Coronary Artery Disease and Type 2 Diabetes: a Proteomic Study

Diabetes Care Volume 43, April 2020 843

Coronary Artery Disease and Giulia Ferrannini,1,2 Maria Laura Manca,3 Marco Magnoni,4 Felicita Andreotti,5 Type 2 Diabetes: A Proteomic Daniele Andreini,6,7 Roberto Latini,8 Attilio Maseri,9 Aldo P. Maggioni,10 Study Rachel M. Ostroff,11 Stephen A. Williams,11 and Ele Ferrannini12 Diabetes Care 2020;43:843–851 | https://doi.org/10.2337/dc19-1902

OBJECTIVE Coronary artery disease (CAD) is a major challenge in patients with type 2 diabetes (T2D). Coronary computed tomography angiography (CCTA) provides a detailed anatomic map of the coronary circulation. Proteomics are increasingly used to improve diagnostic and therapeutic algorithms. We hypothesized that the protein panel is differentially associated with T2D and CAD.

RESEARCH DESIGN AND METHODS In CAPIRE (Coronary Atherosclerosis in Outlier Subjects: Protective and Novel 1Department of Medical Sciences, Postgraduate Individual Risk Factors Evaluationda cohort of 528 individuals with no previous School of Internal Medicine, University of Turin, 2 Turin, Italy cardiovascular event undergoing CCTA), participants were grouped into CAD 2 1 Department of Medicine Solna, Karolinska (clean coronaries) and CAD (diffuse lumen narrowing or plaques). Plasma proteins Institutet, Stockholm, Sweden were screened by aptamer analysis. Two-way partial least squares was used to 3Department of Clinical and Experimental Med- simultaneously rank proteins by diabetes status and CAD. icine, University of Pisa, Pisa, Italy 4IRCCS Ospedale San Raffaele and Universita` Vita-Salute San Raffaele, Milan, Italy RESULTS 5 1 Institute of Cardiology, FPG IRCCS, Catholic Uni- ADOACLRADMTBLCRISK METABOLIC AND CARDIOVASCULAR Though CAD was more prevalent among participants with T2D (HbA1c 6.7 6 1.1%) versity Medical School, Rome, Italy than those without diabetes (56 vs. 30%, P < 0.0001), CCTA-based atherosclerosis 6CentroCardiologicoMonzino,IRCCS,Milan,Italy 7 burden did not differ. Of the 20 top-ranking proteins, 15 were associated with both Cardiovascular Section, Department of Clinical Sciences and Community Health, University of T2D and CAD, and 3 (osteomodulin, cartilage intermediate-layer protein 2, and Milan, Milan, Italy HTRA1) were selectively associated with T2D only and 2 (epidermal growth factor 8Mario Negri Institute of Pharmacological Research- receptor and contactin-1) with CAD only. Elevated renin and GDF15, and lower IRCCS, Milan, Italy 9 adiponectin, were independently associated with both T2D and CAD. In multivariate Heart Care Foundation, Florence, Italy 10ANMCO Research Center, Heart Care Founda- analysis adjusting for the Framingham risk panel, patients with T2D were “pro- tion, Florence, Italy tected” from CAD if female (P 5 0.007), younger (P 5 0.021), and with lower renin 11Clinical Research and Development, SomaLogic levels (P 5 0.02). Inc., Boulder, CO 12CNR Institute of Clinical Physiology, Pisa, Italy CONCLUSIONS Corresponding author: Ele Ferrannini, ferranni@ We concluded that 1) CAD severity and quality do not differ between participants ifc.cnr.it with T2D and without diabetes; 2) renin, GDF15, and adiponectin are shared Received 24 September 2019 and accepted 31 December 2019 markers by T2D and CAD; 3) several proteins are speciﬁcally associated with T2D or 4 This article contains Supplementary Data online CAD; and ) in T2D, lower renin levels may protect against CAD. at https://care.diabetesjournals.org/lookup/suppl/ doi:10.2337/dc19-1902/-/DC1. Cardiovascular (CV) disease has steadily represented the leading cause of death in the © 2020 by the American Diabetes Association. world for the past 15 years, with coronary artery disease (CAD) and stroke being Readers may use this article as long as the work is properly cited, the use is educational and not responsible for a combined 15 million deaths in 2016 (1). Morbidity and mortality of for proﬁt, and the work is not altered. More infor- patients with CAD is considerably higher in the presence of diabetes, accounting for mation is available at https://www.diabetesjournals ;50% of deaths in this population (2). Therefore, detection of CAD and its risk factors .org/content/license. 844 Protein Biomarkers for CAD in Diabetes Diabetes Care Volume 43, April 2020

is crucial to improve both treatment and presence of type 2 diabetes (T2D) and 4) previous documented acute or chronic prevention, especially in patients with CCTA-proven CAD by using an expanded peripheral vascular disease (stroke, tran- diabetes. proteomic platform and bivariate anal- sient ischemic attack, previous revascu- Coronary computed tomography an- ysis of associations. larization); 5) claudication at rest or at giography (CCTA) provides a noninvasive, low-grade effort); and 6) active inflam- highly sensitive anatomic investigation RESEARCH DESIGN AND METHODS matory or neoplastic disease. of the epicardial coronary circulation by Study Design visualizing the coronary artery lumen and The Coronary Atherosclerosis in Outlier CVRF Definition wall with an intravenous contrast agent Subjects: Protective and Novel Individual The conventional CVRFs that are based (3). CCTA is highly accurate in the early Risk Factors Evaluation (CAPIRE) study on the Adult Treatment Panel III and the detection of obstructive atherosclerosis, (ClinicalTrials.gov identifier: NCT02157662) 2013 American College of Cardiology/ defined by invasive coronary angiogra- is part of the Gruppo Italiano per lo Studio American Heart Association guidelines phy, and therefore is being increasingly dellaStreptochinasinell’InfartoMiocardico for CV disease prevention were consid- included in international guidelines for (GISSI) Outlier Project, jointly promoted ered in selecting the study population the diagnostic workup of CAD (4,5). The by the Heart Care Foundation Onlus, (18). CVRFs were defined as follows: extent of atherosclerotic burden can be Italian Association of Hospital Cardiol- family history of ischemic heart disease assessed by several CCTA-derived scores, ogists (ANMCO), and Mario Negri In- (history of early manifestations of ische- which have demonstrated independent stitute of Pharmacological Research. mic heart disease in first-degree relatives long-term predictive power for adverse CAPIRE was designed as a prospective, before 55 years of age for men and cardiac events (i.e., the segment involve- observational study aimed at identifying 65 years for women), arterial hyperten- ment score, the segment stenosis score, possible new mechanisms that promote sion (history of arterial hypertension, and the CT-adapted Leaman score [6]). In or protect against atherothrombosis; in ongoing antihypertensive treatment, or patients with diabetes, the prevalence of its longitudinal phase, participants are recent observation of blood pressure obstructive, CCTA-detected CAD is nearly being followed for 5 years or longer [BP] values .140/90 mmHg), hypercho- 25% (i.e., much higher than in the pop- (17). The study enrolled participants 45– lesterolemia (total serum cholesterol ulation without diabetes) (7). Moreover, 75 years of age, without previous clin- .200 mg/dL or ,200 mg/dL with on- CCTA provides long-term prognostic in- ical manifestations of ischemic heart going lipid-lowering medications), diabe- formation for patients with diabetes, disease (acute myocardial infarction, un- tes (fasting plasma glucose levels .126 allowing their risk stratification and show- stable angina, chronic stable angina, pre- mg/dL or a 2-h value $200 mg/dL by oral ing excellent prognosis when no evidence vious percutaneous or surgical coronary glucose tolerance test or isolated eleva- of atherosclerosis is detected (7,8). revascularization, heart failure), who un- tion of glycated hemoglobin [HbA1c] $6.5% Addressing traditional CV risk factors derwent a 64-slice (or superior) CCTA in or current use of insulin or oral glucose- (CVRFs) has proved advantageous in re- the outpatient clinics of the 11 participat- lowering medications), and cigarette smok- ducing CV mortality in the general pop- ingcentersbecauseofsuspectedCAD.The ing (current cigarette smoking habit or ulation as well as in people with diabetes main indications for CCTA were 1)unin- recent abstention within 1 year). (9–11). Nevertheless, the relationship terpretable, equivocal, or contraindicated Source data for defining CVRFs were between the presence of traditional functional stress test (44% of patients); 2) physical examination, medical records, CVRFs and atherosclerosis development new-onset chest pain syndrome at low- and laboratory tests reported by the is hardly linear, and CV events can occur intermediate pretest likelihood of CAD participant or documented before CCTA. independently of traditional CVRF man- (25% of patients); and 3) other indication After enrollment, a centrally performed agement.Thus, substantial improvement (including evaluation before valve or non- biomarker profile, including lipid profile in CAD prevention is still needed, espe- cardiac surgery, elevated risk profile, ar- and metabolic markers, allowed a refined cially in patients with diabetes (12,13). rhythmias, or atypical symptoms) (31% of assessment of CVRFs, such as diabetes Over the past decade, technology has patients). Patients with T2D were further and dyslipidemia. According to data in provided novel tools for the identifica- selected to be eligible for the CCTA pro- the literature, most patients without any tion of protective and susceptibility fac- cedure if they were in good clinical con- CVRFs or one single risk factor belong tors for both CAD and diabetes. More dition and stable metabolic control. to a risk group of ,10% of events at recently, progress in the technology of Exclusion criteria were 1) CCTA not meet- 10 years according to the Framingham proteomics has provided further oppor- ing the quality control criteria; 2)previous Heart Study, and patients with three or tunities for the development of new CV events (myocardial infarction, unstable more CVRFs belong to a risk group of diagnostic and therapeutic algorithms angina, stable or unstable angina, percu- .20%ofeventsat10years(18).Diabetes (14). High-performance platforms can taneous or surgical coronary revasculari- as a single CVRF was excluded because screen a large number of proteins that zation, heart failure), both clinically evident of evidence of its own higher CV risk for may serve as biomarkers, eventually to and confirmed by clinical and conventional the development of CAD and indepen- be included in multiparametric models of diagnostics; 3) other previous heart disor- dent association with the atherosclerotic risk assessment (15). Proteins may also ders documented or identified at CCTA, burden (18). serve as indicators of therapeutic effec- such as dilated cardiomyopathy (regardless tiveness and can be targeted directly (16). of etiology), obstructive hypertrophic car- CCTA Analysis In the current study, we aimed to identify diomyopathy, atrial fibrillation, myocardi- The CCTA data interpretation in CAPIRE proteins differentially associated with the tis, and inflammatory vascular disease; was performed with advanced plaque care.diabetesjournals.org Ferrannini and Associates 845

assessment. Coronary arteries were di- Laboratory Determinations using the Student t test for independent vided into 16 segments according to the In all participants, a peripheral venous samples; variables with a skewed distri- AHA classification (19). Normal coronary blood sample was taken, mostly under bution (by the Shapiro-Wilk test) were arteries were defined as no atheroscle- fasting conditions (i.e., in the morning compared by Mann-Whitney U tests for rotic plaque detected in any segment after an overnight, 10- to 12-h fast). independent samples. Proportions were within the coronary artery wall or lumen. Following centrifugation, 0.5-mL aliquots compared using a x2 or Fisher exact test, Lumen stenosis was measured for each of serum or plasma were stored at 270°C as appropriate. Differences by diabetes coronary artery plaque detected and in a dedicated biological bank (SATURNE- (NGT or T2D) and CAD (CAD1 or CAD2), graded as normal, nonobstructive (,50%), 1; Mario Negri Institute of Pharmaco- each treated as a nominal variable, were moderate stenosis (50–70%), and severe logical Research). All biomarkers were analyzed by two-way ANOVA; for this stenosis (.70%). High-risk plaque fea- measured in a central laboratory, in a analysis, variables with a skewed distri- tures were also assessed and defined single batch, by personnel unaware of bution were logarithmically transformed as described in a previous report from participant characteristics. Serum creati- for use in ANOVA. CAPIRE (20). Plaque length was also nine, HbA1c, lipids, and plasma metabolites Two-way partial least squares (PLS) recorded. Plaque consistency was as- were measured by standard, automated was used to rank proteins according to sessed using Hounsfield units (HU), and laboratory methods. hs-CRP was measured the strength of their separate association low-attenuation plaque was defined as with an automatic immunoturbidimetric with T2D or CAD. This method has been thepresenceofanyvoxel,30 HU. Total method (Beckman-Coulter, Galway, Ire- shown to be preferable to random forest plaque volume, defined as the entire vol- land). High-sensitivity cardiac troponin T or least absolute shrinkage and selection ume of coronary plaque (including both wasmeasuredonanautomatedplatform operator regression when the number of calcified and noncalcified plaque) was (ECLIA Cobas e 411; Roche Diagnostics, predictors (e.g., proteins) is much larger measured. Low-attenuation plaque vol- Rotkreuz, Switzerland) with a limit of blank than the number of cases and when there ume and noncalcified fibrofatty plaque of 3 ng/L. Plasma levels of pentraxin 3 were is a high degree of multicollinearity in the volume was expressed as the amount of measured by in-house sandwich ELISA data (25). Proteins were ranked according plaque with ,30 HU and between 30 and as previously described (21) to the Variable Importance in Projection 150 HU, respectively (expressed in mm3). (VIP) score. Multivariate logistic regres- Noncalcified plaque volume was consid- Quantification of Proteins in Human sion was carried out by standard methods. ered as plaque volume with attenuation Plasma R and SPSS for Mac Os X software were density ,150 HU (20). Atherosclerosis The method of quantification of proteins used; the statistical significance threshold burden was assessed on a per-patient by modified aptamers has been previ- level was set at P , 0.05. basis using previously validated CT scores: ously described (23,24). In brief, each segment involvement score, segment ste- of ;5,000 individual proteins has its nosisscore,andCT-adaptedLeamanscore, own binding reagent made of chemically RESULTS as previously reported; the higher quartile modified DNA, referred to as modified In the CAPIRE cohort, the percentage of was used as the cutoff (6,20,21). Finally, aptamer. Each plasma sample was in- individuals with T2D (12%) is more than cardiac mass was measured in every cubated with the mixture of modified double the age-standardized national prev- patient. aptamers to generate modified aptamer- alence (4.9% in 2016, https://www.istat.it/ protein complexes. Unbound modified en/archivio/202712), reflecting the ex- Participant Groups aptamers and unbound or nonspecifi- pected enrichment of T2D among indi- On the basis of CCTA, participants were cally bound proteins were eliminated viduals referred to a cardiology center. grouped into CAD2 (clean coronaries) by two bead-based immobilization steps The clinical and metabolic phenotype and CAD1 (coronary atherosclerosis ex- and competition with unlabeled polyan- of the participants with T2D included tended to .5 of the 16 segments ac- ion. After eluting the modified aptamers stronger history of hypertension, hyper- cording to the AHA classification [19] from the target protein, the fluorescently cholesterolemia, and smoking; higher fi [segment involvement score .5], with labeled modi ed aptamers were directly BMI, HbA1c, systolic BP, triglycerides, or without coronary stenosis). The five- quantified on an Agilent hybridization non-HDL cholesterol, and hs-CRP; and coronary-segment involvement cutoff array (Agilent Technologies). Calibrators lower HDL cholesterol (Table 1). On the was chosen to define CAD on the basis were included so that the degree of other hand, independently of T2D, the of previously assessed prognostic value fluorescence was a quantitative reflec- presence of CAD was associated with a and on the results of the COronary CT tion of the availability of the three- higher prevalence of male sex, older age, Angiography EvaluatioN for Clinical Out- dimensional shape-charge epitope on each higher BP, serum creatinine, hs-CRP, and comes: An InteRnational Multicenter Reg- specific protein. Results are expressed as troponin T levels. istry (CONFIRM) registry study (17,22). units of fluorescence intensity (F.I.). Medication use is also shown in Table 1. The final population (n 5 528) con- b-Blockers, ACE inhibitors, angiotensin sisted of 64 participants with T2D and Statistical Analysis II receptor blockers (ARBs), and statins 464 participants with normal glucose Continuous variables are presented as were prescribed more frequently in par- tolerance (NGT). CAD1 was detected mean 6 SD; variables with a skewed ticipants with T2D than in participants in 180 participants, 36 of whom had T2D; distribution are given as median (inter- with NGT. Additionally, the CAD1 group ofthe 348participants withCAD2, 28had quartile range). Continuous variables had a higher frequency of statin use than T2D. with a normal distribution were compared the CAD2 group. 846 Protein Biomarkers for CAD in Diabetes Diabetes Care Volume 43, April 2020

Table 1—Demographic, clinical, and biochemical characteristics CAD2 CAD1 NGT (n 5 320) T2D (n 5 28) NGT (n 5 144) T2D (n 5 36) P1 CAD# P2 T2D# P1 3 P2# Sex (% women) 53 57 19 17 ,0.0001 NS NS Age (years) 58 6 8596 9636 7656 6 ,0.0001 NS NS BMI (kg/m2) 25.7 6 3.9 28. 5 6 3.2 27.6 6 4.5 28.6 6 4.2 NS 0.0006 NS Waist girth (cm) 90 6 18 97 6 21 93 6 27 100 6 20 NS 0.0225 NS

HbA1c (%) 5.3 6 0.5 6.3 6 1.1 5.4 6 0.5 6.7 6 1.1 0.0003 ,0.0001 NS Glucose (mmol/L) 5.29 6 1.37 7.02 6 3.90 5.47 6 1.83 7.38 6 3.47 ,0.0001 0.0492 NS OHA/insulin (n) d 22/3 d 26/3 ddd Familial IHD (%) 29 75 33 50 NS NS NS Hypertension (%) 41 89 60 94 0.0007 ,0.0001 NS Hypercholesterolemia (%) 48 89 55 94 NS ,0.0001 NS Current smoking (%) 18 39 33 36 NS 0.0338 NS Cigarettes (n/day) 2.4 6 5.8 8.8 6 12.4 5.6 6 10.0 5.9 6 14.1 NS 0.0031 0.0072 Systolic BP (mmHg) 126 6 14 132 6 18 132 6 15 136 6 20 0.0082 0.0108 NS Diastolic BP (mmHg) 78 6 8796 10 81 6 7816 8 0.0386 NS NS eGFR (mL z min21 z 1.73 m22)916 12 89 6 13 85 6 16 88 6 14 0.0011 NS NS HDL-C (mg/dL) 55 6 16 46 6 14 45 6 12 46 6 11 0.0013 0.0048 NS Triglycerides (mg/dL) 105 6 76 176 6 129 124 6 69 134 6 69 NS 0.0001 0.0041 Non–HDL-C (mg/dL) 148 6 40 137 6 38 147 6 40 121 6 36 NS 0.0008 NS hs-CRP (mg/dL) 0.26 6 0.40 0.41 6 0.38 0.42 6 0.70 0.61 6 1.00 0.0202 0.0241 NS Troponin T (ng/L) 5.12 6 3.87 6.05 6 3.49 8.54 6 9.97 8.91 6 5.10 0.0002 NS NS b-Blockers, n (%) 67 (21) 12 (43) 34 (24) 17 (47) NS 0.0009 NS ACE inhibitors, n (%) 40 (13) 10 (36) 31 (22) 16 (44) ,0.05 0.0002 NS ARBs, n (%) 31 (10) 7 (25) 28 (19) 11 (31) NS 0.0210 NS Statins, n (%) 64 (20) 17 (61) 42 (29) 27 (75) 0.0417 ,0.0001 NS Diuretics, n (%) 20 (6) 3 (11) 16 (11) 6 (17) NS NS NS Data are mean 6 SD unless otherwise indicated. eGFR, estimated glomerular ﬁltration rate; HDL-C, HDL cholesterol; IHD, ischemic heart disease; OHA, oral hypoglycemic agents. #P1 CAD, effect of CAD; P2 T2D, effect of diabetes; P1 3 P2, T2D 3 CAD interaction.

The results of the proteomic analysis Use of ACE inhibitors or ARBs also was and segment stenosis score), and pla- are given in Table 2 for the 20 top-ranking independently associated with incre- que quality (as indexed by the segment proteins selected by the two-way PLS ased renin signal, yet inclusion in the involvement score, plaque volume and procedure. Fifteen of them were asso- model of these treatments (or statin, length, remodeling index, low-attenuation ciated with both T2D and CAD (with a diuretic, and b-blocker treatment) as plaque, spotty calcifications, and non- significant T2D 3 CAD interaction for additional covariates did not change calcified fibrofatty plaque)das resolved eight of them), while three proteins the association of renin with T2D and by CCTAddid not differ between patients (osteomodulin, cartilage intermediate- CAD. Likewise, adiponectin levels were with and without T2D (Supplementary layer protein 2, and serine protease higher in women and increased with age Table 2). HTRA1) were associated only with T2D, but were significantly reduced in both In a multivariate logistic model adjust- andtwootherproteins(epidermalgrowth T2D and CAD (Supplementary Table 1). ing for the Framingham risk factors factor receptor [EGFR] and contactin-1) The physiological relevance of the ob- (sex, age, hypertension, hypercholester- were selectively associated with CAD. By servedproteomicpatternissupportedby olemia, and smoking), patients with T2D way of example, Fig. 1 plots the median individual-level data. For example, one of were “protected” from CAD if they values of four of the proteins with the proteins uniquely associated with were female (odds ratio [OR] 0.19 the highest VIP score; in participants CAD, EGFR, was reciprocally related to [95% CI 0.04–0.70]), younger (ORSD with T2D and CAD, renin levels were hs-CRP and troponin T concentrations 0.12 [0.012–0.99]), and had lower plasma more than doubledand adiponectin (i.e., established CAD markers) (Fig. 2). renin levels (ORln[F.I.] 0.25 [0.07–0.71]). levels were ;30% reduceddcompared Likewise, one of the proteins selectively with participants with NGT without related to T2D, serine protease HTRA1, CONCLUSIONS CAD. Of note, renin levels were signif- was directly associated with suboptimal To the best of our knowledge, the current icantlylowerinwomenthaninmenand glycemic control, as indexed by HbA1c, study is the first to investigate ;5,000 were directly related to age; yet, in a and serum triglycerides. Of note, myo- plasma proteins in relation to the pres- multivariate regression model adjusting cardial mass, extent and severity of ence/absence of T2D and CAD and in for sex and age, renin was still increased coronary atherosclerotic involvement relation to clinical, metabolic, and angio- in association with both T2D and CAD. (as indexed by lumen stenosis .70% graphic phenotyping. The main conclusion care.diabetesjournals.org Ferrannini and Associates 847

Table 2—Serum proteins with a VIP >3.8 CAD2 CAD1 NGT T2D NGT T2D P1 T2D* P2 CAD* P1 3 P2* Growth differentiation factor 15 19.6 6 7.6 32.5 6 13.8 24.3 6 9.5 33.6 6 12.2 ,0.001 ,0.001 NS Renin 15.8 6 9.6 18.2 6 11.7 20.7 6 12.1 35.5 6 20.5 ,0.001 ,0.001 0.004 Heparan sulfate 6-O-sulfotransferase 1.31 6 0.21 1.17 6 0.20 1.18 6 0.20 1.10 6 0.18 ,0.001 ,0.001 NS Matrix-remodeling–associated protein 8 0.54 6 0.12 0.43 6 0.97 0.47 6 0.10 0.43 6 0.10 ,0.001 NS 0.013 Adiponectin 3.4 6 1.4 2.4 6 0.7 2.7 6 1.1 2.3 6 0.8 ,0.001 0.004 NS Chondroadherin 8.3 6 0. 2.6 6.4 6 2.2 7.4 6 2.4 5.6 6 1.6 ,0.001 0.013 NS Osteomodulin 8.4 6 2.9 5.7 6 2.1 7.3 6 2.7 5.6 6 1.9 ,0.001 NS NS Cartilage intermediate-layer protein 2 2.2 6 0.7 1.6 6 0.5 1.9 6 0.6 1.6 6 0.5 ,0.001 NS NS Serine protease HTRA1 12.0 6 2.7 14.8 6 3.2 13.2 6 3.2 14.5 6 0.3 ,0.001 NS NS EGFR 16.1 6 2.4 15.0 6 2.3 14.6 6 2.3 14.4 6 2.3 NS 0.001 NS Leucine-rich repeat-containing protein 15 1.3 6 0.4 1.0 6 0.3 1.2 6 0.4 0.98 6 0.28 ,0.001 NS 0.008 Neurocan core protein 4.0 6 1.5 2.8 6 1.2 3.3 6 1.2 3.1 6 1.1 ,0.001 NS 0.008 Anthrax toxin receptor 2 1.6 6 0.5 1.3 6 0.4 1.4 6 0.4 1.4 6 0.4 0.008 NS 0.007 Tetranectin 4.8 6 0.7 4.5 6 0.6 4.5 6 0.6 4.2 6 0.5 0.002 ,0.001 NS IGF-binding protein complex subunit 28.8 6 5.1 26.0 6 4.9 26.5 6 5.4 24.6 6 7.2 ,0.001 0.004 NS Contactin-1 32.8 6 6.6 30.9 6 5.9 29.6 6 5.1 28.4 6 4.6 NS 0.001 NS Receptor-type tyrosine-protein phosphatase 5.6 6 1.8 4.6 6 1.3 4.7 6 1.4 4.7 6 1.3 0.015 NS 0.029 Protein FAM177A1 2.8 6 0.7 2.6 6 0.7 2.5 6 0.7 2.2 6 0.5 0.005 ,0.001 NS Coiled domain-containing protein 126 1.1 6 0.3 0.9 6 0.3 0.9 6 0.3 0.9 6 0.3 0.001 NS 0.004 Ubiquitin-conjugating enzyme E2 G2 3.2 6 0.6 3.7 6 1.0 3.5 6 0.7 3.6 6 0.9 0.002 NS 0.017 Data are F.I. units 3 1023.*P1 CAD, effect of CAD; P2 T2D, effect of diabetes; P1 3 P2, T2D 3 CAD interaction. is that the majority of “hit” proteins iden- current study have already been studied Malmo¨ Diet and Cancer-Cardiovascular tifiedbybivariate PLSanalysisaresharedby in both CAD and T2D by investigations Cohort, in which 4,360 participants the diabetes status and the presence of using traditional assay methods. with normoglycemia were followed coronary lesions. Additionally, some inter- In our analysis, GDF15 emerged as the for 19 years, recently showed that esting proteins are uniquely associated highest-ranking protein associated with GDF15 may be useful for the identifica- with T2D or CAD, suggesting diverging both CAD and T2D (Fig. 1). GDF15 is a tion of people with a risk of incident disease pathways and, possibly, different cytokine widely expressed in macro- diabetes (27). High GDF15 levels might therapeutic targets. phages, vascular smooth muscle cells, be a target for further investigations Proteomics offer unique insights into adipocytes, cardiomyocytes, endothe- and/or strengthen the indication for disease pathways, as proteins are the lial cells, fibroblasts, and brain tissues preventive measures. effectors of all biological processes and in response to inflammation, oxidative The second-highest ranking protein their differential expression reflects the stress, and hypoxia. In patients with was renin, with more than twofold higher intricate interplay between environmen- established CAD, GDF15 concentrations levels in patients with both T2D and CAD tal and genetic factors. In fact, the top- are independently associated with age, than in participants with normoglyce- ranking proteins in our predictive model diabetes, smoking, hs-CRP levels, and mia without CAD. Growing evidence are involved in a broad spectrum of renal impairment, all features of our points to plasma renin activity, and processes,includingdifferentiation,extra- CAD1 subgroup. In additiondunlike car- the attendant activation of the renin- cellularmatrixcomposition,metabolicsig- diac troponin, hs-CRP, and natriuretic angiotensin-aldosterone system (RAAS), naling, and ubiquitination. Reflecting this, peptidesdcirculating GDF15 levels re- as a marker of CV risk in patients with for instance,matrix-remodeling–associated main stable over time, therefore repre- hypertension (28). Most drugs targeting protein 8, cartilage intermediate-layer senting an optimal biomarker of chronic the RAAS, either directly or indirectly, protein 2,chondroadherin,leucine-rich CAD burden. Among 14,577 patients have shown a favorable effect in reduc- repeat-containing protein 15, and HTRA1 with stable CAD participating in the Sta- ing morbidity and mortality of patients are all engaged in pericellular microen- bilization of Atherosclerotic Plaque by with CAD and/or heart failure (28). More- vironment regulation. Moreover, some of Initiation of Darapladib Therapy Trial over, in animal models of hypertension, the detected proteins are already known (STABILITY), GDF15 was found to be in- direct renin inhibition improves insulin to have a pathogenetic role in several dependently associated with CV, non-CV, sensitivity and secretion, skeletal mus- clinical conditions; for example, anthrax and cancer mortality. Furthermore, GDF15 cle glucose uptake, and glucose toler- toxin receptor 2 gene mutation causes concentrations were lowest in patients ance (29). Our data on lower renin hyaline fibromatosis syndrome, a rare with NGT, intermediate in patients with levels in women than in men are in recessive autosomal disease. In addition, prediabetes, and highest in patients with line with epidemiological studies, an some of the proteins detected in the T2D (26). Strengthening this finding, the effect attributed to estrogens and 848 Protein Biomarkers for CAD in Diabetes Diabetes Care Volume 43, April 2020

Figure 1—GDF15, renin, adiponectin, and chondroadherin signals in participants with T2D and participants with NGT with or without CAD. Bars are median group values. Statistical analysis is presented in Table 2.

associated with favorable lipid and glu- note, meta-analyses have failed to clearly of transforming growth factor-b sig- cose profiles, less visceral adipose tissue, establish its predictive role in CAD, reflecting naling, in programmed cell death, and in and lower BP in premenopausal women the multitude of factors that influence its modulating the EGFR/Akt pathway (33). (29). Consistent with the above, the circulating levels (30). To our knowledge, Moreover, this proteasedprotective currentstudy identifieda youngerfemale while the association between T2D and against both CAD and diabetes in our with T2D but free of CAD. Importantly, adiponectin is well established (31), datadcontains an IGF-binding domain the association of renin with T2D and no previous study has tested adiponectin potentially interacting with other IGF- CAD was independent of the use of RAAS against anatomical evidence of CAD as binding protein complex subunits, which blockers, which raise renin levels. In done here. Collectively, our data sup- also ranked high in our model. To the best future studies, this finding should be port a protective role of adiponectin of our knowledge, this is the first report confirmed by direct assay of plasma renin against both CAD and T2D. linking serine protease HTRA1 with a concentrations. Encouragingly, one of the proteins clinical metabolic disease. The inverse correlation between adi- selectively associated with T2D, serine One of the proteins selectively asso- ponectin and both CAD and T2D is con- protease HTRA1, correlated directly with ciated with CAD is EGFR, a tyrosine kinase sistent with substantial literature that both HbA1c levels and serum triglycer- widely studied as a therapeutic target has unraveled its antioxidant, anti- ides, which are hallmarks of the diabetic in cancer biology and more recently in inflammatory, antiatherosclerotic, and phenotype. With regard to its specific relation to CV disease (34). However, the insulin-sensitizing effects. Indeed, adipo- function,serineproteaseHTRA1has clinical impact of EGFR activation for nectin is now considered a key protective been suggested to play a role in vascu- cardiac function and atherogenesis regulator of vascular homeostasis, glucose lar abnormalities and angiogenesis remains to be elucidated with targe- metabolism, and lipid oxidation (30). Of (32). It is involved in the inhibition ted clinical studies, since it acts as a care.diabetesjournals.org Ferrannini and Associates 849

Figure 2—Association of EGFR with markers of CAD (hs-CRP and high-sensitivity cardiac troponin T) and of serine protease HTRA1 with markers of diabetes (HbA1c and triglycerides).

crossroad to many signaling pathways third highest-ranking protein in our with T2D with or without CAD was not and its levels are influenced by diverse model, previously showed a direct cor- very different (i.e., familial ischemic concurrent factors (34,35). In our model, relation with albuminuria in participants heart disease, hypertension, hypercho- EGFR was found to be higher in the with diabetes (38). lesterolemia, etc.) (Table 1). This has CAD2 group and, consistently, inversely To qualify our findings, it is important two corollaries. On the one hand, it related to hs-CRP and troponin T, well- to consider that the T2D group included helps justify the finding of similar quan- established CAD markers. This associa- in CAPIRE is not a random sample of a titative and qualitative involvement of tion may contribute to improving our typical T2D populationdas the anthro- the coronary vasculature between pa- understanding of its role in clinical pometrics and metabolic characteristics tients with T2D and participants without settings. would indicatedon several accounts. diabetes (Supplementary Table 1); in less Tetranectin, which is actively involved First, patients were selected for absence selected cohorts, coronary atheroscle- in fibrinolysis as a plasminogen activa- of previous CV events but presence of rosis, when present, is more diffuse tion enhancer, has been suggested as a suspectedCADandanindicationfor and more severe in individuals with dia- biomarker in stable and acute CAD, both CCTA. Second, patients with T2D had betes than in those without diabetes (39). in general populations and in T2D (36). to be older than 45 years and in good On the other hand, individuals with T2D Accordingly, in our screen, tetranectin glycemic control (as attested by their (andNGT)withahighCVriskprofile who was inversely associated with both T2D HbA1c and fasting glucose levels); in have reached their 60s with anatomi- and CAD. Also, we found cartilage in- fact, only six of them were on insulin cally clean coronaries offer the oppor- termediate-layer protein 2, which may treatment, the rest being on oral anti- tunity to begin to search for potentially have a role in both cardiac extracellu- hyperglycemic agents or diet alone. Fi- protective features. Efforts to pheno- lar matrix remodeling and insulin resis- nally, as participants were grouped into type study participants are key when tance (37). An altered level of enzyme CAD1 and CAD2 a posteriori (i.e., after performing proteomic studies, particu- heparan sulfate 6-O-sulfotransferase, the the CCTA), the CV risk profile of patients larly with advanced methodology of 850 Protein Biomarkers for CAD in Diabetes Diabetes Care Volume 43, April 2020

recent introduction, such as aptamer- Author Contributions. G.F. and E.F. researched Guidelines on cardiovascular disease prevention based technology, for which there are data and wrote the manuscript. M.L.M. carried in clinical practice: the Sixth Joint Task Force of out all statistical analyses. M.M., F.A., D.A., R.L., limited data available for replication the European Society of Cardiology and Other A.M., A.P.M., R.M.O., and S.A.W. reviewed/ Societies on Cardiovascular Disease Prevention studies. edited the manuscript and contributed to the in Clinical Practice (constituted by representa- There are several limitations in our discussion. R.L.supervisedthe routinelaboratory tives of 10 societies and by invited experts) study. First, being a cross-sectional in- analyses. R.M.O.and S.A.W. were responsiblefor Developed with the special contribution of the vestigation, it cannot prove causality. the proteomics. E.F. is the guarantor of this work European Association for Cardiovascular Pre- and, as such, had full access to all the data in the Second, the number of participants is vention & Rehabilitation (EACPR). Eur Heart J study and takes responsibility for the integrity of 2016;37:2315–2381 relatively small, especially in the T2D the data and the accuracy of the data analysis. 10. Oellgaard J, Gæde P, Rossing P, et al. Re- group. CAPIRE was not an intervention duced risk of heart failure with intensified mul- trial population but a highly selected References tifactorial intervention in individuals with type cohort that was based on CCTA and 1. World Health Organization. Global Health 2 diabetes and microalbuminuria: 21 years of designed to yield a sizeable number of Estimates 2016: Disease Burdenby Cause, Age, follow-up in the randomised Steno-2 study. – Diabetologia 2018;61:1724–1733 outlier participants (i.e., at the oppo- Sex, by Country and by Region, 2000 2016 [Internet], 2018. Available from https://www 11. Rawshani A, Rawshani A, Franzen´ S, et al. site extremes for the presence of CAD .who.int/healthinfo/global_burden_disease/ Risk factors, mortality, and cardiovascular out- and traditional CVRFs) with the specific estimates/en/index1.html. Accessed 20 Jan- comes in patients with type 2 diabetes. N Engl J purpose of generating hypotheses rather uary 2020 Med 2018;379:633–644 than hypothesis-driven outcomes. To 2. American Diabetes Association. 9. Cardiovas- 12. Rao Kondapally Seshasai S, Kaptoge S, Thompson A, et al.; Emerging Risk Factors Col- achieve this goal, the anatomical pheno- cular disease and risk management: Standards of Medical Care in Diabetesd2018. Diabetes Care laboration. Diabetes mellitus, fasting glucose, type of participants had to be described 2018;41(Suppl. 1):S86–S104 and risk of cause-specific death. N Engl J Med meticulously. Thus, CCTA was the main 3. Mark DB, Berman DS, Budoff MJ, et al.; 2011;364:829–841 entry criterion. American College of Cardiology Foundation 13. Sarwar N, Gao P, Seshasai SR, et al.; Emerg- ing Risk Factors Collaboration. Diabetes mellitus, Nonetheless, we believe this to be the Task Force on Expert Consensus Documents. fasting blood glucose concentration, and risk of fi ACCF/ACR/AHA/NASCI/SAIP/SCAI/SCCT 2010 rst proteomic analysis of a T2D popu- expert consensus document on coronary com- vascular disease: a collaborative meta-analysis of lation in relation to coronary phenotyp- puted tomographic angiography: a report of the 102 prospective studies [published correction ing; accuracy is supported by plausibility American College of Cardiology Foundation Task appears in Lancet 2010;376:958]. Lancet 2010; with published data and by significant Force on Expert Consensus Documents. Circula- 375:2215–2222 14. Lindsey ML, Mayr M, Gomes AV, et al.; individual-level data correlations. Our tion 2010;121:2509–2543 4. Knuuti J, Wijns W, Saraste A, et al.; ESC American Heart Association Council on Func- model highlights the potential role of Scientific Document Group. 2019 ESC Guidelines tional Genomics and Translational Biology, Coun- GDF15, renin, serine protease HTRA1, for the diagnosis and management of chronic cil on Cardiovascular Disease in the Young, EGFR, and other top-ranking proteins in coronary syndromes [published correction ap- Council on Clinical Cardiology, Council on Car- the development of T2D and/or CAD. pears in Eur Heart J 2019:ehz825]. Eur Heart J diovascular and Stroke Nursing, Council on Hy- pertension, and Stroke Council. Transformative Further investigation, including specific 2019:ehz425 5. Fihn SD, Gardin JM, Abrams J, et al.; American impact of proteomics on cardiovascular health mechanistic studies, is required to vali- College of Cardiology Foundation. 2012 ACCF/ and disease: a scientific statement from the date our findings. In addition, direct AHA/ACP/AATS/PCNA/SCAI/STS guideline for American Heart Association. Circulation 2015; quantitation of the more promising pro- the diagnosis and management of patients 132:852–872 tein biomarkers could contribute to a with stable ischemic heart disease: executive 15. Ganz P, Heidecker B, Hveem K, et al. De- summary: a report of the American College of velopment and validation of a protein-based risk deeper understanding of their clinical score for cardiovascular outcomes among pa- fi Cardiology Foundation/American Heart Associ- signi cance and potential therapeutic ation task force on practice guidelines, and the tients with stable coronary heart disease. JAMA implications. Also, follow-up events in American College of Physicians, American Associ- 2016;315:2532–2541 CAPIRE will help with understanding ation for Thoracic Surgery, Preventive Cardiovas- 16. Lopez-Villar´ E, Martos-Moreno GA,´ Chowen whether screening proteomics can siz- cular Nurses Association, Society for Cardiovascular JA, Okada S, Kopchick JJ, Argente J. A proteomic approach to obesity and type 2 diabetes. J Cell ably improve CAD prediction in partic- Angiography and Interventions, and Society of Thoracic Surgeons. Circulation 2012;126:3097– Mol Med 2015;19:1455–1470 ipants with and without diabetes or even 3137 17. Magnoni M, Andreini D, Gorini M, et al.; reduce the indication for CCTA. 6. Mushtaq S, De Araujo Gonçalves P, Garcia- CAPIRE Study Group. Coronary atherosclerosis Garcia HM, et al. Long-term prognostic effect of in outlier subjects at the opposite extremes of coronary atherosclerotic burden: validation of traditional risk factors: rationale and preliminary the computed tomography-Leaman score. Circ results of the Coronary Atherosclerosis in outlier Funding. Funding was provided by the Heart Cardiovasc Imaging 2015;8:e002332 subjects: Protective and novel Individual Risk Care Foundation of the Italian Association of 7. Budoff MJ, Raggi P, Beller GA, et al.; Imaging factors Evaluation (CAPIRE) study. Am Heart J Hospital Cardiologists, Florence, Italy. Council of the American College of Cardiology. 2016;173:18–26 Duality of Interest. F.A. reports receiving con- Noninvasive cardiovascular risk assessment of 18. Expert Panel on Detection, Evaluation, and sultancy/speaker fees from Amgen, Bayer, the asymptomatic diabetic patient: the Imaging Treatment of High Blood Cholesterol in Adults. Boehringer Ingelheim, Bristol-Myers Squibb/ Council of the American College of Cardiology. Executive summary of the third report of the Pfizer,andDaiichiSankyo.A.P.M.reportsre- JACC Cardiovasc Imaging 2016;9:176–192 National Cholesterol Education Program (NCEP) ceiving fees, outside the present work, from 8. Andreini D, Pontone G, Mushtaq S, et al. expert panel on detection, evaluation, and treat- Bayer, Fresenius, and Novartis for participation Prognostic value of multidetector computed ment of high blood cholesterol in adults (adult in study committees. R.M.O. and S.A.W. are tomography coronary angiography in diabetes: treatment panel III). JAMA 2001;285:2486– employeesof SomaLogic Inc.E.F.reports receiving excellent long-term prognosis in patients with 2497 consultancy/speaker fees, outside the present normalcoronaryarteries.DiabetesCare2013;36: 19. Austen WG, Edwards JE, Frye RL, et al. A work, from Boehringer Ingelheim, Eli Lilly, Astra- 1834–1841 reporting system on patients evaluated for cor- Zeneca, and Sanofi. No other potential conflicts of 9. Piepoli MF, Hoes AW, Agewall S, et al.; ESC onaryarterydisease.ReportoftheAdHoc interest relevant to this article were reported. Scientific Document Group. 2016 European Committee for Grading of Coronary Artery care.diabetesjournals.org Ferrannini and Associates 851

Disease, Council on Cardiovascular Surgery, for omics-type data. Adv Appl Bioinform Chem tumor growth by deregulating angiogenesis. American Heart Association. Circulation 1975; 2009;2:57–70 Oncogene 2018;37:4260–4272 51(Suppl.):5–40 26. Desmedt S, Desmedt V, De Vos L, Delanghe 33. Skorko-Glonek J, Zurawa-Janicka D, Koper T, 20. Andreini D, Magnoni M, Conte E, et al.; JR, Speeckaert R, Speeckaert MM. Growth dif- et al. HtrA protease family as therapeutic targets. CAPIRE Investigators. Coronary plaque features ferentiation factor 15: a novel biomarker with Curr Pharm Des 2013;19:977–1009 on CTA can identify patients at increased risk of high clinical potential. Crit Rev Clin Lab Sci 2019; 34. Forrester SJ, Kawai T, O’Brien S, Thomas W, cardiovascular events [published ahead of print 56:333–350 Harris RC, Eguchi S. Epidermal growth factor on 14 August 2019]. JACC Cardiovasc Imaging 27. Bao X, Borne´ Y, Muhammad IF, et al. Growth receptor transactivation: mechanisms, path- DOI: 10.1016/j.jcmg.2019.06.019 differentiation factor 15 is positively associated ophysiology, and potential therapies in the car- 21. Andreini D, Pontone G, Mushtaq S, et al. A with incidence of diabetes mellitus: the Malmo¨ diovascular system. Annu Rev Pharmacol Toxicol – long-term prognostic value of coronary CT an- Diet and Cancer-Cardiovascular Cohort. Diabe- 2016;56:627 653 giography in suspected coronary artery disease. tologia 2019;62:78–86 35. MakkiN,ThielKW,MillerFJJr.Theepidermal JACC Cardiovasc Imaging 2012;5:690–701 28. Bangalore S, Fakheri R, Wandel S, Toklu B, growth factor receptor and its ligands in cardiovascular disease. Int J Mol Sci 2013;14:20597– 22. Min JK, Dunning A, Lin FY, et al.; CONFIRM Wandel J, Messerli FH. Renin angiotensin system 20613 Investigators. Age- and sex-related differences in inhibitorsforpatientswith stablecoronaryartery 36. Rahim MAA, Rahim ZHA, Ahmad WAW,Bakri all-cause mortality risk based on coronary com- disease without heart failure: systematic review MM, Ismail MD, Hashim OH. Inverse changes in puted tomography angiography findings results and meta-analysis of randomized trials. BMJ plasma tetranectin and titin levels in patients from the International Multicenter CONFIRM 2017;356:j4 with type 2 diabetes mellitus: a potential pre- (Coronary CT Angiography Evaluation for Clinical 29. White MC, Fleeman R, Arnold AC. Sex differ- dictor of acute myocardial infarction? Acta Phar- Outcomes: an International Multicenter Registry) of ences in the metabolic effects of the renin- macol Sin 2018;39:1197–1207 23,854 patients without known coronary artery angiotensin system. Biol Sex Differ 2019;10:31 37. Wu T, Zhang Q, Wu S, et al. CILP-2 is a novel disease. J Am Coll Cardiol 2011;58:849–860 30. Woodward L, Akoumianakis I, Antoniades C. secreted protein and associated with insulin re- 23. Gold L, Ayers D, Bertino J, et al. Aptamer- Unravelling the adiponectin paradox: novel roles sistance. J Mol Cell Biol 2019;11:1083–1094 based multiplexed proteomic technology for of adiponectin in the regulation of cardiovascular 38. Teumer A, Tin A, Sorice R, et al.; DCCT/EDIC. biomarker discovery. PLoS One 2010;5:e15004 disease. Br J Pharmacol 2017;174:4007–4020 Genome-wide association studies identify ge- 24. Vaught JD, Bock C, Carter J, et al. Expanding 31. Li S, Shin HJ, Ding EL, van Dam RM. Adipo- netic loci associated with albuminuria in diabe- the chemistry of DNA for in vitro selection. J Am nectin levels and risk of type 2 diabetes: a sys- tes. Diabetes 2016;65:803–817 Chem Soc 2010;132:4141–4151 tematic review and meta-analysis. JAMA 2009; 39. Natali A, Vichi S, Landi P, Severi S, L’Abbate A, 25. Palermo G, Piraino P, Zucht H-D. Perfor- 302:179–188 Ferrannini E. Coronary atherosclerosis in Type II mance of PLS regression coefficients in selecting 32. Klose R, Adam MG, Weis E-M, et al. In- diabetes: angiographic findings and clinical out- variables for each response of a multivariate PLS activation of the serine protease HTRA1 inhibits come. Diabetologia 2000;43:632–641