, Lockwood dysfunction in mice. Biochem Biophys Res Com- regional, and national comparative risk assessment CM. Distribution of the product confidence limits mun 2017;491:890–896