Circulating Hematopoietic Stem/Progenitor Cells are Associated with Coronary Stenoses in Patients with Coronary Heart Disease

Inflammatory cells in atherosclerotic plaque exclusively originate from hematopoietic stem/progenitor cells (HSPCs). In this study, we investigated whether circulating HSPCs frequency related to coronary stenosis in patients with coronary heart disease (CHD). Coronary angiography was performed in 468 participants who were recruited at Cardiology Centre in LuHe Hospital from March 2016 to May 2017. Among these subjects, 344 underwent echocardiography. Mononuclear cells isolated from peripheral blood were stained with an antibody cocktail containing anti-human CD34, anti-human lineage, anti-human CD38, and anti-human CD45RA. Lineage−CD38−CD45RAdimCD34+HSPCs were quantified by flow cytometry. CHD was defined as coronary stenosis ≥50% and the extent of CHD was further categorised by coronary stenosis ≥70%. A p < 0.0031 was regarded statistically significant by the Bonferroni correction. Circulating HSPCs frequency was 1.8-fold higher in CHD patients than non-CHD participants (p = 0.047). Multivariate-adjusted logistic analysis demonstrated that HSPCs was the only marker that was associated with the odds ratio of having mild vs. severe coronary stenosis (2.08 (95% CI, 1.35–3.21), p = 0.0009). Left ventricular ejection fraction was inversely correlated with HSPCs frequency and CRP in CHD patients (p < 0.05 for both). In conclusion, HSPCs frequency in circulation is intimately related to coronary stenoses in CHD patients.

RoC. CRP is a general inflammatory marker related to inflamed atherosclerotic plaque. The area under the curve (AUC) for CRP in the discrimination between stenosis <70% and ≥70% was 0.64 (95% CI, 0.59-0.69) in all subjects. Using CRP as a reference (Fig. 4), the AUC values of HSPCs, white blood cell count and LDL-c were Cardiac function by echocardiography. Finally, we analysed the relationship between these biomarkers and cardiac function. Among all subjects, 344 were examined by echocardiography. Left ventricular ejection fraction, end-systolic diameter and end-diastolic diameter did not differ between non-CHD subjects and CHD patients (Table 4).  Table 3. Multivariable-adjusted associations of coronary occlusion status with biomarkers. All analyses were adjusted for covariables, including age, sex, mean arterial pressure, heart rate, plasma glucose, serum creatinine, plasma glucose, γ-glutamyltransferase, smoking, alcohol intake, history of hypertension (1 or 0), history of diabetes (1 or 0), use of diuretics, inhibitors of the renin-angiotensin system (β-blockers, angiotensin-converting-enzyme inhibitors and angiotensin type-1 receptor blockers), vasodilators (calciumchannel blockers and α-blockers), metformin and statins. With the exceptions of LDL-c and HDL-c, all other biomarkers were additionally adjusted with the total-to-HDL cholesterol ratio. A value of P < 0.0031 was considered significant after the Bonferroni correction. Significance of the associations: † P < 0.01; ‡ P < 0.001; § P ≤ 0.0001.

Figure 3.
We divided the study population of 468 participants using coronary artery stenosis and circulating biomarkers. V-plots were generated for the PLS-DA-derived VIP scores versus the centred and rescaled correlation coefficients. HSPCs, neutrophils, lymphocytes and monocytes are expressed as percentages among white blood cells. HSPCs, CRP, serum triglyceride, fibrinogen and D-dimer were logarithmically transformed in the analysis. VIP represents the importance of each marker in the construction of the PLS factors. The correlation coefficients reflect the association of significant coronary stenosis (≥70%) with each marker. The markers with VIP score ≥1 and VIP score <1 are indicated in red and cyan blue, respectively. Multivariate-adjusted linear regression analysis demonstrated that one-SD increases of circulating HSPCs frequency, serum CRP level, white blood cell count and neutrophil count were associated with 2.33% (95% CI, −4.61% to −0.06%; p = 0.045), 3.04% (95% CI, −5.65% to −0.44%; p = 0.023), 1.92% (95% CI, −3.28% to −0.56%; p = 0.007) and 2.25% (95% CI, −3.58% to −0.09%; p = 0.01) declines of left ventricular ejection fraction in CHD patients, respectively. Furthermore, one-SD increases of circulating HSPCs frequency, white blood cell count and neutrophil count were associated with 2.42 mm (95% CI, 0.03-2.93 mm; p = 0.045), 1.12 mm (0.23-2.01 mm; p = 0.013) and 1.23 mm (95% CI, 0.38-2.09 mm; p = 0.004) increases of end-systolic diameter in the left ventricle in CHD patients, respectively. By contrast, CRP was not associated with end-systolic diameter in the left ventricle (p = 0.26). Moreover, LDL-c was associated with neither ejection fraction nor end-systolic diameter (p ≥ 0.19). Bonferroni correction of 16 biomarkers did not identify any significant association between any biomarker and cardiac function studied.
Besides the biomarkers mentioned above, conventional cardiac injury markers including serum creatinine, brain natriuretic peptide (NT-pro-BNP), lactic dehydrogenase (LDH), α-hydroxybutyrate dehydrogenase (HBDH) and cardiac troponin I (TnI) were included individually in the analysis 18,19 . Serum creatinine and serum lactic dehydrogenase were negatively associated with ejection fraction and positively associated with end-systolic diameter of the left ventricle in CHD patients. Nevertheless, none of the biomarkers was associated with end-dilation diameter of the left ventricle in the patients after adjusting for covariables (p ≥ 0.12). Figure 5 shows the −log10(P) probability plot of the multivariable-adjusted association of various markers with ejection fraction (continuous) or end-systolic diameter (continuous) of the left ventricle in CHD patients.

Discussion
The key findings of this study can be summarised as follows. (1) HSPCs frequency in the peripheral blood was 1.8-fold higher in CHD patients than in non-CHD participants. (2) By multivariate-adjusted logistic analysis, per one-SD increase for each of the following variables, the odds ratio of having CHD vs. not having it was 1.24 for HSPCs (p = 0.017), 1.48 for CRP (p < 0.0001), 1.22 for white blood cell count (p = 0.043) and 1.33 for LDL-c   HSPCs remained significantly associated with the odd ratio of having CHD vs. no CHD and mild vs. severe coronary stenosis, respectively. Atherosclerosis is the underlying pathogenesis of CHD and contributes substantially to cardiac dysfunction and disease progression. Inflammatory cells and pro-inflammatory cytokine and chemokine production from inflammatory cells are hallmarks of atherosclerotic plaque [10][11][12] . The evolution of early/stable plaque to advanced/ vulnerable plaque results from a phenotypic transition from controlled to uncontrolled inflammation 3 . HSPCs give rise to all types of blood cells and are the source of inflammatory cells. They could thus be an important biomarker and intervention target for atherosclerosis.
This study on CHD patients was prompted in part by previous observations that hypercholesterolemia induced by a high-fat diet enhanced the proliferation of HSPCs and their subsequent differentiation to myeloid cells, including neutrophils, whereas the control of HSPCs proliferation and differentiation reversed hypercholesterolemia-induced leukocytosis and reduced atherosclerosis plaque in mice 13,15 . In this study, we reported that circulating HSPCs frequency was tightly associated with coronary stenoses in CHD patients. Compared with CRP, the conventional biomarker of coronary stenosis 20 , HSPCs is more sensitive in assessing the progression from mild stenosis (<70%) to medium stenosis (≥70%) and reflecting left ventricular end-systolic diameter in CHD patients. We further applied a correction for multiple testing for all the biomarkers studied 21 . By Bonferroni correction, HSPCs remained the only marker that was associated with odds ratio of having mild vs. severe coronary stenosis in CHD patients.
GM-CSF and SDF-1α trigger HSPCs mobilisation from bone marrow to peripheral blood [22][23][24] . Moreover, a positive correlation between serum levels of SDF-1αand coronary artery occlusion was reported in CHD patients 25 . In parallel with this, when atheroma cell suspension was obtained from carotid artery, GM-CSF levels were higher in patients with symptomatic plaque than in those with asymptomatic plaque 26 . In this study, we performed a detailed dissection of the relationships between GM-CSF, SDF-1α, HSPCs and coronary stenosis. Circulating HSPCs frequency was found to be positively associated with serum levels of GM-CSF (r = 0.12, p = 0.011) but not with SDF-1α (r = 0.02, p = 0.61). Unlike HSPCs, neither GM-CSF nor SDF-1αwas associated All analyses were adjusted for covariables, including age, sex, mean arterial pressure, heart rate, plasma glucose, serum creatinine, plasma glucose, γ-glutamyltransferase, smoking, alcohol intake, history of hypertension (1 or 0), history of diabetes (1 or 0), use of diuretics, inhibitors of the renin-angiotensin system (β-blockers, angiotensin-converting-enzyme inhibitors and angiotensin type1 receptor blockers), vasodilators (calcium-channel blockers and α-blockers), metformin and statins. With the exceptions of LDL-c and HDL-c, all other biomarkers were additionally adjusted with total-to-HDL-c ratio. Biomarkers in Fig. 3 as well as creatinine kinase (CK), brain natriuretic peptide (NT-pro-BNP), lactic dehydrogenase (LDH), α-hydroxybutyrate dehydrogenase (HBDH) and cardiac troponin I (TnI) were analysed and those with p < 0.05 are named here. with the severity of coronary stenosis in CHD patients. These findings indicate that GM-CSF partially contributed to the increase in HSPCs in the blood. GM-CSF and SDF-1α appear to have weak effects on coronary occlusion. Decreasing LDL-c and elevating HDL-c have been therapeutic targets for attenuating atherosclerosis and treating ischemic heart disease for decades. Recently, meta-analysis of multiple statin trials elegantly demonstrated the beneficial effects of statins in lowering LDL-c and reducing cardiovascular risks 27,28 . Proprotein convertase subtilisin-kexin type 9 (PCSK9) inhibitors are another drug family that can reduce LDL-c. Inhibition of PCSK9 abrogates LDL-receptor degradation and thus enhances LDL uptake for clearance 29 . PCSK9 inhibitors have also been shown to lead to a 60% reduction in LDL-c without any indication of side effects 30,31 . In the current study, we illustrated that LDL-c was increased in CHD patients and positively associated with atherosclerosis-based coronary stenosis even after adjusting for statins and other potential covariables. Our data are in line with other studies supporting the concept that LDL-c is an appropriate target for cardiovascular disease treatment. Paradoxically, a Mendelian randomisation study involving 111,194 individuals from two prospective general populations reported that a low LDL-c level due to genetic variation in PCSK9 and HMGCR is associated with a high risk of neurocognitive diseases including Alzheimer's disease and Parkinson's disease 32 . Therefore, there is a need for further investigation of the optimal therapeutic range of LDL-c for treating cardiovascular disease and protecting against adverse neurocognitive events.
Atherothrombosis is another major cause of coronary occlusion. D-dimer is a product of the degradation of cross-linked fibrin and is thus commonly used as a marker to predict plaque severity based on the Gensini score 33 . In the current study, despite the extensive administration of anti-coagulation drugs, serum D-dimer and fibrinogen did not differ between subjects with and without CHD. In addition, none of these variables had any significant association with the severity of coronary stenosis in CHD patients.
The present study should be interpreted in consideration of its limitations. First, this is a cross-sectional study. Whether HSPCs could predict the outcome of adverse cardiovascular events or the incidence of CHD remains to be proven in longitudinal studies. Recently, Hammadah measured CD34 + cells in CHD patients and found that their low levels in circulation independently predict adverse cardiovascular disease outcomes 34 . A follow-up study of these patients would provide a better understanding of the role of HSPCs in cardiovascular disease outcomes. Second, we did not categorise monocytes into M1 and M2 subtypes or other subgroups. Third, we could not rule out the possibility that the increased HSPCs frequency in peripheral blood was derived from increased HSPCs proliferation or mobilisation from bone marrow into circulation. Fourth, we did not have data on the body mass of the subjects because, when most of them arrived, it was an emergency situation. As we included extensive covariables for adjustment, the impact of body mass index on the analysis should have been limited.
In conclusion, we identified HSPCs as an important marker to assess atherosclerosis-induced coronary stenosis. The level of circulating HSPCs increases in association with the occurrence of CHD and is significantly associated with the progression of mild coronary occlusion to a severe state. The increase of HSPCs in CHD patients has an adverse impact on ejection fraction and is positively associated with end-systolic diameter in the left ventricle. Further studies are required to testify whether HSPCs could be as a novel intervention target for CHD patients.

subjects.
All study procedures complied with the Declaration of Helsinki regarding investigations of human subjects. They received ethical approval from the institutional review boards of both Lu He Hospital and Capital Medical University. All participants provided written informed consent.
From March 2016 to May 2017, 556 patients were enrolled in this study. Their blood pressure was recorded as the mean of three readings and the mean arterial pressure was determined as diastolic pressure plus one-third of pulse pressure. Hypertension was defined as blood pressure of at least 140 mmHg systolic or 90 mmHg diastolic or the use of antihypertensive drugs. Diabetes was defined as plasma glucose of at least 7.0 mmol/L while fasting or of 11.0 mmol/L or more 2 h after an orally administered glucose load of 75 g. Additional characteristics including age, medical history, smoking and drinking habits, and intake of medications were also recorded.
We excluded 88 patients because of no coronary angiography having been performed (n = 40), lack of FACS-based HSPC data (n = 29), missing basic information (n = 18) or values exceeding the mean by three standard deviations (SDs) or more (n = 1). Thus, in total, 468 participants were statistically analysed. Among these CAD patients, 344 were examined by echocardiography. A flowchart of the study is presented in Fig. 1. echocardiography. Echocardiography was performed prior to coronary artery angiography. A single observer performed the echocardiography using a Philips iE33 (Philips, Amsterdam, Netherlands) device and analysed the digitally stored images, averaging three heart cycles, using a workstation running Hina Uses Workstation (version 2.0; Hina, China). Analyses of the echocardiography images were performed by an investigator who was blinded to the identity of the specific groups. Briefly, diastolic left ventricular (LV) function included the peak early (E) and late (A) diastolic velocities and flow duration from the transmitral blood flow Doppler signal, together with left ventricular ejection fraction (LVEF), left ventricular end-diastolic diameter, left ventricular end-systolic diameter, interventricular septal thickness, ventricular septal amplitude, and left ventricular volume including end-systolic volume (ESV) and end-diastolic volume (EDV). LVEF was calculated as follows: experienced cardiologists who were blinded to the group randomisation. In cases of disagreement, consensus was reached by further joint reading.
Biochemical measurement. After overnight fasting, venous blood samples were drawn for measurement of the total and differential white blood cell counts, serum total cholesterol, HDL cholesterol, triglycerides, creatinine, D-dimer, C-reactive protein (CRP), plasma glucose and γ-glutamyltranspeptidase. Estimated glomerular filtration rate (eGFR) was derived from serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation 35 . LDL cholesterol (LDL-c) was computed from serum total and HDL cholesterol (HDL-c) and serum triglycerides by the Friedewald equation 36 . Participants were classified as having dyslipidaemia if at least one of the following criteria was met: total cholesterol higher than 4.9 mmol/L, LDL cholesterol exceeding 3 mmol/L, triglycerides higher than 1.7 mmol/L or HDL cholesterol less than 1.2 mmol/L in women and 1 mmol/L in men 37 . Serum creatinine, high-sensitive C-reactive protein (CRP), cardiac creatinine, lactic dehydrogenase (LDH), α-hydroxybutyrate dehydrogenase (HBDH) and cardiac troponin I (TnI) were measured in the central laboratory at Lu He Hospital.
Colony-forming assays. Mononuclear  Flow cytometry. HSPCs constitute a tiny population among white blood cells. Therefore, to achieve reliable analysis, mononuclear cells were isolated from peripheral blood by Ficoll-based density gradient centrifugation, as described previously 16 . After isolation, cells were stained with an antibody cocktail containing anti-human lineage cocktail APC (BD), anti-human CD38 APC (eBioscience), anti-human CD34PE (eBioscience) and anti-human CD45RA PerCP-Cy5.5 (eBioscience). Dead cells were excluded by their size forward scatter (FSC) and side scatter (SSC) and staining of 7-AAD. Cells stained with isotype antibodies or unstained cells were used as a negative control for FACS analysis. At least 100,000 events were acquired for FACS analysis. Data were acquired using a FACS Canto (BD) and analysed by FlowJo. HSPCs, defined as Lin − CD34 + CD38 − CD45RA dim cells, were counted. The frequency of HSPCs among mononuclear cells was calculated.
enzyme-linked immunosorbent assay. The concentrations of serum N-terminal pro-B-type natriuretic peptide (NT-pro-BNP), stromal-derived factor 1α (SDF-1α), granulocyte-macrophage colony stimulating factor (GM-CSF), apolipoprotein A-1 (apoA-1), apolipoprotein B (apoB) and D-dimer were measured by ELISA, in accordance with the manufacturer's instructions (MLBio, Shanghai, China). Intra-assay coefficients of variation for NT-pro-BNP, SDF-1a, GM-CSF, apoA-1 and apoB were 0.25%, 0.29%, 0.48%, 0.27% and 0.59%, while inter-assay ones were 1.99%, 4.93%, 4.81%, 3.13% and 3.73%, respectively. statistical analysis. For database management and statistical analysis, we used the SAS system, version 9.4 (SAS Institute Inc., Cary, NC, USA). We normalised the distributions of γ-glutamyltransferase, HSPC, serum creatinine, serum triglyceride, fibrinogen and D-dimer by logarithmic transformation. Means were compared using t-test or ANOVA and proportions by Fisher's exact test. Pearson's correlation was used for single association analysis. For t-test, ANOVA, proportions by Fisher's exact test and Pearson's correlation, a p-value < 0.05 was considered statistically significant. Subjects with coronary artery stenosis ≥50% were diagnosed as having CHD. The extent of CHD was further categorised into mild, intermediate and severe groups using the following criteria: all stenosis <70%, one coronary artery stenosis ≥70% or more than one coronary artery stenosis ≥70%, respectively. Thereafter, multivariate-adjusted logistic analysis was performed to evaluate the coronary occlusion severity with each marker independently. The importance of each marker with its association to coronary stenoses(≥70% vs. <70%)) was assessed from the Variable Importance in Projection (VIP) scores of Wold in all study subjects following construction of the partial least square (PLS) factors. We further evaluated the potential of CRP, circulating HSPCs, white blood cell count and LDL-c to discriminate between CHD patients and controls with coronary stenosis ≥70% vs. <70% by constructing receiver operating characteristic (ROC) curves and by calculating the area under them (AUC). The 95% confidence interval (95% CI) of the AUC was calculated by the DeLong method.
Multivariate-adjusted linear regression analysis was performed for the regression of LV function against each marker in CHD patients. While accounting for covariables, we performed regression of the indexes of diastolic LV function on the markers and constructed -log10 probability plots. Bonferroni correction was applied for variants (i.e. biomarkers) that were associated with coronary stenosis or cardiac function index. A P value < 0.0031 (0.05/16) was considered statistically significant. In this study, all analyses included adjustments for covariables, including age, sex, mean arterial pressure, heart rate, plasma glucose, serum creatinine, plasma glucose, γ-glutamyltransferase, smoking, alcohol intake, history of hypertension (1 or 0), history of diabetes (1 or 0), use of diuretics, inhibitors of the renin-angiotensin system (β-blockers, angiotensin-converting-enzyme inhibitors and angiotensin type-1 receptor blockers), vasodilators (calcium-channel blockers and α-blockers), metformin and statins. With the exceptions of LDL-c and HDL-c, all other biomarkers were additionally adjusted for using the total-to-HDL-cholesterol ratio.