Soluble receptor-mediated selective inhibition of VEGFR and PDGFRβ signaling during physiologic and tumor angiogenesis

Abstract

The simultaneous targeting of both endothelial cells and pericytes via inhibition of VEGF receptor (VEGFR) and PDGFβ receptor (PDGFRβ) signaling, respectively, has been proposed to enhance the efficacy of antiangiogenic tumor therapy. Clinical and preclinical modeling of combined VEGFR and PDGFRβ signaling inhibition, however, has used small molecule kinase inhibitors with inherently broad substrate specificities, precluding detailed examination of this hypothesis. Here, adenoviral expression of a soluble VEGFR2/Flk1 ectodomain (Ad Flk1-Fc) in combination with a soluble ectodomain of PDGFRβ (Ad sPDGFRβ) allowed highly selective inhibition of these pathways. The activity of Ad sPDGFRβ was validated in vitro against PDGF-BB and in vivo with near-complete blockade of pericyte recruitment in the angiogenic corpus luteum, resulting in prominent hemorrhage, thus demonstrating an essential function for PDGF signaling during ovarian angiogenesis. Combination therapy with Ad PDGFRβ and submaximal doses of Ad Flk1-Fc produced modest additive antitumor effects; however, no additivity was observed with maximal VEGF inhibition in numerous s.c. models. Notably, VEGF inhibition via Ad Flk1-Fc was sufficient to strongly suppress tumor endothelial and pericyte content as well as intratumoral PDGF-B mRNA, obscuring additive Ad sPDGFRβ effects on pericytes or tumor volume. These studies using highly specific soluble receptors suggest that additivity between VEGFR and PDGFRβ inhibition depends on the strength of VEGF blockade and appears minimal under conditions of maximal VEGF antagonism.

The central role of angiogenesis in promoting tumor progression and metastasis is now well appreciated and is governed by a balance between stimulators and inhibitors of angiogenesis, as proposed by Hanahan and Folkman (1). VEGF-A signaling constitutes the dominant regulatory pathway for developmental and tumor angiogenesis (2, 3). In randomized clinical trials, the humanized monoclonal anti-VEGF-A antibody Avastin provided an overall survival benefit for colon cancer patients when combined with conventional chemotherapy (4), thus validating angiogenesis inhibition as an effective anticancer strategy.

Targeting endothelial cells (ECs) or the principal angiogenic factor VEGF-A alone may not be itself sufficient for maximal antiangiogenic effects (2). Progressive and refractory disease often supervenes in patients, perhaps mediated by VEGF resistance via the production of alternative angiogenic mediators such as FGFs and reactivation of tumor angiogenesis (5–7). Additionally, non-EC types such as mural, stromal, and hematopoietic cells positively contribute to tumor angiogenesis, albeit indirectly (2, 8). Pericytes (PCs), in particular, have been recently appreciated as critical regulators of vessel formation, stabilization, and function (9). The amount of PC coverage of vessels for various tumors ranges from extensive to none (10). Nevertheless, the protective and stabilizing functions of PCs on tumor vasculature have been clearly demonstrated (11–13).

PDGFβ receptor (PDGFRβ) signaling is critical for the recruitment of PCs or perivascular progenitor cells during developmental and tumor angiogenesis (12, 14, 15). EC-derived PDGF-BB, as the classical PDGFRβ ligand, is essential for embryonic PC recruitment (16), and a requisite role of local extracellular PDGF-BB concentration gradients for proper integration of PDGFRβ-positive PCs into the wall of tumor vessels has been demonstrated (17). As an alternative pathway, angiopoietin 1 (Ang1) and its receptor Tie2 have been implicated as mediators of blood vessel stability and mural cell recruitment during embryonic (18, 19) and tumor angiogenesis (20).

The tumor vessel-stabilizing functions of PCs have prompted intensive efforts to target both tumor vasculature PCs and ECs for combinatorial antiangiogenesis therapy. Combination treatment with small molecule receptor tyrosine kinase inhibitors (RTKI) such as SU6668 and Gleevec targeting VEGF and PDGFRβ signaling, respectively (21, 22), produced increased efficacy as compared with either monotherapy at all stages of pancreatic islet cancer and achieved regression of late-stage tumors in the RIP-Tag2 model (23, 24); similar additivity has been reported for the VEGF inhibitor SU10944 and Gleevec (25). Most recently, in human clinical trials, broad-spectrum small molecule RTKIs such as sorafenib and sunitinib targeting both VEGF receptor 1–3 (VEGFR1–3) and PDGFRα/β among others have demonstrated efficacy in renal cell carcinoma, gastrointestinal stromal tumor (GIST), and hepatocellular carcinoma (26–28). However, the use of relatively promiscuous RTKIs does not allow inhibitory effects to be unequivocally ascribed to PDGFRβ inhibition, and highly specific reagents are needed to test the PC hypothesis. Further, the potential ability of VEGF to stimulate endothelial PDGF-BB expression (29) and the extensive reciprocal cross-talk between ECs and PCs (9) suggest that inhibition of these two pathways might not be fully additive or synergistic.

We previously used adenovirus to achieve reversible and systemic expression of soluble VEGFRs in mice, producing stringent VEGF inhibition and potent inhibition of tumor growth and angiogenesis (30, 31). Here, we generated adenoviruses expressing soluble PDGFRβ ectodomains (Ad sPDGFRβ) to inhibit PC-recruitment pathways during physiological and pathological angiogenesis and elucidate the specific contributions of inhibition of VEGFR and PDGFRβ signaling during combination antiangiogenic cancer therapy.

Results

Characterization of Adenoviruses Expressing Soluble PDGFRβ Ectodomains.

We previously demonstrated extensive antiangiogenic and antitumor efficacy after adenoviral expression of soluble ectodomains of the VEGFRs Flt-1 (Ad Flt1) and Flk1 (Ad Flk1-Fc) (30, 31). Accordingly, we constructed an analogous adenovirus expressing a soluble PDGFRβ ectodomain with C-terminal His₆ tag (Ad sPDGFRβ). Single i.v. administration of purified Ad sPDGFRβ (5 × 10⁸ pfu) to adult C57BL/6 mice produced robust plasma expression of the respective ectodomains from hepatic transduction for >30 days (Fig. 1A), consistent with the typical expression kinetics of adenovirus in immunocompetent mice (31). The soluble recombinant PDGFRβ ectodomain was purified from adenoviral supernatants by using the C-terminal His₆ tag and potently inhibited PDGF-BB-induced Akt phosphorylation in NIH 3T3 cells (Fig. 1B) (32). As an additional control, we constructed Ad Tie2-Fc expressing a soluble Tie2 ectodomain fused to IgG2α Fc. Ad Tie2-Fc also produced sustained plasma expression in C57BL/6 mice (Fig. 1C), and the corresponding purified Tie2-Fc from adenovirus supernatant neutralized Ang1-stimulated Akt phosphorylation in human umbilical vein ECs (HUVECs) (Fig. 1D).

Fig. 1.

Characterization of PDGFRβ and Tie2-Fc adenoviruses. Tail vein i.v. administration of Ad sPDGFRβ (A) or Ad Tie2-Fc (C) (10⁹ pfu) to adult mice produced plasma expression of their respective ectodomains for >30 days as determined by Western blotting. Affinity-purified PDGFRβ-His (200 ng/ml) and Tie2-Fc (5 μg/ml) ectodomains from adenoviral supernatants inhibited PDGF-BB- (10 ng/ml) and angiopoietin-1- (Ang1, 200 ng/ml) induced Akt phosphorylation in NIH 3T3 and HUVEC cells, respectively (B and D). IDE,insulin-degrading enzyme.

Ad sPDGFRβ Induces Hemorrhage During Corpus Luteum (CL) Angiogenesis.

The effects of Ad sPDGFRβ on physiologic angiogenesis were validated in a murine ovarian CL angiogenesis model. After ovulation, the basement membrane between the theca and granulosa layers undergoes rapid dissolution, and the thecal vessels invade the avascular granulosum and form a dense vascular network surrounding the luteinized granulosa cells to sustain CL development and function, accompanied by robust endothelial proliferation and PC recruitment (33, 34). Using this model, we administered Ad sPDGFRβ i.v. 24 h before hormonal induction of ovulation [human choriogonadotropin (hCG) injection], and we examined the ovaries 48 h after ovulation. As comparative controls for Tie2 and VEGF blockade, Ad Tie2-Fc and Ad Flk1-Fc were administered in parallel, and Ad Fc encoding an IgG2α Fc fragment was used as negative control.

Ovaries from Ad Fc (Fig. 2A, E, and I) and Ad Tie2-Fc-treated animals (Fig. 2 C, G, and K) appeared grossly and histologically normal, whereas VEGF inhibition with Ad Flk1-Fc resulted in a marked reduction in ovary size (Fig. 2B). These Ad Flk1-Fc-treated ovaries contained mostly small antral follicles, only a very few fully developed CL [Fig. 2F; supporting information (SI) Fig. S1] with occasional central necrosis (Fig. 2J), consistent with previous reports documenting an essential role for VEGF in CL development (35).

Fig. 2.

Effect of Ad sPDGFRβ on CL angiogenesis. Prepubescent female mice were administered the indicated Ads (5 × 10⁸ pfu) 1 day before hCG-induced ovulation. As demonstrated in whole-mount images of ovaries, Ad sPDGFRβ induces prominent CL hemorrhage (D) compared with Ad Fc control-treated animals (A). Histological analysis revealed a large number of arrested follicles displaying large central cavities entirely filled with blood and widespread hemorrhaging throughout the luteal tissue in developed CL only in Ad sPDGFRβ-treated mice (H and L vs. Ad Fc, and E and I). Ad Flk1-Fc (B, F and J) but not Ad Tie2-Fc (C, G and K) suppresses CL development. (Magnification: A–H, ×25; I–L, ×200.)

In contrast, ovaries from the Ad sPDGFRβ treatment group, although comparable in size to the Ad Fc control, displayed prominent multifocal hemorrhage (Fig. 2D). In the Ad sPDGFRβ-treated ovaries, the number of fully developed CL was markedly reduced, and invariably, these CL displayed extensive luteal hemorrhage (Fig. 2 H and L; Fig. S1). Furthermore, numerous blood-filled follicles that had failed to ovulate exhibited extensive hemorrhage in peripheral luteinized granulosa cell layers (Fig. 2H).

Ad sPDGFRβ Potently Suppresses CL PC Recruitment.

Strikingly, Ad sPDGFRβ treatment resulted in a 93% reduction of CL NG2⁺ PC content, without alterations of PC density in the preexisting thecal vasculature (Fig. 3H, L, N, and O). The lack of PC investment in the developing Ad sPDGFRβ CL vasculature was confirmed by using desmin as a second PC marker (Fig. 3O). CL PCs did not display immunoreactivity for smooth muscle actin (SMA) under either basal or inhibited conditions, whereas immunostaining for PDGFRβ demonstrated complete tissue penetration of ovarian tissue with the soluble PDGFRβ ectodomain, allowing for the neutralization of local PDGF-BB gradients (36) (data not shown). Taken together, the quantitative Ad sPDGFRβ-mediated suppression of PC recruitment during physiologic CL angiogenesis demonstrated both the potency of this reagent and the requirement of PDGFRβ-mediated PC recruitment for development of the CL and its associated vasculature. In contrast, oral administration of Gleevec, an Abl-inhibitory RTKI with activity against PDGFRβ (37), engendered smaller effects on CL PC coverage and CL morphology, suggesting lesser PDGFR antagonism (Fig. S2).

Fig. 3.

Suppression of CL PC recruitment by Ad sPDGFRβ. (A–N) Immunofluorescence staining of CL sections for ECs (CD31, green) and PCs (NG2, red) was performed on ovaries obtained from female mice (n = 4) pretreated with the indicated adenoviruses before ovulation. Ad sPDGFRβ potently inhibits the recruitment of NG2-positive PCs to the nascent CL vasculature as compared with Ad Fc control virus, without affecting the preestablished thecal and stroma vasculature (H and N vs. E and M). Ad Tie2-Fc demonstrates minimal effect on PC content (G), whereas Ad Flk1-Fc reduced ECs (B and J), with resultant secondary inhibition of PC recruitment (F). (Magnification: A–L, ×100; M and N, ×400.) (O) Differential effects on ECs and PCs in adenovirus-treated CL. EC (CD31) and PC content (NG2, desmin) were determined from immunostained images by automated pixel quantitation (n = 4 mice per virus). Ad Flk1-Fc potently suppressed endothelial CD31 immunoreactivity with consequent lack of NG2 signal, whereas Ad sPDGFRβ selectively inhibited PC recruitment (, P < 0.02 vs. Ad Fc) with a trend toward suppression of CD31 signal (P > 0.05 vs. Ad Fc). Sections are from the experiment in Fig. 2.

The PC-selective Ad sPDGFRβ phenotype contrasted with VEGF inhibition via Ad Flk1-Fc, which induced near-total suppression of luteal immunoreactivity for both the EC marker CD31 and the PC marker NG2 (Fig. 3 B, F, J, and O). The potent Ad Flk1-Fc inhibition of both ECs and PCs illustrates the strong dependence of PC recruitment on EC function. Ad sPDGFRβ reduced CL CD31⁺ microvessel density by ≈45% (Fig. 3 D and O), suggestive of significant reciprocal regulation of EC by PDGFRβ⁺ PCs. Inhibition of Tie2 signaling by Ad Tie2-Fc reduced CL CD31⁺ microvessel density by 32% without significantly decreasing PC content, further indicating the specificity of the Ad sPDGFRβ phenotype (Fig. 3 C, G, K, and O).

sPDGFRβ Ectodomains Inhibit PC Recruitment to Nascent Tumors.

Recent studies suggested a requisite role of PDGFRβ signaling for mural cell recruitment to the tumor microvasculature (17, 38), although fully conditional strategies were not used. Pretreatment of C57BL/6J mice with Ad sPDGFRβ 24 h before s.c. Lewis lung carcinoma (LLC) implantation dramatically decreased the PC content of tumor microvessels (≈95%), as evidenced by SMA and desmin staining, while producing a more modest ≈20% reduction in CD31⁺ endothelial content (Fig. 4). In contrast, Ad Flk1-Fc tumors displayed a >90% reduction in CD31⁺ microvessel density accompanied by an ≈80% decrease in PC content (Fig. 4), paralleling results in the CL (Fig. 3) and again indicating an obligate suppression of PC after primary EC inhibition. Ad Tie2-Fc reduced tumor CD31⁺ microvessel density by ≈20% without significant effects on PC (Fig. 4).

Fig. 4.

Effect of Ad sPDGFRβ on pathological angiogenesis in a pretreatment LLC model. Adenoviruses (5 × 10⁸ pfu) were administered 1 day before the s.c. implantation of LLCs in C57BL/6J mice. Tumors were excised 12 days after implantation, and immunostaining for CD31 and PC markers α-SMA and desmin was performed (A). Ad sPDGFRβ ectodomains potently suppress the recruitment of PCs to the growing tumor vasculature (n = 4; *, P < 0.03; #, P < 0.01 vs. Ad Fc) (B). (Magnification in A: ×100.)

Comparative and Combinatorial Antitumor Effects of Soluble PDGFRβ and VEGFR Ectodomains.

We further exploited the efficacy and specificity of these adenoviral reagents to test the effects of combined blockade of VEGFRs and PDGFRβ on tumor growth and angiogenesis in preestablished (200-mm³) s.c. tumor models. Two different doses of Ad Flk1-Fc were used: a submaximal dose (7.5 × 10⁷ pfu for T241, 1 × 10⁸ pfu for LLC) and a maximal dose (5 × 10⁸ pfu), associated with <25% and 60–70% tumor inhibition, respectively, as monotherapy. We observed additive effects between submaximal doses of Ad Flk1-Fc and maximal doses of Ad sPDGFRβ in both the LLC and the T241 fibrosarcoma tumor models (Fig. 5A and Fig. S3). However, when the stringency of VEGF inhibition was increased via maximal doses of Ad Flk1-Fc (5 × 10⁸ pfu), no additivity was observed with Ad sPDGFRβ (5 × 10⁸ pfu) in (Fig. 5B and Fig. S4). These results indicated that the beneficial effects of combined VEGF/PDGFRβ inhibition are minimal under conditions of stringent VEGF blockade.

Fig. 5.

Effect of combination antiangiogenic therapy on preestablished tumor growth. (A) Under conditions of suboptimal inhibition of VEGF signaling (Ad Flk1-Fc, 1 × 10⁸ pfu for LLC, 7.5 × 10⁷ pfu for T241), the addition of Ad sPDGFRβ produces additive antitumor effects on LLC (n = 16) and T241 (n = 7) tumor growth vs. Ad sPDGFRβ or Ad Flk1-Fc monotherapies. Final tumor volumes are presented (*, P < 0.02 vs. Ad Fc). (B) Adult mice bearing LLC (n = 7 per group); T241 fibrosarcoma (n = 6/group), or B16Bl6 (n = 7 per group) tumors were treated with a single adenovirus Ad Fc (10⁹ pfu) or with adenovirus combinations, Ad Flk1-Fc/Ad Fc, Ad sPDGFRβ/Ad Fc, or Ad Flk1-Fc/Ad sPDGFRβ (5 × 10⁸ pfu each) simultaneously. In all three tumor models, the combination of stringent VEGF blockade via Ad Flk1-Fc with PC-targeting Ad sPDGFRβ did not augment the inhibition of tumor growth afforded by Ad Flk1-Fc alone. Only the Ad Flk1-Fc-containing therapies significantly reduced tumor volume relative to Ad Fc (P < 0.008 for all three tumor models). *, P < 0.008 vs. Ad Fc. (C) Effect of combination antiangiogenic therapy on tumor vasculature. Endothelial (CD31) and PC (SMA) content was quantitated in LLC tumors treated with the indicated Ads. Ad Flk1-Fc/Ad sPDGFRβ combination therapy does not further increase the reduction of LLC microvessel density achieved by Ad Flk1-Fc treatment (n = 4; *, P < 0.05). (D) Quantitative real-time PCR analysis of PDGF-B mRNA in nascent T241 tumors. Stringent inhibition of VEGF signaling by Ad Flk1-Fc leads to a significant reduction of total tumor PDGF-B mRNA expression, accompanying reduced PC content in Ad Flk1-Fc-treated tumors (n = 4; #, P < 0.03).

Likewise, combination Ad Flk1-Fc and Ad Tie2-Fc treatment or even triple treatment combinations with Ad sPDGFRβ did not elicit additive inhibitory effects on tumor growth as compared with Ad Flk1-Fc monotherapy (Fig. S4). As monotherapy, Ad sPDGFRβ was less efficacious than Ad Flk1-Fc (37–45% versus 62–75% inhibition). Ad Tie2-Fc was ineffective in all three models (Fig. S4), and adenoviruses expressing Tie1, ephrin-B2, or EphB4 ectodomains or full-length Ang1 or Ang2 exhibited modest to no efficacy in LLC (B.Y.Y.T. and C.J.K., unpublished work).

Combination PDGFRβ and VEGFR2 Ectodomain Effects on Tumor Vasculature.

To explore the basis for the lack of additivity between high-dose VEGFR/PDGFRβ inhibition, effects on the tumor vasculature were examined in these preestablished tumor models. Ad Flk1-Fc treatment of LLC reduced CD31⁺ EC by ≈80% (Fig. 5C). Ad sPDGFRβ treatment reduced CD31⁺ content by ≈45% (Fig. 5C), consistent with the notion that the targeting of PC recruitment pathways alone can suppress EC via cross-talk. Paralleling tumor growth, combined Ad Flk1-Fc and Ad sPDGFRβ treatment did not result in additional reduction of CD31⁺ EC content versus Ad Flk1-Fc alone. Although Ad sPDGFRβ reduced SMA⁺ PC coverage by 70%, Ad Flk1-Fc was itself sufficient to decrease PC content by ≈60%, with lack of statistically significant additive effects on PC via the combination (P = 0.59 vs. Ad sPDGFRβ, P = 0.1 vs. Ad Flk1-Fc) (Fig. 5C), providing a mechanistic explanation for the observed lack of additivity on tumor growth.

The Ad sPDGFRβ-induced decreases in ECs coupled with the Ad Flk1-Fc-induced decreases in PCs are consistent with the notion of substantial and mutually reinforcing cross-talk between VEGFR⁺ ECs and PDGFRβ⁺ PCs. These observations also suggest a potential mechanism for the observed lack of additivity on the combination of PDGFRβ and VEGF inhibition because the latter already strongly inhibits PC recruitment. Indeed, in T241 tumors, for which ECs but not tumor cells themselves constitute the predominant source of PDGF-BB (39), Ad Flk1-Fc pretreatment significantly (≈90%) reduced PDGF-B RNA (Fig. 5D). Thus, VEGF inhibition can indirectly inhibit PC recruitment via primary suppression of ECs and accompanying endothelial PDGF-BB production in the absence of PDGFRβ-targeted inhibition. We additionally observed lack of additivity on EC or PC content between triple combinations of Ad Tie2-Fc, Ad sPDGFRβ, and Ad Flk1-Fc vs. Ad Flk1-Fc monotherapy (Fig. S5).

Discussion

The participation of PDGF-BB and its cognate receptor PDGFRβ during embryonic and adult PC recruitment has rendered this pathway an extremely attractive target for antiangiogenic cancer therapy, particularly in concert with VEGF antagonism (23, 24). However, examination of the roles of PDGFRβ signaling in both tumor angiogenesis and normal physiology has been hampered by the lack of facile and robust methods for specific, conditional inactivation of this receptor system in adult organisms. In the current work, we addressed this question by using adenoviral expression of ligand-binding ectodomains of PDGFRβ, first validating bioactivity in a CL angiogenesis assay before combinatorial evaluation with VEGF inhibition in tumor models.

Prior methodology for inhibiting PDGF-BB/PDGFRβ-mediated recruitment of PCs in the adult has included EC-specific deletion of PDGF-B, or fine genetic deletion of the PDGF-B C-terminal heparin-binding retention motif has circumvented embryonic lethality of PDGF-B-null mutants (17, 36, 40, 41). However, these deficits originate in embryogenesis and may not generate fully null phenotypes given the potential for incomplete deletion in the former and neomorphic or hypomorphic effects in the latter. Pharmacologic approaches have included either small molecule inhibitors of PDGFRβ tyrosine kinase activity or anti-PDGFRβ antibodies (12, 23, 42), although these have potential disadvantages of promiscuous action against multiple RTKs in the former and pharmacokinetic limitations with the need for frequent and/or repetitive administration in both cases. Nonetheless, the aforementioned strategies have provided compelling evidence for therapeutic inhibition of PDGFRβ signaling during tumor angiogenesis and for PC developmental hierarchies.

Adenoviral administration of PDGFRβ ectodomains represents an alternative and convenient strategy for adult inactivation, with persistent and high-level circulating transgene levels from single i.v. injection. Inherently, the use of soluble PDGFRβ as decoy, although affording specific blockade of PDGFRβ signaling, does not distinguish between effects mediated by sequestration of either PDGFRβ ligand, PDGF-BB or PDGF-DD (43). Nonetheless, the adenoviral approach affords specific blockade of PDGF-B signaling, and the in vivo efficacy of Ad sPDGFRβ is strongly supported by near-quantitative PC suppression in the physiologic setting of CL angiogenesis. Single i.v. dosing of Ad sPDGFRβ reduced CL PC content by 93%, in contrast to partial inhibition by the multitargeted RTK inhibitor Gleevec. The blood vessel-destabilizing effects of impaired PC recruitment have been previously demonstrated during development and tumor angiogenesis (16, 17), and loss of PCs likewise leads to profuse hemorrhaging during CL formation, although contribution of secondary EC effects cannot be ruled out. PDGF-BB and PDGF-DD are abundantly expressed in the ovary (43, 44), and the prominent hemorrhage in Ad sPDGFRβ-treated CL strongly suggests an essential requirement for PDGFRβ-mediated signaling in providing vessel stability during CL angiogenesis and ovarian physiology. Moreover, our findings establish the CL angiogenesis model as a robust assay system to evaluate PDGF antagonists.

Combinatorial therapy targeting both PDGFRβ and VEGF is thought to produce improved antiangiogenic effects vs. either alone. Indeed, combination treatment with small molecule RTK inhibitors targeting VEGF (SU5416, SU10944) and PDGFRβ (SU6668, Gleevec) exhibited enhanced efficacy in the RIP-Tag pancreatic cancer model (23) and various xenograft models (25), although these relatively promiscuous RTKIs did not allow inhibitory effects to be unequivocally ascribed to VEGF or PDGFRβ inhibition.

Here, the potent and specific inhibition engendered by Ad sPDGFRβ afforded an opportunity to test the effects of combined PDGFRβ/VEGFR inhibition on tumor angiogenesis. Consistent with prior results with small molecule RTKIs (23, 25), we observed additive effects when combining Ad sPDGFRβ with submaximal doses of Ad Flk1-Fc. However, combined high-grade treatment with Ad sPDGFRβ and Ad Flk1-Fc did not produce appreciable synergistic antitumor effects when tumor burden was used as an endpoint. Stringent VEGF inhibition by Ad Flk1-Fc alone decreased PC number to a similar extent as Ad sPDGFRβ, suggesting a basis for the lack of additional effects from Ad sPDGFRβ. We further present evidence in the T241 fibrosarcoma model for a mechanism in which VEGF inhibition with primary suppression of ECs can suppress EC production of PDGF-BB with secondary consequences of impaired PC recruitment.

The current data do not exclude the potential for additivity in human patient populations. Ad sPDGFRβ enhanced the efficacy of submaximal doses of Ad Flk1-Fc (Fig. 5), a scenario that may parallel the clinical situation where near-complete inactivation of VEGF signaling may be difficult to achieve. We also cannot exclude class-specific additive effects in the RIP-Tag vs. subcutaneous models used herein, given the different mechanisms of action between small molecule inhibitors and soluble receptors. Interestingly, although the tumors used in the current studies express VEGF in the tumor parenchyma (Fig. S6), Ad sPDGFRβ suppresses CD31⁺ microvessel density more strongly than Ad Flk1-Fc in LLC variants in which VEGF is produced solely in tumor PCs and not the tumor parenchyma (B.S. and D.M.M., unpublished work). Ad sPDGFRβ exhibits strong activity in the B16Bl6 and CT26.CL25 metastasis models, suggesting the potential utility of PDGFRβ inhibitors such as orally bioavaliable small molecules for suppression of hematogenous metastases (F.K. and C.J.K., unpublished results).

ECs and PCs exhibit extensive bidirectional paracrine cross-talk, wherein endothelium-derived PDGF-BB acts on PCs, whereas reciprocal PC secretion of VEGF-A and Ang-1 stimulates ECs (16, 19, 45). In the current studies, we provide supporting in vivo data whereby VEGF inhibition elicited primary effects on ECs that were obligately accompanied by substantial decrements in PC content. Conversely, although selective PDGF inhibition with Ad sPDGFRβ primarily affected PCs, lesser although detectable secondary decreases in the endothelium were noted. Selective manipulation of the endothelium and PCs by VEGF and PDGFRβ inhibition, respectively, as manifested by the highly specific soluble receptor approach described herein, appears intrinsically self-limited by reciprocal cross-talk and indirect inhibition of the other cell type, limiting the efficacy of combination therapy particularly when maximal blockade is achieved.

Currently, significant interest exists in dual targeting of VEGF and PDGFRβ in cancer patients, given the encouraging clinical activity of multitargeted small molecule RTK inhibitors such as sorafenib and sunitinib that inhibit VEGFR2 and PDGFRβ, among other kinases (26–28). The promiscuous nature of these agents likely contributes to their unquestioned clinical efficacy but at the same time complicates the definition of relative contributions from inhibition of diverse vascular and nonvascular RTK targets. Within the vascular RTKs, the present studies emphasize the substantial cross-talk between VEGF and PDGFRβ signaling during their action on tumor ECs and PCs. Accordingly, our studies suggest that as increasingly efficacious VEGF inhibitors are developed, additive benefits from superimposed PDGFRβ antagonism may diminish, given that primary EC suppression via VEGF blockade can be itself sufficient to inhibit PDGF-BB production and PC recruitment in parallel. We further anticipate that the Ad sPDGFRβ reagent will greatly facilitate the continued investigation of adult physiologic and pathophysiologic functions of the PDGF family.

Methods

Construction of Adenoviruses.

PDGFRβ ectodomain cDNA (corresponding to amino acids 1–527) was amplified from embryonic day 12.5 mouse embryo cDNA with C-terminal His₆ epitope tag, sequenced, and cloned into the E1 region of E1⁻E3⁻ Ad strain 5 by homologous recombination followed by Ad production in 293 cells and CsCl gradient purification of virus as described (31, 46). The construction of Ad Flk1-Fc encoding the entire murine Flk1 ectodomain fused to IgG2α Fc and Ad Fc encoding IgG2α Fc has been described previously (31). Ad Tie2-Fc expresses the entire cognate murine ectodomains followed by fusion to C-terminal murine IgG2α Fc.

Adenovirus Administration and Detection of Plasma Transgene Expression.

Adenoviral administration and analysis of transgene expression were performed as described in ref. 31. Antibodies can be found in SI Methods.

In Vitro Analysis of Soluble Ectodomain Function.

Inhibition of Akt phosphorylation by purified soluble Tie2 and PDGFRβ ectodomains was demonstrated by using HUVECs and NIH 3T3 cells, respectively. NIH 3T3 cells were incubated in the presence or absence of 200 ng/ml soluble PDGFRβ ectodomain for 10 min before stimulation with PDGF-BB (10 ng/ml, R&D Systems). For HUVECs, cells were incubated with 150 ng/ml Ang1 in the presence or absence of excess soluble Tie2 protein (5 μg/ml) for 15 min. Inhibition of Akt phosphorylation was determined by Western blotting of whole-cell extracts with primary antibody specific to phosphorylated Ser-473 Akt antibody (Cell Signaling Technology). Equal loading was determined by Western blotting with primary antibodies specific to GAPDH or insulin-degrading enzyme.

CL Angiogenesis Model.

Three- to 4-week-old female C57BL/6 mice were first injected i.p. with 5 international units of pregnant mare serum gonadotropin (Sigma) followed by the i.v. administration of 5 × 10⁸ pfu of Ad Fc, Ad Flk1-Fc, Ad Tie2-Fc, or Ad sPDGFRβ 24 h later. One day after adenovirus infection, ovulation was induced by treatment with 5 international units of hCG (Sigma). Forty-eight hours after hCG injection, ovaries were harvested and processed for hematoxylin and eosin histological analysis according to standard protocols. Immunofluorescence analysis was performed as described below. Gleevec (100 mg/kg, Novartis) was administered twice daily by oral gavage from the day after pregnant mare serum gonadotropin treatment until the harvesting of the ovaries.

Tumor Modeling.

Tumor modeling was performed as described in ref. 31. For a detailed description, see SI Methods.

Immunofluorescence Staining and Vessel Quantification.

Immunoflurescence staining and vessel quantification were performed according to standard protocols. For details and antibody information, see SI Methods.

Analysis of PDGF-B mRNA Expression.

PDGF-B mRNA was quantified by SYBR Green real-time PCR. For details, see SI Methods.