Circulating protein biomarkers of pharmacodynamic activity of sunitinib in patients with metastatic renal cell carcinoma: modulation of VEGF and VEGF-related proteins

Sixty-three patients with metastatic RCC and prior treatment with first-line cytokine-based therapy (IFN-a, IL-2) were enrolled in this phase II study. The primary endpoint of the trial was objective response rate, as summarized above [14]. (For a complete listing of eligibility criteria, please see prior publication.)

The study was approved by the institutional review board at each of the seven participating centers and was performed in accordance with the Declaration of Helsinki and Good Clinical Practice Guidelines.

The starting dose of sunitinib was 50 mg/day, administered using the 4/2 schedule. It was self-administered orally once daily without regard to meals. Intrapatient dose escalation in increments of 12.5 mg/day (up to 75 mg/day) was permitted in the absence of treatment-related toxicity. Dose reduction for toxicity was allowed to 37.5 mg/day and then to 25 mg/day, depending on severity of toxicity.

Plasma concentrations of sunitinib and its active metabolite, SU12662, were determined on days 1 and 28 of cycles 1 to 4. Plasma concentrations of both were determined predose by a liquid chromatography/mass spectrometry method at BASi (West Lafayette, IN), with a lower limit of detection of 0.1 ng/mL [23].

Predose plasma samples were collected on days 1 and 28 of each cycle for assessment of soluble proteins that may be correlates of angiogenic activity and/or pharmacodynamic inhibition of VEGF receptor-mediated signaling [24‐26]. Each of the soluble proteins was analyzed with enzyme-linked immunosorbent assay (ELISA) kits (all manufactured by R&D Systems, Minneapolis, MN). The (VEGF)-A ELISA assay measures the VEGF-A165 and VEGF-A121 isoforms. The PlGF assay primarily measures PlGF-1. sVEGFR-2 was quantified with an ELISA that measures the extracellular (soluble) domain of VEGFR-2 [27]. Similarly, an ELISA component kit that measures the extracellular (soluble) domain of VEGFR-3 was employed. Both the sVEGFR-2 and sVEGFR-3 assays are calibrated against recombinant proteins consisting of the full-length extracellular domain of the respective receptors. No cross-reactivity or interference is detected between the two receptors in the ELISA assays. Though the structural details of sVEGFR-2 and sVEGFR-3 remain to be established, plasma-derived sVEGFR-2 has been reported to be heavily glycosylated and to have a molecular weight of ~90 kDa when de-glycosylated; the size is similar to that or insect-derived recombinant sVEGFR-2, implying that endogenous sVEGFR-2 may be similar in structure to recombinant versions of the VEGFR-2 extraceullar domain [27]. All ELISA assays were run under Good Laboratory Practice conditions, and performance specifications of each ELISA were validated for their intended purpose, as per established guidelines [28].

Protein plasma concentration data and correlations with drug levels and response data were analyzed with Microsoft Excel, S-Plus Version 6.2, and Spotfire Decision Site; comparison results from Student's t-test or Wilcoxon rank sum tests with P < 0.05 were considered statistically significant. All P-values reported are results of two-sided tests. In hierarchical clustering analysis, all values were log2-transformed prior to unsupervised two-dimensional clustering (Spotfire Decision Site).

Sixty-three patients were treated with sunitinib. The median age was 60 years, and 55 patients (87%) had clear-cell histology [14]. The objective response rate, the primary endpoint of the study, was 40% (For a full listing of baseline characteristics and summary of efficacy and safety findings, please see prior publication.)

Patients achieved and maintained steady-state trough plasma concentrations (C_trough) of sunitinib and its active metabolite throughout the dosing periods for multiple cycles. The median total drug C_trough (sunitinib and SU12662 combined) at steady state in all patients was 84.3 ng/mL [14], which is greater than the 50 ng/mL shown to inhibit target RTKs in preclinical models [3]. Drug concentrations increased during the on-drug periods and decreased during the off periods, with no accumulation observed, across dosing cycles.

Significant changes (p < 0.0001) in the mean plasma levels of VEGF, sVEGFR-2 and PlGF were noted within each dose cycle, as reported in brief elsewhere [14] and as summarized in Table 1 for the first treatment cycle. Baseline (cycle 1, day 1) values for VEGF and sVEGFR-2 were in the assay detection range for all cases; however, in the case of PlGF, only 17 baseline readings (~27%) were measured as above the lower limit of quantitation for the PlGF ELISA assay (at day 28 of cycle 1, 53 of 56 (~95%) cases were in the detectable range). Due to the prevalence of readings below the level of quantitation for PlGF, for all subsequent data analysis such readings were assigned a value equal to the lowest detectable concentration standard (26.2 pg/mL). At the end of the first cycle, VEGF and PlGF levels increased greater than 3-fold relative to baseline (the cycle 1, day 1 timepoint, which is prior to the first sunitinib dose) in 24 of 54 cases (44%) and 22 of 55 cases (40%), respectively. Levels of sVEGFR-2 were decreased by at least 30% in 50 of 55 cases (91%) during the same time frame, and by at least 20% in all cases. For each of these markers, levels tended to return to near-baseline after the 2-week off-treatment period between cycles. This suggests a drug-dependent effect, as illustrated by the patterns of mean and median ratios to baseline as shown for VEGF and sVEGFR-2 (Figure 1a and 1b). Similarly, the results for PlGF clearly indicate a pharmacodynamic effect of sunitinib treatment (Figure 1c), and suggest that PlGF is a biomarker of interest for this agent; however, since most baseline samples (and day 1 samples from subsequent cycles) were below the lower limit of quantification, precise calculation of ratios to baseline was not feasible in most cases.

In addition to the three angiogenesis-related proteins described above, we also assessed the plasma levels of sVEGFR-3, an apparent circulating variant of VEGFR-3. Assessment of sVEGFR-3 via ELISA was initially selected based on a result of proteomic screening of plasma samples, using an antibody microarray-based multiplex assay [29] in which sVEGFR-3 was identified as a potential pharmacodynamic marker of sunitinib activity [8]. Confirmatory ELISA assessments revealed that significant changes (p < 0.0001) in the mean plasma levels of sVEGFR-3 were noted within each dose cycle (Table 2). Baseline values for sVEGFR-3 were in the assay detection range for all cases; however, at day 28 of the first cycle, sVEGFR-3 levels were reduced to less that the lower limit of quantitation in 12 of 56 cases (21%). Readings below the quantitation limit were frequently observed at day 28 of subsequent cycles as well. For all subsequent data analysis such readings were assigned a value equal to the concentration of the lowest detectable quality control standard (21,200 pg/mL). Levels of sVEGFR-3 were decreased by at least 30% in 48 of 55 cases (87%) during the first cycle, and by at least 20% in all but 2 cases. For each of these markers, levels tended to return to near-baseline after the 2-week off-treatment period between cycles. This suggests a drug-dependent effect, similar to that seen with the other proteins, as illustrated by the patterns of mean and median ratios to baseline for sVEGFR-3 (Figure 2). In the case of VEGF, the magnitude of ratios to baseline at cycle 1, day 28 was higher on average in cases with VEGF baseline values less than the overall median (68.2 pg/mL) compared to cases with higher baseline values (5.7 vs. 3.1; P = 0.039). No significant difference was observed in similar analysis for sVEGFR-2 and sVEGFR-3; this was not done for PlGF due to the detection limitations at baseline for this analyte.

Exploration of the potential relationships between an indicator of sunitinib exposure (trough drug levels) and the modulation of plasma protein levels was performed. Linear regression analysis revealed a modest correlation between total drug trough drug levels (C_trough of sunitinib and SU12662, combined) at the end of the first two cycles and the change in mean sVEGFR-2 plasma levels at the end of cycle 1, relative to baseline (Figure 3a). Similar analysis of the change in sVEGFR-3 levels revealed a trend towards greater sVEGFR-3 reduction at higher trough levels (Figure 3b); however, the trend was weaker than that observed for sVEGFR-2. Such a linear trend was not as apparent in similar analysis for VEGF and PlGF modulation (data not shown); however, analyses of the fold-change in mean plasma levels of VEGF and PlGF at different ranges of trough levels indicated a threshold effect with higher mean fold-changes found for trough levels between 50 and 90 ng/mL as compared to those <50 ng/mL, with a further increase at levels >90 ng/mL (Figure 4). This observation suggests a threshold effect for maximal induction of VEGF and PlGF, although there is still considerable overlap in the range of fold-changes across the different trough exposure categories.

As summarized above, significant clinical benefit was achieved in this study [14]. To assess potential correlations between the biomarker proteins and objective tumor response, the primary endpoint of the clinical study, a comparison of the biomarker results was performed between the 25 patients with RECIST-defined PR and 32 of the 38 patients with SD or progressive disease (PD). (All cases with baseline and at least one post-treatment timepoint were included). Significantly larger proportional changes in VEGF, sVEGFR-2, and sVEGFR-3 levels were observed in patients exhibiting objective tumor response compared with those exhibiting SD or PD (i.e., 'non-PR') (Table 3); PlGF values were not evaluated in this analysis. The significance was most apparent when only on-treatment ratios (i.e., day 28 of the first 3 cycles) were compared. Analysis in Table 3 is limited to the first 3 treatment cycles, since several patients who did not achieve PR were not treated beyond 3 cycles. It should be noted that while overall, there appeared to be a greater pharmacodynamic effect in the PR group, the patterns of change for these proteins were not discrete enough to allow for definitive response prediction on an individual basis. It is likely that changes in circulating protein levels (detectable here at the first post-treatment sample collection, on day 28) preceded tumor size reductions in at least some cases, since the median time to first observation of partial response was 2.3 months [14]. Baseline plasma levels of these proteins did not differ significantly between the PR and non-PR response groups in this analysis.

For the purpose of illustrating how the patterns of each of these endpoints may be inter-related, hierarchical clustering analysis was performed on the fold-changes for each soluble protein (cycle 1, day 28 over baseline), on C_trough levels, and on binary scores assigned for best objective tumor response (Figure 5). This 'heat map' depiction indicates that VEGF and PlGF inductions often occur in parallel, though not necessarily to the same extent of magnitude (with the caveat that baseline PlGF levels are often not quantifiable with the ELISA method); similarly, sVEGFR-2 and sVEGFR-3 reductions tend to occur in parallel, but not to the same extent. In this illustration, clustering analysis, using categorical values to represent best objective response outcomes for each patient, suggests that a greater proportion of PR cases are grouped in the range where greater biomarker changes are observed in cycle 1.