Plasma levels of soluble VEGF receptor isoforms, circulating pterins and VEGF system SNPs as prognostic biomarkers in patients with acute coronary syndromes

The Coronary Disease Cohort Study (CDCS) recruited 2067 patients after admission to Christchurch or Auckland City Hospitals with a diagnosis of ACS, from July 2002 to January 2009. Inclusion criteria included ischemic discomfort plus one or more of ECG changes (ST-segment depression or elevation of ≥0.5 mm, T-wave inversion of ≥3 mm in ≥3 leads, or left bundle branch block), elevated levels of cardiac markers, a history of coronary disease, age of ≥65 years, and a history of diabetes or vascular disease [24]. Patients with serious co-morbidity that reduced their life expectancy to < 3 years (e.g. end-stage renal failure, cancer) were excluded from the study. Recruitment included a wide spectrum of age, both sexes and patients with established risk factors for CHD such as hypertension and diabetes. Plasma was collected at a baseline clinic, a median of 32 days following index admission for ACS. Demographic and clinical data was collected at baseline including blood pressure, ECG, echocardiography, family and personal medical history, height, weight, and medication prescribed. Plasma samples were assayed for natriuretic peptides and other analytes. Patients were followed for a median of 5.04 years. Patients attended follow-up clinics 3–5 months and 12–14 months post-onset of ACS, and participants completed questionnaires at two and three years post-discharge. Ethnicity was self-declared and categorized as Maori/Pacific Islander, European, Other or Unknown. Standardized transthoracic echocardiography was performed at baseline and at each follow-up clinic using a GE Vivid 3 (GE Medical Systems) ultrasound system at Christchurch Hospital and Philips ATL HDI 5000 and ie33 (Philips Healthcare, Bothell, WA) ultrasound systems at the University of Auckland as described previously [21, 23]. The study was approved by the New Zealand (NZ) Multi-Region Ethics Committee and all participating patients provided written, informed consent.

Clinical events were determined from recruitment questionnaires, planned follow-up clinic visits, consultation of patient notes, the NZ Ministry of Health and hospital Patient Management System databases, linked through the National Health Index (NHI) number for each patient. Survival times were calculated from the date of index admission. The investigation conforms to the principles outlined in the Declaration of Helsinki and Title 45, U.S. Code of Federal Regulations, Part 46.

In the Christchurch subgroup of the cohort collateral vessels were graded according to the Rentrop classification: 0: no filling of any collateral vessels, 1: filling of side branches of the artery to be perfused by collateral vessels without visualization of the epicardial segment; 2: partial filling of the epicardial artery by collateral vessels; and 3: complete filling of the epicardial artery by collateral vessels [29], while blinded to other clinical data. Coronary artery anatomy, severity of coronary stenoses and the myocardium at risk were assessed according to the Brandt score [30] in the same subgroup.

Plasma samples were collected and stored at − 80°C. sFlt-1 and sKDR were analysed using chemiluminescent quantitative sandwich enzyme immunoassays (R&D Systems, Minneapolis) detection limits, sFlt-1 3.5 pg/mL and sKDR 4.6 pg/mL. Samples from patients recruited earliest at Christchurch Hospital, for which sufficient plasma was available, were chosen for assay of sFlt-1, sKDR and pterins to maximise length of follow-up and therefore accumulated numbers of clinical events and the strength of survival analyses. Circulating levels of BNP and N-terminal proBNP (NT-proBNP) were assayed as previously described [31]. HPLC measurement of pterins was performed using a Shimadzu, Kyoto, Japan, Sil-20A HPLC with auto-sampler, RF-20Axls fluorescence detector. A 10 μL plasma sample was injected onto a Luna 5 μm SCX 100 Å, 250 × 4.6 mm column with a mobile phase of 100% 20 mM ammonium phosphate (pH 2.5) pumped at 1 mL/min as previously described [32]. Neopterin was detected by fluorescence at wavelength emission of 438 nm and excitation of 353 nm, whereas 7,8 NP was detected following oxidation to neopterin by iodide. All analysis of plasma analytes was conducted in duplicate.

Extraction of genomic DNA for genotyping was performed as described previously [23]. DNA samples were genotyped for the rs2071559 (T-604C), rs2305948 (G1192A), rs1870377 (T1719A) polymorphisms in the VEGFR-2 receptor gene (encoding KDR) and rs748252 (C8764T) and rs9513070 (A189427G) in the VEGFR-1 receptor gene (encoding Flt-1). The SNPs from VEGFR-1 were chosen with reference to dbSNP and International HapMap Project data, as at the time of initiating this study there were no prior reports of genetic association studies of polymorphisms in VEGFR-1 and CVD. Genotyping was performed by real-time PCR, using allele-specific TaqMan genotyping probes (Applied Biosystems) in 5 μL reaction volumes in 384-well plates, including 1× Roche LightCycler 480 Probes Master mix and 100 ng of genomic DNA in a Roche LC480 (Roche Diagnostics, Auckland).

Univariate analyses to test for associations between SNP genotype and demographic, analyte levels and echocardiographic measurements were performed using χ² and ANOVA tests. Skewed data were log-transformed before analysis and geometric means with 95% confidence intervals reported and adjusted for age, and the time between index admission and baseline sampling. The survival of stratified groups was compared using Kaplan-Meier analysis and the log-rank test. Independent associations between genotype and survival were tested using Cox proportional hazards multivariate analysis including the following established predictors; age, gender, previous MI, Type 2 diabetes, baseline creatinine, physical activity and NTproBNP levels as previously justified [23, 31]. Multivariate linear regression models were based on covariates showing univariate association with sFlt-1. The study had power to detect a hazard ratio (HR) of > 1.7 as statistically significant (two tailed α < 0.05, 90% power) in the CDCS cohort for analysis of those patients assayed for sFlt-1. Levels of sFlt-1 at baseline were assayed in baseline plasma samples from 513 patients selected from the CDCS cohort. Samples from patients with a baseline plasma sample available at the earliest recruitment dates were chosen to maximise number of events on follow up for inclusion in survival analyses (recruited between July 2002 and August 2007). Levels of KDR were assayed in baseline plasma samples from 211 patients, in whom sFlt-1 levels were available (recruited between July 2002 and February 2004). Levels of pterins were assayed in 144 baseline plasma samples from patients in whom sFlt-1 levels were available (recruited between September 2006 and May 2007). In the genetic association part of the study there was 80% power to detect HRs > 1.4 with rare homozygote groups having a frequency of 20% of total cohort. An additive genetic model was used unless stated otherwise. All analyses were performed using SPSS version 22 (IBM, Armonk, USA). Statistical significance was set at the 5% level (p < 0.05).

Baseline features of the CDCS cohort are summarized in Table 1. Genotypes were obtained from DNA samples from patients from the cohort for the VEGFR-1 gene SNPs (C8764T rs748252 [n = 2027], A189427G rs9513070 [n = 2048]) and for the VEGFR-2 gene SNPs (T-604C rs2071559 [n = 2050], G1192A rs2305948 [n = 2066], T1719A rs1870377 [n = 2042]) as shown in Table 2. The genotype distributions conformed to the Hardy-Weinberg equilibrium (p ≥ 0.998).

Levels of sFlt-1 at baseline were assayed in baseline plasma samples from 513 patients selected from the CDCS cohort. Baseline characteristics of this group are summarised and compared to the remainder of the cohort in Additional file 1: Table S1. Patients with a baseline plasma sample available from the earliest recruitment dates (between July 2002 and August 2007) were chosen to maximise the number of events during follow up for inclusion in survival analyses. Baseline levels of sFlt-1 had a geometric mean of 105 pg/mL (95% CI 102–110 pg/mL, CV = 52.8%) and were weakly correlated with age (n = 513, r = 0.167, p < 0.001), troponin I (n = 509, r = 0.267, p < 0.001), NTproBNP (n = 513, r = 0.216, p < 0.001) and BNP levels (n = 513, r = 0.181, p = 0.003) at baseline. A weak inverse correlation of sFlt-1 with LVEF (n = 498, r = − 0.144, p < 0.001) was observed. Levels of sFlt-1 differed significantly (p = 0.035) between patients who were current drinkers of alcohol (n = 306, mean = 106 pg/mL) and non-drinkers (n = 143, mean = 101 pg/mL). Using a multivariate linear regression model, age, troponin I, and time to plasma sampling (all p < 0.001) were independently associated with sFlt-1 levels.

Levels of KDR, assayed in baseline plasma samples from 211 patients, in whom sFlt-1 levels were available, had a geometric mean of 10,500 pg/mL (95% CI 10200–10,900 pg/mL, CV = 24.1%) at baseline, with an inverse correlation with sFlt-1 (n = 185, r = − 0.184, p = 0.027) and age (n = 211, r = − 0.153, p < 0.037). KDR was significantly higher (p = 0.025) in drinkers (Current Drinkers: n = 135, mean = 11,200 ± 221 pg/mL; Ex-/Non-drinkers: n = 76, mean 10,300 ± 295 pg/mL). As preliminary analysis provided no indication that KDR was associated with any parameter of clinical interest, the decision was made to assay no further samples for KDR.

Investigation of the association of VEGFR receptor gene SNPs with baseline characteristics revealed that the genotype of theVEGFR-2 SNP rs1870377 A allele was significantly associated with higher levels of sFlt-1 and lower levels of sKDR at baseline (Table 3, Fig. 1). These associations did not appear to be influenced by any differences in allele frequencies between European and non-European ethnic groups (Additional file 1: Table S2). There were no significant associations between the other SNPs assayed and either sFlt-1 or KDR levels.

Levels of pterins (neopterin, n = 144; 7,8 NP, n = 135; total pterins, n = 138) were assayed in baseline plasma samples from among patients in whom sFlt-1 levels were available. Baseline characteristics of this group are summarised and compared to the remainder of the cohort in Additional file 1: Table S3. Levels of total pterins ranged between 16.58 nM and 624.6 nM, mean of 94.4 ± 9.1 nM and a standard deviation of 107.08 nM. Total pterins trended higher in females (78.5 [62.7–98.4] nM) than males (60.1 [50.9–70.8] nM)(p = 0.058); and lower in current drinkers of alcohol (n = 138, drinkers 56.8 [49.1–65.7], ex−/non-drinkers 78.0 [62.0–98.1] nM, p = 0.020), but the latter association was lost after adjustment for patient age and gender. 7,8 NP levels had a geometric mean of 37.7 (31.4–45.3) nM, and neopterin 17.8 (15.9–20.0) nM. Neopterin was weakly positively correlated with NTproBNP (n = 143, r = − 0.167, p = 0.034).

Baseline levels of plasma sFlt-1 were associated with all-cause mortality (p < 0.001) (Fig. 2). In multivariate survival analysis, sFlt-1 predicted mortality independent of age, gender, NT-proBNP, creatinine, troponin I, physical activity score, alcohol consumption, history of previous MI, diabetes, ethnicity, ACS diagnosis and time to sampling (p = 0.023) (Table 4). This was also true for the endpoint of deaths attributable to cardiovascular (CV) causes (CV deaths n = 87, above median sFlt-1 mortality = 21.6%, below median sFlt-1 mortality = 12.6% p = 0.029). ROC analysis is also presented in Additional file 2: Figure S1 comparing sFlt-1 and NTproBNP as predictors of mortality at 5 years of follow-up. The VEGFR2 SNP rs1870377 was associated with reduced risk of HF readmission (AA v TA/TT, n = 2033, events = 379, hazard ratio = 0.74, 95% CI 0.58–0.95, p = 0.019, adjusted for age, time to sampling, gender, and ethnicity).

All-cause mortality (n = 138) was greater for those with above-median compared with below-median total pterin levels (Fig. 3.; 33.3% versus 10.1%, respectively; p = 0.005) and greater for patients with above-median 7,8 NP (28.2% versus 15.3%, p = 0.026). In a Cox’s proportional hazards model total pterin levels were predictive of death independent of age, time to sampling, sFlt-1 and NT-proBNP levels (Table 5).

Levels of sFlt-1 at baseline were not significantly correlated with Brandt score (sFlt-1 n=337, r = 0.078 p = 0.156; sKDR n = 124, r = -0.201, p = 0.026). BNP and NT-proBNP levels at baseline were associated with Brandt Score (BNP n = 1068, r = 0.206, p < 0.001; NTproBNP n = 1068, r = 0.222, p < 0.001). The VEGFR-1 SNP rs748252 TT genotype group trended towards higher age-adjusted Brandt score than other genotype groups (Table 3) and this appeared to be consistent across ethnic groups (Additional file 1: Table S2). Rentrop classification for extent of angiographically apparent coronary collateral vessels was not significantly associated with levels of soluble VEGF receptor, pterins or any of the five SNPs genotyped for this study.