Introduction
The importance of angiogenesis in the growth and progression of advanced renal cell carcinoma (RCC) has been confirmed in the recent past by the success of direct or indirect antiangiogenic treatments in improving the outcome of patients with this disease [
1‐
7]. Although prognostic criteria have been identified which enable classification of patients with advanced RCC into good, intermediate and poor risk groups [
8,
9], the benefits of antiangiogenic therapy have in many cases been shown to span two or more of these groups [
1,
2,
4,
6,
7]. With an increasing range of agents approved for the treatment for advanced RCC, the discovery of biological markers that reliably predict and help to monitor response to a given agent would assist clinicians in devising individual patient treatment strategies.
Sunitinib malate (SUTENT
®) is an oral multitargeted tyrosine kinase inhibitor with potent activity against vascular endothelial growth factor receptor (VEGFR)-1, -2 and -3, platelet-derived growth factor receptor (PDGFR)-α and -β, stem cell factor receptor (KIT) and other receptor tyrosine kinases, which demonstrates antiangiogenic and antitumor activities [
10‐
12] and is approved for the treatment of advanced RCC. In a randomized, multicenter phase III trial, sunitinib showed superior progression-free survival (PFS; the primary endpoint) to interferon-alpha (IFN-α) as first-line therapy of metastatic RCC (mRCC) with median PFS of 11 versus 5 months (
P < 0.001) [
4]. The objective response rate (ORR) was also significantly higher in the sunitinib arm (ORR 47 vs 12 %;
P < 0.001), while median overall survival (OS) was 26.4 and 21.8 months in the sunitinib and IFN-α arms, respectively (
P = 0.051) [
4,
13].
Here, we have investigated potential biomarkers of sunitinib and IFN-α efficacy in a subset of patients enrolled in this phase III trial through the assessment of plasma levels of four soluble proteins that are closely linked to the angiogenesis process: VEGF-A, VEGF-C, a soluble extracellular fragment of VEGF receptor-3 (sVEGFR-3) and interleukin-8 (IL-8). VEGF-A is an endothelial cell-specific mitogen that is upregulated in hypoxia through stabilization of the transcription factor HIF-1α [
14]; this member of the VEGF family is the primary ligand for VEGFR-2 and has been shown to mediate angiogenesis in a variety of animal models through binding to this receptor [
14]. Early studies on VEGF-C and its receptor VEGFR-3 implicated these proteins in the regulation of lymph vessel formation in the adult, but more recent research indicates an additional role in mediating angiogenesis in a wide range of solid tumors [
15]. VEGFR-3 is highly expressed in angiogenic sprouts in a variety of in vivo mouse models and genetic or antibody targeting of this receptor inhibits angiogenesis [
16], while VEGF-C induces angiogenesis in the mouse corneal pocket assay [
17]. IL-8 is a proinflammatory cytokine with proliferative and migratory activities in a variety of cell types, including tumor cells and endothelial cells [
18], which stimulates angiogenesis in vivo [
19,
20]. In addition to their roles in the angiogenesis process, these four proteins were selected for the present study on the basis of the results of prior correlative biomarker studies in RCC, suggesting prognostic value for circulating VEGF-A [
21,
22], possible predictive or prognostic value for circulating VEGF-C and sVEGFR-3 in patients treated with sunitinib [
23] and an association between elevated tumor IL-8 messenger ribonucleic acid (mRNA) expression and advanced disease [
24].
Discussion
In this study, we have investigated the plasma pharmacodynamics of a panel of circulating proteins linked to the mechanism of action of sunitinib (VEGF-A, VEGF-C, sVEGFR-3), as well as IL-8, a potential mediator of resistance to VEGFR inhibition, in a subset of patients in a randomized phase III study comparing sunitinib and IFN-α as first-line treatment for mRCC. In addition, we have explored possible associations between baseline levels of these proteins, or changes from baseline at each time point, and clinical outcome. Significant and consistent changes from baseline levels were seen for plasma VEGF-A, IL-8 and sVEGFR-3 in the sunitinib arm and for plasma IL-8 in the IFN-α arm. In both treatment arms, baseline levels of plasma VEGF-A, VEGF-C and IL-8 were significantly associated with PFS or OS, while baseline plasma sVEGFR-3 was significantly associated with PFS and OS in the sunitinib arm only. No significant and consistent correlations were seen between plasma protein changes from baseline and clinical outcome in either treatment arm.
Our findings provide additional support for the hypothesis that circulating VEGF-A is prognostic for OS in RCC, with low baseline concentrations of VEGF-A correlating with longer OS in both sunitinib and IFN-α arms in the present study. Consistent with these results, Peña et al. [
21] observed that low baseline serum VEGF-A levels in the placebo arm correlated with longer OS by univariate analysis in a placebo-controlled phase III study of sorafenib, a multitargeted kinase inhibitor with potent activity against the 3 VEGF receptors. As observed here, significance was not seen by multivariate analysis when other circulating biomarkers were included. Also, no correlations were observed between the change in biomarker levels (from baseline to week 3 or 12) and outcome (PFS or OS) in sorafenib-treated patients. With respect to PFS in sunitinib-treated RCC patients, our finding of an association between low baseline VEGF-A and prolonged PFS is similar to that reported by Porta et al. [
27] in a biomarker study from the sunitinib expanded access program; assessment of potential correlations with OS was not included in that study. Although Rini et al. [
23] did not observe a correlation between baseline VEGF-A levels and PFS in a phase II study of sunitinib in mRCC patients refractory to the anti-VEGF-A antibody bevacizumab, this negative result may have been influenced by prior VEGF-A pathway inhibition by bevacizumab in this patient population, or by the confounding effects of residual bevacizumab in plasma samples on antibody-based VEGF-A measurement by ELISA. In that phase II sunitinib study, baseline VEGF-A concentrations were inversely correlated with time since final bevacizumab treatment. Other studies in RCC patients have linked high baseline VEGF-A concentrations with shorter survival time, higher clinical stage and higher tumor grade [
28‐
30] and, in some cases, have identified baseline VEGF-A as an independent prognostic factor for PFS and/or OS [
22,
31].
The phase II sunitinib RCC study cited earlier that involved patients refractory to bevacizumab [
23] demonstrated an association between low baseline sVEGFR-3 concentrations and prolonged PFS, consistent with the association reported here for the sunitinib arm. The absence of a significant association between sVEGFR-3 and either PFS or OS in the IFN-α arm of the present study, although trending toward significance for OS, suggests that sVEGFR-3 may be more predictive of sunitinib efficacy than prognostic in RCC, but more research is necessary to address this question. Early reports implicated VEGFR-3 exclusively in the process of lymph vessel production (lymphangiogenesis), but more recent studies have shown that this receptor for VEGF-C and VEGF-D (but not VEGF-A) is expressed both in tumor lymph vessels and in tumor endothelium in a variety of malignancies [
32‐
34]. Although lymphangiogenic activity appears to be relatively low in clear cell RCC [
35,
36], our correlative findings for sVEGFR-3 suggest a possible role for the inhibition of lymphangiogenesis in the clinical activity of sunitinib in this disease, in addition to antiangiogenesis. Further research is required to distinguish between these possibilities. Also of interest is the recent paper by Garcia-Donas et al. [
37], which reported strong associations between two VEGFR3 polymorphisms and PFS in RCC patients treated with sunitinib. Taken together, these findings strongly implicate VEGFR-3 as a potential target for sunitinib that may contribute to efficacy in patients with mRCC.
IL-8 (CXCL8) is a CXC family chemokine that activates multiple signaling pathways, increases proliferation and survival of both endothelial and tumor cells and facilitates the migration of these cell types [
18]. In addition, IL-8 possesses potent proangiogenic activity in vivo [
19,
20]. Tumor IL-8 expression is upregulated in RCC and has been associated both with more advanced disease and with poor survival [
24,
38,
39]. Consistent with these published reports, the present findings suggest that high baseline plasma levels of IL-8 are associated with poor prognosis in RCC; plasma IL-8 concentrations correlated inversely with OS in both treatment arms and correlated inversely with PFS in the IFN-α arm, remaining an independent predictor of OS in the sunitinib arm. Studies with the potent VEGF receptor inhibitor pazopanib have also shown that IL-8 is prognostic for outcome in RCC, as well as predictive of response [
40,
41]. Preclinically, IL-8 has recently been shown to mediate sunitinib resistance in animal models of RCC [
42]. These investigators found that sunitinib-resistant renal tumors were more highly vascularized than those that were sensitive and hypothesized that the tumors had escaped from the antiangiogenic effects of sunitinib by activation of a VEGF/VEGFR-independent mechanism. Screening of xenograft-bearing mice for changes in plasma levels of 89 angiogenic factors revealed that human IL-8 levels were significantly elevated in sunitinib-resistant mice, while levels of human VEGF-A (and other factors) were unchanged. Furthermore, neutralization of IL-8 activity partially restored sunitinib sensitivity in these preclinical models. In a small prospective study, the same authors also found that baseline IL-8 expression was significantly higher in tumor specimens from RCC patients with intrinsic resistance to sunitinib than in tumor specimens from patients who did not progress on treatment. Overall, preclinical and clinical investigations have provided in vivo evidence of a role for IL-8 as a mediator of tumor progression in RCC and as a possible mechanism of innate or acquired resistance to antiangiogenic therapy in this disease.
Plasma IL-8 levels were significantly elevated above baseline at all time points in both arms of this randomized study. Hypoxia has been shown to upregulate IL-8 expression in human rhabdomyosarcoma cell lines, in a manner that is independent of HIF-1α activity [
43]. A similar mechanism might account for plasma IL-8 induction in the present study, even in the presence of VEGF receptor blockade in sunitinib-treated patients. Although hypoxia also induces IL-8 expression in cultured endothelial cells [
44], this response is mediated by VEGF-A and is therefore unlikely to account for elevated plasma IL-8 levels following sunitinib treatment. In vivo evidence for hypoxia-induced IL-8 expression was obtained in D-12 melanoma xenografts, in which IL-8 expression was seen in vascular hot spots that were associated with hypoxic foci [
45]. In support of a role for hypoxia in IL-8 induction in the IFN-α arm, IFN-α has been shown to possess antiangiogenic activity [
46,
47], in addition to its immunomodulatory properties. However, the finding that plasma VEGF-A levels did not significantly change in response to IFN-α treatment, whereas a marked and significant increase in VEGF-A from baseline levels was seen at the end of each treatment period in the sunitinib arm, suggests that a mechanism other than hypoxia induction may be involved in plasma IL-8 induction in IFN-α-treated patients. Clearly, further research is required to explore possible mechanism(s) underlying plasma IL-8 pharmacodynamics in the present study.
Accumulating data suggests an influence of germline polymorphisms on RCC patient efficacy and safety when receiving targeted anti-VEGF or VEGFR2 tyrosine kinase inhibitor therapies. Specifically, publications have cited polymorphisms in the VEGF-A gene [
48] or VEGFR-3 as associating with clinical outcome [
49]. However, the absence of concordance of findings, different SNPs assayed and endpoints measured preclude use of germline polymorphism data for patient selection at the current time. Large multicenter prospective RCC studies in which baseline characteristics, clinical outcomes and SNPs are measured will allow for unequivocal assignment of utility of SNPs for patient selection.
A number of limitations apply to the present investigation. Firstly, only selected sites participated in the exploratory biomarker component of this pivotal phase III study. As a result, samples sizes in the sunitinib and IFN-α arms (33 and 30, respectively, at baseline) were small in relation to patient numbers in the corresponding ITT populations (less than 10 % in each case). The power to detect significant correlations between plasma proteins and clinical outcome was thus markedly reduced, and the possibility of type II errors was markedly increased, when compared with a biomarker analysis involving all patients on study. Nonetheless, a number of significant associations with PFS or OS were observed for soluble proteins in pretreatment plasma samples, and it should be noted that baseline characteristics and clinical outcome (PFS, OS) for the biomarker-evaluated patient subsets in the two treatment arms were a good representation of their respective full study populations. Another limitation of this study is that we have focused on a small panel of circulating proteins, yet there are many proteins not directly linked to sunitinib’s molecular mechanism, of which IL-8 is but one example, that have the potential to predict resistance to sunitinib therapy in RCC based on a known regulatory role in angiogenesis. Circulating biomarkers, including plasma proteins, have the distinct practical advantage of relatively facile sampling and quantitative analysis when compared with tumor tissue-based end points. However, the search for circulating proteins with utility in clinical decision-making for any approved agent in RCC, or indeed any other tumor type, has proven largely unsuccessful, likely in part because the data collected often reflect systemic processes to a greater extent than the relevant tumor biology. Finally, we have interpreted significant associations that were obtained for OS in both treatment arms to suggest prognostic rather than predictive value for a biomarker. However, since the efficacy of IFN-α in RCC may be mediated in part by angiogenesis inhibition, prediction of survival for such a biomarker may not be entirely independent of treatment modality.
In summary, the present investigation provides evidence that plasma concentrations of VEGF-A and IL-8 may be prognostic for OS in mRCC, with high levels being unfavorable. In addition, low plasma levels of the soluble form of VEGFR-3 may predict improved outcome for RCC patients receiving sunitinib, suggesting that inhibition of angiogenesis and/or lymphangiogenesis mediated by this VEGF receptor family member may contribute to the efficacy of this potent multitargeted tyrosine kinase inhibitor. Further predictive biomarker research is clearly warranted in mRCC, not only for sunitinib but also for other VEGF pathway inhibitors and for agents targeting other pathways.