Diabetes Care Volume 40, March 2017 375

Novel Protein Glycan–Derived Carlos Lorenzo,1 Andreas Festa,2 Anthony J. Hanley,3,4 Marian J. Rewers,5 Markers of Systemic Inf lammation Agustin Escalante,1 and Steven M. Haffner1§ and C-Reactive Protein in Relation to Glycemia, Insulin Resistance, and Insulin Secretion

Diabetes Care 2017;40:375–382 | DOI: 10.2337/dc16-1569

OBJECTIVE N-acetylglucosamine/galactosamine (GlycA) and sialic acid (GlycB) moieties of glycosylated serum proteins are nonspecific measures of inflammation, but con- clusive data on their relationship with insulin resistance or insulin secretion are missing. Therefore, we aimed to examine the relation of GlycA, GlycB, and C-reactive protein (CRP) to direct measures of insulin sensitivity (insulin sensitivity index [SI]) and insulin secretion (acute insulin response [AIR]).

RESEARCH DESIGN AND METHODS PATHOPHYSIOLOGY/COMPLICATIONS This study used cross-sectional analyses and included 1,225 participants with and without type 2 diabetes in the Insulin Resistance Atherosclerosis Study (IRAS). SI and AIR were measured using the frequently sampled intravenous glucose toler- ance test, and GlycA and GlycB were measured using nuclear magnetic resonance spectroscopy. 1Department of Medicine, University of Texas Health Science Center, San Antonio, TX RESULTS 2Eli Lilly and Company, Vienna, Austria 3 GlycA and GlycB had a strong correlation with CRP (r =0.60[P < 0.001] and r = 0.46 Department of Nutritional Sciences and Dalla Lana School of Public Health, University of Tor- [P < 0.001], respectively). In a linear regression model with both GlycA and CRP as onto, Toronto, Ontario, Canada independent variables, GlycA (b31SD,20.04 6 0.02; P < 0.01) and CRP (20.06 6 4Leadership Sinai Centre for Diabetes, Mt. Sinai 0.02; P < 0.001) were independently associated with S even after adjusting for Hospital, Toronto, Ontario, Canada I 5 demographics, smoking, physical activity, plasma glucose, and BMI. However, Barbara Davis Center for Diabetes, University Colorado School of Medicine, Aurora, CO neither CRP nor GlycA had an independent relationship with AIR. Corresponding author: Andreas Festa, andreasfesta@ CONCLUSIONS icloud.com. fl Received 26 July 2016 and accepted 6 December GlycA may complement CRP in evaluating the relationship between in ammation, 2016. glucose tolerance, and insulin resistance. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/ C-reactive protein (CRP) concentration, a marker of chronic subclinical inflamma- suppl/doi:10.2337/dc16-1569/-/DC1. tion, is related to insulin resistance (1) and has been identified as a risk factor for C.L. and A.F. contributed equally to this study. cardiovascular disease (CVD) (2). CRP concentration may be clinically relevant; it has §Retired. been shown to improve prediction algorithms for the stratification of individuals © 2017 by the American Diabetes Association. according to the risk of future CVD (3). Readers may use this article as long as the work fi is properly cited, the use is educational and not Glycosylation, the most common posttranslational protein modi cation, modu- for profit, and the work is not altered. More infor- lates protein function (4,5). Most acute-phase proteins, released from the liver mation is available at http://www.diabetesjournals during an inflammatory response, are enzymatically glycosylated and circulate .org/content/license. 376 Glycans and Insulin Resistance and Secretion Diabetes Care Volume 40, March 2017

at concentrations high enough to be based studies (the San Antonio Heart place at the University of Southern Cal- measurable via proton nuclear mag- Study and the San Luis Valley Diabetes ifornia (Los Angeles). Plasma insulin netic resonance (NMR) spectroscopy (6,7). Study, respectively). A total of 1,625 indi- concentration was measured using the N-acetylglucosamine/galactosamine viduals were enrolled in the IRAS (56% dextran-charcoal radioimmunoassay. (GlycA) and sialic acid (GlycB) moieties women) from among the 3,416 contacted CRP was measured by in-house ultra- of glycosylated serum proteins are non- (response rate of 48%). The examinations sensitive competitive immunoassay (anti- specificmeasuresofinflammation. Both began in October 1992 and were com- bodies and antigens from Calbiochem), GlycA and GlycB have a strong relation- pleted in April 1994. The IRAS protocol with an interassay coefficient of varia- ship with CRP (7,8). Increased GlycA has was approved by local institutional review tion of 8.9% (1). GlycA and GlycB were been associated with prevalent CVD risk committees, and all participants provided measured using NMR spectroscopy factors including smoking, diabetes, hyper- written informed consent. (LipoScience Inc., Raleigh, NC). GlycA tension, dyslipidemia, and obesity (8). Pro- GlycA and GlycB were measured in and GlycB signals in plasma arise from spectively, GlycA has been associated with 1,489 participants (561 non-Hispanic circulating acute-phase proteins (6,18). incident coronary heart disease and CVD whites, 429 African Americans, and The concentration of glycoproteins re- events independent of conventional risk 499 Hispanics). These participants did sponsible for the GlycA and GlycB signals factors in the Women’sHealthStudyand not differ from those with missing infor- was estimated to be ;13 mg/mL in nor- the Multi-Ethnic Study of Atherosclerosis mation on GlycA and GlycB (n =136)in mal human plasma (18). Fibrinogen, a1- (MESA), respectively (8,9). In the Women’s terms of adiposity, insulin resistance, acid glycoprotein, a1-antichymotrypsin, Health Study, the relation of GlycA to CVD and plasma concentrations of glucose, a1-antitrypsin, haptoglobin, and comple- was comparable to that of CRP (8). How- lipids, and CRP (P . 0.21 for all compar- ment C3 contributed significantly to the ever, the association between GlycA and isons). The present report includes data increase in GlycA in chronic systemic in- CVD was no longer significant after con- from 1,225 participantsd947 individu- flammation (18). Blood samples were trolling for CRP (8). als without diabetes and 278 patients stored at 270°C until analysis (approxi- CRP has been established as a risk factor with type 2 diabetes who were not tak- mately 18 years later). Ritchie et al. (19) for incident type 2 diabetes (10,11). GlycA ing any glucose-lowering drugsdin order already proved the stability of GlycA in also predicts future development of dia- to exclude any potential drug-specific samples stored for more than10 years. betes (12,13), but conclusive data on the confounding on key outcome measures, Fasting and 2-h plasma glucose concen- relation of GlycA and GlycB to insulin re- including markers of inflammation. Thus, trations were used to define categories of sistance or insulin secretion are missing. patients with diabetes were newly diag- glucose tolerance: normal (NGT), fasting It is therefore of interest to determine nosed or were treated with diet and/or glucose ,5.6 mmol/L and 2-h glucose whether the relation of GlycA and GlycB exercise. ,7.8 mmol/L (n = 455); isolated impaired to insulin resistance and insulin secretion fasting glucose (IFG), fasting glucose 5.6– has utility similar or complementary to Measurements 6.9 mmol/L and 2-h glucose ,7.8 mmol/L conventional inflammatory markers such Age, sex, ethnicity, smoking, physical (n = 188); isolated impaired glucose toler- as CRP. Thus, we examined the relation activity, menopausal status, and phar- ance (IGT), fasting glucose ,5.6 mmol/L of GlycA, GlycB, and CRP to measures of macologic treatment (glucose-lowering and 2-h glucose 7.8–11.0 mmol/L (n = insulin resistance and insulin secretion in agents and estrogen and progesterone 99); and IFG/IGT, fasting glucose 5.6– participants of the Insulin Resistance medications) were gathered by trained 6.9 mmol/L and 2-h glucose 7.8–11.0 Atherosclerosis Study (IRAS). In the personnel. Anthropometric measure- mmol/L (n = 205). Diabetes was defined IRAS, a frequently sampled intravenous ments were carried out using standard- as fasting glucose $126 mg/dL and/or glucose tolerance test (FSIGTT) was ad- ized protocols. The IRAS protocol required 2-h glucose $200 mg/dL (n = 278). ministered in all participants to obtain two visits, approximately 4 h each, 1 week HOMA of insulin resistance (HOMA-IR) direct measures of insulin sensitivity apart. Participants were asked before was calculated according to Matthew’s and insulin secretion: the insulin sensitiv- eachvisittofastfor12h,toabstainfrom formula: fasting insulin (mIU/mL) 3 fasting ity index (SI) and acute insulin response heavy exercise and alcohol for 24 h, and to glucose (mmol/L) 4 22.5. Normal weight, (AIR), respectively. refrain from smoking the morning of the overweight, and obesity were defined examination. as BMI ,25, 25–29.9, and $30 kg/m2, RESEARCH DESIGN AND METHODS A 75-g oral glucose tolerance test was respectively. Cigarette smoking was cat- Subjects administered to assess glucose tolerance egorized as nonsmoker and low- and The design and methods of the IRAS have status during the first visit. During the sec- high-degree smokers (0, 1–9, and $10 been described in detail (14). Briefly, the ond visit, insulin sensitivity and insulin se- cigarettes/day, respectively). study was conducted at four clinical cen- cretion were determined using an FSIGTT ters. At centers in Oakland and Los An- (15,16). Insulin sensitivity, expressed as Statistical Analyses geles, California, non-Hispanic whites and the SI, was calculated using mathematical The analysis was carried out using SAS African Americans were recruited from modeling methods (MINMOD version 3.0, statistical software (version 9.2; SAS In- Kaiser Permanente, a nonprofit health 1994) (17). Acute insulin response was stitute Inc., Cary, NC). Differences in maintenance organization. Centers in San calculated as the mean plasma insulin markers of inflammation in participants Antonio, Texas, and San Luis Valley, Colo- concentration 2 and 4 min after the ad- categorized by age, sex, ethnicity, smok- rado, recruited non-Hispanic whites and ministration of glucose. Laboratory anal- ing, BMI, and glucose tolerance status Hispanics from two ongoing population- yses of plasma glucose and insulin took were determined by one-way ANCOVA. care.diabetesjournals.org Lorenzo and Associates 377

The strength of the relationship be- had more adiposity and insulin resis- a linear increase of all three markers by tween inflammatory markers and be- tance (as measured by fasting insulin BMI category. tween inflammatory markers and other concentration, HOMA-IR, and SI), and We generated three models that had metabolic variables was assessed using higher AIR than non-Hispanic whites. each inflammatory marker as the de- Pearson correlation coefficients. Corre- Total energy expenditure was higher pendent variable and age, sex, ethnicity, lation coefficients were compared using among Hispanics compared with non- clinic, current smoking, and glucose tol- the Steiger t test. Multiple linear regres- Hispanic whites, but a smaller propor- erance categories as independent vari- sion analysis was used to examine the tion of Hispanics engaged in vigorous ables (Fig. 1). There were no ethnic relation of demographic and metabolic physical activity. Total energy expendi- differences in GlycA and GlycB, but variables to each of the inflammatory ture and vigorous physical activity were CRP was higher among both Hispanics markers (dependent variable) and to de- similar among African Americans and (P = 0.017) and African Americans (P = termine the proportion of the variance non-Hispanic whites. Hispanics had 0.017). All three inflammatory markers (R2) that each of the models was able to higher CRP, GlycA, and GlycB than non- were higher among women compared explain. The relation of markers of in- Hispanic whites. However, African Amer- with men (P , 0.01 for all three markers), flammation to insulin resistance or BMI icans had higher CRP, similar GlycA, and and in isolated IGT compared with NGT (dependent variable) was also examined lower GlycB. (P , 0.01 for all three markers). However, by multiple linear regression analysis to CRP, GlycA, and GlycB by age, sex, none of the inflammatory markers was as- account for the effect of demographics ethnicity, BMI, and glucose tolerance sociated with isolated IFG. CRP levels and and other metabolic variables. Log- categories are shown in Supplemen- GlycAwerealsoelevatedintype2diabe- transformed values of insulin, HOMA- tary Fig. 1. In addition to the ethnic dif- tescomparedwithisolatedIGT(P , 0.001 IR, AIR, and CRP were used in all analyses ferences in CRP, GlycA, and GlycB, as and P = 0.015, respectively), but GlycB was to meet the normality assumptions of presented above, none of these inflam- not increased. the tests. We also used the log transfor- matory markers was related to age, but Pearson correlation coefficients relat- mation of (SI + 1) given that some partic- all three were higher among women ing markers of subclinical inflammation ipants had an SI of zero. We considered a compared with men (P , 0.001 for all to relevant metabolic variables are P value ,0.050 statistically significant. three markers). All three markers of sub- shown in Supplementary Table 1. GlycA clinical inflammation were elevated in and GlycB were highly correlated (r = RESULTS isolated IGT, IFG/IGT, and type 2 diabe- 0.74; P , 0.001). GlycA (r =0.60;P , Table 1 presents demographic and met- tescomparedwithNGT(P , 0.001 0.001) and GlycB (r = 0.46; P , 0.001) abolic characteristics of the three ethnic for all comparisons), but none was in- had a strong relationship with CRP. All groups. Hispanics and African Americans creased in isolated IFG. Also, there was three inflammatory markers had direct

Table 1—Demographic and metabolic variables by ethnicity Non-Hispanic whites (n = 477) African Americans (n = 340) Hispanics (n =408) Female sex 50.9 57.9† 58.8† Menopausal status 36.9 42.4 44.0 Taking estrogen and/or progesterone 20.8 14.4‡ 12.3‡ Type 2 diabetes 21.2 28.5† 19.6 Smokers 12.6 16.8 22.4§ Age (years) 56.2 (49–64) 55.2 (48–62) 54.4 (47–62)‡ BMI (kg/m2)28.2(24.5–31.1) 30.3 (26.3 – 33.2)§ 29.2 (25.6–31.7)† Waist circumference (cm) 92.0 (83.1–100.4) 93.5 (84.5–101.3) 92.4 (83.3–100.0) Total energy expenditure (kcal/kg/year) 14,602 (12,853–15,458) 14,324 (12,651–15,026) 15,098 (13,104–16,346)† Vigorous activity per week ,1 time 16.9 13.6 19.5 1 time 6.0 3.0 4.0 .1 time 16.1 11.2 9.7‡ Fasting glucose (mg/dL) 109.4 (92–113) 114.7 (95–122)† 106.2 (91–110) 2-h glucose (mg/dL) 154.6 (105–177) 165.7 (106–202)† 153.3 (102–182) Fasting insulin (mU/mL)* 12.6 (8.0–18.0) 15.2 (10.5–22.0)§ 15.3 (10.0–23.0)§ HOMA-IR* 3.29 (1.97–5.16) 4.18 (2.48–6.42)§ 3.93 (2.41–6.36)§ 24 –1 21 21 SI (3 10 min z mU z mL )* 2.64 (0.75–3.06) 2.18 (1.51–2.92)§ 2.25 (1.49–3.32)§ AIR (mU/mL)* 35.9 (20.5–62.2) 44.3 (22.6–83.1)§ 50.9 (28.5–88.2)§ CRP (mg/L)* 1.70 (0.72–3.53) 2.25 (0.95–5.53)§ 2.32 (1.17–5.10)§ GlycA (mmol/L) 349.5 (305.3–392.7) 359.5 (304.3–400.8) 365.5 (317.3–404.6)‡ GlycB (mmol/L) 85.8 (67.2–101.2) 81.6 (63.1–96.6)† 99.6 (76.6–120.2)§ Data are percentages or mean (25th–75th percentiles). *Values log-transformed then back-transformed P for test of difference between minority populations and non-Hispanic whites; †P , 0.05; ‡P , 0.01; §P , 0.001. 378 Glycans and Insulin Resistance and Secretion Diabetes Care Volume 40, March 2017

Figure 1—CRP levels and GlycA and GlycB NMR signals (dependent variables). Age, sex, ethnicity, clinic, current smoking, and glucose tolerance categories were included as independent variables in all three models. AA, African American; Hisp, Hispanic; iIFG, isolated impaired fasting glucose; iIGT, isolated impaired glucose tolerance; NHW, non-Hispanic white; Ref., reference category.

relationships with measures of adipos- none was related to AIR after adjusting for drugs on subclinical inflammation in ity, plasma glucose concentrations, and SI.Therewasnosignificant ethnic interac- women. Menopausal status was not asso- insulin resistance, and inverse correla- tion for the relation of CRP, GlycA, and ciated with CRP (b =0.08[95%CI20.13, tions with SI and total energy expendi- GlycB to measures of adiposity, plasma 0.30]), GlycA (5.56 [28.85, 19.97]), or ture. Correlations involving measures of glucose, and insulin resistance/sensitivity GlycB (0.85 [24.75, 6.46]) after adjusting adiposity and insulin resistance were (P for interaction .0.05). GlycA had a for age, estrogen/progesterone drugs, relatively strong. The relation of CRP to more robust correlation with CRP, plasma ethnicity, clinic, smoking, physical activity, SI (r = 20.41; P , 0.001) was stronger glucose, and measures of adiposity and 2-h glucose, BMI, and SI. However, taking than that of GlycA (r = 20.33, P , 0.001) insulin resistance than GlycB (P , 0.05 estrogen and/or progesterone was inde- and GlycB (r = 20.29; P , 0.001). All for all comparisons). pendently related to all three: CRP (b = three inflammatory markers were more We used multiple linear regression 0.64 [95% CI 0.47, 0.80]; P , 0.001), GlycA relatedto2-hglucosethantofastingglu- analysis to examine the effect of meno- (16.76 [5.45, 28.07]; P = 0.004), and GlycB cose (P , 0.001 for all comparisons), and pausal status and estrogen/progesterone (4.54 [0.14, 8.94]; P =0.043). care.diabetesjournals.org Lorenzo and Associates 379

The relation of demographics, estrogen/ markers were independently related to linking inflammation and metabolic dis- progesterone drugs, cigarette smoking, both SI and BMI (models 1–4). In a model ease are still incompletely understood physical activity, 2-h glucose, BMI, and SI that had both CRP and GlycA as indepen- (20). Our results also indicate that the re- (independent variables) to GlycA, GlycB, dent variables (model 5), CRP and GlycA lation of CRP to insulin resistance is stron- or CRP (dependent variable) was further had an independent relationship with SI, gerthanthatofGlycAandGlycB. examined by multiple linear regression but only CRP had an independent associa- GlycA, a composite signal that can analysis (Table 2). Sex, estrogen/ tion with BMI. be detected by NMR spectroscopy, progesterone drugs, smoking more represents a subset of acute-phase than 10 cigarettes a day, .1 episode of CONCLUSIONS reactants, including a1-acid glyco- vigorous physical activity per week, BMI, Our study has the following findings: 1) protein, haptoglobin, a1-antitrypsin, a and SI had a strong independent associ- Adiposity and SI have independent rela- 1-antichymotrypsin, and transferrin (7), ation with CRP, GlycA, and GlycB. Other tionships with CRP concentration and and has recently been characterized as an independent relationships (Hispanic eth- GlycA and GlycB NMR signals. 2)Both inflammatory biomarker with analytic nic origin and 2-h glucose with CRP) were CRP and GlycA demonstrate a statistically and clinical attributes that may comple- relatively weak. These models explained independent relation to SI,suggestingthat ment or provide advantages over existing 34.0, 24.1, and 28.5% of the variability of GlycA may reflect an inflammatory path- clinical markers of systemic inflammation CRP, GlycA, and GlycB, respectively. In way distinct from the pathway related to (7). CRP, an acute-phase protein synthe- similar analyses limited to premeno- CRP. 3)Allthreeinflammatory markers are sized by the liver mainly as a result of in- pausal women aged 40 to 49 years who morerelatedto2-hglucosethantofasting terleukin-6 stimulation, also undergoes were not taking estrogen/progesterone glucose. 4) GlycB has weaker relationships glycosylation with glycan attachments, drugs and age-matched men, CRP (b = with CRP and measures of insulin resis- which may vary under acute inflammatory 0.22 [95% CI 0.02, 0.42]; P = 0.028) and tance and adiposity than GlycA. conditions (23), but CRP contributes only GlycA (21.0 [6.06, 36.0]; P = 0.006) were CRP and GlycA have an independent negligibly to the GlycA signal (7). In a pro- elevated in women. However, GlycB was relationship with insulin resistance (Table spective study of initially healthy women, not significantly different (3.75 [22.88, 3, model 5), but not with insulin secretion baseline GlycA was associated with inci- 10.37]; P = 0.267). (Supplementary Table 1). The association dent CVD, consistent with a possible role Because of the strong relationship of inflammatory markers with insulin for protein glycans in inflammation and that CRP, GlycA, and GlycB had with resistance has been demonstrated in CVD (8). The association between GlycA measures of insulin resistance and adipos- cross-sectional studies (1), and results and CVD events was comparable to that ity, we further assessed the relation of of prospective studies indicate a role of CRP, and was attenuated after adjust- each inflammatory marker (independent of inflammatory processes in the path- ing for CRP. Interestingly, an analysis of variable) to SI or BMI (dependent variable) ophysiology of type 2 diabetes and CVD follow-up time in this long-term study after adjusting for demographic variables, (1–3,11,20–22). These findings have led revealed a relation of GlycA to incident estrogen/progesterone drugs, smoking, to attempts to improve metabolic disease CVD that was independent of CRP in the physical activity, 2-h glucose, and BMI (or by way of anti-inflammatory interven- first 6 years of the study, but not there- SI) (Table 3). All three inflammatory tions (22); however, the exact mechanisms after (median follow-up 17.2 years).

Table 2—Multiple linear regression analysis with GlycA or GlycB NMR signals or CRP levels as the dependent variable Log CRP GlycA GlycB Intercept 0.42 (0.26, 0.58)‡ 333.03 (321.72, 344.34)‡ 79.04 (74.59, 83.49)‡ Age (3 1SD) 0.05(20.01, 0.10) 1.21 (22.54, 4.95) 1.37 (20.11, 2.84) Sex§ Women not taking hormones vs. men 0.24 (0.12, 0.36)‡ 24.08 (15.85, 32.31)‡ 4.24 (1.01, 7.48)* Women taking hormones vs. men 0.89 (0.73, 1.04)‡ 41.30 (30.70, 51.86)‡ 9.30 (5.14, 13.46)‡ Ethnicity African American vs. NHW 0.09 (20.06, 0.25) 20.58 (211.12, 9.95) 22.14 (26.28, 2.00) Hispanic vs. NHW 0.23 (0.07, 0.39)† 6.33 (24.65, 17.30) 2.78 (21.53, 7.10) Smoking (cigarettes/day) 1–9 vs. none 20.08 (20.31, 0.15) 6.35 (29.73, 22.43) 20.65 (26.97, 5.67) $10 vs. none 0.32 (0.16, 0.49)‡ 35.12 (23.61, 46.64)‡ 11.32 (6.79, 15.84)‡ Vigorous activity (times/week) 1 vs. ,1 0.02 (20.15, 0.19) 1.78 (29.75, 13.30) 21.31 (25.85, 23.22) .1vs.,1 20.17 (20.29, 20.05)† 210.16 (218.41, 21.92)* 24.03 (27.28, 20.79)* 2-h glucose (3 1 SD) 0.11 (0.04, 0.17)† 3.48 (20.94, 7.89) 1.14 (20.59, 2.88) BMI (3 1 SD) 0.35 (0.28, 0.41)‡ 14.97 (10.63, 19.31)‡ 5.01 (3.30, 6.71)‡

Log SI (3 1SD) 20.21 (20.29, 20.14)‡ 213.75 (218.77, 28.72)‡ 24.30 (26.28, 22.32)‡ R2 for the model (%) 34.0 24.1 28.5 Data are b (95% CI) unless otherwise indicated. NHW, non-Hispanic white. *P , 0.05; †P , 0.01; ‡P , 0.001; §Hormones indicate estrogen and/or progesterone medications. 8 lcn n nui eitneadSecretion and Resistance Insulin and Glycans 380

Table 3—Multiple linear regression analysis relating GlycA and GlycB NMR signals and CRP concentrations to BMI or SI Model 1 Model 2 Model 3 Model 4 Model 5

Log SI as the dependent variable* Intercept 0.82 (0.75, 0.90)‡ 0.81 (0.73, 0.88)‡ 0.80 (0.73, 0.87)‡ 0.81 (0.73, 0.88)‡ 0.80 (0.72, 0.87)‡ Age (3 1SD) 20.05 (20.07, 20.02)‡ 20.04 (20.06, 20.02)‡ 20.04 (20.07, 20.02)‡ 20.04 (20.07, 20.02)‡ 20.04 (20.06, 20.02)‡ Sex Women not taking hormones vs. men 0.13 (0.07, 0.18)‡ 0.14 (0.08, 0.19)‡ 0.15 (0.09, 0.20)‡ 0.13 (0.08, 0.18)‡ 0.15 (0.09, 0.20)‡ Women taking hormones vs. men 0.05 (20.02, 0.12) 0.12 (0.05, 0.19)§ 0.09 (0.02, 0.16)† 0.07 (0.003, 0.14)| 0.12 (0.05, 0.19)‡ Ethnicity African American vs. NHW 20.11 (20.18, 20.04)§ 20.10 (20.17, 20.03)§ 20.11 (20.18, 20.04)§ 20.11 (20.18, 20.04)§ 20.10 (20.17, 20.03)§ Hispanic vs. NHW 20.06 (20.13, 0.01) 20.04 (20.11, 0.03) 20.05 (20.12, 0.02) 20.05 (20.12, 0.02) 20.04 (20.11, 0.03) Smoking (cigarettes) 1–9 vs. none 20.01 (20.12, 0.09) 20.02 (20.12, 0.08) 20.01 (20.11, 0.10) 20.01 (20.12, 0.09) 20.01 (20.11, 0.09) $10 vs. none 20.03 (20.11, 0.04) 20.01 (20.08, 0.07) 0.00 (20.07, 0.08) 20.01 (20.08, 0.07) 0.01 (20.06, 0.08) Vigorous activity (times/week) 1 vs. ,1 0.08 (0.003, 0.15)| 0.07 (20.003, 0.14) 0.08 (0.004, 0.15)| 0.07 (20.001, 0.15) 0.07 (20.003, 0.14) .1 vs. ,1 0.08 (0.03, 0.13)§ 0.07 (0.01, 0.12)| 0.07 (0.01, 0.12)| 0.07 (0.02, 0.12)| 0.06 (0.01, 0.11)| 2-h glucose (3 1 SD) 20.25 (20.27, 20.22)‡ 20.23 (20.26, 20.21)‡ 20.24 (20.26, 20.21)‡ 20.24 (20.27, 20.22)‡ 20.23 (20.26, 20.21)‡ BMI (3 1 SD) 20.22 (20.24, 20.19)‡ 20.19 (20.21, 20.16)‡ 20.20 (20.22, 20.17)‡ 20.20 (20.23, 20.18)‡ 20.18 (20.21, 20.16)‡ Log CRP (3 1 SD) – 20.08 (20.11, 20.05)‡– –20.06 (20.09, 20.03)‡ GlycA (3 1 SD) ––20.07 (20.10, 20.05)‡–20.04 (20.07, 20.01)§ GlycB (3 1 SD) –––20.06 (20.09, 20.03)‡– R2 for the model (%) 47.9 49.4 48.8 48.7 49.7 BMI as the dependent variable† Intercept 28.75 (27.88, 29.62)‡ 29.22 (28.39, 30.06)‡ 289.12 (28.26, 29.98)‡ 29.08 (28.22, 29.95)‡ 29.29 (28.44, 30.12)‡ Age (3 1SD) 20.80 (21.08, 20.51)‡ 20.79 (21.07, 20.52)‡ 20.79 (21.06, 20.51)‡ 20.82 (21.10, 20.54)‡ 20.79 (21.06, 20.52)‡ Sex Women not taking hormones vs. men 1.66 (1.04, 2.29)‡ 1.12 (0.52, 1.73)‡ 1.25 (0.63, 1.88)‡ 1.48 (0.86, 2.10)‡ 1.05 (0.44, 1.66)‡ Women taking hormones vs. men 0.50 (20.32, 1.31) 20.87 (21.69, 20.06)| 20.12 (20.94, 0.70) 0.19 (20.62, 1.00) 20.92 (21.74, 20.11)| Ethnicity African American vs. NHW 0.62 (20.19, 1.43) 0.45 (20.33, 1.22) 0.60 (20.19, 1.40) 0.67 (20.13, 1.47) 0.46 (20.32, 1.23) Hispanic vs. NHW 0.33 (20.51, 1.17) 20.04 (20.85, 0.77) 0.23 (20.60, 1.06) 0.23 (20.60, 1.07) 20.04 (20.85, 0.77) Smoking (cigarettes) 0–9 vs. none 21.04 (22.28, 0.19) 20.83 (20.201, 0.35) 21.10 (22.31, 0.11) 20.99 (22.21, 0.22) 20.87 (22.04, 0.08) $10 vs. none 21.17 (22.05, 20.28)| 21.55 (22.39, 20.70)‡ 21.63 (22.51, 20.75)‡ 21.49 (22.37, 20.61)‡ 21.65 (22.50, 20.79)‡ Care Diabetes Vigorous activity (times/week) 1 vs. ,1 0.11 (20.78, 0.99) 0.03 (20.82, 0.88) 0.08 (20.79, 0.95) 0.14 (20.73, 1.02) 0.02 (20.82, 0.87) .1 vs. ,1 20.77 (21.40, 20.14)| 20.45 (21.06, 0.16) 20.59 (21.21, 0.04) 20.62 (21.24, 0.01) 20.77 (21.40, 20.14)| 2-h glucose (3 1 SD) 20.08 (20.42, 0.26) 20.24 (20.57, 0.09) 20.12 (20.46, 0.21) 20.11 (20.45, 0.22) 20.24 (20.57, 0.09) oue4,Mrh2017 March 40, Volume Log SI (3 1 SD) 23.00 (23.35, 22.65)‡ 22.42 (22.77, 22.07) ‡ 22.69 (23.04, 22.34)‡ 22.78 (23.13, 22.43)‡ 22.39 (22.74, 22.03)‡ Log CRP (3 1 SD) – 1.73 (1.41, 2.03)‡– –1.57 (1.21, 1.93)‡ GlycA (3 1 SD) ––1.06 (0.75, 1.36)‡–0.30 (20.04, 0.65) GlycB (3 1 SD) –––0.94 (0.62, 1.26)‡– R2 for the model (%) 31.3 37.5 33.8 33.1 37.6

Data are b (95% CI) unless otherwise indicated. *Basic model: age, sex, ethnicity, clinic, smoking, 2-h glucose, and BMI; †Basic model: age, sex, ethnicity, clinic, smoking, 2-h glucose, and SI; |P , 0.05; §P , 0.01; ‡P , 0.001. NHW, non-Hispanic white. care.diabetesjournals.org Lorenzo and Associates 381

We also found that all three inflam- African American women. In the IRAS, HL-47889, HL-47890, HL-47892, HL-47902, matory markers are more related to 2-h both African Americans and Hispanics HL-55208, and R01-HL-58329) and the General glucose than to fasting glucose. In mul- have higher CRP levels than non- Clinic Research Centers Program (grants NCRR GCRC, M01 RR431, and M01 RR01346). tivariate analyses, there is a closer rela- Hispanic whites. However, only Hispanic Duality of Interest. No potential conflicts of tionship of CRP to glycemia compared ethnic origin is associated with an in- interest relevant to this article were reported. with that of GlycA and GlycB (Table 2). creased CRP concentration independent Author Contributions. C.L. and A.F. analyzed Taking these data together, one might and interpreted data, drafted the manuscript, of smoking, adiposity, and insulin resis- fi fl and gave nal approval to the version to be speculate that CRP and GlycA re ect tance. Our results also indicate that published. A.F. critically revised the manuscript distinct inflammatory processes, which there are no significant ethnic differ- for important intellectual content. A.J.H. inter- may affect insulin resistance and subse- ences in GlycA and GlycB after adjusting preted data, critically revised the manuscript for quently incident type 2 diabetes. Fur- for multiple risk factors, including insulin important intellectual content, and gave final ther research, including “omics”-based resistance (Table 2). approval to the version to be published. M.J.R. and A.E. acquired data, critically revised the analyses, may help identify targets in- Smokers, women, and sedentary indi- manuscript for important intellectual content, volved in this pathophysiological cascade; viduals tend to have elevated CRP and gave final approval to the version to be those targets may be amenable to thera- (33–38). In European studies, CRP did published. S.M.H. conceived and designed the peutic intervention. Impaired insulin secre- not differ between men and women af- study, drafted the manuscript, interpreted data, tion, by contrast, may not contribute as ter taking into consideration the effect critically revised the manuscript for important intellectual content, and gave final approval to much as insulin resistance, or its contribu- of estrogen on CRP (39,40). This sug- the version to be published. C.L. and A.F. are tion may be missed by epidemiology, be- gests that no sex-specificcutpointfor guarantors of this work and, as such, had full cause proteins reflecting the metabolism CRP is indicated to assess CVD risk (40). access to all the data in the study and take re- of large organs (such as liver and adipose However, higher CRP has been described in sponsibility for the integrity of the data and the accuracy of the data analysis. tissue) may yield levels of biomarkers suf- women not taking estrogen in U.S. popula- ficientlyhightobedetectableinserum, tions (33–35).IntheIRAS,bothfemalesex fl References whereas markers re ecting solely islet and the intake of estrogen/progesterone ’ fl 1. Festa A, D Agostino R Jr, Howard G, Mykkanen¨ cell in ammation (and hence affecting in- drugs (along with smoking and a lack of L, Tracy RP, Haffner SM. Chronic subclinical in- sulin secretion) might not reach detectable vigorous physical activity) are indepen- flammation as part of the insulin resistance syn- serum concentrations (20). dently associated with elevated CRP. Fe- drome: the Insulin Resistance Atherosclerosis CRP, GlycA, and GlycB are associated male sex, estrogen/progesterone drugs, Study (IRAS). Circulation 2000;102:42–47 with adiposity even after adjusting for 2. Ridker PM, Rifai N, Rose L, Buring JE, Cook smoking, and lack of vigorous physical NR. Comparison of C-reactive protein and low- insulin resistance and glucose tolerance. activity are also independently associated density lipoprotein cholesterol levels in the pre- A decrease of CRP levels with lifestyle with elevated GlycA and GlycB. diction of first cardiovascular events. N Engl interventions (weight loss and physical Strengths of this study include 1)a J Med 2002;347:1557–1565 activity) has previously been demon- large, well-described, multiethnic popula- 3. Ridker PM, Paynter NP, Rifai N, Gaziano JM, strated in individuals with and without Cook NR. C-reactive protein and parental history tion, 2) a direct measure of insulin resis- improve global cardiovascular risk prediction: the diabetes (24–26). The CRP concentra- tance and insulin secretion, and 3)useof Reynolds Risk Score for men. Circulation 2008;118: tion decrease with weight loss may be an accurate methodology for the assess- 2243–2251, 4p following 2251 related in part to the regulation of insulin ment of inflammatory markers and insulin 4. Ohtsubo K, Marth JD. Glycosylation in cellu- resistance and the inflammatory re- lar mechanisms of health and disease. Cell 2006; sensitivity. Limitations of the study include 126:855–867 sponse by macrophages and T cells in ad- 1) the cross-sectional nature of the analy- 5. Lauc G. Sweet secret of the multicellular life. ipose tissue (27,28). Our results suggest sis, making conclusions related to cause Biochim Biophys Acta 2006;1760:525–526 that BMI has a relationship with all three and effect difficult, and 2) the absence of 6. Bell JD, Brown JC, Nicholson JK, Sadler PJ. “ ” inflammatory markers that is at least as upstream markers of inflammation (such Assignment of resonances for acute-phase robust as that of waist circumference glycoproteins in high resolution proton NMR as interleukin-6). Interleukin-6 was not ex- spectra of human blood plasma. FEBS Lett (Supplementary Table 1). This suggests amined because our study is not designed 1987;215:311–315 that the production of CRP and glycosyla- to evaluate mechanisms. 7. Otvos JD, Shalaurova I, Wolak-Dinsmore J, tion of acute-phase proteins by the liver is In summary, GlycA was related to in- et al. GlycA: a composite nuclear magnetic res- determined by overall adiposity rather onance biomarker of systemic inflammation. sulin resistance and other features of – than visceral adiposity in particular, unlike Clin Chem 2015;61:714 723 the metabolic syndrome, independent 8. Akinkuolie AO, Buring JE, Ridker PM, Mora S. lipoproteins and insulin resistance. of CRP, indicating that GlycA may repre- A novel protein glycan biomarker and future Several studies have reported ethnic sent an inflammatory pathway distinct cardiovascular disease events. J Am Heart Assoc differences in CRP concentration, but from the CRP-related pathway. Further 2014;3:e001221 9. Duprez DA, Otvos J, Sanchez OA, Mackey RH, only among women (29–31). In these research in the field may help clarify our studies, results were adjusted for demo- Tracy R, Jacobs DR Jr. Comparison of the pre- understanding of the inflammatory dictive value of GlycA and other biomarkers of graphics, cardiovascular risk factors, pathophysiology of type 2 diabetes and inflammation for total death, incident cardio- and/or adiposity, but none of them CVD, and the ability of GlycA to improve vascular events, noncardiovascular and non- fl took into consideration the effect of in- prediction models for CVD. cancer in ammatory-related events, and total sulin resistance. In another study that cancer events. Clin Chem 2016;62:1020–1031 10. Pradhan AD, Manson JE, Rifai N, Buring JE, took into consideration insulin sensitiv- Ridker PM. C-reactive protein, interleukin 6, and ity determined by an FSIGTT (32), there Funding. ThisworkwassupportedbytheNational risk of developing type 2 diabetes mellitus. wasnoexcessCRPconcentrationin Heart, Lung, and Blood Institute (grants HL-47887, JAMA 2001;286:327–334 382 Glycans and Insulin Resistance and Secretion Diabetes Care Volume 40, March 2017

11. Festa A, D’Agostino R Jr, Tracy RP, Haffner SM; 21. Wang X, Bao W, Liu J, et al. Inflammatory (from the Women’s Health Study). Am J Cardiol Insulin Resistance Atherosclerosis Study. Elevated markers and risk of type 2 diabetes: a systematic 2004;93:1238–1242 levels of acute-phase proteins and plasminogen review and meta-analysis. Diabetes Care 2013; 32. Hyatt TC, Phadke RP, Hunter GR, Bush NC, activator inhibitor-1 predict the development of 36:166–175 Munoz~ AJ, Gower BA. Insulin sensitivity in type 2 diabetes: the insulin resistance atheroscle- 22. Donath MY. Targeting inflammation in the African-American and white women: associa- rosis study. Diabetes 2002;51:1131–1137 treatment of type 2 diabetes: time to start. Nat tionwithinflammation. Obesity (Silver Spring) 12. Akinkuolie AO, Pradhan AD, Buring JE, Rev Drug Discov 2014;13:465–476 2009;17:276–282 Ridker PM, Mora S. Novel protein glycan side- 23. Gornik O, Lauc G. Glycosylation of serum 33. Ford ES, Giles WH, Myers GL, Mannino DM. chain biomarker and risk of incident type 2 di- proteins in inflammatory diseases. Dis Markers Population distribution of high-sensitivity abetes mellitus. Arterioscler Thromb Vasc Biol 2008;25:267–278 C-reactive protein among US men: findings 2015;35:1544–1550 24. Belalcazar LM, Reboussin DM, Haffner SM, from National Health and Nutrition Examination 13. Connelly MA, Gruppen EG, Wolak- et al.; Look AHEAD Research Group. A 1-year Survey 1999–2000. Clin Chem 2003;49:686–690 Dinsmore J, et al. GlycA, a marker of acute phase lifestyle intervention for weight loss in individuals 34. Khera A, McGuire DK, Murphy SA, et al. glycoproteins, and the risk of incident type 2 with type 2 diabetes reduces high C-reactive protein Race and gender differences in C-reactive pro- diabetes mellitus: PREVEND study. Clin Chim levels and identifies metabolic predictors of change: tein levels. J Am Coll Cardiol 2005;46:464–469 Acta 2016;452:10–17 from the Look AHEAD (Action for Health in Diabe- 35. Albert MA, Ridker PM. C-reactive protein 14. Wagenknecht LE, Mayer EJ, Rewers M, et al. tes) study. Diabetes Care 2010;33:2297–2303 as a risk predictor: do race/ethnicity and gen- The Insulin Resistance Atherosclerosis Study 25. Haffner S, Temprosa M, Crandall J, et al.; der make a difference? Circulation 2006;114: (IRAS) objectives, design, and recruitment re- Diabetes Prevention Program Research Group. e67–e74 sults. Ann Epidemiol 1995;5:464–472 Intensive lifestyle intervention or metformin on 36. Frohlich¨ M, Sund M, Lowel¨ H, Imhof A, 15. Bergman RN, Finegood DT, Ader M. Assess- inflammation and coagulation in participants Hoffmeister A, Koenig W. Independent associa- ment of insulin sensitivity in vivo. Endocr Rev with impaired glucose tolerance. Diabetes tion of various smoking characteristics with 1985;6:45–86 2005;54:1566–1572 markers of systemic inflammation in men. Re- 16. Steil GM, Volund A, Kahn SE, Bergman RN. 26. Esposito K, Pontillo A, Di Palo C, et al. Effect sults from a representative sample of the gen- Reduced sample number for calculation of in- of weight loss and lifestyle changes on vascular eral population (MONICA Augsburg Survey sulin sensitivity and glucose effectiveness from inflammatory markers in obese women: a ran- 1994/95). Eur Heart J 2003;24:1365–1372 the minimal model. Suitability for use in popu- domized trial. JAMA 2003;289:1799–1804 37. Kuller L, Tracy R, Shaten J, Meilahn EN. Re- lation studies. Diabetes 1993;42:250–256 27. Xu H, Barnes GT, Yang Q, et al. Chronic in- lationship of C-reactive protein and coronary 17. Pacini G, Bergman RN. MINMOD: a com- flammation in fat plays a crucial role in the de- heart disease in the MRFIT nested case-control puter program to calculate insulin sensitivity velopment of obesity-related insulin resistance. study. Multiple Risk Factor Intervention Trial. and pancreatic responsivity from the frequently J Clin Invest 2003;112:1821–1830 Am J Epidemiol 1996;144:537–547 sampled intravenous glucose tolerance test. 28. Wu H, Ghosh S, Perrard XD, et al. T-cell accu- 38. Ohsawa M, Okayama A, Nakamura M, et al. Comput Methods Programs Biomed 1986;23: mulation and regulated on activation, normal T cell CRP levels are elevated in smokers but unre- 113–122 expressed and secreted upregulation in adipose lated to the number of cigarettes and are de- 18. Smith KD, Pollacchi A, Field M, Watson J. tissue in obesity. Circulation 2007;115:1029–1038 creased by long-term smoking cessation in male The heterogeneity of the glycosylation of 29. Carroll JF, Fulda KG, Chiapa AL, et al. Impact smokers. Prev Med 2005;41:651–656 alpha-1-acid glycoprotein between the sera of race/ethnicity on the relationship between 39. Imhof A, Frohlich¨ M, Loewel H, et al. Distri- and synovial fluid in rheumatoid arthritis. Bi- visceral fat and inflammatory biomarkers. Obe- butions of C-reactive protein measured by high- omed Chromatogr 2002;16:261–266 sity (Silver Spring) 2009;17:1420–1427 sensitivity assays in apparently healthy men and 19. Ritchie SC, Wurtz¨ P, Nath AP, et al. The bio- 30. Kelley-HedgepethA,Lloyd-JonesDM, women from different populations in Europe. marker GlycA is associated with chronic inflam- Colvin A, et al.; SWAN Investigators. Ethnic dif- Clin Chem 2003;49:669–672 mation and predicts long-term risk of severe ferences in C-reactive protein concentrations. 40. Ledue TB, Rifai N. Preanalytic and analytic infection. Cell Syst 2015;1:293–301 Clin Chem 2008;54:1027–1037 sources of variations in C-reactive protein mea- 20. Donath MY, Shoelson SE. Type 2 diabetes as 31. Albert MA, Glynn RJ, Buring J, Ridker PM. surement: implications for cardiovascular dis- an inflammatory disease. Nat Rev Immunol C-reactive protein levels among women of var- ease risk assessment. Clin Chem 2003;49: 2011;11:98–107 ious ethnic groups living in the United States 1258–1271