Prospective Exploratory Analysis of Angiogenic Biomarkers in Peripheral Blood in Advanced NSCLC Patients Treated With Bevacizumab Plus Chemotherapy: The ANGIOMET Study

Finding angiogenic prognostic markers in advanced non-small-cell lung cancer is still an unmet medical need. We explored a set of genetic variants in the VEGF-pathway as potential biomarkers to predict clinical outcomes of patients with non-small-cell lung cancer treated with chemotherapy plus bevacizumab. We prospectively analyzed the relationship between VEGF-pathway components with both pathological and prognostic variables in response to chemotherapy plus bevacizumab in 168 patients with non-squamous non-small-cell lung cancer. Circulating levels of VEGF and VEGFR2 and expression of specific endothelial surface markers and single-nucleotide polymorphisms in VEGF-pathway genes were analyzed. The primary clinical endpoint was progression-free survival. Secondary endpoints included overall survival and objective tumor response. VEGFR-1 rs9582036 variants AA/AC were associated with increased progression-free survival (p = 0.012 and p = 0.035, respectively), and with improved overall survival (p = 0.019) with respect to CC allele. Patients with VEGF-A rs3025039 harboring allele TT had also reduced mortality risk (p = 0.049) compared with the CC allele. The VEGF-A rs833061 variant was found to be related with response to treatment, with 61.1% of patients harboring the CC allele achieving partial treatment response. High pre-treatment circulating levels of VEGF-A were associated with shorter progression-free survival (p = 0.036). In conclusion, in this prospective study, genetic variants in VEGFR-1 and VEGF-A and plasma levels of VEGF-A were associated with clinical benefit, progression-free survival, or overall survival in a cohort of advanced non-squamous non-small-cell lung cancer patients receiving chemotherapy plus antiangiogenic therapy.

Finding angiogenic prognostic markers in advanced non-small-cell lung cancer is still an unmet medical need. We explored a set of genetic variants in the VEGF-pathway as potential biomarkers to predict clinical outcomes of patients with non-small-cell lung cancer treated with chemotherapy plus bevacizumab. We prospectively analyzed the relationship between VEGF-pathway components with both pathological and prognostic variables in response to chemotherapy plus bevacizumab in 168 patients with nonsquamous non-small-cell lung cancer. Circulating levels of VEGF and VEGFR2 and expression of specific endothelial surface markers and single-nucleotide polymorphisms in VEGF-pathway genes were analyzed. The primary clinical endpoint was progression-

INTRODUCTION
Lung cancer is one of the most frequent malignancies, and presently the leading cause of cancer-related deaths in Europe (1). Approximately 85% of lung cancer cases are non-small-cell lung cancer (NSCLC), and most NSCLC patients (70%) are diagnosed with advanced-stage disease (i.e., stages IIIB/IV) at presentation (2). A crucial aspect for solid tumor growth is vascularization, and various tumors have been found to produce angiogenic factors themselves or benefit from vascularization induced by inflammatory mediators (3). The major regulator of angiogenesis is vascular endothelial growth factor (VEGF) (4), whose overexpression seems to play a most relevant role in malignant phenotype of solid tumors, including NSCLC (5). Carboplatinbased chemotherapy with bevacizumab has become a standard therapy for eligible NSCLC patients after the results obtained in trials (6)(7)(8)(9) and observational studies (10)(11)(12). However, careful consideration of the results within the study populations points to the need of determining prognostic markers to select those patients that might achieve greater benefit of treatment with these schemes.
Serum levels of VEGF isoforms and their receptors have been used as prognostic markers and to monitor response to chemotherapy or anti-angiogenic agents (5,13). In fact, their levels before therapy initiation correlate to prognosis in NSCLC (14)(15)(16)(17), but clinical biomarkers and these characteristics alone have been insufficient to predict the course of the disease and response to therapy (16,17). Variants within VEGF-related genes seem to regulate their transcription (18,19), and several single-nucleotide polymorphisms (SNPs) have been identified, some of them influencing levels of VEGF isoforms or of their receptors in plasma (18,(20)(21)(22)(23)(24). Genotype variants in proteins of the VEGF pathway have been shown to impact patient outcomes, however with inconsistent results (25)(26)(27)(28)(29)(30)(31)(32), and SNPs are therefore still subject of study as potential prognostic biomarkers. The possible role as biomarkers in NSCLC has also been widely studied in specific endothelial surface proteins CD31, CD34, CD133, and CD146 (33,34).
Kristen Rat Sarcoma viral oncogene (KRAS) is a well-known driver of NSCLC for which no targeted therapy has been developed yet. KRAS mutations have been classically defined as a negative prognostic factor in terms of progression-free survival (PFS) and overall survival (OS), but results are heterogenous and the clinical significance remains controversial, also in patients that receive platinum-based chemotherapy (35). However, SNPs in KRAS have been related to relapse-free survival in NSCLC (36) and PFS in colorectal cancer (37), and results suggest further research in their potential as biomarkers in NSCLC.
Therefore, the main objective of this clinical study was to increase insight in the correlation between selected molecular biomarkers in genes coding for VEGF-A, VEGF receptors 1 and 2 (VEGF-R1 and VEGFR-2), and KRAS, circulating levels of angiogenic mediators and expression of endothelial markers, and the clinical response to the combined treatment of carboplatin, paclitaxel, and bevacizumab in advanced NSCLC patients.
The authors present the following article in accordance with the STROBE reporting checklist.

MATERIALS AND METHODS
ANGIOMET (NCT01814163) was a case-only observational exploratory, post-authorization study. The study has a prospective and multicenter scheme in advanced NSCLC of non-squamous histology, treated in first line with carboplatinpaclitaxel-bevacizumab conducted between February 2011 and February 2013. It was designed to investigate the relationship between genotypes and circulating levels of selected angiogenic mediators and the clinical outcomes and response to this treatment scheme. Twenty hospitals of the public healthcare system across Spain participated. any part of the work are appropriately investigated and resolved. The trial was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The protocol was approved by the Ethics Committee of Hospital General Universitario de Alicante and was accepted by each participating center. All patients provided written consent for inclusion.
Patients aged over 18 years with advanced non-resectable NSCLC, metastatic or recurrent, and not previously treated with chemotherapy were invited to participate if the treating specialist considered the study therapy as the most appropriate. Those with squamous or non-measurable tumors [according to RECIST 1.1 criteria (38)] or in whom peripheral blood samples could not be obtained were excluded. Detailed inclusion and exclusion criteria are provided in the Supplementary Appendix. Recruitment period was 12 months, and the patients were followed for 24 months. These patients received standard therapy of a combination of carboplatin (AUC 6), paclitaxel (200 mg/m 2 ), and bevacizumab (15 mg/kg) every 21 days for a total of six cycles unless there was evidence of disease progression or intolerance to treatment.

Sample Collection and RNA/Protein Measurement
Peripheral blood samples were collected before the first chemotherapy cycle and after three treatment cycles in tubes containing EDTA as anticoagulant (BD Vacutainer ® , USA) and Blood RNA tubes (PAXgene ® , USA). Samples were stored at 4°C until DNA, RNA, or plasma extraction was performed.
Blood samples were sent to a reference laboratory within 24 h of blood collection and subject to RNA or plasma extraction. Two centrifugation steps were performed to obtain plasma (10 min at 1,100 g at room temperature and a second centrifugation of the supernatants for 10 min at 2,000 g at RT to eliminate any possible cell fragments). Plasma aliquots were immediately stored at −80°C until further analysis. Circulating levels of VEGF-A and VEGFR-2 in plasma were assayed as previously described (39); double sandwich ELISA (Duo Set, R&D Systems) was used, in which the lower limit of detection for VEGF-A was 31.2 pg/ml, and 15.6 pg/ml for VEGFR-2.
RNA was isolated using the PreAnalytiX blood RNA kit (Qiagen, Valencia, CA, USA) and quantified by spectrophotometry in a NanoDrop2000c device (ThermoScientific, USA). Retrotranscription was performed with 500 ng of RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, USA). Expression of CD31, CD34, CD133, and CD146 markers was determined by real-time quantitative PCR using primers and probes designed with Taqman ® technology (Gene Expression Assays, Applied Biosystems). Expression of target genes was normalized against endogenous expression of a combination of two genes, GAPDH and CDKN1B, as a reference. cDNA sample from a known cell line was used as a calibrator to minimize inter-trial variability.
To analyze differences between baseline and post-treatment samples, the circulating levels of VEGF-A and VEGFR-2 and the expression of endothelial markers (CD31, CD34, CD133, and CD146) were showed as ratios, calculated as the fraction between baseline and post-treatment measurements. Results were dichotomized in "high" and "low." To evaluate the association with survival times and response to treatment, Chi-square test or Fisher's Exact test were used as appropriate.

SNP Selection and Genotyping
Molecular analysis of 10 SNPs in the genes coding for VEGFR-1, VEGFR-2, VEGF-A, and KRAS was performed by means of realtime PCR. The SNPs analyzed were VEGFR-1 rs7996030, Details and rationale for their inclusion in molecular analysis are provided in the supplementary material (Table S1).
Genomic DNA was extracted from the buffy coat fraction (EDTA tubes) using QIAamp DNA blood Mini Kit (Qiagen, Valencia, CA, USA). Real-time PCR reactions were carried out on 20 ng of DNA using predesigned assays (TaqMan ® SNP Genotyping Assay, Applied Biosystems) and master mix containing DNA polymerase, dNTPs, and buffer (TaqMan ® genotyping master mix, Applied Biosystems ® ) in a final volume of 5 µl. A positive and a negative control were included for each SNP analyzed and each reaction plate. All samples were tested in duplicate. The PCR conditions were as follows: 95°C for 1 min, 40 cycles of 95°C for 10 s, 60°C for 1 min. Genotypes were discriminated using probes labeled with FAM and VIC fluorophores in a single multiplex reaction. Fluorescence intensity readings of the probes were measured by the ABI 7900 instrument, and the genotypes were assigned using the SDS 2.4 Software (ABI). All blood samples were centrally analyzed at the Research Institute of Hospital General Universitario de Valencia (Valencia, Spain).

Endpoints and Variables
Correlation between angiogenic markers and PFS was the primary clinical endpoint (defined as the time from treatment initiation until disease progression or death, whichever occurred first). OS (defined as the time from diagnosis until death from any cause or last clinical follow-up) at 12 and 24 months and response to treatment (classified using the RECIST 1.1. criteria in complete response, partial response, stable disease, and disease progression) were among the secondary endpoints. The absolute and variable frequencies of the responses according to the treatment are presented through contingency tables.

Statistical Methods
Sample size was estimated considering feasibility in the 20 participating hospitals, given the recruitment period, and set in 200 patients. Descriptive analysis was performed with all variables, which were summarized as mean and standard deviation (SD), median and interquartile range (IQR), or frequency and proportion as appropriate. OS and PFS curves were plotted according to the Kaplan-Meier method, and differences between groups were assessed using the log-rank test. Association of SNPs, circulating levels of angiogenic mediators, and expression of endothelial markers with PFS and OS was analyzed by means of univariate Cox regression models. Multivariate Cox analysis was performed including all statistically significant variables from the univariate analyses. Clinical response to treatment was evaluated by the Chi-square test or Fisher's Exact test when appropriate. The tests were analyzed using Stata v16.1. software (College Station, TX: StataCorp LLC). The level of significance was set to a = 0.05.

Patient Characteristics
Of the 201 patients initially included in the study, two of them did not start treatment (consent was withdrawn in one case, and another experienced rapid deterioration); therefore, 199 received the planned combination of carboplatin, paclitaxel, and bevacizumab (CPB). Median age was 62 years (range 54-67), and 139 (70%) were male. Performance status at baseline was ECOG 0 or 1 in the majority (96.2%), and most of them were current smokers or ex-smokers (85.3%). The most frequent histological type was adenocarcinoma (89.1%). Regarding tumor characteristics ( Table 1), all patients presented a median of two target lesions of mean size 8.41 ± 5.63 cm, and 91.5% presented also non-target lesions. One hundred and seventy-four patients presented two or more metastases, of which pulmonary lymph nodes (65.2%) and lung metastases (50.2%) were the most common (Table S2).

Clinical Results
One hundred and sixty-eight patients had baseline valid blood samples and complete data for the variables of interest and outcomes analyzed. The median OS for these patients was 14.6 months (range 11.7-16.2) ( Figure 1A), and the median PFS was 7.1 months (range 6.1-7.9) ( Figure 1B). The OS was 80% at 6 months and 58% at 1 year. Regarding PFS, it was 59% at 6 months, and 18% after 1 year. One patient achieved complete response, 86 achieved partial response (43.2% of the overall sample and 51.2% among those that received at least three cycles), and 20 (10.1%) progressed during initial therapy ( Figure 2). All patients were followed up until death, abandonment, or end of study. Of the 199 treated patients, 11 (5.5%) died during initial therapy and 142 (71.3%) during the follow-up period. Adverse events and other safety outcomes are shown in Supplementary Appendix and Tables S3-S5.

Genetic Variants and Outcomes
Molecular analysis of genetic variants was performed in DNA samples obtained from peripheral blood, and the frequencies obtained are detailed in Table 2. Table 3 shows the results on the association between SNPs in VEGFA, VEGFR1, VEGFR2, and    Correlation between analyzed polymorphisms and PFS or OS in our cohort is shown in Table 4. The genetic variant in VEGFR-1 rs9582036 was significantly associated to OS and PFS and VEGF-A rs3025039 was significantly associated to OS ( Figure 3). Patients harboring the AA/AC alleles in VEGFR-1 rs9582036 had increased OS compared with those with CC allele, with median OS of 17.4, 13.7 and 10.0 months, respectively. Similarly, patients with the VEGF-A rs3025039 SNPs, TT and CT presented half the risk of death than those patients with the CC allele (HR TT : 0.45, p = 0.049; HR CT: 0.54, p = 0.068). On the other hand, the PFS was significatively increased in patients with VEGFR-1 rs9582036 AA/AC alleles (HR AA : 0.47, p = 0.012; HR AC : 0.54, p = 0.035). No association was found in any of the SNPs studied in KRAS and OS or PFS ( Table 4).

Circulating Levels of VEGF-A and VEGFR-2 and Clinical Outcomes
We evaluated baseline VEGF-A and VEGFR-2 circulating levels as possible biomarkers of clinical outcomes such as PFS, OS, or treatment response to CPB (Table S6). By using the median of each soluble marker as a cutoff, continuous variables were dichotomized as "high" (values above the median) and "low" (values equal or less than the median). Our results showed that patients within the "high" VEGF-A group had significantly shorter PFS [5.9 vs. 8.4 months, p = 0.036] than patients within the "low" VEGF-A group. A similar trend (nonstatistically significant) was observed in OS (12.0 vs. 16.0 months, p = 0.095). Survival analysis was performed also for VEGFR-2, and in this case, no significant association was found between basal levels of this biomarker and PFS or OS.
When we analyzed the possible changes in circulating levels of VEGF-A and VEGFR-2 between pre-and post-treatment values (expressed as a ratio and using a cutoff value = 1), we found no significant association with risk of death, progression, or response to treatment either in the group with a ratio below 1 or in the group with a ratio above 1 (data not shown).

Expression of Endothelial Markers and Prognosis
No correlations were found between expression levels of CD31, CD34, CD146, or CD133 at baseline, post treatment, or when ratios baseline/post-treatment (cutoff value = median) were analyzed and correlated with PFS, OS, or response to treatment in this study (Table S6 and data not shown).

Multivariate Analysis
A multivariate Cox regression model for PFS and OS was built using clinical variables (PS-ECOG, gender, histology, smoking habit) and the analytical variables that were found significant in the univariate analysis. Regarding PFS, this analysis revealed that only the VEGFR-1 rs9582036 and baseline circulating levels of VEGF-A were independent prognostic variables (p = 0.012 and p = 0.042, respectively) in our patient population. Moreover, ECOG-PS, VEGFR-1 rs9582036, and VEGF-A rs3025039 were found as independent markers for OS (p = 0.004, p = 0.018 and p = 0.007, respectively) in our study.

DISCUSSION
There is a strong unmet need for new biomarkers that can predict clinical responses to various therapies in patients with NSCLC, facilitating a more individualized treatment. During the last years, several studies have evaluated different molecules related to VEGF pathway in relation to clinical outcomes in patients with NSCLC treated with CPB regimen, but until now, there are no validated predictive biomarkers in the clinical setting (40). This prospective study contributes to the body of evidence demonstrating the potential of some biomarkers, including SNPs in angiogenic genes and circulating levels of VEGF, as prognostic factors in NSCLC. Our results show that both VEGFR-1 rs9582036 and VEGF-A rs3025039 were independent prognostic markers for OS. Moreover, shorter PFS correlates with higher baseline plasma levels of VEGF-A and the presence of the CC allele in VEGFR-1 rs9582036. Then, we propose the analysis of variants in VEGFR-1 and VEGF-A as potential biomarkers in patients with NSCLC treated with CPB. Several studies have linked rs9582036 VEGFR-1 polymorphism and clinical outcomes in bevacizumab-treated patients with different kinds of cancer (36,41,42). The CC allele of rs9582036 variant was predictive of shorter survival in pancreatic and colorectal cancer (41,42). Regarding NSCLC patients, the study performed by Glubb et al. (36) in 2015 showed that AA variants of rs9582036 were associated with longer relapse-free survival independent of the treatment, and were proposed as a prognostic biomarker in stage I-III NSCLC (36). Our results found that the CC rs9582036 allele has a negative predictive effect in patients with advanced NSCLC treated with chemotherapy plus bevacizumab, while the AA and AC variants are associated with an increase in PFS and OS. Taken together, these data suggest that carriers of the CC allele of the VEGFR-1 variant rs9582036 may be less responsive to angiogenesis inhibition having a detrimental effect on survival. However, one should consider that the VEGFR-1 variants found in our study were detected in peripheral blood, while the SNPs reported in the study by Glubb et al. (36) were found in somatic tumor cells. Considering all these pieces of evidence supporting the role of the rs9582036 polymorphism in VEGFR-1 in treatment response and outcome in NSCLC, more studies are needed to understand the molecular mechanism of VEGFR-1 genetic variation in response to VEGF blockage. Additionally, we found that patients harboring both the TT and CT alleles in rs3025039 VEGF-A had reduced their risk of death by approximately half compared to CC. Again, there is some controversy in the literature regarding the effect of this SNP in lung cancer clinical outcomes. While patients with advance NSCLC treated with chemoradiotherapy showed no significant association with survival (43), a trend toward improved survival was observed in early-stage NSCLC patients treated with surgical resection, in concordance with our findings (30). In line with these results, Chen et al. reported that patients with the TT allele treated with chemotherapy presented lower risk of death from all causes (44).
Only VEGF-A rs833061 polymorphism was found to be significantly associated with response to treatment in our study. Previous reports have already demonstrated a potential role of this particular SNP in the metastatic capacity of many tumors, including NSCLC (26,30,(45)(46)(47). Regarding its impact on the therapeutic efficiency of bevacizumab plus chemotherapy, our analysis showed that more than 60% of the rs833061 CT carriers had a clinical benefit of the treatment. That agrees with the exploratory study performed by Pallaud et al. in 2014 (27), showing increased best overall response to first-line bevacizumab plus chemotherapy treatment in patients harboring this allele. These data provide strong evidence of the predictive value of rs833061 SNP as potential biomarker for response to anti-VEGF therapy.
We also investigated whether expression profiles of circulating VEGF-A and VEGFR-2 could be of clinical interest for finding predictors of clinical outcomes to CPB in NSCLC. We found that high pre-treatment VEGF-A plasma levels were associated with less favorable prognosis, increasing the risk of progression. In recent years, some studies have also explored the usefulness of measuring the levels of angiogenic factors as prognostic or predictive markers in NSCLC (25-28, 30, 31, 36, 48). Sanmartıń et al. identified a signature of angiogenic factors related with NSCLC outcome (48). Particularly, patients with high levels of VEGF-A and low expression of VEGF-B and VEGF-D had worse OS and lower relapse-free survival. A metanalysis including 74 studies and 7,631 patients found that VEGF-A overexpression was an independent prognostic factor in early-stage NSCLC (49). Regarding the role of circulating levels of VEGF-A, there are evidences showing an association between higher levels of this angiogenic mediator and worse clinical outcomes in NSCLC (49,50). All together, these results suggest that high levels of VEGF-A could be associated with negative outcomes in NSCLC. Since VEGF-A is a target of bevacizumab treatment, one possible explanation of the effects of higher  concentrations of VEGF-A in peripheral blood is directly related to a lower therapeutic effect of this drug. Evaluation of VEGF levels in lung cancer development and in response to therapy might suppose an important tool to better understand its prognostic impact. The limitations of this study include its exploratory nature, since the sample size followed the feasibility criteria and is limited. Also, the SNP panel used is limited, and it should be extended in order to test other SNPs, specifically in VEGF-A, that have shown an association with clinical outcomes in other cancers. On the other hand, this is one of the few studies that prospectively evaluated the effect of different angiogenesis biomarkers in liquid biopsies in the NSCLC setting, accurately collecting information and eliminating any recall bias.
The results of this prospective study support previous findings regarding the influence of SNPs on angiogenic VEGF genes, and the circulating levels of VEGF on the clinical outcomes of patients with non-squamous NSCLC receiving antiangiogenic treatment such as bevacizumab. In most cases, consistent results across studies are lacking, warranting further research on the subject before molecular biomarkers can be useful in selecting patients with NSCLC and being included in treatment algorithms for antiangiogenic agents. However, important advances are being made in other cancer types, including breast cancer, in which a recent study has found an association of independent prognostic factors with specific treatments and has weighted them by the outcome category (51).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Hospital General Universitario de Alicante. The patients/participants provided their written informed consent to participate in this study.