Contribution of plasma levels of VEGF-A and angiopoietin-2 in addition to a genetic variant in KCNAB1 to predict the risk of bevacizumab-induced hypertension

Abstract

Bevacizumab-induced hypertension poses a therapeutic challenge and identifying biomarkers for hypertension can enhance therapy safety. Lower plasma levels of VEGF-A, angiopoietin-2, and rs6770663 in KCNAB1 were previously associated with increased risk of bevacizumab-induced hypertension. This study investigated whether these factors independently contribute to grade 2–3 bevacizumab-induced hypertension risk in 277 cancer patients (CALGB/Alliance 90401). Multivariable analyses assessed the independent association of each factor and hypertension. Likelihood ratio test (LRT) evaluated the explanatory significance of combining protein levels and rs6770663 in predicting hypertension. Boostrap was employed to assess the mediation effect of protein levels on the rs6770663 association with hypertension. Lower protein levels and rs6770663 were independently associated with increased hypertension risk. Adding rs6770663 to protein levels improved the prediction of hypertension (LRT p = 0.0002), with no mediation effect observed. Protein levels of VEGF-A, angiopoietin-2 and rs6770663 in KCNAB1 are independent risk factors and, when combined, may improve prediction of bevacizumab-induced hypertension. ClinicalTrials.gov Identifier: NCT00110214.

Introduction

Hypertension induced by vascular endothelial growth factor (VEGF) pathway inhibitors like bevacizumab during cancer treatment remains a challenge for optimal delivery of therapy and poses a threat to patient health and quality of life, especially when severe (grade 3), which occurs in 7–28% of patients [1]. The discovery of biomarkers that could be easily assessed in blood has the potential for improving the safety of VEGF-pathway inhibitors, which are highly used to treat several types of cancers.

We have demonstrated in a previous publication that lower plasma levels at baseline of VEGF-A, angiopoietin-2, and VCAM-1 were associated with a higher risk of bevacizumab-induced hypertension [2]. The levels of 17 different circulating proteins associated with the function of the vasculature were measured in plasma samples of cancer patients collected before treatment with bevacizumab and, among the 17 proteins evaluated, VEGF-A, angiopoietin-2, and VCAM-1 showed to act independently, protecting patients from the risk of bevacizumab-induced hypertension [2].

Moreover, we have also discovered that the intronic genetic variant rs6770663 in the KCNAB1, which codes for the K⁺ voltage activated channel subfamily 1 subunit β1 (Kvβ1.1–1.3), increases the risk of bevacizumab-induced hypertension [1]. The G allele rs6770663 in KCNAB1 was associated with an increased risk of bevacizumab-induced hypertension in the largest GWAS meta-analysis of bevacizumab-induced toxicities, which included more than 1000 patients. The G allele of rs6770663 in KCNAB1 was present in 8–9% of patients and the association with increased risk of bevacizumab-induced hypertension was further validated in an external cohort of breast cancer patients treated with bevacizumab [1, 3].

It is currently unknown whether the protein plasma levels of VEGF-A, angiopoietin-2, and VCAM-1 and the genetic variant rs6770663 in the KCNAB1 contribute independently to the risk of hypertension. It is predicted that the serum response factor (SRF), a cardiac-enriched transcription factor, is more likely to bind the major allele A than the minor allele G of rs67706633 [4], possibly activating the transcription of KCNAB1. It was postulated that patients with the G allele of rs6770663 have a lower activation of KCNAB1 and, hence, impaired activation of the K+ voltage-activated channel, which leads to an increased vasoconstriction [1]. It is plausible to question whether high levels of VEGF-A, Ang-2, and VCAM-1 could stimulate vasodilation, thereby counteracting the vasoconstrictive effects resulting from the genetic variation, and whether lower levels of proteins could (or could not) contribute additively to the vasoconstriction and endothelium dysfunction caused by the lower activation of KCNAB1. We hypothesize that the risk of hypertension is higher in patients who have both reduced plasma levels and the genetic variant compared to patients who have either one of these factors.

Methods

We have used data from metastatic castration-resistant prostate cancer patients treated with bevacizumab from a randomized, double-blind phase III clinical trial from the Cancer and Leukemia Group B (CALGB, now part of the Alliance for Clinical Trials in Oncology, Alliance), CALGB 90401 [5]. In CALGB 90401, metastatic castration-resistant prostate cancer patients received placebo or bevacizumab 15 mg/kg with docetaxel 75 mg/m2 and prednisone 5 mg on day 1 of a 21-day cycle. More details on patient eligibility, characteristics, stratifications, and treatments can be found in the publications reporting the results on outcome from the clinical trial [5]. For this study, only patients with genetically determined European ancestry (as determined in previous studies [6]) were included. Each participant signed an IRB-approved, protocol-specific informed consent document in accordance with federal and institutional guidelines.

Bevacizumab-induced hypertension

According to the pre-specified criteria described in the trial protocol, hypertension was recorded according to the Common Toxicity Criteria for Adverse Events (CTCAE) version 3.0 by the Alliance Statistics and Data Center. Grade 2 hypertension is defined as recurrent or persistent (≥ 24 h) or symptomatic increase by > 20 mm Hg (diastolic) or > 150/100, whereas grade 3 hypertension is defined as the same blood pressure levels as grade 2 but requiring more than one drug or more intensive therapy than previously administered. Blood pressure was measured on day 1 of each cycle. Only hypertension with attribution of “possibly related” or higher was included, according to standardized criteria. Patients enrolled in the trial should have had no uncontrolled hypertension before receiving bevacizumab.

Measurement of plasma proteins

Peripheral venous blood was collected in EDTA tubes before the first dose of bevacizumab. Plasma levels of VEGF-A, VCAM-1, and angiopoietin-2 were measured before treatment with bevacizumab using the SearchLight (now Quanterix) multiplex platform. The tubes were centrifuged twice within 30 min of the collection at 2500 g for 15 min and double-spun, platelet-poor plasma samples were stored at −80 °C until analysis. Plasma samples were thawed on ice, centrifuged at 20,000 g for 5 min to remove precipitates, and subsequently loaded onto SearchLight plates following manufacturers’ instructions, as previously described [7, 8].

Genotyping of rs6770663 in KCNAB1

Germline DNA was obtained from peripheral blood, and the genotyping platform used was Illumina HumanHap610-Quad. Additional information on the QC procedures can be found in the individual publications of the GWAS data of CALGB 90401 [5].

Data analysis

We performed univariate and multivariable analyses of plasma levels of VEGF-A, angiopoietin-2, and VCAM-1 and rs6770663 (A > G) in KCNAB1 in relation to the risk of grade 2–3 hypertension (CTCAE v3.0, no grade 4 observed) in patients treated with bevacizumab. The model included plasma levels of VEGF-A, angiopoietin-2, and VCAM-1 dichotomized (below vs. above an optimal cut-point previously determined [2]) and rs6770663 (AG vs. AA, no GG genotype observed). The likelihood ratio test [9] was performed to evaluate the explanatory significance of adding rs6770663 to plasma levels of VEGF-A and angiopoietin-2 in predicting the risk of grade 2–3 hypertension. The bootstrap method was employed for mediation analysis to assess the extent to which plasma protein levels mediate the genetic variant association with hypertension. The difference between the prevalence of proteins below the cut-point and rs6770663 in patients with grade 0–1 versus grade 2–3 bevacizumab-induced hypertension was assessed by the chi-square test.

Results

A total of 277 patients treated with bevacizumab from CALGB 90401 had protein samples and genetic data available and were included in the study. The CONSORT diagram presented in Fig. 1 provides detailed information about the number of patients included. In sum, out of the 1050 patients randomized in CALGB 90401, 773 patients were excluded from the study, including 176 patients excluded due to lack of consent, 129 patients excluded because protein’s sample were not available, 380 patients excluded because they were in the placebo arm and 88 patients excluded because they were of no European ancestry. The median age of patients was 68.7 ± 8.2 years.

Fig. 1

CONSORT diagram for CALGB 90401.

Out of the 277 patients included, grade 2 bevacizumab-induced hypertension was recorded in 19 (6.9%) patients, and grade 3 bevacizumab-induced hypertension was recorded in 22 (7.9%) patients. The minor allele frequency (MAF) of rs6770663 was 0.08.

Lower levels of VEGF-A (p = 0.004, OR = 2.98, 95% CI: 1.45–6.37), angiopoietin-2 (p = 0.013, OR = 2.48, 95% CI: 1.22–5.20), and the AG genotype of rs6770663 (p = 0.003, OR = 3.56, 95% CI: 1.51–8.74) were independently associated with an increased risk of grade 2–3 bevacizumab-induced hypertension in the multivariable analysis. Levels of VCAM-1 (p = 0.5645, OR = 0.81, 95% CI: 0.40–1.64) were not associated with an increased risk of grade 2–3 bevacizumab-induced hypertension (Table 1). The likelihood ratio test indicated that when rs6770663 was added to levels of VEGF-A and angiopoietin-2, there was an improvement in the explanatory power in predicting the risk of grade 2–3 hypertension (p = 0.0002). The mediation analysis indicates a lack of mediation of plasma protein levels on the genetic variant association with hypertension (total effect p-value of 0.012 and average causal mediation effect p-value of 0.944).

Table 1 Univariate and multivariable models of plasma levels of VEGF-A, angiopoietin-2, and VCAM-1 and rs6770663 (A > G) in KCNAB1 in relation to the risk of grade 2–3 hypertension.

In a comparison of the prevalence of patients with angiopoietin-2 and VEGF-A with levels below the cut-point and with the AG genotype of rs6770663 in patients with grade 0–1 versus grade 2–3 hypertension, we observed that patients with levels of angiopoietin-2 or VEGF-A below the cut-point, or the AG genotype of rs6770663 had similar ORs for grade 2–3 bevacizumab-induced hypertension (2.53 [95% CI: 1.26–5.08], 1.96 [1.00–3.84], and 2.38 [1.11–5.09]) (Table 2). The AG genotype of rs6770663 was associated with levels of VEGF-A above the cut-point, which might indicate that the protective effect of higher levels VEGF-A is not enough to attenuate the increased risk of bevacizumab-induced hypertension attributed by the AG genotype of rs6770663. No association between the AG genotype of rs6770663 and levels of angiopoietin-2 was observed (Fig. 2).

Table 2 Prevalence of patients with angiopoietin-2 and VEGF-A with level below the cut-points and with AG genotype of rs6770663 in KCNAB1 for patients with grade 0–1 hypertension and grade 2–3 hypertension.

Fig. 2

Frequency of angiopoietin-2 and VEGF-A by genotype of rs6770663 in KCNAB1.

Discussion

Herein we have demonstrated that lower circulating levels of VEGF-A, angiopoietin-2, and the AG genotype of rs6770663 in KCNAB1 are independently linked to an increased risk of grade 2–3 bevacizumab-induced hypertension. The assessment of protein levels and the genetic variant better anticipated the risk of bevacizumab-induced hypertension than each factor alone. The absence of a mediation effect indicates the possible additive effect of protein levels on the genetic variant in predicting the risk of bevacizumab-induced hypertension. While these markers have previously been linked individually to bevacizumab-induced hypertension, it’s crucial to recognize that drug-induced hypertension is a multifaceted condition. By considering the independent and additive contributions of these markers, we gain valuable insights into a composite risk factor associated with hypertension, enhancing predictive accuracy and deepening our understanding of drug-induced hypertension’s underlying pathophysiology.VEGF-A, angiopoietin-2, and KCNAB1 have consistent biology associated with hypertension. VEGF-A is the key regulator of endothelial function [10], and angiopoietin-2 is a potent inducer of vascular permeability [11]. The new mechanism proposed by our group is that low circulating levels of these two proteins might interfere with endothelial function. This mechanism is partly supported by previous evidence of reduced VEGF-A concentrations associated with dysfunctional endothelium in preeclampsia [12], which interferes with the capacity to maintain normal blood pressure.

KCNAB1 encodes for a subunit of the K⁺ channel, which gene deletion was associated with elevated blood pressure [13]. From a mechanistic standpoint detailed previously [1], the G allele of rs6770663 can lead to lower activation of KCNAB1, increasing the risk of hypertension in patients treated with bevacizumab. The independent and possibly additive effect of having lower levels of VEGF-A and angiopoietin-2 with the lower activation of KCNAB1 might further increase the risk of bevacizumab-induced hypertension.

The identification of biomarkers that can contribute to the prediction of the risk of drug-induced hypertension in patients treated with VEGF-pathway inhibitors can have significant implications in the risk assessment of these drugs. The same blood sample collected from patients before initiating treatment with bevacizumab and other VEGF-pathway inhibitors can be used to measure the plasma levels of VEGF-A and angiopoietin-2 and to genotype rs6770663 in KCNAB1. These markers can help identify patients who are at a higher risk of developing drug-induced hypertension and healthcare providers can tailor their treatment plans accordingly, potentially adjusting dosages or employing preventive measures to mitigate the risk of hypertension-related complications. We encourage further studies to validate these findings and to identify other risk factors that might improve the prediction of the model.