Multiple circulating proangiogenic factors induced by sunitinib malate are tumor-independent and correlate with antitumor efficacy

Abstract

Cancer patients treated with antiangiogenic multitargeted receptor tyrosine kinase (RTK) inhibitors show increased levels of plasma VEGF and placental growth factor and decreased levels of soluble VEGF receptor-2, thus implicating these overall changes as a possible class effect of such drugs and raising the possibility of their exploitation as surrogate biomarkers for pharmacodynamic drug activity/exposure and patient benefit. A postulated mechanism for these changes is that they are tumor-dependent, resulting from drug-induced decreases in vascular function, increases in tumor hypoxia, and changes in hypoxia-regulated genes. However, here we report that an identical pattern of change is observed in normal nontumor-bearing mice treated with SU11248/sunitinib, a small-molecule inhibitor of VEGF and PDGF RTKs. The changes were dose-dependent, plateaued after 4 days of consecutive treatment, reversed after discontinuation of therapy, and correlated with antitumor activity. Altered protein expression was found in a broad variety of tissues, and dose-dependent elevations were observed of several plasma proteins previously unassociated with this class of inhibitor, including G-CSF, SDF-1α, SCF, and osteopontin. Our results suggest that observed sunitinib-induced molecular plasma changes, including those both directly and indirectly targeted by drug, represent a systemic tumor-independent response to therapy and may correlate with the most efficacious antitumor doses, potentially having utility for defining the optimal biologic dose range for this drug class but not as predictive markers of tumor response or clinical benefit. They may also be relevant to drug-associated toxicities, drug resistance, and observed rapid tumor (re)growth seen after cessation of therapy.

The increasing clinical use of antiangiogenic drugs, both approved and investigational, has revealed a number of important recurring needs and findings. With respect to the former, the growing need to discover and validate surrogate biomarkers to help determine the optimal therapeutic dose, predict which patients are most likely to achieve clinical benefit, indicate emerging resistance, or foretell a particular toxic side effect are all well known (1, 2). With respect to recurrent findings, toxicities such as hypertension and proteinuria are very common side effects of treatment with drugs that target VEGF or VEGF receptors (VEGFR), including VEGFR-2, whether antibodies or small-molecule receptor tyrosine kinase (RTK) inhibitors (RTKIs) (3). For RTKIs, another common feature is a triad of molecular changes involving circulating proteins, namely, increased levels of plasma VEGF and placental growth factor (PlGF), another member of the VEGF family that binds to VEGFR-1, along with reduced levels of soluble VEGFR-2 (sVEGFR-2) (4). These changes have now been noted with three different VEGF RTKIs, including SU11248/sunitinib, AG-013736/axitinib, PTK787/vatalanib, and at least eight more reported at recent meetings, including BAY-43–9006/sorafenib, such that this set of changes can now be considered a “class” effect of small-molecule antiangiogenic RTKIs [supporting information (SI) Table 1].

The biologic significance of these changes is unknown. For example, on the basis of results of a limited phase I trial testing PTK787, there was an initial trend noted that showed increases in plasma levels of VEGF occurred mainly in responding colorectal carcinoma patients but not in nonresponding patients (5). It was suggested that this was likely a consequence of the biologic (antiangiogenic) activity of the drug on the tumor, namely, induction of greater levels of tumor hypoxia, which is a major inducer of VEGF (and PlGF), as a result of the drug's ability to suppress tumor blood flow, perfusion, and possibly microvessel numbers as well (5). However, subsequent randomized trials with a larger number of patients using sunitinib did not show a correlation between increases of VEGF or PlGF and patient response/clinical benefit in renal cell carcinomas (6) or patients with refractory gastrointestinal stromal tumors (7).

In all of the aforementioned studies, the results were obtained in cancer patients; there were no normal healthy volunteers. Because VEGF in cancer-bearing hosts can be derived not only from tumor cells but also in substantial quantities from various host cells (8, 9), it remains formally possible that the sources of the increases in VEGF and PlGF are mainly from normal host cells. If so, this could help explain the lack of correlation with tumor response, or clinical benefit. To examine the effects of an antiangiogenic drug in normal vs. tumor-bearing hosts for inducing such molecular changes in plasma VEGF and PlGF (and sVEGFR-2), as well as some of the possible consequences of such changes, we decided to evaluate the effects of sunitinib in human tumor xenograft-bearing and normal mice using a discontinuous clinical-like schedule of drug administration. This approach provides an opportunity to examine the reversibility and durability of any observed changes.

We now report that the clinical pattern of plasma VEGF elevation can be observed in tumor-bearing mice, suggesting at least part of the observed clinical changes could be derived from the tumor. However, we also noted that similar VEGF elevations, along with increases in PlGF and decreases in sVEGFR-2, could be remarkably duplicated in normal mice free of any tumor. The changes plateaued after several days of treatment, reversed within a few days of stopping drug administration, and were dose-dependent, with the observed changes correlating with antitumor efficacy. Unexpectedly, we also observed a number of additional dose-dependent elevations in numerous plasma proteins previously unassociated with this class of inhibitor that suggest systemic collateral or indirect drug effects. Interestingly, these include elevated levels of several cytokines or chemokines that have all been implicated as having proangiogenic activities. For example, granulocyte colony-stimulating factor (G-CSF), stromal derived factor 1α (SDF-1α), and osteopontin, as well as stem cell factor (SCF), the ligand for the sunitinib-targeted c-kit, were all up-regulated. Our results suggest that clinically observed sunitinib-induced changes in plasma molecular proteins, including those both directly and indirectly targeted by drug, correlate with the most efficacious antitumor doses and have potential utility for defining the optimal biologic dose range for this class of therapeutics, independent of tumor burden. Additionally, given the pleiotropic functions of these proteins, including proangiogenic activity, it is conceivable they could have an impact on a number of important aspects of tumor–host biology and drug action, including toxicity, drug resistance, and primary or metastatic tumor growth.

Results

Dose-Dependent Increase of Plasma VEGF After Sunitinib Treatment Is Both Tumor- and Host-Derived.

To decipher the source of the VEGF elevation after drug administration, human PC-3 cells overexpressing human VEGF₁₆₅ were grown s.c. in SCID mice and treated for 7 days with varying doses of sunitinib. Levels of circulating human (tumor-derived) VEGF₁₆₅ measured in mouse plasma were found to be elevated 4-fold at dose levels of 60-mg/kg-per-day and >7-fold at 120-mg/kg-per-day sunitinib when standardized for tumor burden and compared with vehicle-treated controls, indicating that elevation of VEGF levels in plasma may originate, at least partially, from tumor tissue (Fig. 1A). Surprisingly, measurement of the same plasma for mouse (host-derived) VEGF by mouse-specific ELISA demonstrated 6- and 15-fold increases in VEGF at sunitinib doses of 60 and 120 mg/kg per day, respectively, when standardized for tumor burden and compared with vehicle-treated controls, indicating a tumor-independent contribution to observed VEGF elevations (Fig. 1B). Bevacizumab, a monoclonal antibody against human VEGF, can bind to free human VEGF and limit its detectability by ELISA-based methods (10, 11). We therefore administered bevacizumab in combination with 120-mg/kg-per-day doses of sunitinib to ensure species specificity of the VEGF ELISA measurements. Fig. 1 shows bevacizumab to limit detectability of human VEGF but not mouse VEGF, confirming the absence of cross-reactivity between kits. Taken together, these results demonstrate that dose-dependent elevation in plasma VEGF after sunitinib treatment is both tumor cell- and host-derived.

Fig. 1.

Host (mouse) and tumor (human) VEGF levels rise in tumor-bearing mice treated with sunitinib. (A) Plasma taken from SCID mice bearing s.c. human PC3 tumors overexpressing human VEGF₁₆₅ were measured for human (tumor-derived) VEGF by ELISA and found to be elevated after 7 days of oral administration of sunitinib at varying doses (15, 30, 60, and 120 mg/kg, respectively). (B) The same plasma also showed elevated mouse (host-derived) VEGF levels. Bevacizumab was administered (300 μg per mouse) every 3 days alone and in combination with sunitinib and limited human VEGF detectability, confirming species specificity of the mouse ELISA (A and B). y axis represented as percent of control with original values represented as pg/ml of VEGF standardized to excised tumor weight (grams) (see SI Text). n = 3 or 4 for all groups. Symbols and bars, mean ± SD.

Tumor-Independent Modulation of VEGF, sVEGFR-2, and PlGF by Sunitinib Is Reversible and Coincides with Optimal Antitumor Dose Range.

It has been reported that administration of sunitinib to renal cell carcinoma (6) patients or patients with refractory gastrointestinal stromal tumors (7) over a 4-week on, 2-week off schedule results in cyclical increases (on therapy) and decreases (off therapy) of plasma VEGF levels, whereas the opposite cyclical pattern is seen with sVEGFR-2 (6, 12). We tested whether changes in VEGF and sVEGFR-2 levels, as well as PlGF [also noted to be elevated (6)], are also reversible in nontumor-bearing mice during and after cessation of treatment with sunitinib. Daily oral doses of 7.5, 15, 30, 60, and 120 mg/kg sunitinib were administered to Balb/C mice for two 7-day courses of sunitinib treatment separated by a dosing break. Drug-dosing choice was based on prior reports demonstrating the optimal preclinical dose of sunitinib for antitumor efficacy in mice to be between 40 and 80 mg/kg per day (13). These findings have been confirmed in studies with mice bearing s.c. tumor xenografts, including human A431, rat C6, and human Colo205 cells treated with sunitinib (SI Fig. 6). For our studies, the aforementioned doses were chosen for the following reasons: the 60-mg/kg-per-day dose (an intermediate between 40 and 80 mg/kg per day) represents a confirmed efficacious antitumor dose, the 7.5- to 30-mg/kg-per-day doses test a subefficacious dose range, and the 120-mg/kg-per-day dose tests the effects of further elevated administration of the drug. Blood samples were taken before treatment (Day 1), 24 h after initiation of treatment (Day 31), after two 7-day cycles (Days 38 and 66 for cycle 1 and cycle 2, respectively), and after a 2-week break (day 55). Plasma VEGF and PlGF exhibited marked elevations in the 60- and 120-mg/kg-per-day groups immediately after completion of both cycles. Levels of plasma VEGF and PlGF returned to baseline after cessation of treatment (Fig. 2 A and B). Conversely, sVEGFR-2 levels decreased in 60- and 120-mg/kg-per-day groups immediately after completion of both treatment cycles and returned to baseline after treatment cessation (Fig. 2C). These results in mice recapitulate respective patterns of VEGF, PlGF, and sVEGFR-2 changes in patients treated with sunitinib, including reversal of drug-induced changes off therapy, but indicate that these effects do not depend on the presence of tumor. In addition, levels of VEGF, PlGF, and sVEGFR-2 remain unchanged in mice treated with 7.5-, 15-, and 30-mg/kg-per-day doses, demonstrating that the modulation of VEGF, PlGF, and sVEGFR-2 levels coincide remarkably well with the previously determined optimal biological dose range for antitumor efficacy.

Fig. 2.

Sunitinib-mediated modulations of VEGF, PlGF, and sVEGFR-2 are reversible. Balb/C mice were treated with 7.5, 15, 30, 60, and 120 mg/kg sunitinib daily for two 7-day cycles separated by a break period. Blood was taken retroorbitally before treatment, 24 hours after treatment initiation, after both 7-day treatment cycles, and after the break period. Increased VEGF (A) and PlGF (B), along with decreased sVEGFR-2 (C), were seen after both cycles in 60 and 120 mg/kg groups but returned to baseline after treatment cessation. R/O = Retro-Orbital bleed, C/P = Cardiac Puncture bleed, Tx = Treatment. n = 5 for all groups. Symbols and bars, mean ± SD.

Plateaus of VEGF, PlGF, and sVEGFR-2 After Daily Sunitinib Administration.

To determine the kinetics of sunitinib-dependent changes, Balb/C mice were administered daily doses of 60 mg/kg sunitinib with concurrent daily analysis of selected plasma protein levels in alternating mouse groups. After 7 days, treatment was stopped in a subset of groups to assess the kinetics of protein change reversal. Peak levels of VEGF and PlGF, along with the sVEGFR-2 nadir, were observed after 4 days of drug administration (Fig. 3A, B, and C, respectively). Variability in the levels of VEGF and PlGF was seen after 4 days, whereas decreased levels of sVEGFR-2 were maintained during treatment. Return of plasma VEGF and PlGF levels to baseline levels occurred within 1–2 days, whereas sVEGFR-2 levels returned to baseline after 4–5 days.

Fig. 3.

Plateaus of VEGF, PlGF, and sVEGFR-2 after sunitinib treatment. Daily retroorbital blood measurements were taken from alternating groups of Balb/C mice treated with 60-mg/kg-per-day sunitinib to determine kinetics of protein level changes. Plateaus in increased VEGF (A) and PlGF (B), along with decreased sVEGFR-2 (C), were seen after 4 days of treatment and reversed after 2 (VEGF and PlGF) and 5 days (sVEGFR-2) when treatment was stopped. n = 3 for all groups. Symbols and bars, mean ± SE.

Multiple Plasma Proteins Modulated by Sunitinib Are Tumor-Independent and Coincide with Optimal Biological Dose Range.

To further assess levels of other proteins potentially altered after treatment with sunitinib, we used ELISA-based analysis on plasma derived from tumor-free Balb/C mice treated with sunitinib for 7 days at doses of 30, 60, and 120 mg/kg. Proteins tested include ligands for receptors inhibited by sunitinib, i.e., VEGF₁₆₅ (VEGFR-2, VEGFR-1), SCF (c-kit), PlGF (VEGFR-1), and PDGF-AB/PDGF-BB (PDGFR). Additionally, plasma was tested for levels of soluble receptors and other proteins potentially affected directly or indirectly by sunitinib, including sVEGFR-2, soluble tunica internal endothelial cell kinase 2 (sTIE-2), IL-6, SDF-1α, erythropoietin (EPO), G-CSF and osteopontin (OPN). Results show that plasma VEGF levels increased significantly at 60 mg/kg-per-day and 120 mg/kg per day but not at the 30-mg/kg-per-day dose (Fig. 4A). Plasma levels of SDF-1α significantly increased at 30, 60, and 120 mg/kg per day, whereas levels of SCF rise and sVEGFR-2 levels fall significantly at the 60- and 120-mg/kg-per-day doses (Fig. 4 B–D). Significant elevations were seen in G-CSF, sTIE-2, and OPN at the 120-mg/kg-per-day dose (Fig. 4 E–G). In contrast, no significant changes in plasma levels of IL-6 (Fig. 4I), PDGF-AB (Fig. 4J), or PDGF-BB (data not shown) were observed at any dose level of sunitinib tested. Fluctuations were seen in EPO levels at all doses, with significant increases in the 60-mg/kg-per-day dose group; however, low detectable levels of protein in all groups as measured by ELISA made observation of an overall trend difficult (Fig. 4H). Taken together, these results demonstrate dose-dependent modulation of VEGF, SDF-1α, SCF, sVEGFR-2, G-CSF, sTIE-2, and OPN plasma levels consistent with the previously determined optimal biological dose range for sunitinib. Other proteins potentially modulated by sunitinib were screened by comparing plasma taken from Balb/C mice treated with 120 mg/kg per day for 7 days to vehicle controls using a protein array designed to detect angiogenesis related proteins ranging from those with antiangiogenic activity to those with proangiogenic activity (SI Fig. 7A). Elevations in G-CSF and lack of change in IL-6 levels were observed in confirmation of previous ELISA results. Changes in leptin and IFN-γ were observed, indicating potential candidates for future study of dose-related changes (SI Fig. 7B).

Fig. 4.

Dose-dependent changes of multiple plasma proteins after sunitinib treatment in nontumor-bearing mice. Plasma of Balb/C mice treated with 30, 60, or 120 mg/kg sunitinib for 7 days were analyzed for (A) VEGF, (B) SDF-1α, (C) SCF, (D) sVEGFR-2, (E) G-CSF, (F) sTIE-2, (G) OPN, (H) EPO, (I) IL-6, and (J) PDGF-AB. All values are represented as percent of control. See Experimental Procedures for baseline levels for vehicle treated groups. Significant differences from control: *, 0.01<P < 0.05; **, 0.001<P < 0.01; ***, P < 0.001. n = 4 for all groups. For graphing simplicity, the x axis of E and J is shared by A–D and F–I, respectively. Symbols and bars, mean ± SD.

Modulation of VEGF and VEGFR-2 Levels in Mouse Tissues After Sunitinib Treatment.

To identify the possible source(s) of VEGF and sVEGFR-2 changes seen in the plasma, experiments were performed in nontumor-bearing mice treated with sunitinib to assess protein levels of VEGF and full-length VEGFR-2 in various tissues. Protein lysates were derived from homogenized tissues taken from Balb/C mice treated for 7 days with 120 mg/kg sunitinib and compared with lysates from control mice. VEGF was not significantly changed in the lung but was found to be significantly elevated in the liver, heart, spleen, kidney, and skin and decreased in the bone marrow (Fig. 5A). Levels of VEGFR-2 were found to decrease in the liver, heart, spleen, kidney, bone marrow, and the skin but not significantly in the lung (Fig. 5B). Concurrent studies with lysates derived from mice treated with vehicle and sunitinib (120 mg/kg per day) for 7 days were subjected to SDS/PAGE and Western blot analyses to confirm decreased VEGFR-2 expression in both spleen and heart (SI Fig. 8 A and B). Taken together, these results suggest that the rise in VEGF and the decrease in sVEGFR-2 seen after sunitinib treatment are the result of cellular expression level changes in multiple organs, likely contributing to the overall level changes seen in the plasma. As a further test to determine a potential source for these protein changes after drug administration, we tested the effect of sunitinib on whole blood extracted from mice to establish whether circulating cells, including platelets, monocytes, neutrophils, circulating endothelial cells, etc., known to influence angiogenic protein expression, could alter the levels of proteins described in Fig. 4. Previous studies established that the levels of circulating sunitinib 24 h after a 40 mg/kg dose were in the range of 0.01–1 μM sunitinib (13). We therefore incubated samples of whole blood with a range of sunitinib concentrations of 0.01, 0.1, 1, and 10 μM for 24 h at 4°C and tested plasma for these molecular markers. SI Fig. 9 shows that plasma levels of VEGF, sVEGFR-2, SDF-1α, SCF, G-CSF, OPN, EPO, PDGF-AB, and sTIE-2 did not change significantly, suggesting that the changes in these proteins seen are not the result of direct effects of sunitinib on circulating blood cells.

Fig. 5.

VEGF increase and VEGFR-2 decrease in tissue lysates after sunitinib treatment. Lysates made from homogenized tissues of Balb/C mice treated with 120 mg/kg sunitinib for 7 days and tested by ELISA. (A) Levels of VEGF significantly increased in liver, heart, spleen, kidney, and skin but not in bone marrow (BM) or lung. (B) Levels of VEGFR-2 in the same lysates show a significant decrease in expression in the liver, heart, spleen, kidney, BM, and skin. Significant differences from control: *, 0.01 < P < 0.05; **, 0.001 < P < 0.01; ***, P < 0.001. n = 4 for all groups. Values are shown as fold change compared with control and represent pg/ml of protein standardized to milligrams of total protein added to ELISA. Symbols and bars, mean ± SD.

Lack of Correlation Between Sunitinib-Modulated Protein Changes and Tissue Hypoxia.

To further study the potential underlying mechanism governing these drug-induced changes, we examined whether changes in the microenvironment, such as tissue hypoxia and/or increased expression of hypoxia-regulated genes such as hypoxia inducible factor 1 (HIF-1α), may contribute to the observed molecular modulation of plasma proteins after sunitinib treatment. Spleen samples taken from Balb/C mice treated for 7 days with 120-mg/kg-per-day sunitinib were subjected to immunofluorescent staining with antibodies to hypoxiprobe-1 and HIF-1α and found to have no differences after treatment (SI Fig. 10 A and B). Additionally, nuclear extracts were prepared from additional tissues, including heart (SI Fig. 10C), as well as spleen, liver, and kidney (data not shown), and subjected to SDS/PAGE and Western blotting analysis to measure HIF-1α expression. The lack of differences between treated and untreated groups suggests localized sunitinib-induced hypoxia or expression level changes in hypoxia-regulated genes such as HIF-1α are not likely the source of the molecular changes observed in the plasma.

Discussion

Our results show that the sunitinib-induced reversible increases in circulating plasma VEGF reported in renal cell carcinoma (6) and gastrointestinal stromal tumor patients (7) can be reproduced in tumor-bearing mice, but importantly, also in tumor-free mice. Furthermore, these induced tumor-independent changes in VEGF, as well as increases in PlGF and decreases in sVEGFR-2, are dose-dependent and correlate with antitumor activity of the drug. Additional dose-dependent changes were also found in other plasma markers, including cytokines and chemokines that bind to receptors not known to be targeted by the drug, e.g., G-CSF, SDF-1α, and osteopontin. Finally, based on studies with VEGF and VEGFR-2, it would appear that the observed changes are a consequence of a systemic multiorgan response to drug treatment. Although our studies presented here involve only one drug, the results likely generally apply in varying degrees to many other similar drugs that have been shown in cancer patients to induce elevations of plasma VEGF (and PlGF in many cases) and decreases in sVEGFR-2 (see SI Table 1).

There are a number of potential implications these results may have with respect to the clinical use of antiangiogenic RTKIs. First, they provide an explanation of why plasma marker changes such as increases in VEGF or PlGF have not been found to be predictive of tumor response or clinical benefit (6, 7). Our results would suggest that drug treatment, even if devoid of significant antitumor/antiangiogenic activities in a patient, may lead to extratumoral increases in plasma VEGF and PlGF, likely to a similar extent in all treated patients and therefore potentially masking any differences in responding/nonresponding patients. However, the changes may be useful in other ways, such as drug activity markers, because the changes are reflective of systemic drug exposure. As such, a second implication relates to the determination of an optimal biological dose range for a drug like sunitinib, which may be assisted by dosing to maximum increases of plasma VEGF/PlGF (and decreases in sVEGFR-2) in a conceptually similar manner as using dose-limiting toxicities, e.g., monitoring neutrophil counts to define a maximum tolerated dose (MTD) for certain chemotherapeutic drugs. In this regard, our results suggest that the optimal biological dose of sunitinib, and likely other antiangiogenic RTKIs, may coincide with the MTD. However, that these plasma marker changes and others we have detected occur in the same dose range could be indicative of systemic exposure to drug and represent a “threshold” for efficacy. Examining whether this threshold is reached in patients could be exploited as an indicator of dosing within an efficacious antitumor range, independent of disease stage or type. Such a threshold dose, if monitored in the clinic, could serve as a guide for further dose escalation, with current toxicity profiles remaining in place, or have utility in clinical trials testing lower doses of sunitinib administered in a continuous fashion (clinicaltrials.gov; Identifier: NCT00338884).

Given the breadth and diversity of the plasma marker changes we have detected and their multiorgan origin, there are several possible important implications for antiangiogenic therapy using drugs such as sunitinib that should be considered for future studies. The first is whether any of the changes may be relevant to a particular toxicity commonly associated with the use of such drugs, such as hypertension, extreme fatigue, diarrhea, nausea, and hand-foot syndrome, among others (3, 6, 12). If so, one or more of the changes might be exploited as a predictive marker for such toxicities. This would be of particular interest with respect to sunitinib-induced neutropenia and thrombocytopenia. It is possible that such myleosuppressive effects may persist, even if similar elevations in G-CSF levels are observed, because of sunitinib activity against Kit and FLT3, both of which play critical roles in several aspects of hematopoiesis (14), as well as possible inhibition of VEGF-dependent intracellular activation of various bone marrow cells, including hematopoietic stem cells (15).

A second implication concerns acquired resistance to such drugs and/or rapid relapse seen off therapy either when using discontinuous dosing schedules or when therapy is terminated altogether (16). With respect to acquired resistance, Casanovas et al. (17) recently reported that the antitumor efficacy of DC101, an anti-VEGFR-2 monoclonal antibody, began to wane in treated mice within 1 month of continuous therapy. Loss of response was associated with up-regulation of several proangiogenic growth factors, apparently secondary to DC101-induced tumor hypoxia, one of which, basic FGF, was found responsible for resumption of tumor angiogenesis and tumor growth (17). Our results suggest that a similar process can also occur independently of tumor with drug-induced up-regulation of multiple proangiogenic growth factors. In this regard, a number of studies have shown that angiogenesis can be induced or amplified by G-CSF (18), SCF (19), SDF-1α (20), and osteopontin (21), by diverse mechanisms, including mobilization of circulating bone-marrow-derived proangiogenic cell populations (22), such as endothelial progenitor cells (23). Such effects could conceivably contribute to the rapid vascular “rebound” in mouse tumors observed within 1 week of terminating therapy with an RTKI similar in nature to sunitinib (16), results that could clearly have implications for whether a continuous vs. discontinuous schedule of treatment using such drugs would be superior. In this regard, discontinuous schedules in particular may inadvertently contribute to the “conditioning” of certain organ environments for metastatic tumor growth. Likewise, the results may be pertinent to situations where drugs such as sunitinib are used for neoadjuvant therapy of early-stage primary resectable tumors.

With respect to the broad protein profile changes seen after sunitinib treatment, elucidating the underlying mechanism(s) may prove difficult, particularly because multiple targets and target-dependent mechanisms may play a role. Although we can broadly classify the markers tested here as being directly target-related (such as VEGF, PlGF, SCF, or sVEGFR-2) and indirectly or collaterally affected (such as SDF-1α, G-CSF, OPN, and sTIE-2) by sunitinib, that they may potentially interact and influence each other makes distinguishing the nature of the underlying change inherently difficult. VEGF and SDF-1α, for example, can regulate each other (24, 25); PlGF can influence VEGF and VEGFR-2 expression (26); G-CSF can alter SDF-1α (27) and VEGF levels (28); and there may be other levels of complexity. One potential underlying mechanism we studied is hypoxia. Because a number of the observed proteins, including VEGF, VEGFR-2, PlGF, OPN, SDF-1α, and PDGF-BB, are known to be regulated by changes in the microenvironment, such as hypoxia/HIF-1α (29), we investigated whether hypoxia may contribute to molecular modulation of plasma proteins after sunitinib treatment seen in various tissues. However, such studies yielded no supporting evidence of differential expression levels (SI Fig. 10). Perhaps studies with sunitinib in transgenic animals, such as mice heterozygous for a null (knockout) allele at the Hif1a locus encoding HIF-1α (30), may be useful in addressing this question. However, it may be that more complex mechanisms are the basis of this phenomenon.

In summary, our results provide a previously undescribed perspective regarding the properties of certain antiangiogenic drugs to induce multiple changes in circulating plasma proteins and how such changes may be exploited as pharmacodynamic activity markers. It will be of considerable interest to evaluate in detail the spectrum and extent of such changes with other antiangiogenic drugs, including both small-molecule RTKIs and antibodies that target VEGF or VEGFRs. In this regard, we have previously reported that anti-VEGFR-2 antibodies can dose-dependently increase circulating plasma VEGF levels in normal mice by what appears to be a fundamentally different mechanism, in that a single injection of the drug can induce such changes within 2 hours (31), whereas sunitinib treatment produces plasma VEGF plateaus after several days. We have also begun to analyze plasma marker changes in normal mice using other antiangiogenic drugs. For example, we have found that nontumor-bearing mice receiving vandetanib (AZD6474/Zactima) also produced dose-dependent elevations in VEGF, PlGF, SDF-1α, and G-CSF, and declines in sVEGFR-2 (data not shown). As well, we found that cetuximab (Erbitux), an antagonizing antibody against EGF receptor, can cause dose-dependent increases in tumor-derived TGFα (A. Mutsaers, personal communication). Taken together, these studies confirm other VEGF RTKIs can have similar effects and suggest that such changes may also apply to antibodies and RTKIs in other drug classes as well, including those targeting EGF receptor. Interestingly, in similar studies with PTK787 and SU5416, neither were found to induce significant changes except for high doses, too toxic for repeated administration (32). It is of interest that both these drugs did not induce clinical benefit in randomized phase III clinical trials in contrast to sunitinib, perhaps an indication that the induced changes in plasma markers we have described could reflect the relative potency of such drugs.

Materials and Methods

Drugs.

SU11248/sunitinib malate (Sutent, Pfizer) was suspended in carboxymethylcellulose vehicle formulation, containing carboxymethylcellulose sodium (0.5% wt/vol), NaCl (1.8% wt/vol), Tween 80 (0.4% wt/vol), benzyl alcohol (0.9% wt/vol), and reverse osmosis deionized water (added to final volume) and adjusted to pH 6.0. Drug aliquots were prepared once weekly and kept in the dark at 4°C. Mice were gavaged once daily. Bevacizumab (Avastin) was generously provided by Genentech and diluted in sterile PBS before i.p. injection of 300 μg per mouse twice weekly (9).

Animals and Cell Lines.

The human epidermoid carcinoma A431, human colorectal adenocarcinoma Colo205, human prostate cancer PC-3, and rat glioma C6 cell lines were obtained from the American Type Culture Collection (Manassas, VA) and maintained in either DMEM or RPMI medium 1640 with 5% FBS (Gibco Invitrogen, Burlington, ON, Canada) while incubated at 37°C and 5% CO₂ in a humidified incubator. Cells were injected (2 × 10⁶ cells) s.c. into the flank of 6- to 8-week-old female CB-17 SCID mice (Charles River Breeding Laboratory, Montreal, QC, Canada). PC-3 cells were transfected to overexpress human VEGF₁₆₅, designated “PC3^VEGF-HIGH” (kind gift of G. Francia, Sunnybrook Hospital, Toronto, ON, Canada) and were used to ensure detectability of the human (tumor-derived) VEGF by ELISA in mouse plasma. Experimental details with PC3^VEGF-HIGH cells are detailed in SI Text. Institutional guidelines were strictly followed for maintenance of animals and end point of tumor studies. Female age-matched Balb/C mice (Charles River Breeding Laboratory) were used for nontumor studies.

Plasma Collection.

Blood samples, obtained from the retroorbital sinus or by cardiac puncture of mice under anesthesia with isofluorane, were collected in Microtainer (Becton Dickinson, Franklin Lakes, NJ) plasma separating tubes and centrifuged at 4°C. Plasma was immediately pooled, aliquoted, frozen, and stored at −70°C until assayed. All capillary tubing, syringes, and needles used for bleeding were first rinsed with heparin to avoid any clotting.

ELISA Analysis.

Levels of murine VEGF, sVEGFR-2, PlGF, SDF-1α, OPN, SCF, IL-6, PDGF-AB, PDGF-BB, EPO, sTIE-2, and G-CSF were assessed by using commercially available mouse sandwich ELISAs (R&D Systems, catalog nos. MMV00, MVR200, MP200, MCX120, MCK00, M6000B, MHD00, MBB00, MEP00, MTE200, and MCS00, respectively), following the manufacturer's instructions with the exceptions outlined in SI Text. Detection of human VEGF in the plasma of human tumor-bearing mice was performed by using the human VEGF ELISA kit (R&D Systems, catalog no. DVE00), following the manufacturer's instructions. For comparison purposes, protein levels in Fig. 2 are represented as percent of control. Average plasma levels for vehicle-treated Balb/C mice used as controls for this study are outlined in SI Text.

Cell Lysis and Protein Extraction.

Tissues extracted at time of death were rinsed with ice-cold PBS and then sectioned with scalpel. Fragments were washed once with ice-cold PBS and centrifuged at 100 × g for 10 min. For whole-cell lysates, tissue fragments were then resuspended in cold lysis buffer (20 mM Tris, pH 7.5/137 mM NaCl/100 mM NaF/10% glycerol/1% Nonidet P-40/1 mM Na₂VO₄, supplemented with 1 mM PMSF/10 μg/ml aprotinin/10 μg/ml leupeptin) as described (4). Details for lysate preparation and protein quantification are outlined in SI Text.

Statistical Analysis.

Results were subjected to statistical analysis by using the GraphPad Prism software package, Ver. 4.0 (GraphPad, San Diego, CA). One-way ANOVA was followed by the Student–Newman–Keuls test; two-way ANOVA was followed by the Bonferroni test, and nonlinear regression curves fit using the Boltzmann method. The levels of significance were set at *, 0.01 < P < 0.05; **, 0.001 < P < 0.01; ***, P < 0.001.

Additional Methods are described in SI Text.

Abbreviations

RTK

receptor tyrosine kinase

RTKI

RTK inhibitor

PlGF

placental growth factor

VEGFR-1/2

VEGF receptor-1/2

sVEGFR

soluble VEGFR

HIF

hypoxia-inducible factor.