The significance of combining VEGFA, FLT1, and KDR expressions in colon cancer patient prognosis and predicting response to bevacizumab.

Targeting angiogenesis through inhibition of the vascular endothelial growth factor (VEGF) pathway has been successful in the treatment of late stage colorectal cancer. However, not all patients benefit from inhibition of VEGF. Ras status is a powerful biomarker for response to anti-epidermal growth factor receptor therapy; however, an appropriate biomarker for response to anti-VEGF therapy is yet to be identified. VEGF and its receptors, FLT1 and KDR, play a crucial role in colon cancer progression; individually, these factors have been shown to be prognostic in colon cancer; however, expression of none of these factors alone was predictive of tumor response to anti-VEGF therapy. In the present study, we analyzed the expression levels of VEGFA, FLT1, and KDR in two independent colon cancer datasets and found that high expression levels of all three factors afforded a very poor prognosis. The observation was further confirmed in another independent colon cancer dataset, wherein high levels of expression of this three-gene signature was predictive of poor prognosis in patients with proficient mismatch repair a wild-type KRas status, or mutant p53 status. Most importantly, this signature also predicted tumor response to bevacizumab, an antibody targeting VEGFA, in a cohort of bevacizumab-treated patients. Since bevacizumab has been proven to be an important drug in the treatment of advanced stage colon cancer, our results suggest that the three-gene signature approach is valuable in terms of its prognostic value, and that it should be further evaluated in a prospective clinical trial to investigate its predictive value to anti-VEGF treatment.

Introduction

Angiogenesis plays a critical role in the progression of colon cancer, and the vascular endothelial growth factor (VEGF) pathway has been shown to play a central role in controlling tumor angiogenesis.1 VEGF ligands, such as VEGFA, VEGFB, VEGFC, and VEGFD have different affinities toward, and activation potencies at, different VEGF receptors, including VEGFR1 (FLT1), VEGFR2 (KDR), and VEGFR3 (FLT4).2 VEGFA has been shown to activate both FLT1 and KDR,2 and activation of this signaling pathway has led to vascularization and angiogenesis in malignant tissue to support its growth and survival.3 VEGFA has been shown to be overexpressed in colorectal cancer specimens, especially in those tumors with secondary metastases, including liver or lymph node metastases.4–7 Its mRNA expression in tumor as well as plasma is higher in colorectal cancer patients than in non-diseased patients.8 Despite this, VEGF expression has not been shown to be predictive of response to bevacizumab.9 In addition, VEGFA was shown to be a significant predictor for morbidity in a colorectal cancer cohort.10

Expression of FLT1, also known as VEGFR1, was shown to be higher in primary tumors and nodal metastases of colorectal carcinomas, and its overexpression was associated with lymphovascular invasion and local disease recurrence, but not with survival.11 Moreover, the serum level of soluble FLT1 has also been shown to be significantly higher in patients with colorectal cancer compared to non-diseased individuals.12 Likewise, KDR, also known as VEGFR2, has been shown to be overexpressed in metastatic colon cancer.6 Overexpression of its phosphorylated version was associated with larger tumor diameter and poor histological differentiation,10 while several single nucleotide polymorphisms of KDR have also been identified that were postulated to modulate recurrence of colorectal cancer in patients receiving chemotherapy.13 Further investigation on various angiogenic factors in a large colorectal cancer patient cohort has further suggested that the expression of angiogenic factors can be associated with tumors of more aggressive phenotypes.14

Inhibition of the VEGF pathway using antibodies, recombinant fusion proteins, or tyrosine kinase inhibitors has been shown to be a successful approach in prolonging survival of patients with stage IV colorectal cancer.15 Bevacizumab is a humanized monoclonal immunoglobulin G1 antibody against soluble VEGFA that has been shown to be an effective and tolerable therapy that improves survival in colorectal cancer patients.16 In a phase III trial (AVF2107g), patients with metastatic colon cancer who received bevacizumab in addition to chemotherapy (irinotecan, bolus fluorouracil, and leucovorin) had significantly longer progression-free and overall survivals than patients who received chemotherapy only.17 Another phase III trial has also confirmed the efficacy of bevacizumab in treatment of colorectal cancer.18 Recent advances in the targeting of angiogenesis in colorectal cancer include the approval of novel agents,19,20 the demonstration of the importance of continuous inhibition of the VEGF pathway,19,21,22 and the investigation of the efficacy and safety of bevacizumab in the elderly.23 While the success of anti-EGFR (epidermal growth factor receptor) antibody therapy in advanced stage colorectal cancer patients has been improved by selecting patients based on their NRas status in addition to their KRas status,24–26 no such predictive biomarker exists for improving the success of antibody targeting of angiogenesis.

The prognostic value of individual angiogenic factors in colorectal cancer has been limited; for more comprehensive prognostication, co-consideration of multiple angiogenic factors and receptors may be more valuable.5 The survival rate of colorectal cancer patients was shown to be poor when their tumor cells overexpressed VEGFA, and their tumor-associated endothelial cells overexpressed KDR.7 Antiangiogenic therapy is also a successful strategy in the treatment of lung cancer patients;27,28 we recently reported that a VEGFA-FLT1-KDR gene signature had strong prognostic significance in lung cancer patients.29 In the present study, we investigated the prognostic significance of the combined mRNA expressions of VEGFA, FLT1, and KDR in colorectal cancer using colon cancer datasets available in the Gene Expression Omnibus (GEO) database.

Materials and methods

Extraction of clinical and microarray gene expression data from colon cancer patient datasets

Three colorectal cancer patient datasets, GSE14333,30 GSE14538,31 and GSE4096732 were identified in the GEO database; datasets compiled using the HG-U133 microarray platform, which comprised ≥200 patients for whom survival data were available in the GEO database were included in this study. A separate colorectal cancer patient cohort, GSE19862, which contained information on patient response to bevacizumab treatment, was also analyzed. Microarray gene expression data were retrieved from the data matrixes deposited to the GEO database by the original authors. R scripting was used to extract the expression values of a small number of genes (probesets) of interest, and the clinical data from the data matrixes were downloaded from GEO, as previously described.33

Correlations of gene expression levels and clinical data

All statistical analyses were performed using SPSS version 19.0. Expression levels were divided into high and low levels using the median expression level as the cut-off point for Kaplan–Meier survival analysis. Results were compared by log-rank test. Univariate Cox regression analysis was used to correlate gene expression levels and patient survival. Patients were divided into three groups based on the expression levels of VEGFA and FLT1; the VEGFA-FLT1-low group consisted of patients who expressed both genes below median levels (low); the VEGFA-FLT1-high group consisted of patients who expressed both genes at above median levels (high); the VEGFA-FLT1-intermediate group consisted of the remaining patients, who expressed either gene at above median level, and the other below median. Similar stratification was performed for the VEGFA-KDR grouping. Patients were also divided into four groups based on the expression levels of VEGFA, FLT1, and KDR depending on the number of genes expressed at above the median. The survival time of patients stratified by this grouping method were analyzed by Kaplan–Meier analysis and Cox regression.

Results

The association between mRNA expression of VEGFA, FLT1, and KDR, and colon cancer patient survival

A high level of expression of VEGFA in both GSE14333 (P=0.052) and GSE17538 (P=0.006) was associated with poorer disease-free survival (Figure 1A and B). Similarly, a high level of expression of FLT1 (VEGFR1) was associated with poorer disease-free survival in both cohorts (P=0.016 for GSE14333 and P=0.075 for GSE17538; Figure 1C and D). However, the association between KDR (VEGFR2) expression and disease-free survival was less pronounced in these two datasets (Figure 1E and F).

Figure 1

The association between VEGFA, FLT1, and KDR mRNA expression and colon cancer patient survival.

Notes: Kaplan–Meier analyses for VEGFA mRNA expression in (A) GSE14333 and (B) GSE17538. Kaplan–Meier analyses for FLT1 mRNA expression in (C) GSE14333 and (D) GSE17538. Kaplan–Meier analyses for KDR mRNA expression in (E) GSE14333 and (F) GSE17538.

Abbreviation: VEGFA, vascular endothelial growth factor a.

The prognostic significance of VEGFA-FLT1 and VEGFA-KDR combinations

The prognostic significance of VEGFA, FLT1, or KDR was not consistently significantly demonstrated in the two colon cancer datasets. In an attempt to boost their prognostic power, we assessed co-overexpression of the ligand, VEGFA, and its receptors, FLT1 or KDR. As expected, patients whose primary tumors expressed high levels of both VEGFA and FLT1 had a significantly poorer survival than patients who had low expression of both VEGFA and FLT1 in both colon cancer datasets (P=0.008 and P=0.004, respectively; comparison for all three groups, P=0.004 and P=0.026, respectively, for GSE14333 and GSE17538; Figure 2A and B). Similarly, patients whose primary tumors expressed high levels of both VEGFA and KDR had a significantly poorer survival than patients who had low expression of both VEGFA and KDR in both colon cancer datasets (P=0.011 and P=0.001, respectively; comparison for all three groups, P=0.053 and P=0.008, respectively, for GSE14333 and GSE17538; Figure 2C and D).

Figure 2

The association between combination of VEGFA/FLT1 and VEGFA-KDR mRNA expression and colon cancer patient survival.

Notes: Kaplan–Meier analyses for VEGFA and FLT1 mRNA expression in combination in (A) GSE14333 and (B) GSE17538. Kaplan–Meier analyses for VEGFA and KDR mRNA expression in combination in (C) GSE14333 and (D) GSE17538.

Abbreviation: VEGFA, vascular endothelial growth factor a.

The prognostic significance of the VEGFA-FLT1-KDR combination

Since the combinations of VEGFA-FLT1 and VEGFA-KDR were shown to be more consistent in terms of prognostication than when these genes were considered individually, we further investigated whether combination assessment of expression of the three genes, VEGFA-FLT1-KDR, could be a better prognostic indicator. In GSE14333, patients whose tumors expressed all three genes at a low level had a mean disease-free survival of 101 months (95% confidence interval [CI] =86–116 months), which is significantly longer than patients whose tumors expressed all three genes at a high level (mean disease-free survival =72 months, 95% CI =54–90 months). As shown in Figure 3A, the VEGFA-FLT1-KDR gene signature was significantly associated with disease-free survival in colon cancer patients (P=0.014). Similar results were obtained using the GSE17538 dataset. Patients whose tumors expressed all three genes at a low level had a mean disease-free survival of 104 months (95% CI = 95–113 months), while those who expressed all three genes at a high level had a mean disease-free survival of 71 months (95% CI =57–86 months). As shown in Figure 3B, the VEGFA-FLT1-KDR gene signature was again significantly associated with disease-free survival in this patient cohort (P=0.024).

Figure 3

The association between the combination of VEGFA/FLT1/KDR mRNA expression and colon cancer patient survival.

Notes: Kaplan–Meier analyses for VEGFA, FLT1, and KDR mRNA expression in combination in (A) GSE14333 and (B) GSE17538.

Abbreviation: VEGFA, vascular endothelial growth factor a.

The impact of genetic alterations on the prognostic significance of the VEGFA-FLT1-KDR combination

In another independent dataset, GSE40967, which has available clinicopathological parameters, we sought to investigate whether the association between the three-gene signature and disease-free survival retained relevance in patients with specific genetic alterations. The clinical characteristics for patients in this cohort are listed in Table 1. In the GSE40967 colon cancer patient cohort, patients whose tumors had a high level of expression of the three-gene signature again had a significantly shorter disease-free survival than those patients whose tumors had a low level of expression of the three-gene signature (P=0.048; Figure 4A). Interestingly, this VEGFA-FLT1-KDR signature and survival association was only significant in patients with proficient mismatch repair (MMR) status (P=0.031; Figure 4B), but not in those with deficient MMR status (P=0.657; data not shown). Likewise, the three-gene signature was a potent prognostic indicator in patients with wild-type KRas status (P=0.001; Figure 4C), but not in those with mutant KRas status (P=0.848; data not shown). Similarly, the three-gene signature was also significantly associated with survival in patients with mutant p53 status (P=0.038; Figure 4D), but not in those with wild-type p53 status (P=0.501; data not shown).

Table 1

Patient clinical and pathologic features for GSE40967

	Number of cases	Percent	Median (range)
Age	565		68.1 (22–97) years
Stage
0	4	0.7
1	33	5.8
2	264	46.6
3	205	36.2
4	60	10.6
Sex
Female	256	45.2
Male	310	54.8
Location
Proximal	224	39.6
Distal	342	60.4
Colon cancer subtype
C1 (chromosome instability–immune-down subtype)	116	20.5
C2 (dMMR subtype)	104	18.4
C3 (KRas mutant subtype)	75	13.3
C4 (cancer stem cell subtype)	59	10.4
C5 (chromosome instability–Wnt-up subtype)	152	26.9
C6 (chromosome instability–normal-like subtype)	60	10.6
MMR status
Proficient	444	78.4
Deficient	75	13.3
Missing	47	8.3
BRaf status
Wild-type	461	81.4
Mutant	51	9
Missing	54	9.5
KRas status
Wild-type	328	58
Mutant	217	38.3
Missing	21	3.7
p53 status
Wild-type	161	28.4
Mutant	190	33.6
Missing	215	38

Abbreviations: MMR, mismatch repair; dMMR, deficient mismatch repair.

Figure 4

The association between the combination of VEGFA-FLT1-KDR mRNA expression and colon cancer patient survival in cohorts stratified based on genetic aberration.

Notes: Kaplan–Meier analyses for VEGFA, FLT1, and KDR mRNA expression in combination in GSE40967 in (A) the whole cohort, (B) patients with proficient mismatch repair, (C) patients with wild-type KRas status, and (D) patients with mutant p53 status.

Abbreviations: VEGFA, vascular endothelial growth factor a; mRNA, messenger ribonucleic acid; pMMR, proficient mismatch repair.

The predictive value of the three-gene signature to treatment of bevacizumab

The ability of the three-gene signature to predict response to bevacizumab was tested in a cohort of colon cancer patients who had received bevacizumab treatment and for whom data regarding response to the therapy were known. We found that none of the patients who responded well to bevacizumab expressed the three-gene signature at a low level, while 71% of those patients who did not respond to bevacizumab expressed the three-gene signature at a low level; however only 14% of patients who did not respond to bevacizumab expressed this three-gene signature at a low level (chi-square test, P=0.02; Figure 5).

Figure 5

The association between the combination of VEGFA-FLT1-KDR mRNA expression and response to bevacizumab in colon cancer patients in the GSE19862 dataset.

Abbreviation: VEGFA, vascular endothelial growth factor a.

Discussion

Anti-angiogenic agents, such as bevacizumab, aflibercept, and regorafenib play an important role in the treatment of colon cancer. Importantly, continuous inhibition of the VEGF pathway has been shown to prolong survival in colon cancer patients,19–21 further reinforcing angiogenesis as an essential process in tumor progression and validating angiogenesis as an appropriate target for cancer therapy. Recent advances in molecular testing have led to the identification of a subgroup of patients, whose tumors have KRas or NRas mutations, for whom anti-EGFR antibodies, including cetuximab and panitumumab, are not a viable treatment option.24,34 However, no advances to identify suitable (or unsuitable) candidates for anti-angiogenesis treatment have been made. Our previous report showed that combining VEGFA, FLT1, and KDR expression could be a better prognostic indicator in lung cancer compared to when these factors were considered individually.29 In the present study, we have shown that using similar methodology, this three-gene signature is a robust prognostic indicator in colon cancer. Most importantly, this three-gene signature has been shown to predict the response to bevacizumab, an anti-angiogenic agent. Since VEGFA activates both FLT1 and KDR,2 the co-overexpression of the ligand and the two receptors may be required for the activation of the pathway. In addition, as shown in Figure 3, the consideration of all three genes together can better stratify patients into smaller groups with a better defined prognosis, such that patients with all three genes expressed at high levels had the poorest prognosis compared to those with all three genes expressed at a low level, who had the best prognosis. Our results warrant further study on combining VEGFA, FLT1, and KDR expression profiling as a tool for predicting of response to anti-angiogenic agents in colorectal cancer patients. Since patient treatment was heterogeneous in the patient cohorts investigated in the present study, a prospective clinical trial is needed to test the clinical applicability for the three-gene signature as a prognostic and predictive biomarker for colorectal cancer patients.

Our results on the impact of genetic background on the prognostic significance of this three-gene signature for colon cancer are interesting and should be further investigated. MMR deficiency is an important prognostic determinant for colon cancer.35 A recent study has shown that bevacizumab may impose different effects on patients with MMR-deficient tumors compared to MMR-proficient tumors;36 in the current study, we show that the three-gene signature significantly predicted survival in patients with tumors proficient in MMR, but not in those with tumors deficient in MMR.

KRas status is an important biomarker for anti-EGFR therapy; however, it failed to offer predictive value for the anti-angiogenic agent, bevacizumab.37 As anti-EGFR agents are ineffective in colon cancers bearing KRas mutations, anti-angiogenic agents are the most common treatment option for this type of patients. In the present study, we found that the three-gene signature was highly prognostic in patients with wild-type KRas, but not in those with KRas mutation. Bevacizumab and cetuximab have both been shown to have similar efficacy in patients with KRas wild-type tumors in terms of response rate, progression-free survival, and overall survival,38–40 suggesting both anti-VEGF or anti-EGFR are viable options for colon cancer patients with wild-type KRas. The three-gene signature, when verified in a prospective trial, may be able to further stratify colon cancer patients with wild-type KRas in the future.

Previous reports investigating the association between mutational status of p53 and response to bevacizumab were contradictory; some reports showed that p53 status did not affect the efficacy of bevacizumab,41,42 while one report showed that patients with p53-positive (expression) tumors had a significantly longer survival when treated with bevacizumab with chemotherapy;43 another report showed that the progression-free survival was significantly longer with bevacizumab-containing regimens in patients with mutant p53 tumors, but not in those with wild-type p53 tumors.44 In the present study, our results show that the three-gene signature is prognostic only in tumors with mutant p53, but not in those with wild-type p53, suggesting that VEGFA-mediated angiogenesis may play an important role in a p53 mutant genetic background. Further analysis is required to confirm the prognostic and predictive value of p53 in colon cancers treated with bevacizumab.

Conclusion

In conclusion, our results suggest that the three-gene signature may be a potential prognostic indictor for colon cancer and a potential predictive biomarker for response to anti-angiogenic agents such as bevacizumab. Importantly, the prognostic significance of this three-gene signature has also been demonstrated in lung cancer, in which the efficacy of anti-angiogenic agents has been shown. Since anti-angiogenic agents have been shown to be active also in other cancer types, further investigation into the prognostic significance of this three-gene signature in other types of cancer is warranted.