Partial p53 reactivation is sufficient to induce cancer regression

The tumor suppressor gene TP53 encodes the p53 transcription factor, which regulates target genes in multiple pathways to counteract the expansion of malignant cells [1‐3]. p53 thereby poses a major barrier to tumor development, explaining why tumorigenesis strongly selects for cells with inactivated p53 [3]. Loss of p53 function (LOF) can result from p53 degradation or sequestration by viral oncoproteins or cellular inhibitors like Mdm2 or Mdmx [4, 5]. In approximately 50% of all tumors, LOF is caused by somatically acquired genetic alterations affecting the TP53 gene directly [5]. This includes large deletions encompassing the entire gene as well as small non-sense, frameshift or missense mutations, which can differ massively in their functional impact [5, 6]. Different from other tumor suppressor genes, most TP53 mutations are missense mutations, explained in part by dominant-negative and tumor-promoting neomorphic (gain-of-function, GOF) properties of at least some of the mutant proteins [7‐10]. Of note, full p53 activity is essential for optimal tumor suppression as even a partial loss of p53 function increases the cancer risk in mice and causes hereditary cancer susceptibility in humans [11‐17].

Importantly, restoration of p53 in p53-deficient tumor cells was found to be detrimental, usually resulting in cell death or loss of proliferative capacity due to senescence or differentiation [18]. For example, in p53ER^TAM mice with tamoxifen-switchable p53 activity, EμMyc-driven Burkitt-like lymphomas develop in the p53-off state, but undergo rapid regression when p53 is switched on [19]. This was observed even when p53 inactivation was not the tumor-initiating driver lesion and acquired only later during tumor progression [20]. Together with other studies in independent mouse cancer models, this firmly established that tumor cells can become addicted to the loss of p53 [21‐24]. Moreover, these studies demonstrated the therapeutic potential of p53 reactivation and provided critical support for the development of p53-reactivating drugs as cancer therapeutics [25, 26].

Meanwhile a growing number of p53 reactivating compounds has been developed. While Mdm2 and Mdmx inhibitors are being evaluated for treatment of tumors with wild-type p53, diverse strategies have been proposed to target tumors with TP53 mutations [25, 26]. As straightforward as reactivation may seem, as challenging it turns out in practice. The most common TP53 mutations either destroy DNA contact residues or destabilize the thermodynamically labile p53 DNA binding domain and cause its denaturation at normal body temperature [27]. The most direct reactivation approaches aim to refold the pool of mutant p53 proteins, that have accumulated in the tumor cells, into a native conformation [28, 29]. However, as more than 2000 different mutant proteins have been identified in cancer patients, a universal reactivation strategy is unrealistic. As such, many reactivation compounds target only single or groups of mutants [30]. For example, PhiKan083 specifically targets a surface crevice created by the Y220C mutation [31]. Metallochaperones increase the intracellular availability of zinc to rescue folding of zinc-binding site mutants like R175H [32]. Arsenic trioxide (ATO) binds to a cryptic allosteric site formed by Arsenic-coordinating cysteines and stabilizes the native fold of a subset of p53 mutants [33]. The clinically most advanced compound APR-246 (PRIMA-1^Met, Eprenetapopt), is converted to the thiol-reactive metabolite methylene quinuclidinone (MQ) and proposed to alkylate and thereby reactivate structural mutants, but also, in a mechanistically poorly understood manner, some DNA contact mutants [34‐36]. In combination with azacytidine, APR-246 showed promising therapeutic responses in phase II studies involving myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) patients [37, 38]. However, MQ also reacts with numerous other cellular cysteines, forming a thiol-bound drug reservoir, and through depletion of glutathione (GSH) and inhibition of the GSH and thioredoxin antioxidant systems exhibits cytotoxic activities [39, 40]. Other reactivating compounds also display considerable p53-independent toxicity that certainly adds to, if not determines, the therapeutic effects [30]. This raises the critical question: how much of the observed cytotoxicity is due to p53 reactivation [18]? Last but not least, the mutant p53 refolding activity of various reactivating compounds differs by at least two orders of magnitude not only between compounds but also between individual mutants [33]. Together these findings indicate that pharmacological mutant p53 reactivation is continuously improving, but is currently still far from optimal. This brings up an even more important question: to what extent does p53 have to be restored [18]? Is full reactivation required to achieve a therapeutic response or is partial reactivation already sufficient?

Here, we have modelled partial p53 reactivation using a conditional knock-in mouse with inducible expression of the partial LOF variant p53^E177R (short: E177R), which corresponds to human E180R. The E177R mutation disrupts an intermolecular salt-bridge with R178 (human R181) that is required for cooperative DNA binding [41‐43]. The decrease in cooperativity reduces DNA binding and target gene activation [15, 44, 45]. As a result, E177R causes cancer susceptibility and cooperates with various oncogenes to promote cancer development, including lung and pancreatic adenocarcinoma and various types of leukemia and lymphoma [15, 43, 46, 47]. Here, we report that in p53-deficient leukemia and lymphoma models the effects of E177R expression are entirely opposite: E177R inhibits cell proliferation and viability and induces cancer regression. p53-deficient cancer cells prove to be so addicted to the absence of p53 that the residual activity of E177R is detrimental for cancer growth. As such, the study provides genetic proof-of-concept that incomplete p53 reactivation can make a therapeutic impact.

Previous studies have indicated that many p53-deficient tumors regress when p53 is restored, implying that they are strongly addicted to the absence of p53 activity [19‐22, 24]. However, it has remained unclear how much p53 activity is required to trigger tumor regression [18]. Given the yet limited pharmacological abilities to fully reactivate a p53 mutant [25, 26], we aimed to analyze whether partial p53 reactivation may suffice to induce a therapeutic effect. To reproducibly reactivate p53 to a defined suboptimal degree, we chose to genetically limit p53 activity using the previously well-characterized partial LOF variant E177R [15]. This mutation does not affect any DNA contacting residues and maintains a normally folded DNA binding domain [41]. Nevertheless, it is impaired in DNA binding because it fails to form a strong intermolecular salt-bridge which is crucial to stabilize the tetrameric protein complex on DNA [43]. Previous DNA binding studies in vitro and in vivo have characterized the DNA binding defect as a global reduction in DNA binding across the entire target gene spectrum [44, 45]. By limiting global DNA binding in a genetically fixed manner, the E177R mutation is suitable to model the consequences of an incompletely reactivated p53 mutant.

Even though the E177R mutation reduced DNA binding globally, we observed transactivation of the typical p53 target genes (Fig. 1e) and induction of a broad range of effector programs including cell cycle arrest, senescence, differentiation and apoptosis (Fig. 2). While transactivation of antiproliferative target genes like p21/Cdkn1a is in line with the described phenotype of E177R knock-in mice and explains induction of cell cycle arrest, senescence and differentiation [15], activation pro-apoptotic target genes was rather unexpected, because E177R and several other related cooperativity mutants are generally characterized by an apoptosis defect [43, 57‐59]. However, the ability of p53 to induce apoptosis is highly context-dependent and modulated by the extent and dynamics of p53 protein accumulation as well as the intrinsic apoptosis threshold of the cell and its level of mitochondrial priming [60‐64]. It is therefore conceivable that sustained high-level expression of a DNA-binding impaired p53 mutant can trigger apoptosis especially in cell types with an intrinsically low apoptosis threshold [65, 66]. In line, massive constitutive stabilization of the E177R mutant protein by knockout of Mdm2 was shown to trigger lethal apoptosis in highly proliferative embryonic tissues [65], which have a lower apoptosis threshold than most adult tissues due to Myc-mediated mitochondrial priming [65, 66]. Similarly, tumors in E177R mice commonly display constitutive stabilization of the E177R mutant protein, and DNA damage in such tumors was reported to induce pro-apoptotic p53 target genes and render them sensitive to chemotherapy [47]. In all our leukemia and lymphoma models, the E177R protein was readily detectable by immunohistochemistry (Figs. 3e, 4a, 5f), showing a staining pattern similar to human tumors with a massively stabilized mutant p53 protein. Moreover, p53-deficient leukemias and lymphomas commonly upregulate p19Arf [20, 47], which can sequester Mdm2, thereby contribute to the massive E177R accumulation, overcome the apoptosis-deficiency and enable tumor cell killing by E177R.

In vivo, additional non-cell-autonomous p53 effects might contribute to eradication of tumor cells [21, 67, 68]. In liver carcinomas, for example, wild-type p53 restoration triggers infiltration by innate immune cells like macrophages, neutrophils and lymphocytes that support clearance of senescent tumor cells [21]. While we have not detected changes in macrophages, we have observed lymphocytic infiltration upon E177R activation (Fig. 4), which makes it tempting to speculate that immune infiltration induced by p53 in a non-cell-autonomous manner might contribute to tumor regression in vivo. In summary, the presented data indicate that E177R is directly capable of inducing apoptosis in p53-deficient leukemia and lymphoma cells and provide an explanation for the observed cancer regression. Whether and how immune cells contribute remains to be investigated.

We observed cancer regression not only in AML but also in lymphoma models, including thymic T cell lymphomas that spontaneously develop in p53-deficient mice. While non-reactivated mice mostly succumbed to thymic lymphoma, reactivated mice at the time of sacrifice more often presented with other types of cancer, often sarcomas (Fig. 6e and f), suggesting that lymphomas are more vulnerable to p53 reactivation than other cancer entities. A differential sensitivity of sarcomas and lymphomas was also reported upon restoration of wild-type p53, where restoration in lymphomas caused widespread apoptosis, whereas sarcomas showed a delayed anti-proliferative response with features of senescence [22]. The higher sensitivity of hematopoietic cancers is not entirely unexpected as already the normal bone marrow displays an exquisite vulnerability to elevated p53 activity which is at least partially explained by mitochondrial priming of the hematopoietic compartment [61, 66, 69‐72]. Moreover, this is in line with clinical studies on p53/mutp53-reactivating compounds like Mdm2-inhibitors or eprenetapopt (APR-246), which have reported clinical responses mostly in patients with hematological cancers [37, 38, 73].

Even though most animals demonstrated strong responses to partial reactivation, none of the animals was cured and all relapsed eventually. The cause of relapse differed between the different models. Relapsed EμMyc-driven lymphomas showed a high percentage of p53-negative tumor cells (Fig. 5c), suggesting that not all lymphoma cells had recombined the LSL-E177R allele and therefore escaped due to a technically inefficient reactivation. An alternative explanation would be a secondary loss or inactivation of the E177R mutant, for example, by LOH. In contrast, relapsed AML mice showed homogeneous high-level expression of the E177R mutant, indicating that some tumor cells eventually adapt and tolerate E177R expression. Of note, these observations are not unique to partial p53 reactivation and similar findings have been reported in the previous studies with wild-type p53 restoration. All EμMyc lymphomas with tamoxifen-inducible p53ER^TAM activity relapsed after reactivation, losing either p53ER^TAM expression or deleting its upstream activator p19ARF [19]. Responses to partial reactivation are therefore mostly transient, calling for synergistic approaches such as DNA damaging chemotherapeutics that might help to boost p53 activity to obtain longer-lasting remissions. In AML, even a transient response might be efficient, by inducing the clinical remission needed for a bone marrow transplant as the final curative treatment.

Tamoxifen, used to induce E177R-mediated tumor regression in our study, is an estrogen receptor (ER) antagonist. As estrogens vary between sexes and interfere with the p53 pathway [74], the anti-estrogenic tamoxifen activity might have contributed to the p53 reactivation response in a sex-dependent manner. In the AML model, both the LSL and KO leukemia cells were derived from female embryos and 86% of all recipient mice were also female. The improvement in survival upon tamoxifen treatment of LSL-E177R AML mice remained highly significant when excluding male mice from the survival analysis (Supplemental Fig. S3). A gender-effect as the cause for the differential in vitro and in vivo response to tamoxifen treatment can therefore be excluded. However, due to the small number of male recipient mice, we could not analyze whether the reactivation response is different in male AML mice. Both lymphoma models comprised comparable numbers of male and female mice and the reactivation responses were similar in both sexes (Supplemental Fig. S1 and S2). In the transplanted Myc lymphomas, reactivation was also independent of the sex of the lymphoma donor (Supplemental Fig. S1). Together, these analyses indicate no confounding sex-dependent effects in the hematopoietic cancer models studied here. This is in line with the clinical use of tamoxifen primarily for the treatment of hormone-dependent ER+ breast and endometrial cancer. In other ER-negative cancer types, anti-tumor effects of tamoxifen require doses 4–8 fold higher than necessary for ER inhibition [75], suggesting that these effects are hormone-independent.

Of note, tumor cells can not only become addicted to the loss of wild-type p53 activity, they can also become dependent on neomorphic GOF properties of the p53 mutant. Refolding a GOF mutant would therefore not only restore some degree of wild-type function but simultaneously deprive tumor cells of survival-promoting GOF effects. As such, it has been demonstrated that loss of GOF activities by therapeutic ablation of the mutant protein can be sufficient to induce cancer regression, even in the absence of any restoration of wild-type function [76, 77]. By switching p53-null tumor cells to E177R, we can exclude that the responses observed in our cancer models are mediated by a loss of GOF properties. However, it is reasonable to assume that an additional loss of GOF would synergize with restoration of wild-type function and could further enhance the extent and duration of tumor regression. It is therefore expected that a partial p53 reactivation would be even more effective in tumors bearing GOF missense mutants rather than pure LOF mutants.

We demonstrate that a p53 variant, which has partially lost the activity of wild-type p53 and by itself increases cancer susceptibility and promotes oncogene-driven development of leukemia and lymphoma, induces cancer regression in vivo when introduced into p53-deficient leukemia and lymphomas. This not only provides compelling evidence that p53 mutants function in a highly context-dependent manner, being pro-tumorigenic in one setting and tumor suppressive in another, it also highlights how sensitive cancer cells respond to changes in p53 activity and that changes in activity are more important than the absolute activity level. Even a slight increase in p53 activity resulting from switching a p53-null into a partially active p53 allele can suffice to induce tumor regression. As such, our findings will encourage ongoing research into mutant p53-reactivating drugs by providing genetic proof-of-principle evidence that incomplete reactivation of p53 mutants can elicit a beneficial therapeutic response.