Partial p53 reactivation is sufficient to induce cancer regression

Impaired p53 function is one of the central molecular features of a tumor cell and even a partial reduction in p53 activity can increase the cancer risk in mice and men. From a therapeutic perspective it is noteworthy that tumor cells often become addicted to the absence of p53 providing a rationale for developing p53 reactivating compounds to treat cancer patients. Unfortunately, many of the compounds that are currently undergoing preclinical and clinical testing fail to fully reactivate mutant p53 proteins, raising the crucial question: how much p53 activity is needed to elicit a therapeutic effect? We have genetically modelled partial p53 reactivation using knock-in mice with inducible expression of the p53 variant E177R. This variant has a reduced ability to bind and transactivate target genes and consequently causes moderate cancer susceptibility. We have generated different syngeneically transplanted and autochthonous mouse models of p53-deficient acute myeloid leukemia and B or T cell lymphoma. After cancer manifestation we have activated E177R expression and analyzed the in vivo therapy response by bioluminescence or magnetic resonance imaging. The molecular response was further characterized in vitro by assays for gene expression, proliferation, senescence, differentiation, apoptosis and clonogenic growth. We report the conceptually intriguing observation that the p53 variant E177R, which promotes de novo leukemia and lymphoma formation, inhibits proliferation and viability, induces immune cell infiltration and triggers cancer regression in vivo when introduced into p53-deficient leukemia and lymphomas. p53-deficient cancer cells proved to be so addicted to the absence of p53 that even the low-level activity of E177R is detrimental to cancer growth. The observation that a partial loss-of-function p53 variant promotes tumorigenesis in one setting and induces regression in another, underlines the highly context-specific effects of individual p53 mutants. It further highlights the exquisite sensitivity of cancer cells to even small changes in p53 activity and reveals that changes in activity level are more important than the absolute level. As such, the study encourages ongoing research efforts into mutant p53 reactivating drugs by providing genetic proof-of-principle evidence that incomplete p53 reactivation may suffice to elicit a therapeutic response.

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][8][9][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][12][13][14][15][16][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, EμMycdriven 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][22][23][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][35][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][42][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.
For the EμMyc lymphoma model, double-heterozygous Trp53 LSL-E177R/+ ; Rosa26 CreERT2/+ females were bred with EμMyc transgenic males. Lymphomas from EμMyc; Trp53 LSL-E177R/LSL-E177R+ offspring mice with or without CreER T2 were used for transplantation into 129/B6 F1 hybrid recipient mice as described [50]. After disease onset was confirmed by palpation of enlarged mandibular, axillary or subiliac lymph nodes, mice were treated 1 week with daily i.p. injections of 1 mg tamoxifen. Mice were euthanized when pre-defined humane endpoint criteria were reached. Control experiments for Cremediated toxicity were performed with EμMyc; Rosa-26 CreERT2/+ lymphomas. Both male and female donor and recipient mice were used.
For the spontaneous T-lymphoma model, both male and female Trp53 LSL-E177R/LSL-E177R ; Rosa26 CreERT2 mice were examined with magnet-resonance tomography using 7 T Clinscan 70/30 USR (Bruker) as described [51]. T2 weighted sequences triggered on respiration in transverse and coronal orientation were used for anatomical imaging of the thymus. The total measurement time was approximately 28 min per mouse. Tumor size was measured using RadiAnt DICOM Viewer and tumor volume was calculated with the ellipsoid formula V = 4/3*π*abc. Trp53 LSL/LSL mice without Cre were used as a control cohort. The first imaging was performed at the age of 120 days or earlier for mice that showed clinical symptoms of thymic lymphoma (weight loss, hunchback posture, shortness of breath).
For reactivation of E177R expression via Cre-mediated recombination in vivo, mice were injected i.p. for 7 consecutive days with 100 μl of 10 mg ml − 1 tamoxifen (Sigma) in sterile corn oil. For mock treatment, mice were injected with 100 μl of corn oil only.

PCR and RT-PCR
PCR detection of the LSL cassette and Cre-mediated recombination in the targeted Trp53 allele was done as described [15]. For reverse-transcription real-time PCR (RT-qPCR) RNA was isolated from cells or tissue samples using the RNeasy Mini Kit (Qiagen) and cDNA was generated with the SuperScript VILO cDNA Synthesis Kit (Invitrogen). Gene expression was analyzed on a LightCycler 480 (Roche) using SYBR Green (Thermo Fisher Scientific) and primers specific for mouse Trp53, Cdkn1a/p21, Ccng1, Bbc3/Puma, β-Actin/Actb [50]. Data were evaluated by the ΔΔCt method using β-actin gene expression for normalization.

Chromatin immunoprecipitation (ChIP)
AML cells were treated and fixed after 48 h with freshly prepared 18.5% (w/v) paraformaldehyde (PFA) for 10 min at RT, aiming at a final concentration of 0.88% (v/v) PFA. Crosslinking of DNA and proteins was terminated by quenching unreacted PFA by addition of glycine to 125 mM end concentration and further incubation for 5 min at RT. Cells were pelleted by centrifugation at 300 x g for 10 min at 4 °C, washed twice with ice-cold PBS and lysed at a concentration of 2 × 10 7 cells ml − 1 in SDS lysis buffer containing protease inhibitor. 250 μl lysate per tube was sonicated to shear DNA to a fragment size of 200-1000 bp, using the Sonicator Bioruptor Twin UCD-400 (Diagenode) for 5 cycles of 30 s ON/ 30 s OFF. After sonication, cell debris was pelleted by centrifugation for 10 min at 10,000 x g at 20 °C. Supernatant containing the sheared chromatin was aliquoted á 100 μl and either frozen at − 80 °C for later use or directly processed. For each antibody, one 100 μl aliquot was used and diluted 1:10 with dilution buffer. Pre-clearing was performed for 1 h at 4 °C using 50 μl blocked Protein G-coupled Sepharose beads (GE Healthcare, 1:1 slurry with 20% EtOH) per sample. Supernatant of pelleted beads (3000 x g, 1 min, 4 °C) was transferred into a new tube and 1% was removed as input DNA and stored at 4 °C. Samples were rotated over night at 4 °C with 2.5 μg antibody: p53 (FL393, Santa Cruz sc-6243) or isotype control (E5Y6Q, Cell Signaling #61656). For precipitation, protein-DNA complexes were bound to blocked beads (50 μl per sample) for 4 h at 4 °C. Beads were washed once with each Low Salt, High Salt and LiCl Immune Complex washing buffers and twice with TE buffer. Crosslinking was reverted by incubating immunoprecipitated samples and inputs at 99 °C for 10 min with 100 μl of 10% (w/v) Chelex 100 solution. Proteins were digested with proteinase K for 30 min at 55 °C and this enzyme was subsequently inactivated at 99 °C for 10 min. DNA was eluted in two steps with ddH 2 O, first with 55 μl and then with 100 μl each time by centrifugation (1 min at 12,000 x g). Analysis of bound DNA fragments were performed using qPCR with 1 μl DNA per reaction using primers specific for p53 response elements in the Cdkn1a/p21, Bax and Bbc3/Puma genes.

Statistical analysis
For statistical analysis the GraphPad Prism 8 software was used. Graphs show mean values obtained with n technical or biological replicates, and error bars in all figures represent standard deviation (SD), unless indicated otherwise. Two groups were tested for statistically significant differences by a two-sided unpaired t-test or, if not normally distributed, by a Mann-Whitney test. Multiple groups were tested by 1way ANOVA in conjunction with a Dunnett's multiple comparisons test. P-values of 1way ANOVA and selected pairwise comparisons are reported in the respective figures. Three or more groups that have been split on two independent variables (here treatment and genotype) were analyzed by 2way ANOVA in conjunction with Sidak's multiple comparisons test. P-values for the interaction effect between the two independent variables and P-values of selected pairwise comparisons are reported in the figures. Survival data were analyzed with pairwise Log-rank (Mantel-Cox) tests. A p-value < 0.05 was used as level of significance.

Leukemia model for partial p53 reactivation
To model the therapeutic effect of partial p53 reactivation, we have used a previously described genetic Cre-mediated recombination approach to activate a conditional p53 allele in p53-deficient tumor cells [22]. In contrast to earlier studies based on restoration of wildtype p53 [19,21,22], we have used a conditional knock-in allele for the p53 partial loss-of-function variant E177R (Trp53 LSL-E177R ) in combination with the Rosa26 CreERT2 transgene, which constitutively expresses tamoxifen-regulated CreER T2 [15,22] (Fig. 1a). The inducible CreER T2 recombinase can be activated at will by tamoxifen in vivo, or 4-hydroxytamoxifen (4OHT) in vitro, to excise the lox-stop-lox (LSL) cassette that silences expression of the Trp53 LSL-E177R locus, thus triggering expression of the E177R protein (Fig. 1b). We transduced fetal liver cells from Trp53 LSL-E177R;LSL-E177R ;Rosa26 CreERT2 mice with two bicistronic retroviral constructs -one expressing Nras G12D and firefly luciferase, the second AML1/ETO9a and EGFP, and transplanted them into lethally irradiated recipient mice as previously described [46] (Fig. 1a). The mice developed acute myeloid leukemia within 1-2 months. As a control for p53-independent toxicity of tamoxifen and DNA damage induced by Cre, we simultaneously generated CreER T2 -expressing AML with a constitutive p53 knock-out (KO).

Partial p53 reactivation inhibits leukemia proliferation and viability
p53-deficient leukemia cells were explanted and further propagated ex vivo. In LSL-E177R;CreER T2 (short: LSL) leukemia cells, 4OHT induced excision of the lox-stoplox (LSL) cassette and resulted in expression of E177R at the mRNA and protein level (Fig. 1c-e). This was accompanied by significantly enhanced expression of p53 target genes such as the cyclin-dependent kinase inhibitor p21/ Cdkn1a. Of note, we also observed significant induction of several pro-apoptotic target genes including Bax, Bbc3/Puma and Pmaip1/Noxa (Fig. 1e). As E177R has previously been reported to be unable to bind and transactivate pro-apoptotic target genes [15], we also analyzed binding of E177R to chromatin (Fig. 1f ). We not only detected significant levels of E177R binding to the p21/ Cdkn1a promoter, but also to the Bax promoter (Fig. 1f ) Puma promoter that reached statistical significance only in a pairwise comparison (Fig. 1f ). Notably, no target gene activation was observed in control KO;CreER T2 (short: KO) leukemia cells (Fig. 1e). Consistent with previous reports demonstrating that E177R is proficient in regulation of cell cycle and senescence [15], activation of E177R inhibited cell cycle progression, resulting in a strongly reduced number of cells in S phase, which was not observed upon tamoxifen treatment of control KO cells (Fig. 2a). At later time points, we observed a massive accumulation of senescent cells marked by expression of senescence-associated beta-galactosidase selectively in tamoxifen-treated LSL but not KO cells (Fig. 2b). Moreover, as induction of differentiation is a recognized tumor-suppressive mechanism exerted by wild-type p53 [21,54], we tested for signs of differentiation. We observed an increase of the lineage-positive population after 7 days of 4OHT treatment, in particular an enrichment of the population of CD11b-positive cells indicating myeloid differentiation (Fig. 2c). Consistent with robust transactivation of pro-apoptotic target genes (Fig. 1e, f ), we also noticed a significant drop in viability of LSL leukemia cells after 4OHT treatment and flow cytometry revealed induction of the apoptosis markers annexin V and cleaved-caspase 3 after expression of E177R (Fig. 2d). Importantly, activation of E177R abolished the clonogenicity of leukemia cells in 3D culture, whereas KO cells remained unaffected excluding p53-independent toxicity of 4OHT or Cre as a cause (Fig. 2e). Together, these experiments demonstrate that genetic reactivation of E177R induces multiple tumor-suppressive programs in AML cells that induce cell cycle withdrawal or cell death and prevent clonogenic expansion.  Fig. 3 a). After detection of bioluminescence in tubular bones, sternum and spleen as a sign of successful leukemia engraftment, each cohort was randomly subdivided into two groups which were injected for 1 week daily with either tamoxifen (Tam) or solvent (mock) as control. As revealed by BLI and monitoring of clinical symptoms, leukemia quickly progressed in both mock-treated genotypes (median survival for LSL 25 days, for KO 23 days) (Fig. 3b). When the LSL group was treated with tamoxifen a strong clinical response was observed: BLI showed a gradual decrease of luciferase signal to almost background levels within 2 weeks of treatment (Fig. 3a). Median survival was extended more than twice in comparison to the control group and reached 57 days, 3/12 mice survived for > 100 days and 2 for almost 200 days (Fig. 3b). Importantly, tamoxifen administration had only minor effect on survival of mice with KO leukemia, which continued to progress under treatment (Fig. 3a), and provided only a subtle survival advantage (median survival 30 days, Fig. 3b), underscoring the role of E177R as the driver of the therapeutic response.

Partial p53 reactivation induces leukemia regression
To confirm that the therapy response is indeed dependent on E177R expression, we generated an isogenic LSL-E177R;CreER T2 AML cell clone with stable Cas9 expression and subsequently transduced this clone with lentiviruses expressing either a Trp53-targeting (sg-p53) or a non-targeting scrambled sgRNA (sg-scr). Both sg-p53 and sg-scr AML cell types were transplanted into sublethally irradiated recipient mice and, following successful engraftment, treated with tamoxifen as described above. As indicated by BLI and survival analysis, sg-p53 AML did not respond to tamoxifen, whereas the same treatment promoted regression of sg-scr leukemia and resulted in a strong survival benefit (median survival 37 and 95 days, respectively) ( Fig. 3c and d). Immunohistochemistry confirmed the absence of p53 protein in sg-p53 AML after tamoxifen administration, whereas sg-scr leukemia samples collected after disease relapse were  (Fig. 3e). Together, these results indicate that E177R is mediating the reactivation response and strongly suggest that even a partial restoration of p53 activity can provide a significant therapeutic effect in vivo.
The remarkable efficiency of reactivation therapy prompted us to investigate the mechanisms that underlie the E177R-mediated response. Using immunohistochemistry, we analyzed tissue samples collected at different time-points after therapy start. As expected, we observed fast accumulation of p53 protein in LSL-tumors after the start of daily tamoxifen injections (Fig. 4a). Interestingly, all relapsed tumors retained a high expression level of E177R protein (Fig. 3e), indicating that relapse is not driven by incomplete Cre-mediated E177R activation or secondary loss of E177R expression. It rather suggests that some leukemia cells tolerate the residual tumor suppressive activity of E177R and eventually adapt to its presence. Coherent with the loss of the luciferase signal in BLI, E177R induction led to a gradual elimination of GFP-positive leukemia cells (Fig. 4a). Consistent with E177R being proficient in cell cycle inhibition and senescence induction, we detected a marked drop in proliferation in tamoxifen-treated leukemia, as indicated by reduced number of BrdU-positive cells (Fig. 4a). In parallel, we observed a significant accumulation of senescent cells ( Fig. 4a and b). However, while cell cycle arrest and senescence can decelerate or even stop tumor growth, these processes did not fully explain the rapid removal of leukemia cells, which became detectable as early as 3 days and was clearly evident 7 days after the first tamoxifen administration (Fig. 4a). As we had observed induction of apoptosis following E177R induction in vitro, we therefore also analyzed apoptosis levels immunohistochemically in tissue samples by TUNEL assay and observed a significant increase in tamoxifen-treated, but not in mock-treated, tumors ( Fig. 4a and c). Thus, similar to the in vitro results, in vivo induction of E177R also triggered a broad spectrum of tumor suppressive programs, including cell cycle arrest, senescence and apoptosis.
Restoration of wild-type p53 in liver cancer was shown to trigger an infiltration by multiple components of the immune system which contributed to clearance of senescent tumor cells and accelerated cancer regression [21]. Using CD3 as a surrogate marker, we observed significantly increased staining at day 3 in tamoxifen-, but not mock-treated tumors (Fig. 4a and d), suggesting that immune cells might contribute to E177R-mediated leukemia regression.

Partial p53 reactivation induces regression of Myc-driven Burkitt-like lymphoma
Inspired by the results in the AML model, we decided to test whether partial restoration of p53 function is an effective treatment in other hematological malignancies such as lymphoma. We first studied EμMyc-driven B cell lymphoma, which mimics the t (8;14)(q24;q32) translocation in human Burkitt lymphoma and which is facilitated by the E177R mutation [15,48]. We crossed male EμMyc mice with female Trp53 LSL-E177R/+ ;Rosa26 CreERT2 mice to generate littermates that undergo LOH and develop p53-deficient LSL-E177R lymphomas with and without expression of CreER T2 . These p53-deficient lymphoma cells were transplanted into immunocompetent recipient mice. After successful engraftment of the lymphoma, detected by palpable lymph nodes, all animals were treated with tamoxifen for 7 consecutive days (Fig. 5a). In the group of mice with CreER T2 , tamoxifen treatment activated E177R and triggered a pronounced therapeutic response, as suggested by a significantly extended survival compared to tamoxifen-treated mice with LSL-lymphomas lacking CreER T2 (median survival 26 vs. 16 days, Log-rank (Mantel-Cox) test P < 0.0001; Fig. 5a). Excluding a confounding effect of CreER T2 , no difference in survival was observed in mice lacking an inducible p53 allele (Fig. 5b). Although tamoxifen also functions as an antiestrogen, we observed no gender-dependent differences in the p53 reactivation response (Supplemental Fig. S1).
Expression of E177R protein was readily detected by immunohistochemistry in tumors 2 days after the therapy start accompanied by strong inhibition of cell proliferation, evident from the massive reduction in BrdU-positive cells (Fig. 5c and d). Similar as in the AML model, we again detected a massive increase in apoptosis, which culminated at 2 days after therapy start and then decreased to background levels ( Fig. 5c and e). Notably, p53-deficient lymphomas are highly invasive and infiltrate into various nonlymphoid organs as evidenced by periportal invasion and spreading of lymphoma cell clusters throughout the liver parenchyma [55]. Interestingly, Fig. 4 In vivo response of p53-deficient AML to E177R activation. a Mice transplanted with LSL AML cells were sacrificed before and day 1, 3, 7 and 11 after start of 1 week of daily treatment with Tam or corn oil (mock). Samples were analyzed by immunohistochemistry for p53 E177R    we observed robust E177R expression and apoptosis induction also in periportal lymphoma infiltrates (Fig. 5f ), suggesting that partial reactivation is equally effective in a metastatic setting. At later time-points, the number of p53-positive cells dropped and relapsed tumors were mostly negative for p53 staining with only a few remaining small clusters of p53-positive cells (Fig. 5c). This suggests that recombination in this model is either incomplete or recombined LSL-E177R becomes inactivated by, for example, LOH, so that residual or emerging p53-deficient lymphoma cells escape and mediate relapse.

Partial p53 reactivation induces regression of spontaneous T-cell lymphomas in p53-null mice
As a second lymphoma model, we decided to use thymic T-cell lymphoma that accounts for the majority of all cancers detected in p53-null mice and spontaneously develops early within the first 6 months of age [49,56]. We generated Trp53 LSL-E177R/LSL-E177R mice which coexpressed the Rosa26 CreERT2 transgene (short: p53 LSL/ LSL ;CreER T2 ). We used Cre-negative mice that cannot induce E177R expression upon tamoxifen administration as a constitutive p53-null control group (short: p53 LSL/ LSL ). In our previous experiments, we estimated the median survival of p53-deficient mice to 5-6 months and therefore expected that the majority of animals should have developed tumors by 4 months of age. As soon as animals reached 4 months of age, both groups were treated 1 week daily with tamoxifen and monitored for survival. This approach allowed us to avoid a potential bias caused by p53-independent effects of tamoxifen. If re-expression of E177R can block or slow down progression of such tumors, it should result in enhanced survival. Indeed, we observed a significantly longer median survival in the E177R reactivation group (64 vs. 37 days survival after therapy end, Fig. 6a).
From this experiment, however, we could not conclude whether re-expression of E177R prevented tumor formation, limited cancer progression or induced regression of established tumors. To address this question, we examined p53 LSL/LSL animals with and without CreER T2 by MRI. Included were 4-months old or younger mice with clinical signs of thymoma (breath shortening, hunchback posture, weight loss). Animals with thymus enlargement were treated with tamoxifen as described above and monitored by MRI for up to 4 weeks. At the start of therapy both cohorts had a similar average tumor volume (Fig. 6b). In conformity with survival data, imaging showed that all p53 LSL/LSL animals without Cre showed progressive tumor growth under tamoxifen (Fig. 6b-d). In stark contrast, we observed a clear reduction of thymus size in 7 of 9 p53 LSL/LSL mice with Cre 1-2 weeks after E177R activation (Fig. 6b-d). The response to tamoxifen was similar in mice of both sexes (Supplemental Fig. S2). Moreover, all control animals reached the humane endpoint and had to be sacrificed within 2 weeks after therapy start whereas 7/9 mice with E177R activation survived for more than 4 weeks. Thus, E177R expression elicited in B-and T-cell lymphoma models a similar therapeutic response as in AML, indicating that even partial restoration of p53 function is able to induce regression in different hematopoietic cancers.
Of note, after initial remission thymic lymphomas regrew in virtually all mice. Nevertheless, many of these animals did not die of thymic lymphoma but were sacrificed because of other cancers, mostly sarcomas in varying locations and some splenic B cell lymphomas ( Fig. 6e and f ). These were not detected as the major tumor before treatment, but apparently continued to grow despite E177R activation, suggesting that not all types of cancer might be equally vulnerable to partial p53 activation.

Discussion
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][20][21][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 proapoptotic target genes was rather unexpected, because E177R and several other related cooperativity mutants are generally characterized by an apoptosis defect [43,[57][58][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][61][62][63][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 wildtype 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][70][71][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.

Conclusions
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 p53null 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.