Endogenous PCSK9 may influence circulating CD45neg/CD34bright and CD45neg/CD34bright/CD146neg cells in patients with type 2 diabetes mellitus

Protease proprotein convertase subtilisin/kexin type 9 (PCSK9) is a regulator of LDL cholesterol clearance and has been associated with cardiovascular risk. PCSK9 inhibitors increase in vivo circulating endothelial progenitor cells (EPCs), a subtype of immature cells involved in ongoing endothelial repair. We hypothesized that the effect of PCSK9 on vascular homeostasis may be mediated by EPCs in patients with or without type 2 diabetes mellitus (T2DM). Eighty-two patients (45 with, 37 without T2DM) at high cardiovascular risk were enrolled in this observational study. Statin treatment was associated with higher circulating levels of PCSK9 in patients with and without T2DM (p < 0.001 and p = 0.036) and with reduced CD45neg/CD34bright (total EPC compartment) (p = 0.016) and CD45neg/CD34bright/CD146neg (early EPC) (p = 0.040) only among patients with T2DM. In the whole group of patients, statin treatment was the only independent predictor of low number of CD45neg/CD34bright (β = − 0.230; p = 0.038, adjusted R2 = 0.041). Among T2DM patients, PCSK9 circulating levels were inversely related and predicted both the number of CD45neg/CD34bright (β = − 0.438; p = 0.003, adjusted R2 = 0.173), and CD45neg/CD34bright/CD146neg (β = − 0.458; p = 0.002, adjusted R2 = 0.191) independently of age, gender, BMI and statin treatment. In high-risk T2DM patients, high endogenous levels of PCSK9 may have a detrimental effect on EPCs by reducing the endothelial repair and worsening the progression of atherothrombosis.

anti-PCSK9 antibody inhibits markers of inflammation and atherosclerosis and interestingly, leads to an increase in circulating endothelial progenitor cells (EPCs) and angiogenic cells in a mouse model 9 and recently in patients with cardiovascular disease 10 .
This mechanism may be particularly important since EPCs have been defined as "biomarkers" of endothelial function 11 , have a role in ongoing endothelial repair 12 and their reduction has been linked to CV outcomes 13 . Indeed, after adjustment for relevant variables, increased levels of EPCs were associated with a reduced risk of death from CV causes, first major CV event, revascularization, and hospitalization 14 .
EPCs are damaged in disease. In patients with T2DM, EPCs exhibit impaired proliferation, adhesion, and incorporation into vascular structures 15 and in patients at high risk for CV events EPCs show higher senescence. This may lead to EPCs depletion, thus impairing the repair of vessel walls and favouring the progression of CV disease 16 .
CD34 is a common marker for diverse progenitors, including hematopoietic stem and progenitor cells, vascular endothelial progenitors 17 , cardiomyocytes and smooth muscle cells 18 . Human CD34 pos cells are provided with vascular regenerative capacity and proangiogenic potential in vivo 19 , and their depletion is now considered a significant contributor to the impaired coronary endothelial dysfunction 20 and altered cardiovascular homeostasis in diabetes. A useful marker to identify non hematopoietic stem and progenitor cells is the absence of the hematopoietic antigen CD45. The robust proliferative potential and endothelial colony forming capacity of CD45 neg /CD34 pos was confirmed in vitro 21 . During the clinical onset of T2DM, a progressive reduction of CD34pos progenitor cells has been observed 22 .
However, CD34 is also expressed on mature circulating endothelial cells (CECs), and thus additional and appropriate antigens are required to discriminate between CECs and EPCs. CD146 is a marker that may further help the characterization. Shim et al. identified two EPC subpopulations based on CD146, which is not expressed on early outgrowth EPCs (stemming from the bone marrow), whereas late outgrowth EPCs do express CD146 23 . While the phenotype of mature CECs (CD45 neg /CD34 bright /CD146 pos ) is well established 24,25 , the antigen pattern identifying EPCs has not been clarified yet. In any case, the circulating population of CD45 neg /CD34 bright / CD146 neg cells results enriched in the earlier EPCs 23,26 . Another subpopulation of CD34 pos cells enriched in EPCs co-expresses CD309 or kinase insert domain receptor (KDR) or vascular endothelial growth factor receptor-2 (VEGFR2) 27 . Several authors have defined late circulating EPCs as CD34 pos CD309 pos cells, proving that cells characterized by this phenotype are able to stimulate angiogenesis in vivo 27 . Confirming these findings, Sandhu et al. 28 also demonstrated that, in parallel with the progression toward the endothelial cell maturation lineage, CD309 and other endothelial markers increase their expression levels. Taking into account these results we may define the CD45 neg /CD34 bright /CD146 neg subset as the early EPC compartment, while the expression of CD309 within the CD45 neg /CD34 bright may be associated with the late EPC subtype.
We hypothesized that the effect of PCSK9 on vascular homeostasis may be mediated by EPCs. Thus, the aim of this study was to evaluate the relationship between circulating levels of PCSK9 and circulating EPCs as assessed by the number of CD45 neg /CD34 bright (total EPC compartment), CD45 neg /CD34 bright /CD146 neg (early EPC) and CD45 neg /CD34 bright /CD309 pos (late EPC) and on CEC, as reflected by CD45 neg /CD34 bright /CD146 pos , and their mutual relationship, in patients at risk of CV disease, with or without T2DM.

Results
Baseline characteristics. Eighty-two patients were enrolled, 45 with and 37 without T2DM. Baseline characteristics of patients are reported in Table 1.
As expected, patients were significantly different for fasting plasma glucose (p < 0.001), HbA1c (p < 0.001) and diabetes specific treatment such as metformin (p < 0.001) and PPAR-gamma agonists (p = 0.012). Notably, median HbA1c in T2DM patients was 6.8% (51.0 mmol/mol), reflecting good glycemic control in the group of patients with diabetes included in the study.

Repeatability of the outcome measurement.
To verify the intra-subject reproducibility of the main outcome measured, we re-assessed EPC number after 24 h in 9 out of 82 patients analyzed. We found no significant difference (p = 0.573) between the two consecutive days measurements (data not shown).
Thirty-six patients were on statin treatment, 25 with and 11 without T2DM. Patients on statin treatment had a higher prevalence of diabetes, previous myocardial infarction (MI), lower cholesterol and higher transaminase levels. Otherwise, their clinical characteristics were superimposable. Patients on statin treatment, both with and without T2DM, showed higher levels of PCSK9 as compared to their counterparts not on statins (p < 0.001 and p = 0.036 respectively) ( Fig. 2A). Of note, the association of statins with PCSK9 was independent of type of statin used (P = 0.86 for difference among types), of statin dose (P = 0.10 for difference between high versus moderate/ low intensity) and of achievement of LDL target (< 70 mg/dL, according to latest ESC guidelines) (P < 0.0001 for the association of statins with PCSK9 after adjustment for LDL target reached (yes versus no)). Conversely,     To assess whether cell counts may be further related to statin dose or intensity, we stratified patients accordingly. Namely, seven patients were on high-intensity statins, and 29 on moderate/low-intensity statins. We were not able to assess any relationship between statin dose or intensity and the level of any of the EPC phenotypes analysed. Among individuals with statins, CD45 neg /CD34 bright (p = 0.072), CD45 neg /CD34 bright /CD146 neg (p = 0.051) and CD45 neg /CD34 bright /CD309 pos levels were not associated neither with statins type (P = 0.51, P = 0.84, and P = 0.43, respectively) nor with dose of statin (P = 0.22, P = 0.29, P = 0.58) and for difference between high versus moderate/low-intensity, respectively.
Interestingly, among subjects not on statin therapy, no difference was detected in either PCSK9 levels or EPCs number between patients with vs. those without T2DM (Fig. 2), while significantly reduced CD45 neg /CD34 bright (p = 0.035) and CD45 neg /CD34 bright /CD146 neg (p = 0.032) were observed in patients with diabetes vs. patients without diabetes only among statin-treated subjects (Fig. 2B,C).
Only among patients not on statins, we observed a positive correlation between PCSK9 and total cholesterol (Rho = 0.369, p = 0.012) and LDL cholesterol (Rho = 0.470 p = 0.001) (Supplementary Fig. 1) but not with HDL or triglycerides (data not shown). Neither PCSK9 nor EPC number were related to fasting plasma glucose or HbA1c (data not shown).
Since we have observed that statins affect both PCSK9 levels and CD45 ne g/CD34 bright and CD45 neg /CD34 bright / CD146 neg , we made a sub-analysis as a function of statin treatment. Interestingly, when dividing patients according to statin treatment, we found that the inverse correlations between PCSK9 levels and CD45 neg /CD34 bright and CD45 neg /CD34 bright /146 neg remained only in patients with diabetes on statin treatment (Rho = − 0.454, p = 0.022 and Rho = − 0.553, p = 0.004 respectively) (Fig. 4).

Effect of previous myocardial infarction (MI).
Interestingly, patients with a previous MI (n = 12) showed significantly higher levels of plasma PCSK9 (p = 0.007), and significantly lower number of CD45 neg / CD34 bright /CD146 neg (p = 0.031). Again, in patients with a previous MI, the prevalence of ongoing statin treat-

Discussion
In this cross-sectional study, we evaluated and compared the circulating levels of PCSK9, the number of several EPC phenotypes and their mutual relationship between patients with T2DM and without T2DM, otherwise comparable for most of the clinical characteristics. The main findings of our study are: 1. Statin treatment is associated with higher circulating levels of PCSK9 in both groups, and with reduced early EPC number only among patients with T2DM; 2. PCSK9 and EPC number are comparable between patients with vs. without T2DM, when considering subjects not on statins. In contrast, among statin-treated patients, both the whole CD45 neg /CD34 bright subset as well as the early EPC compartment are reduced in patients with T2DM vs. patients without T2DM. 3. Only in the group of T2DM patients, higher circulating PCSK9 is associated with reduced number of early EPCs and CD45 neg /CD34 bright cells. This inverse correlation is not observed in patients without T2DM or in the subgroup of patients not on statins, suggesting that in subjects with diabetes and in particular in statin-treated subjects, enhanced circulating PCSK9 may impair early EPC number, reflecting cardiovascular homeostasis and regenerative capacity. Of note, CEC (CD45 neg /CD34 bright /CD146pos) levels were not related to statin therapy, despite acknowledged pleiotropic effects of statins on vascular integrity 29 . In our patients we observed higher levels of PCSK9 in patients treated with statins, both with and without T2DM. This is in line with the observation that statins can directly upregulate the expression of PCSK9 30,31 . The observed trend toward higher circulating PCSK9 levels in T2DM patients is likely to be driven by the higher prevalence of statin treatment among T2DM (56% vs. 30%), rather than by an influence of glucose metabolism on PCSK9. The relationship between PCSK9 and glucose homeostasis is complex and controversial in the literature 32 .
Mendelian randomization studies have shown that PCSK9 genetic variants associated with lower LDL cholesterol are also associated with circulating higher fasting glucose concentration and an increased risk of Type 2 diabetes 33 . In experimental mice models, PCSK9 deficiency increased LDLR expression and cholesterol esters accumulation in pancreatic islets, which impairs insulin secretion 34 , although this is likely a local effect, rather than a consequence of circulating PCSK9. The potential impact of PCSK9 on insulin secretion was confirmed in a human biobank, where a PCSK9 46L variant was associated with beta cell dysfunction but not with insulin resistance (HOMA-IR) 35 . Vice versa, insulin is known to induce PCSK9 expression 36 . Recently, a novel crosstalk signal between glucose and cholesterol homeostasis via ChREBP-mediated PCSK9 regulation has been described 37 . www.nature.com/scientificreports/ In our study, lack of correlation between fasting plasma glucose or HbA1c and PCSK9 does not support this hypothesis, at least in our cohort.
Additionally, only in patients not in treatment with statins, we found a direct correlation between PCSK9 levels and total cholesterol and LDL. It is known that statin therapy increasing PCSK9 levels and reducing cholesterol levels disrupt their direct correlation 38 .
The effect of statins on EPCs is not well characterized, due to considerable heterogeneity in patient population, cardiovascular risk profile, statin regimens and markers used for EPCs characterization in the diverse studies. Several studies reported a significant increase in EPCs following statin regimen 29,39,40 while others reported a decrease 41 or no change 42 . Even within patients with chronic coronary artery disease (CAD), divergent results were observed, with significantly increased 39 or reduced 41 EPCs counts observed in statin-treated patients vs. matched controls in stable CAD, and an increase 40 or no alterations despite high-dose statins 42 after PCI. Inconsistency may be attributed to the diverse phenotypical characterization of EPC, or to dose and duration of statin treatment. Consistent with our findings, early EPCs were found to be significantly higher in patients not treated with statins, whereas late EPCs were significantly higher in statin-treated patients 43 . In addition, long-term statin therapy maintained late EPCs in circulation which may promote neovasculogenesis 44 .
Length of statin treatment seems to be an additional variable: in stable CAD patients, increased circulating EPC were observed as soon as 1 week since first statin intake, with plateauing after 3-4 week 39 ; in a prospective analysis, initiation of statin therapy significantly diminished the number of EPCs after 3 but not after 1 month 41 , suggesting EPC impairment with chronic statin use. Consistently, a short-term statin discontinuation increases EPCs in T2DM patients 45 . The observation that statin treatment is associated with reduced number of CD45 neg / CD34 bright and CD45 neg /CD34 bright /146 neg only among patients with T2DM, suggests that an interaction exists between diabetes and statins. Our study comprised chronic statin users (at least three months), although we cannot exclude suboptimal adherence. Finally, the effect of statins on EPC seems to be dose-dependent, since statin reloading in moderate statin users was associated with increased EPC levels 40,46 .
Interestingly, among the 12 patients with previous MI, PCSK9 and early EPC (CD45 neg /CD34 bright /146 neg ) are higher and lower, respectively, as compared to patients without previous MI. Again, these differences may be largely attributed to the higher prevalence of statin treatment among those with a history of previous MI (75% vs. 38%). However, our results are in line with previous observations of Laugsand reporting a 47% higher MI risk in patients in the highest quartile of circulating PCSK9 47 . Hill et al. reported a negative correlation between EPCs, measured by colony forming units, and Framingham risk score in 45 men with various degrees of cardiovascular risk. They also reported a positive correlation between CFU and brachial flow-mediated dilation, a measure of endothelial function 48 Consistently, a meta-analysis of 35 randomized controlled trials found that therapy with PCSK9 inhibitors was associated with a lower rate of MI (2.3% versus 3.6%; odds ratio [OR]: 0.72 [95% confidence interval (CI), 0.64-0.81]; p < 0.001) 49 . Likewise, early EPCs reduction in patients with a previous MI may be a biomarker of endothelial damage and reflect early EPC exhaustion, since early EPC can migrate to the foci of ischemia to promote the repair of the injured organs 50 . In our cohort, we did not observe any differences in CD45 neg /CD34 bright in patients with vs without a previous MI. However, this phenotype includes both EPCs and CECs, the latter characterized as CD45 neg /CD34 bright /CD146 pos23 . Indeed, after excluding the CECs, the remaining CD45 neg /CD34 bright /146 neg cells, reflecting early outgrowth EPCs 23 , are depleted in patients with a previous MI. Again, it is conceivable that an alteration of early/late EPC balance exists in relation to cardiovascular burden.
Fewer early EPC colonies and higher late EPC colonies were produced in patients with CAD than in control subjects without CAD 51 . The interpretation of the above-mentioned results may have been biased by lack of adjustment for ongoing treatment potentially affecting EPC, including statins.
We hypothesized that EPC impairment may be mediated, at least in part, by PCSK9. Notably, in patients with T2DM, considered as a whole and in those treated with statins, we found an inverse correlation between PCSK9 levels and both CD45 neg /CD34 bright and CD45 neg /CD34 bright /146 neg phenotypes. The link between PCSK9 and EPCs has been previously observed. In a study published by Chao et al. in patients with peripheral artery disease, high plasma levels of PCSK9 were associated with dysfunction in EPCs 52 . However, in the same study this observation was not paralleled by a reduction in the number of EPCs, at least using the CD34 pos /KDR pos phenotype. A direct effect of PCSK9 levels on Sca-1/VEGF-R2 EPCs has been observed in a mouse model in which the administration of anti-PCSK9 antibodies increased the number of circulating EPCs 9 , although this cells have be recently recharacterized as B2 lymphocyte upon deeper analysis 53 . Our results add further information to expand this hypothesis in humans. Statin treatment further increases PCSK9. Notably, EPCs impairment is prominent in patients with the highest levels of PCSK9, namely T2DM patients on statin treatment. Whether PCSK9′s influence on EPC is dependent on underlying metabolic abnormality is still uncharacterized. It is conceivable that PCSK9 may impair EPC by modifying insulin secretion and metabolic control, or vice versa that the metabolic derangement of diabetes may alter PCSK9 levels, in turn affecting EPC number. Whichever the underlying mechanisms, our findings of a putative selective effect of PCSK9 on EPC number only in patients with diabetes may provide a mechanistic explanation for the results of the ODYSSEY OUTCOMES trial, a randomised, double-blind, placebo-controlled trial performed in patients on high-intensity statin-treatment, showing that the anti-PCSK9 antibody alirocumab produced about twice the absolute reduction in CV events among patients with diabetes as in those without diabetes 54 . Within diabetes pharmacotherapy, a number of drugs besides statins are know to influence EPC counts 55 . ACE inhibitors have been shown to stimulate EPCs 56 Ang II potentiates VEGF-induced human EPCs proliferation 56 . Angiotensin II receptor antagonists increase the number of regenerative EPCs in patients with T2DM 57 . Among anti-hyperglycemic agents, rosiglitazone facilitates angiogenic progenitor cell differentiation toward endothelial lineage 58 . The dipeptidyl peptidase-4 (DPP-4) inhibitor sitagliptin increases circulating EPCs in T2DM patients with concurrent upregulation of SDF-1alpha 59 . Treatment with exenatide, but not with liraglutide, is able to increase the number of circulating EPCs, possibly through an antioxidative/ antiinflammatory effect 60  www.nature.com/scientificreports/ cardiovascular protection elicited by SGLT2is should be mediated by other mechanisms 61 . The actual contribution of EPC modulation to the pleiotropic cardioprotective effects of these medications remains unknown 62 . In our study, none of the ongoing medications apart from statins appeared to influence either EPC phenotype.
Limitations include the small sample to test the hypothesis of an interaction between diabetes and statins on early EPC number. However, this pilot study is hypothesis-generating for a previously unappreciated effect of statins, to be confirmed on adequately sized samples. Another limitation is cross-sectional nature of the study. An intervention study assessing the effect of statin treatment on plasma PCSK9 and early EPC number in patients with and without diabetes would have yielded definitive evidence.
Strengths include balance between the groups in terms of clinical characteristics, despite lack of randomization, and the method used to assess EPCs. Indeed, polychromatic flow cytometry used to enumerate and characterize EPCs has a standardized, high sensitive, flexible, and able to quickly analyse thousands of events and multiple parameters at the same time 24,25 .
In conclusion, we unravelled, in patients with T2DM in good glycemic control, already treated with the state-of-the-art strategies for CV prevention (100% on ASA, 55% on statins, 85% antihypertensives), an inverse correlation between circulating PCSK9 and early EPC number, with those on statins showing the highest PCSK9 levels paralleled by the most impaired EPC number. The relatively small sample size and the cross-sectional nature of this study do not allow us to confirm the cause-and-effect relationship between plasma PCSK9 and EPC number, nor a direct influence of statins on this biochemical and cell derangement. However, these findings highlight a piece of the pathophysiology underlying the "residual risk" of high-risk patients optimally treated with current preventive strategies, and suggest that early EPC impairment may be reverted by PCSK9 inhibitors, thus providing an interesting mechanistic explanation for the cardiovascular benefit of this class of drugs and a further indication for the patient with diabetes on top of statins.

Materials and methods
Patients recruitment. Forty-five T2DM patients (25 male, median age 68 years), with or without vascular disease, were enrolled at the Diabetes Clinic of Chieti University Hospital. Moreover, we studied 37 patients (22 male, median age 66 years) without T2DM, comparable for demographic, anthropometric and clinical characteristics, with particular reference to cardiovascular risk factors and concurrent treatments, referred to our Clinic by general practitioners. Each subject signed written informed consent to participate, and the Protocol was approved by the Ethics Committee of the University of Chieti (Prot.1129 18.07.2013).
T2DM diagnosis was made according to the ADA criteria (fasting plasma glucose ≥ 126 mg/dL or 2-h plasma glucose ≥ 200 mg/dL during OGTT or HbA1c ≥ 6.5 or a random plasma glucose ≥ 200 mg/dL) 63 . All the patients were in treatment with low-dose aspirin (100 mg/die) for cardiovascular prevention. Exclusion criteria were: uncontrolled hypertension, uncontrolled dyslipidemia, significant comorbidities such as kidney or liver disease, pregnancy or lactation, chronic inflammation, cigarette smoking; clinically significant cardiac and/or pulmonary insufficiency; history of malignant neoplasms (diagnosed and treated within the past 5 years); history of malabsorption; regular (daily) alcohol consumption; regular (i.e., more than 3 days per week) non-steroidal anti-inflammatory drug intake. Type 1 diabetes was excluded by islet autoantibodies evaluation (anti-glutamic acid decarboxylase, islet cell cytoplasmic, and IA-2 antibodies), in the presence of any of the following: family history of type 1 diabetes, age lower than 40 years, lean phenotype, early requirement for insulin therapy. No patient was diagnosed as having MODY (Maturity Onset Diabetes of the Young).
This study was performed under the Good Clinical Practice regulations (Good Clinical Practice for Trial on Medicinal Product-CPMP/European Commission-July 1990; Decreto Ministeriale 27.4.1992-Ministero della Sanità) and the Declaration of Helsinki (Hong Kong 1989). By signing the protocol, the participants in the study committed to adhere to local legal requirements. All methods were performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all participants and/or their legal guardians. PCSK9 levels. Since PCSK9 has a diurnal rhythm 64 , all samples were collected at 8 a.m. after an overnight fasting. Blood collected into EDTA containing vacuum tubes (vacutainer, Becton Dickinson) was centrifuged at 1200×g for 10 min at RT to separate plasma. Plasma was aliquoted in small volumes and frozen at − 80 °C. PCSK9 levels were measured with commercial enzyme-linked immunosorbent assays (ELISA) kit (#DPC900, R&D) according to the Manufacturer's instructions. Within-assay and between-assay coefficient of variations were below 7%.
Circulating endothelial cells. The analysis of circulating endothelial cells was carried out by plychromatic flow cytometry on peripheral blood samples as already reported 25 . Briefly, 20 × 10 6 leukocytes/sample underwent an erythrocyte-lysis step (45 mL of Pharm Lyse solution-BD Biosciences-for 15 min at RT, under agitation) and then centrifuged (400g, 10 min, room temperature). Once washed (2 mL of Stain Buffer, BD Biosciences), samples were centrifuged and stained using 1 µM Syto16 (Thermo Fisher Scientifc, Eisai, Medipost-US) and a lyophilized cocktail of reagents (BD Biosciences; cat. 623920) 25 . Samples were incubated in the dark for 30 min at 4 °C, washed, centrifuged, re-suspended in 1.5 mL of FACSFlow (BD Biosciences), and finally 2-4 × 10 6 events/ sample with lymph-monocyte were acquired by flow cytometry (FACSCanto, FACSAria, BD Biosciences). A threshold combination was used on Forward Scatter (FSC) and Fluorescein isothiocyanate (FITC-Syto16) channels. Compensations were calculated using CompBeads (BD Biosciences) and single stained fuorescent cells. Carryover between samples was prevented by appropriate instrument cleaning at the end of each sample acquisition. CD45 neg /CD34 bright , CD45 neg /CD34 bright /CD146 neg , CD45 neg /CD34 bright /CD146 pos and CD45 neg /CD34 bright / CD309 pos phenotypes were analysed (Fig. 5). Briefly, events displaying the typical lymph-monocyte morphology were first selected in a forward scatter (FSC) versus side scatter (SSC) plot (Fig. 5a) www.nature.com/scientificreports/ excluded on the basis of their positivity to 7-AAD (Fig. 5b) and nucleated events (DNApos) were gated (Fig. 5c). The aforementioned three gates were intersected and cells resulting from this logical combination, characterized by lymph-monocyte morphological features, alive and nucleated, were then analysed for their phenotypes. Only non-hematopoietic CD45 neg cells were further analysed. The whole CD34positive cell compartment displayed different levels of CD34 surface expression and two subpopulations were identified: CD34 positive cells and a CD34 bright (Fig. 5d). CD45 neg /CD34 bright cell population was then analysed for CD146 and CD309 expression, on CD146/CD34 (Fig. 5e) and CD309/CD34 dot plots (Fig. 5f), respectively, and compared with the respective Figure 5. Events displaying the typical lympho-monocyte morphology were first selected in a forward scatter (FSC) versus side scatter (SSC) dot-plot (a). Dead cells were excluded on the basis of their positivity to 7-AAD (b) and nucleated events (Syto16 + , c) were gated. Regions identified in (a-c) were logically intersected and cells resulting from this combination (lympho-monocyte morphological features, alive and nucleated), were then analysed for their phenotypes. Two subpopulations were identified on a CD45/CD34 dot-plot: CD34 positive cells (hematopoietic stem cells) and a CD34 bright cells (d). The CD45 neg /CD34 bright cell population was then analysed for CD146 and CD309 expression, on CD146/CD34 and CD309/CD34 dot plots, respectively (e,f), and compared with the respective control tube dot-plots, containing the isotype control of the anti-CD146 and anti-CD309 in combination with all the remaining reagents (g,h). www.nature.com/scientificreports/ control tube dot plots, containing the isotype control of the anti-CD146 and anti-CD309 in combination with all the remaining reagents (Fig. 5g,h). The flow cytometry method here described has been previously standardized and published 24,25 . All antibodies and reagents were titrated under assay conditions. The antibody specificity and the gating strategy were defined using fluorescence minus one controls (FMO), as recommended 65 . Subpopulation numbers and abundance were calculated by a dual-platform counting method using the lymphocyte subset as a reference population as reported 24 .
Statistical analysis. The primary outcome measure is the number of CD45 neg /CD34 bright cells. Our sample size has power = 80% (α = 0.05) to detect a difference greater than 60% of the standard deviation in the primary outcome between patients with diabetes and patients without diabetes, and power = 80% (α = 0.05) to detect an interaction effect between diabetes status and statins use with an effect size > 0.30. Comparisons of variables between groups were performed by X 2 tests or Mann-Whitney U tests. Spearman rank correlation test was used to assess relationships among variables. Stepwise multiple linear regression analysis was performed to assess variables independently associated with EPCs. Covariates included in the multiple regression models were selected on the basis of their significance on univariate analysis and their clinical relevance to the outcome of interest. They included diabetes, statin treatment, PCSK9 levels, age and gender, BMI.
Only 2-tailed probabilities were used for testing statistical significance, and p < 0.05 was considered statistically significant. All calculations were carried out using SPSS (SPSS, Chicago, IL, USA). Sample size and power analysis have been conducted by using of GPower 66 .