RETRACTED ARTICLE: Effects of HDM2 antagonism on sunitinib resistance, p53 activation, SDF-1 induction, and tumor infiltration by CD11b⁺/Gr-1⁺ myeloid derived suppressor cells

One of the major determinants of the response to angiogenesis inhibitors is the p53 status of the tumor cells. Yu et al, for example, showed in 2002 that tumors derived from p53^(+/+) HCT116 colorectal carcinoma cells were far more sensitive to VEGF receptor targeted therapy than tumors generated from isogenic p53^(-/-) cells [1]. This differential sensitivity to treatment correlated with the in vitro susceptibility of the tumor cells to the pro-apoptotic effects of hypoxia. Since the publication of these data over a decade ago, the known range of biologic effects regulated by p53 has expanded well beyond cell cycle control and the expression of pro-apoptotic genes to include such diverse functions as the suppression of angiogenesis [2]. It is possible that the differential sensitivity of p53^(-/-) and p53^(+/+) HCT116 tumors to VEGF receptor-targeted therapy is due to an ability of p53 to complement the effects of VEGF receptor inhibition on the tumor microcirculation.

Although the advent of small molecule inhibitors of VEGFR2 has vastly improved the treatment of patients with renal cell carcinoma (RCC), the response to these agents is generally short-lived [3]. The mechanisms by which tumors ultimately manage to evade the effects of these agents are numerous and only partly understood [3‐5]. One such mechanism involves the production of chemokines (e.g. SDF-1, CSF-1, IL-8) that either drive angiogenesis directly or recruit macrophages and other myeloid lineage cells, including CD11b⁺/Gr-1⁺ myeloid-derived suppressor cells (MDSCs), from the bone marrow into tumor tissue [5‐11]. These cells produce a variety of factors that promote tumor growth, invasiveness, angiogenesis, and immunosuppression [10‐13]. p53 has been shown to suppress the expression of SDF-1 [14, 15]. Otherwise, little is known about how the p53 status of a tumor might affect the extent to which tumors are infiltrated by MDSC or the facility with which they develop resistance to VEGF-targeted therapy.

Another mechanism by which p53 suppresses angiogenesis is through the induction of genes that modify the extracellular matrix (ECM). Angiogenesis is negatively regulated, for example, by several ECM-resident peptides (e.g. endostatin, canstatin, arresten) which interact with integrin receptors on the surface of endothelial cells and suppress their proliferation, survival, and motility [16, 17]. These peptides are all derived from the noncollagenous (NC1) domains of certain types of collagen through the action of proteases such as MMP9. The genes encoding the collagen α chains (e.g. COL4A1) from which these angiostatic peptides are derived as well as that encoding the prolyl hydroxylase needed for the post-translational modification and stabilization of collagen [i.e. α(II) PH] are direct p53 transcriptional targets [18, 19]. p53 activation might therefore be expected to suppress the tumor microvasculature through the enhanced production of these peptides. As an illustration of this point, the production of arresten, an angiostatic collagen fragment processed from α1 collagen IV, is markedly diminished in p53^(-/-) tumor cells and its overexpression has been shown to retard tumor growth and limit angiogenesis [19]. The role played by these collagen-derived peptides in the regulation of angiogenesis in RCC and the extent to which their production is regulated by p53 is unknown.

p53 levels are generally low in unstressed cells as a result of HDM2-dependent ubiquitination and proteasomal degradation [20]. p53 can be activated as a result of phosphorylation of any of several sites in its N-terminal domain, which dissociates p53 from HDM2 and enhances its stability [21]. Several of the kinases capable of phosphorylating p53 (e.g. ATM) are redox-sensitive and capable of activating p53 in the setting of hypoxia [22]. The p53 gene is intact (i.e. neither deleted, mutated, nor methylated) in most RCC [23]. One might therefore expect p53 to be activated in RCC subjected to the stress of angiogenesis inhibition. Several factors, however, limit the extent, duration, and biological consequences of p53 activation in these cells. RCC, for example, generally fail to express p53-dependent genes in response to DNA damage, presumably due to high constitutive NF-κB activity [24‐26]. The transcriptional activity of p53 is also limited by a member of the POK family (KR-POK) frequently overexpressed in RCC [27]. This protein physically interacts with p53 and with the transcriptional corepressors NCoR and BCoR, resulting in reduced histone H3 and H4 acetylation at the promoters of certain p53-dependent genes (e.g. p21^waf1/CDKN1A). These signaling aberrations suggest that p53 might not be able to contribute to the suppression of angiogenesis or any other biological process in RCC, despite the integrity of the p53 gene. Hammond et al, however, have pointed out that many of the functions of p53 in the setting of hypoxia are due to transcriptional repression rather than activation [28‐31]. The anti-angiogenic effects of p53, for example, are in part due to the repression of the miR-17-92 microRNA family [32] and possibly to SDF-1 [14, 15] and it is unclear how these functions would be affected by constitutive NF-κB activity or KR-POK expression.

Several drugs that inhibit HDM2 are in preclinical or Phase I trials [33‐35]. These drugs offer distinct advantages over conventional chemotherapy in that they are able to activate p53 in genetically permissive tumor cells without inducing DNA damage. The studies described in this paper were undertaken to assess the effects of HDM2 blockade alone and in conjunction with VEGF-targeted therapies on p53 function, tumor growth, and angiogenesis in RCC.

Despite the numerous constraints on p53 function in RCC [24‐27], sunitinib treatment does induce the expression of several p53-dependent genes (e.g. NOXA, HDM2, p21^waf) in RCC xenografts. The induction of these genes is, however, limited to the interval during which tumor growth is suppressed and is attenuated once resistance develops. Although several factors have been shown to block p53 transcriptional activity in RCC, these are for the most part stable genetic alterations (e.g. KR-POK expression) that are not known to be subject to regulation by hypoxia or other metabolic changes that occur during treatment with angiogenesis inhibitors.

The factor(s) responsible for the transient activation and subsequent inactivation of p53 transcriptional activity during the course of treatment with sunitinib are unknown but at least one well-characterized p53 transcriptional suppressant (i.e. HDMX) appears to be temporally linked to p53 function in our xenograft models and may therefore be a candidate. Unlike its binding partner HDM2, HDMX is not regulated by p53 [37]. HDMX is constitutively expressed in RCC xenografts but vanishes with the initiation of sunitinib treatment – along with the appearance of p21^waf. HDMX reappears with the development of resistance, in association with the down modulation of p21^waf. These temporal associations strongly implicate HDMX as the factor responsible for the failure of p53 to maintain p21^waf1 expression. The reappearance of HDMX during sunitinib treatment also explains why the suppression of p53 with an shRNA affected the response of 786-0 xenografts to sunitinib only during the first few days of treatment. As shown in Figure 6A, the xenografts in which p53 activation is not impeded characteristically stall for several days during sunitinib treatment but subsequently catch up with those in which p53 expression is suppressed. These data are consistent with the inactivation of p53 function by HDMX.

HDMX is physically associated with HDM2 and drugs that block the interaction between HDM2 and p53 such as MI-319 also interfere to some extent with the ability of HDMX to suppress p53 transcriptional activity. The fact that MI-319 maintains p21^waf levels during sunitinib treatment suggests that the factor responsible for limiting p53 nuclear function most likely interacts with HDM2. This consideration, in addition to the temporal linkage between HDMX expression and the absence of p21^waf, supports the hypothesis that HDMX is the dominant regulator of p53 nuclear function during sunitinib treatment and possibly a major factor in the development of drug resistance.

Despite the induction of p21 (Figure 2B), MI-319 treatment does not have a consistent effect on tumor cell proliferation as determined by Ki67 staining (Figure 3). In A498 xenografts, for example, MI-319 neither retards proliferation when administered as a single agent or when given concurrently with sunitinib. In 786-0 xenografts, however, the addition of MI-319 suppresses the increase in proliferation induced by sunitinib treatment. The enhanced tumor cell proliferation induced by sunitinib is presumably the result of tumor hypoxia, which has been reported to enhance proliferation in other tumor models [38].

The anti-angiogenic effect of sunitinib is augmented by the concurrent administration of MI-319. One mechanism by which MI-319 might further limit angiogenesis is the suppression of the influx of MDSC that generally occurs in response to sunitinib treatment. This suppressive effect is particularly obvious in 786-0 xenografts, from which MDSC are virtually excluded by MI-319 treatment. The means by which MI-319 (and p53) inhibit MDSC trafficking from the bone marrow to tumor tissue is not known but may involve the suppression of chemokine (e.g. SDF-1) production by tumor cells and stromal elements that were rendered hypoxic by the disruption of the tumor vasculature. The ability of MI-319 to suppress both baseline and sunitinib-induced SDF-1 expression in our RCC xenografts is consistent with the known ability of p53 to suppress SDF-1 expression [14, 15]. Although our data suggest that the suppression of SDF-1 may account for the diminution in the influx of MDSC observed in the tumor infiltrates of mice treated with MI-319, it is possible that other factors that are both hypoxia-inducible and suppressed by p53 will be identified that might contribute to the anti-angiogenic effects of the drug.

Our observation that sunitinib treatment increases MDSC infiltration of RCC xenografts is at odds with several previous reports showing that the drug limits the expansion of these cells and enhances immune function [39‐43]. Most of these earlier reports, however, were based on analyses of peripheral blood or splenocytes. Ko et al, for example, showed that RCC patients have increased numbers of MDSC in the peripheral blood and that sunitinib treatment results in a decline in their numbers [39]. Sunitinib treatment consistently reduces MDSC accumulation in the spleens of tumor-bearing mice in several tumor models [40]. However, this effect on splenic MDSC did not extend to the tumor microenvironment, where MDSC continued to accumulate with the expected deleterious effect on T cell function, regardless of treatment. These site-specific effects may be attributable to the cytokine GM-CSF, which is capable of rendering MDSC resistant to the effects of sunitinib [40].

Our studies suggest that sunitinib can actually increase the influx of MDSC into tumor tissue in some circumstances. This result may be unique to VHL-deficient RCC and dependent on the severity of the hypoxia induced in these tumors by VEGF-targeted agents. To the extent that this is the case, one would expect that these tumors would abundantly produce SDF-1 and other HIF-dependent chemokines (which recruit MDSC) in response to sunitinib treatment. The suppressive effects of sunitinib on MDSC accumulation and function are thought to be mediated through the inhibition of STAT3 and c-kit [41, 42]. It is possible that hypoxia-induced chemokine production within tumor tissue may in some circumstances trump these inhibitory effects of sunitinib, resulting in an increase in MDSC infiltration.

Another mechanism by which p53 regulates angiogenesis is through the induction of α(II) PH and the deposition of anti-angiogenic collagen fragments (e.g. arresten, endostatin, canstatin) in the ECM [16‐19]. Several previous studies have in fact suggested that this is one of the dominant mechanisms by which tumor angiogenesis and growth are suppressed by p53 [19]. Our data clearly establish that p53 activation is essential for the deposition of endostatin and arresten in 786-0 xenografts (Figure 6A) and in this respect, our results corroborate the results of Assadian et al, who demonstrated a similar requirement for p53 in the production of arresten by HCT116 cells [19]. Despite the absolute requirement for p53, however, sustained p53 activation does not appear to be essential to maintain endostatin and arresten levels in RCC xenografts during sunitinib treatment. We have not been able to demonstrate a significant decline in endostatin or arresten levels after the initial induction despite the apparent loss of p53 transcriptional activity (i.e. the disappearance of p21^waf1) during treatment. Indeed, endostatin and arresten levels remain nearly unchanged with the development of sunitinib resistance, when p21^waf1 is no longer detectable. Nor have we been able to demonstrate any enhancement in endostatin or arresten deposition by the addition of MI-319 to the treatment regimen, although HDM2 antagonism is essential for the maintenance of p21^waf1 expression. Collectively, these data suggest that although the failure to express α(II) PH and to deposit angiostatic collagen fragments (e.g. endostatin, arresten) in the ECM might account for the faster growth and more vigorous angiogenesis observed in p53^(-/-) tumors, changes in endostatin or arresten levels are not a factor in the development of sunitinib resistance in p53-WT RCC nor in the enhanced suppression of angiogenesis and tumor growth resulting from the concurrent administration of MI-319 with sunitinib.

We have demonstrated that treatment of mice bearing RCC xenografts with VEGF-targeted agents results in p53 activation, the biological effects of which are quickly undermined with the onset of drug resistance, possibly due to the induction of the p53 antagonist HDMX. We have further shown that the HDM2/HDMX antagonist MI-319 maintains p53 function during treatment and delays/prevents the emergence of resistance. These data suggest that the evasion of p53 function is an essential element in tumor escape from the effects of VEGF-targeted therapy. The effects of MI-319 appear to be at least in part due to the ability of the drug to suppress the influx of MDSC into the tumor, which may in turn be due to its ability to block the production of chemokines such as SDF-1 that are otherwise induced in the setting of hypoxia. The potential utility of a combination of an HDM2 antagonist with sunitinib may not be limited to RCC. For example, in a recent study by Henze et al, the HDM2 antagonist Nutlin-3 was shown to augment the apoptotic response of imatinib-resistant gastrointestinal stromal tumor (GIST) cell lines to sunitinib [44]. In this case, however, the effects of sunitinib were most likely attributable to its ability to inhibit c-kit rather than its antiangiogenic effects. Collectively, these data provide a strong rationale for the concurrent use of HDM2 antagonists as adjuncts to VEGF receptor inhibitors in the management of metastatic RCC and other tumor types.

Cell lines and reagents. The human RCC cell lines 786-0 and A498 were obtained from ATCC and maintained in RPMI-1640 (Lonza) and Eagle minimal essential medium (ATCC), respectively containing 10% fetal bovine serum (USA Scientific), 2 mM glutamine and 50 μg/ml gentamycin at 37°C in 5 percent CO₂. The MI-319 was provided by Ascenta Therapeutics (Malvern, PA) and Sanofi-Aventis (Paris, France).