p53 Drives Lung Cancer Regression through a TSC2/TFEB-dependent Senescence Program

p53 activation drives LUAD regression through a complex cascade of events, starting with mTORC1 inhibition and TFEB activation, which in turn triggers autophagy, senescence, macrophage recruitment, and cancer cell phagocytosis.

Abstract

Pharmacologic restoration of p53 tumor suppressor function is a conceptually appealing therapeutic strategy for the many deadly cancers with compromised p53 activity, including lung adenocarcinoma. However, the p53 pathway has remained undruggable, partly because of insufficient understanding of how to drive effective therapeutic responses without toxicity. In this study, we use mouse and human models to deconstruct the transcriptional programs and sequelae underlying robust therapeutic responses in lung adenocarcinoma. We show that p53 drives potent tumor regression by direct Tsc2 transactivation, leading to mTORC1 inhibition and Transcription factor EB (TFEB) nuclear accumulation, which in turn triggers lysosomal gene expression programs, autophagy, and cellular senescence. Senescent lung adenocarcinoma cells secrete factors to recruit macrophages, precipitating cancer cell phagocytosis and tumor regression. Collectively, our analyses reveal a surprisingly complex cascade of events underlying a p53 therapeutic response in lung adenocarcinoma and illuminate targetable nodes for p53 combination therapies, thus establishing a critical framework for optimizing p53-based therapeutics.

Significance:

Cancer therapies based on targeting the p53 pathway remain elusive. To address this gap, we unravel the detailed sequence of events governing p53-induced tumor regression in lung adenocarcinoma. These analyses reveal a TSC2–mTORC1–TFEB axis underlying p53-driven senescence and tumor regression, which suggests new strategies to perfect p53-based combination therapies for lung adenocarcinoma.

Introduction

Lung cancer is the leading cause of cancer deaths worldwide, and lung adenocarcinoma is the most prevalent type of lung cancer (1). Lung adenocarcinoma originates in the alveoli of the lungs, the sacs that mediate gas exchange, predominantly from alveolar type 2 (AT2) cells (2, 3). Oncogenic mutations in genes such as KRAS, which encodes the KRAS signaling protein, induce the transformation of AT2 cells to a tumorigenic state (reviewed in ref. 4). However, studies in mouse models have shown that AT2 cells with an oncogenic Kras mutation alone rarely progress beyond adenomas (5). In contrast, the combined expression of oncogenic Kras and inactivation of the p53 tumor suppressor gene in mice drives advanced lung adenocarcinomas. Indeed, inactivation of TP53 is observed in ∼50% of human lung adenocarcinomas and is associated with poor prognosis and therapeutic responses (1, 6, 7). Notably, restoration of wild-type (wt) p53 expression in established Kras^G12D;p53-deficient lung adenocarcinomas in mice induces tumor regression (8, 9). These findings suggest that loss of p53 not only promotes tumor development but is also required for the maintenance and progression of established tumors. Hence, pharmacologic restoration of p53 constitutes an attractive approach for treating human lung adenocarcinoma.

Although half of human cancers, including lung adenocarcinoma, display mutations in TP53, most commonly missense mutations, the other half retains the wt TP53 gene (10, 11). In the latter case, the negative regulators of p53, MDM2 and MDM4, which cooperate to inhibit p53 transcriptional activity and to subject p53 to ubiquitin-mediated proteolysis, are commonly overexpressed in cancer to dampen p53 tumor suppressor activity (12). Translational efforts have focused on finding approaches to reactivate p53, either by reactivating mutant p53 molecules pharmacologically or liberating wt p53 (referred to as p53^wt hereafter) from MDM2/4 (13, 14). However, due to challenges in identifying molecules with adequate efficacy and without associated toxicity, there are still no standard-of-care strategies to target the p53 pathway in cancer treatment (15). Moreover, little emphasis has been placed on using these approaches in epithelial cancers such as lung adenocarcinoma (16). The recent success in identifying inhibitors of the previously deemed “undruggable” cancer driver KRAS (17) provides new inspiration for similarly targeting p53. Gaining a detailed understanding of the molecular pathways through which p53 drives tumor regression is critical for the thus-far elusive goal of developing logical and reliable strategies to treat cancer through p53 restoration.

p53 is a transcriptional activator that suppresses cancer through the transcriptional induction of a network of target genes involved in restraining various protumorigenic cellular processes (18). In response to cellular stress signals, such as oncogenic stimuli, p53 activates genes that inhibit cellular expansion by inducing cell-cycle arrest, cellular senescence, and apoptosis (reviewed in ref. 19). p53 also regulates other cellular functions, such as metabolism and differentiation, which contribute to its tumor suppression function (20–22). Indeed, we recently showed that p53-mediated suppression of lung adenocarcinoma in mice and humans occurs by p53 driving the differentiation of Kras^G12D-expressing AT2 cells into AT1-like cells, the thin specialized cells that mediate gas exchange in alveoli (20). Although previous studies established that restoration of p53^wt in mouse models of lung adenocarcinoma can induce significant lung adenocarcinoma regression, the molecular pathways underlying these effects were not explored in detail (8, 9). Elucidating the transcriptional programs and associated sequelae underlying tumor regression upon p53 restoration will provide new insights into therapeutic strategies for lung adenocarcinoma. Notably, as highlighted for RAS (23), the more that is understood about a therapeutic response pathway, the greater the opportunity to rationally identify key nodes for combination therapies in an effort to optimize therapies, anticipate therapeutic resistance, or mitigate side effects triggered by these agents.

Here, we sought to understand the molecular pathway underlying a p53 therapeutic response in lung adenocarcinoma by leveraging a panel of knock-in mice expressing p53 transcriptional activation domain (TAD) mutants with a range of transactivation properties, including mice with alterations in the first (p53^L25Q;W26S, referred to as p53^25,26), second (p53^F53Q;F54S, referred to as p53^53,54), or both TADs (p53^{L25Q,W26S,F53Q,F54S}, referred to as p53^25,26,53,54; ref. 24). This allelic series of p53 mutants has allowed us to define the transcriptional programs through which p53 restoration promotes tumor regression in Kras^G12D-driven mouse lung adenocarcinoma models at a detailed molecular level. We thus uncovered an unexpectedly complex tumor regression pathway in which p53 induces Tsc2, a negative regulator of mTORC1, leading to the activation of the master lysosomal biogenesis factor transcription factor (TFEB) and consequent induction of a lysosome/autophagy gene expression program, followed by cellular senescence, culminating in macrophage-mediated lung adenocarcinoma cell clearance. This detailed deconstruction of the molecular cascade driving tumor regression following p53 reactivation reveals multiple nodes that could be targeted in conjunction with p53 restoration, thus laying the groundwork for improving p53-based therapies.

Results

Restoration of the p53^53,54 Mutant Induces Potent Lung Adenocarcinoma Regression

To understand the mechanisms by which p53 restoration can induce tumor regression in Kras^G12D-driven lung adenocarcinoma, we leveraged our panel of p53 TAD mutant knock-in mice expressing mutants in TAD1 (p53^25,26), TAD2 (p53^53,54), or both TAD1 and TAD2 (p53^25,26,53,54; Supplementary Fig. S1A). Previous characterization of the p53^25,26,53,54 mutant revealed a complete deficit in transcriptional activation potential and tumor suppression, underscoring the importance of p53 transcriptional activity for tumor suppressor function (24, 25). The p53^25,26 TAD1 mutant, meanwhile, was greatly impaired in transactivation of many target genes but retained the ability to activate a small subset of largely noncanonical p53 target genes and to suppress the development of multiple tumor types, leading to the idea that noncanonical p53 target genes underlie tumor suppression (24). Finally, our more recent studies characterized p53^53,54 as a “super–tumor suppressor” in both lung adenocarcinoma and pancreatic ductal adenocarcinoma (PDAC), and this activity tracked with the capacity to hyperactivate a subset of p53 target genes (20, 24, 26). We analyzed these p53 alleles in mouse cohorts carrying two flippase (FLP) recombinase–regulated alleles: one Kras^G12D allele silenced by an upstream transcriptional stop element flanked by Frt recombination sites (Frt-Stop-Frt or FSF) and one p53 conditional knockout allele flanked by Frt sites (27, 28). The other p53 allele in these cohorts was a Cre-activatable wt or TAD mutant p53 allele silenced by an upstream transcriptional stop element flanked by loxP recombination sites [Lox-Stop-Lox (LSL)], which acts as a p53 null (p53^null) allele in the absence of Cre (Fig. 1A). These cohorts also carried a Rosa26^CreER (where CreER indicates Cre estrogen receptor) allele (29), which allows ubiquitous Cre expression and activation upon tamoxifen treatment. The introduction of FLP recombinase–expressing adenoviruses induces Kras^G12D expression and deletion of the p53^Frt allele to generate a completely p53^null state, thereby driving lung adenocarcinoma initiation and progression (Fig. 1B). After tumor development, we were then able to evaluate the effect of restoration of p53^wt or p53 TAD mutants by activating the CreER with tamoxifen to induce the expression of the various p53 alleles. Mice without a Rosa26^CreER knock-in allele and injected with tamoxifen served as p53^null controls.

Figure 1.

p53^53,54 restoration induces potent lung adenocarcinoma regression. A, Mouse cohorts used in the study and configuration of mutant alleles. FLP recognition target (Frt) sites can be recombined by FLP recombinase. loxP recombination sites can be recombined by Cre recombinase. n indicates the number of mice in each cohort. B, Lung cancer study schematic. Indicated mice were infected with Ad-CMV-FLPo virus to initiate lung adenocarcinoma. After 16 weeks, mice were treated with tamoxifen for 3 consecutive days per week for 2 weeks. Lungs were collected 2 weeks after the first dose of tamoxifen. C, Representative H&E images of lung adenocarcinoma burden in each cohort. Scale bar, 5 mm. D, Quantification of lung tumor burden (percentage of lung tumor area relative to total lung area) for each mouse cohort (n = 6 for p53^null, n = 9 for p53^wt-restored, n = 7 for p53^53,54-restored, n = 5 for p53^25,26-restored, and n = 8 for p53^25,26,53,54-restored). P values were calculated by one-way ANOVA with Dunnett’s post hoc testing for pairwise comparisons. E, The tumors from mouse cohorts in B were analyzed and graded by GLASS-AI. The proportions of each tumor grade relative to total tumor area in mice of each genotype were compared by a two-way ANOVA with Tukey post hoc test for pairwise comparisons. F, Representative IF images of p53 and BrdU costaining in lung adenocarcinomas from each cohort (n = 3 mice for all cohorts). Arrows indicate examples of BrdU+p53+ cells. DAPI marks nuclei. Scale bar, 50 μm. G, Quantification of proliferation in mouse lung adenocarcinoma samples. The percentage of p53-positive cells that were BrdU-positive per 200× field of lung adenocarcinoma from each mouse cohort was quantified (n = 3 mice for all cohorts; 4–5 fields per mouse were analyzed). H, Representative IF images of p53 and AGER costaining in lung adenocarcinomas from each cohort (n = 3 mice for all cohorts). DAPI marks nuclei. Scale bar, 50 μm. I, AGER quantification in mouse lung adenocarcinoma samples. The percentage of p53-positive cells that were AGER-positive per 200× field of lung adenocarcinomas from each mouse cohort (n = 3 for all cohorts; 2 fields of each mouse were analyzed) was quantified. Scale bar, 50 μm. P values were calculated by one-way ANOVA with the Dunnett post hoc test for pairwise comparisons.

Expression of the Kras^G12D mutant allele and deletion of the p53^Frt allele were achieved by intratracheally transducing 8- to 12-week-old Kras^FSF-G12D;p53^LSL-X/Frtor Kras^FSF-G12D;p53^LSL-X/Frt;Rosa26^CreER mice (where X denotes different reactivatable p53 alleles) with adenoviral (Ad)-Cytomegalovirus (CMV)-FLPo (where FLPo represents mouse codon-optimized FLP). Lungs collected from a subset of mice at 16 weeks showed that tumors were already at an advanced adenocarcinoma stage (Supplementary Fig. S1B) with efficient deletion of the p53^Frt allele confirmed by IHC staining (Supplementary Fig. S1C). The remaining mice were subjected to reactivation of p53^wt or p53 TAD mutants at 16 weeks by delivering tamoxifen to the mice. Lungs were collected 2 weeks later for tumor burden quantification and further analyses. Successful restoration of different p53 variants was confirmed by immunostaining (Supplementary Fig. S1C; also see below). Quantification of tumor burden revealed that restoration of p53^wt significantly reduced tumor burden compared with the p53^null mice (Fig. 1C and D). In contrast, restoration of the transcriptionally dead p53^25,26,53,54 mutant failed to diminish tumor burden relative to the p53^null mice, indicating that p53 transcriptional activity is critical for inhibiting tumor growth upon reactivation. Surprisingly, the p53^25,26 mutant, which can suppress tumor development in various mouse cancer models (24, 25), also failed to diminish tumor burden. These findings suggest that the limited transcriptional activity displayed by this mutant is insufficient for p53 to inhibit tumor burden in an acute restoration context. Notably, mice with restoration of p53^53,54 exhibited the lowest tumor burden among all the mice. These findings suggest that the hyperactivity of this mutant extends beyond initial tumor suppression to a therapeutic context of tumor growth inhibition upon reactivation. Moreover, the decreased tumor burden in mice 18 weeks after tumor induction (2 weeks after restoration of p53^53,54 at 16 weeks) compared with mice before restoration at 16 weeks after tumor induction suggests that p53^53,54 restoration drives lung adenocarcinoma regression (Supplementary Fig. S1D). Beyond affecting tumor burden, histopathologic analysis demonstrated that restoration of p53^53,54 and p53^wt was associated with diminished local infiltration of lung adenocarcinoma cells into the stroma relative to the expression of other mutants (Supplementary Fig. S1E and S1F). Moreover, necrosis, a feature of more aggressive lung tumors in humans (30), was widely seen in p53^null tumors or tumors with p53^25,26 or p53^25,26,53,54 restoration but rarely seen in tumors with p53^wt or p53^53,54 restoration (Supplementary Fig. S1E).

To compare the effects of p53^wt and p53^53,54 restoration on tumors at a more detailed level, we analyzed tumor grade using a recently developed machine learning tool [Grading of Lung Adenocarcinoma with Simultaneous Segmentation by Artificial Intelligence (GLASS-AI; ref. 31)]. This pipeline leverages a neural network trained on histologic images of mouse lung adenocarcinoma to segment and grade lung tumors to reveal intratumor heterogeneity with higher resolution and less subjectivity than with manual analysis. Tumors were classified into grades 1 to 4 based on histologic appearance, with increasing numbers representing more malignant grades. GLASS-AI analysis indicated that restoration of p53^53,54 resulted in more lung adenocarcinomas categorized as grade 3 than with restoration of p53^wt or with p53 inactivation, suggesting that the p53^53,54 tumors, in general, are at a less advanced stage than p53^wt-restored and p53^null tumors (Fig. 1E). With GLASS-AI, it is also possible to visualize the overall distribution of intratumor heterogeneity by examining the proportion of each tumor grade within a tumor (Supplementary Fig. S2A). Among all the tumors detected, the restoration of p53^53,54 and p53^wt resulted in fewer tumors with grade 4 regions than in p53^null mice. Moreover, among the tumors classified as grade 4, a larger percentage of lower-grade (grades 1–3) tumor regions were detected in the p53^53,54- or p53^wt-restored tumors than in p53^null tumors. Together, the observations that p53 reactivation reduces both tumor burden and grade of malignancy suggest that p53 acts both by promoting lung adenocarcinoma regression and by preventing tumor progression. The fact that this hyperactive p53 variant drives more potent responses than p53^wt underscores its utility as a tool for deconstructing p53 functions critical for therapeutic responses and, therefore, for developing novel therapeutic strategies.

p53^53,54 Induces Antiproliferative Cellular Responses

To gain insight into the underlying cellular mechanisms by which p53 restoration dampens lung adenocarcinoma growth, we first sought to assess how p53 affects cellular proliferation and apoptosis. To this end, we performed immunostaining for bromodeoxyuridine (BrdU) incorporation, after BrdU pulse-labeling, to assess cell division and for cleaved caspase 3 (CC3) to measure apoptosis. We found that the restoration of p53^wt or p53^53,54 significantly reduced cell proliferation relative to p53^25,26 or p53^25,26,53,54 restoration (Fig. 1F and G). Apoptosis was occasionally observed in the tumors with reactivation of p53^wt and p53^53,54 but not in the p53^null tumors (Supplementary Fig. S2B and S2C). As we previously observed that p53 promotes AT2 to AT1 differentiation during tumor suppression (20), we also assessed the ability of each p53 variant to drive expression of AT1 differentiation markers. We found that the restoration of either p53^wt or p53^53,54 promoted efficient expression of the AT1 cell markers Advanced Glycosylation End-product Specific Receptor (AGER) and Podoplanin (PDPN) in lung adenocarcinomas, in contrast to the restoration of p53^25,26 or p53^25,26,53,54, which failed to significantly upregulate AT1 marker expression (Fig. 1H and I; Supplementary Fig. S3A). The capacity of specific p53 variants to drive AT1-like differentiation correlated closely with the ability of these p53 variants to block tumor growth after restoration, suggesting that the ability of p53 to inhibit cell-cycle progression and promote AT1-like differentiation contributes to the inhibition of tumor progression after reactivation. Indeed, costaining for BrdU and AGER in mouse lungs after p53^53,54 restoration showed that cells with robust AGER expression failed to display BrdU positivity, consistent with cell-cycle exit being required for differentiation (Supplementary Fig. S3B).

To develop a tractable system for delineating transcriptional programs through which p53 drives tumor regression, we next sought to generate a cellular model that reflected the in vivo phenotypes we observed in an in vitro context. To this end, we generated cell lines derived from Kras^LA2/+; p53^LSL-X/LSL-X mice (where X denotes different reactivatable p53 alleles), in which spontaneous recombination of a latent Kras^G12D allele drives lung adenocarcinoma (Fig. 2A; ref. 32). In this model, the lung tumors that form and the cell lines derived from them are effectively p53^null because of the presence of the upstream stop cassette silencing p53. Importantly, we can control the temporal restoration of p53^wt or the TAD mutants in the cell lines through the transduction of CMV-Cre adenoviruses (Ad-Cre). Infecting with Ad-Empty provides a p53^null control. After infecting the lung adenocarcinoma cell lines with Ad-Cre, we noted efficient reactivation of p53^wt or TAD mutants in almost 100% of cells, validating this as a system to examine the transcriptional programs and cellular responses triggered by p53 reactivation (Fig. 2B; Supplementary Fig. S4A and S4B). By comparing Ad-Cre– and Ad-Empty–infected cells, we found that p53^wt and p53^53,54 reactivation inhibited lung adenocarcinoma cell proliferation after 48 hours, whereas p53^25,26 or p53^25,26,53,54 expression did not affect proliferation (Fig. 2B and C), consistent with our in vivo findings. Although p53^wt and p53^53,54 reactivation similarly dampened cell proliferation after 48 hours, p53^53,54 reactivation induced more robust inhibition of cell proliferation than p53^wt 24 hours after reactivation (Supplementary Fig. S4C). In addition to cell-cycle arrest, restoration of p53^wt or p53^53,54 in lung adenocarcinoma cells induced AT1 cell differentiation markers, as we observed in mouse lung adenocarcinomas (Supplementary Fig. S4D). Thus, these cell lines mimic the p53-related biology that we observed in mice. Moreover, these findings together suggest that restoration of p53^wt and p53^53,54 triggers proliferative arrest, accompanied by AT1 differentiation, which likely contributes to tumor regression.

Figure 2.

p53^53,54 restoration induces potent arrest and hyperactivates p53 target genes in lung adenocarcinoma cells. A, Schematic of lung adenocarcinoma cell generation and use. Cell lines were derived from mice harboring the Kras^LA2 allele, which spontaneously recombines to express KRAS^G12D, and are homozygous for LSL p53 alleles (wt or different TAD mutants). Mice were aged until spontaneous lung tumor formation, at which time cell lines were established. In experiments, cell lines were either transduced with Ad-Empty (to retain p53^null status) or Ad-Cre (to restore p53 expression, including p53^wt, p53^53,54, p53^25,26, and p53^25,26,53,54). B, Representative IF images of p53 and BrdU costaining 48 hours after Ad-Cre transduction into lung adenocarcinoma cell lines. DAPI marks nuclei. Separated channels are shown on the right and the merge on the left. Scale bar, 50 μm. C, Effects of p53 mutants on proliferation. Proliferation was assessed by quantifying BrdU incorporation in Ad-Empty– and Ad-Cre–infected cell lines, using data from B. In Ad-Empty–infected cells (n = 3 cell lines for each genotype), the percentage of cells (defined by DAPI staining; n = 200 cells) that were BrdU-positive was quantified. In the Ad-Cre–infected cell lines (n = 3 for each genotype), the percentage of p53-positive cells (n = 200 cells) that were BrdU-positive was quantified. D, PCA of RNA-seq samples from lung adenocarcinoma cell lines of the indicated genotypes (n = 3 or 4 cell lines for each genotype). E, Heatmap showing the expression of p53-dependent genes (differentially expressed between p53^wt-restored and p53^null RNA-seq samples) in lung adenocarcinoma cells of all genotypes (n = 3 or 4 for each genotype). F, qRT-PCR of a panel of classical and noncanonical p53 target genes in lung adenocarcinoma cells of all genotypes, normalized to β-actin and graphed relative to levels in p53^null cells.

Restoration of p53^53,54 Drives Autophagy and Lysosome Programs

To elaborate on the transcriptional programs underlying the potent tumor regression capacity of p53^53,54, we next performed bulk RNA sequencing (RNA-seq) on Kras^G12D;p53^{LSL-wt/LSL-wt}, Kras^G12D;p53^{LSL-53,54/LSL-53,54}, Kras^G12D;p53^{LSL-25,26/LSL-25,26}, and Kras^G12D;p53^{LSL-25,26,53,54/LSL-25,26,53,54} cell lines 48 hours after Ad-Cre transduction when p53 protein accumulation was robust (Supplementary Fig. S4B). Ad-Empty–infected cells provided p53^null controls. Principal component analyses (PCA) of RNA-seq samples revealed that p53^null, p53^25,26-restored, and p53^25,26,53,54-restored samples clustered together, whereas p53^53,54- and p53^wt-restored samples were more distinct, with p53^53,54-restored samples grouping the farthest from p53^null samples (Fig. 2D; Supplementary Table S1). Using a list of p53-regulated genes derived by comparing p53^wt and p53^null samples (Supplementary Fig. S5A; Supplementary Table S2), we assessed how the TAD mutants affected the expression of these genes. We discovered that p53^53,54 upregulated a subset of p53-activated genes to a greater extent than p53^wt, corroborating the notion that this mutant is hyperactive for the activation of some p53 target genes (Fig. 2E; Supplementary Fig. S5B; Supplementary Table S3). We also found that p53^25,26 does not robustly activate most p53-regulated genes but still displays transcriptional activity on some target genes, consistent with previous analyses in other cell types (Supplementary Fig. S5C; Supplementary Table S4; ref. 24). Finally, p53^25,26,53,54 transcriptionally behaved essentially as a p53^null allele (Supplementary Fig. S5D; Supplementary Table S5). qRT-PCR analysis of the expression of a panel of largely noncanonical p53 target genes that we previously identified (24) supported these conclusions (Fig. 2F). Overall, the transcriptional activation potential of these p53 TAD mutants aligns well with their tumor regressor capacities: p53^53,54 displays transcriptional hyperactivity relative to p53^wt, whereas p53^25,26 and p53^25,26,53,54 are severely or completely compromised, respectively, in transcriptional activation function.

As restoration of p53^53,54 inhibited lung adenocarcinoma progression better than restoration of p53^wtin vivo, we sought to dissect the mechanisms underlying this phenotype by comparing the transcriptional programs induced by restoration of p53^wt and p53^53,54. p53^wt expression resulted in the upregulation of 2,549 genes compared with p53^null cells, including well-known p53 targets, such as Mdm2 and Ccng1 (Supplementary Fig. S5A). Approximately 27% of these upregulated genes were associated with p53 chromatin immunoprecipitation (ChIP) peaks identified in our mouse embryonic fibroblast (MEF) ChIP sequencing (ChIP-seq) dataset (33), suggesting that p53 directly regulates these genes (Supplementary Fig. S5E; Supplementary Table S6). Functional annotation of genes upregulated in p53^wt cells compared with p53^null cells revealed a strong enrichment of autophagy and lysosome signatures (Fig. 3A). In contrast, genes upregulated in p53-deficient cells were enriched for pathways involved in cell-cycle progression, consistent with the result that p53^null cells are more proliferative than p53^wt-restored cells (Supplementary Fig. S5F; Supplementary Table S7; ref. 33). p53^53,54 restoration resulted in the induction of 3,987 genes, the most of any p53 allele (Supplementary Fig. S5B), consistent with its transcriptional hyperactivity. Twenty-five percent of p53^53,54 upregulated genes were p53 bound in ChIP-seq data (Supplementary Fig. S5G; Supplementary Table S8; ref. 33). Gene Ontology analyses to define signatures induced in p53^53,54 cells compared with p53^null cells uncovered a strong enrichment of macroautophagy, autophagy, and lysosome genes (Fig. 3B). To better understand why restoration of p53^53,54 induces stronger tumor regression, we compared the expression of the macroautophagy, autophagy, and lysosome gene sets in p53^null, p53^wt-restored, and p53^53,54-restored lung adenocarcinoma cells. This analysis confirmed hyperactivation of many of these genes by p53^53,54 relative to p53^wt, especially genes involved in the initiation of macroautophagy/autophagy, such as Atg5 and Atg12, and lysosome-associated genes, including Ctsd and Atp6v0d1 (Fig. 3C and D; Supplementary Fig. S6A).

Figure 3.

p53^53,54 restoration hyperactivates autophagy and lysosomal gene expression programs. A, Bar graph depicting the top 10 Gene Ontology terms generated by Metascape and their respective P values for genes activated in p53^wt-restored cells relative to p53^null cells. P values were calculated by the hypergeometric distribution. B, Bar graph depicting the top 10 Gene Ontology terms generated by Metascape and their respective P values for genes hyperactivated in p53^53,54-restored cells relative to p53^null cells. P values were calculated by the hypergeometric distribution. C, Heatmap of genes from the macroautophagy gene list (GO list: 0016236) that are induced in RNA-seq from p53^53,54-restored lung adenocarcinoma (LUAD) cells relative to p53^null lung adenocarcinoma cells (n = 3 for each genotype). D, Heatmap of genes from the lysosome gene list (GO list: 0005764) that are induced in RNA-seq from p53^53,54-restored lung adenocarcinoma cells compared with p53^null lung adenocarcinoma cells (n = 3 for each genotype). E, Representative immunoblot for p53 and LC3-I/LC3-II in indicated cell lines after infection with Ad-Empty or Ad-Cre for 72 hours and treatment of some samples with 0.1 μmol/L bafilomycin (BafA1) for 2 hours (n = 3). GAPDH serves as a loading control. F, Representative immunoblot for p53 and LAMP1 in indicated lung adenocarcinoma cell lines after infection with Ad-Empty or Ad-Cre for 96 hours (n = 3). GAPDH serves as a loading control. G, (Top) Representative IF images of LAMP1 in indicated lung adenocarcinoma cell lines after infection with Ad-Empty or Ad-Cre for 96 hours (n = 2 cell lines for each genotype) and (bottom) representative images of LAMP1 IF in p53^X-restored lung adenocarcinoma samples in vivo (where X = wt; 25,26; 53,54; or 25,26,53,54; n = 3 mice for each genotype). DAPI marks nuclei. Scale bar, 50 μm. H, Quantification of the total LAMP1 area per cell for each cell line in G (n = 30 images for each genotype). P values were calculated by a one-way ANOVA with the Dunnett post hoc test for pairwise comparisons.

To validate the autophagy and lysosomal gene expression programs induced by p53^53,54 and p53^wt, we next examined these processes at the cellular level. To examine autophagy, we performed immunoblot analysis for the canonical autophagy marker LC3-II in lung adenocarcinoma cells after restoration of p53^wt or p53 TAD mutants, with or without bafilomycin treatment. Bafilomycin inhibits autophagic flux and thus reveals whether increased LC3-II reflects enhanced autophagy or a block in autophagic flux (34). We found that restoration of p53^53,54 induced the highest LC3-II levels among the four genotypes and that levels were further increased by bafilomycin treatment, indicating that p53^53,54 induces autophagy (Fig. 3E). Restoration of p53^53,54 also induced strikingly high levels of LAMP1, a lysosome marker, both in lung adenocarcinoma cells in vitro and in lung adenocarcinomas in vivo (Fig. 3F–H; Supplementary Fig. S6B). Moreover, immunofluorescence (IF) staining for LAMP1 indicated that p53^53,54 and p53^wt restoration in lung adenocarcinoma cells drove significantly higher lysosomal content than that observed in p53^25,26-restored, p53^25,26,53,54-restored, or p53^null lung adenocarcinoma cells (Fig. 3G and H; Supplementary Fig. S6B). Together, these findings suggest that p53 drives lysosomal and autophagy programs during tumor regression.

Restoration of p53^53,54 Promotes TFEB Nuclear Translocation in Lung Adenocarcinoma

To understand how p53^53,54 and p53^wt induce macroautophagy and lysosomal programs at the transcriptional level, we sought to identify transcription factors driving these programs. We first queried whether these genes were directly regulated by p53. We overlapped the autophagy/macroautophagy and lysosomal gene signatures with a list of p53-bound genes identified by ChIP-seq. We found that ∼20% of the genes induced in these signatures are also bound by p53^wt, suggesting that p53 directly induces some of these genes but also exerts significant indirect effects on their expression (Supplementary Fig. S6C).

To understand how p53 might indirectly modulate the expression of autophagy and lysosome genes, we tested whether this might be via TFEB, the basic loop–helix transcription factor known to regulate both lysosomal biogenesis and macroautophagy genes, and which is regulated at the level of nuclear localization (reviewed in ref. 35). Using a list of 471 direct human TFEB target genes identified by genomic approaches (36), we focused on those genes with known roles in lysosomal function. We found that most of these genes are induced by both p53^wt and p53^53,54, with some showing hyperactivation by p53^53,54, in our lung adenocarcinoma cells (Supplementary Fig. S7A). Next, by overlapping the autophagy/macroautophagy and lysosomal gene signatures with the 443 mouse orthologues of the 471 human TFEB target gene list, we found that ∼15% to 40% of these genes are directly regulated by TFEB and that these genes are completely different from the genes that are bound by p53, suggesting that p53 also exerts indirect effects on autophagy/lysosome gene expression through TFEB (Supplementary Figs. S6C and S7B). As TFEB is activated by shuttling from the cytoplasm to the nucleus, we examined TFEB localization in lung adenocarcinoma cells of different genotypes. Subcellular fractionation of cells with restoration of either p53^wt or other p53 TAD mutants confirmed that p53^53,54-restored cells have the highest levels of total and nuclear TFEB (Fig. 4A). Moreover, IF analysis revealed that in contrast to p53^25,26- and p53^25,26,53,54-restored cells, which exhibited cytoplasmic TFEB, p53^53,54 restoration drove high levels of nuclear TFEB both in vitro in lung adenocarcinoma cells and in vivo in lung adenocarcinomas (Fig. 4B and C; Supplementary Fig. S8A). Although TFEB upregulation in the nucleus was limited 48 hours after p53^wt restoration (Fig. 4B and C), it became more prominent in these cells 72 hours after restoration (Supplementary Fig. S8B). Thus, p53 promotes robust nuclear TFEB activity, which then stimulates lysosome and autophagy gene expression programs.

Figure 4.

p53^53,54 restoration promotes TFEB nuclear translocation via Tsc2 activation. A, Immunoblot analysis of TFEB and p53 in cytoplasmic and nuclear fractions from indicated cell lines after infection with Ad-Empty or Ad-Cre for 48 hours. Markers of the nuclear fraction and cytoplasmic fraction are lamin A/C and GAPDH, respectively (T, total cell lysate; C, cytoplasmic fraction; N, nuclear fraction; n = 4 replicates). B, Representative IF images of p53 and TFEB in indicated lung adenocarcinoma cell lines. p53^LSL-wt and p53^LSL-53,54 cells were assessed 48 hours after the Ad-Cre infection. Arrows indicate examples of overlap of p53 and TFEB in the nucleus. DAPI marks nuclei. Scale bar, 50 μm. C, Quantification of nuclear TFEB in indicated cell lines. The percentage of p53-positive cells with TFEB staining in the nucleus and in the cytoplasm per 200× field was quantified (n = 3 cell lines for each genotype and 1–2 images for each cell line). P values were calculated by one-way ANOVA with the Dunnett test with multiple comparisons. D, Representative immunoblots for p53, TSC2, p-4EBP1, and 4EBP1 in indicated cell lines 48 hours after Ad-Cre transduction. GAPDH serves as a loading control. E, Normalized gene expression of indicated genes from RNA-seq samples in lung adenocarcinoma cells of all genotypes (n = 3 or 4 for each genotype). F, (Top) MEF ChIP-seq data showing p53-bound peak in the Tsc2 gene. The p53 response element is shown with the nucleotides matching the consensus motif in capital letters and the core motifs in red. There is a three-base pair spacer underlined between the two half sites. (Bottom) ChIP analysis of p53 binding to the Tsc2 response element, relative to input, in doxorubicin-treated E1A;HRas^G12V MEFs. IgG IP serves as a control. G, Representative IF images of p53 and BrdU costaining in indicated cells. Arrows indicate examples of BrdU+p53+ cells in samples with restoration of p53^wt or p53^53,54. Arrows indicate examples of BrdU+ cells in p53^null cells. DAPI marks nuclei. Scale bar, 50 μm. H, Quantification of BrdU-positive cells in indicated cell lines. The percentage of BrdU-positive cells among all cells per 200× field was quantified (n = 2 cell lines for each genotype and 6–10 images for each cell line).

TSC2 and TFEB Are Important for p53-Mediated Proliferative Arrest

TFEB translocation has been shown to be triggered by mTORC1 inhibition (37). Consistent with p53 acting through mTORC1, we found that p53^wt inhibited mTORC1 signaling and that p53^53,54 did so more robustly, as evidenced by reduced levels of p-4EBP1 (Fig. 4D). To understand how p53 promoted mTORC1 inhibition, we explored known negative regulators of mTORC1. By querying our RNA-seq data, we found that restoration of p53^wt and p53^53,54 leads to the induction of specific genes encoding known mTORC1 inhibitors, including Ddit4I, Sesn2, and Tsc2 (Fig. 4E; ref. 38). Notably, p53^53,54 displayed a dramatically enhanced ability to activate Tsc2 relative to p53^wt. This observation, coupled with the fact that Tsc2 is a direct p53 target gene, as seen in ChIP-seq data and confirmatory ChIP assays in both MEFs and embryonic stem (ES) cells (Fig. 4F; Supplementary Fig. S8C), suggests that the enhanced tumor suppressive activity of p53^53,54 may be via TSC2. Indeed, we found that Tsc2 knockdown impeded the downregulation of mTORC1 upon p53^53,54 expression and greatly attenuated the cell-cycle arrest driven by reactivation of p53^53,54 (Fig. 4G and H; Supplementary Fig. S8D; also see below), suggesting that the ability of TSC2 to inhibit mTORC1 is important for p53^53,54-mediated cell-cycle arrest. Tfeb knockdown similarly diminished p53^53,54-driven cell-cycle arrest (Fig. 4G and H). Thus, p53^53,54 induces cell-cycle arrest in a manner dependent on TSC2 and TFEB.

Restoration of p53^53,54 Triggers Robust Senescence

It has been reported that autophagy promotes oncogene-induced senescence (39). Given the potent autophagy-inducing activity of p53^53,54, we tested whether p53^53,54 might limit tumor cell growth by inducing cellular senescence. Staining for senescence markers—senescence-associated β-galactosidase (SA-β-gal) and p21—showed that restoration of p53^53,54 induced robust senescence both in lung adenocarcinoma cells in vitro and in lung adenocarcinoma tumors in vivo, whereas restoration of p53^wt only induced weak senescence. Moreover, no senescent cells were observed with restoration of any other p53 TAD mutant (Fig. 5A and B).

Figure 5.

p53^53,54 restoration drives a potent senescence response in lung adenocarcinoma. A, Representative images of SA-β-gal staining and p21 IHC on in vivo and in vitro samples of each genotype (n = 3 mice and 2 cell lines for each genotype). Scale bar, 50 μm. B, Quantification of analysis from A showing the percentage of SA-β-gal–positive cells among all cells per 100× field. C, Representative immunoblots for p53, TSC2, p-S6, S6, p21, and TFEB in indicated cell lines after siControl/siTSC2/siTFEB transfection and Ad-Empty/Ad-Cre infection (n = 2 cell lines for each genotype). GAPDH serves as a control and is shown for each separate blot performed. D, Representative images of SA-β-gal staining for p53^null and p53^53,54-restored cells after siControl/siTSC2/siTFEB transfection (n = 2 cell lines for each genotype). E, Quantification of analysis from D showing the percentage of SA-β-gal–positive cells among all cells per 100× field (n = 15 images for each genotype). F, Representative immunoblots for p53, ATG5, p62, and p21 in indicated cell lines after siControl/siATG5 transfection and Ad-Empty/Ad-Cre infection (n = 3). GAPDH serves as a loading control and is shown for each separate blot performed. G, Representative images of SA-β-gal staining for p53^null and p53^53,54-restored cells after siControl/siATG5 transfection (n = 3 cell lines for each genotype). H, Quantification of analysis from G showing the percentage of SA-β-gal–positive cells among all cells per 100× field. P values were calculated by one-way ANOVA with the Tukey post hoc test for pairwise comparisons.

We next sought to delineate the signaling pathway through which p53^53,54 drives senescence. To this end, we first queried the importance of the TSC2–mTORC1–TFEB axis for p53^53,54-induced senescence. Consistent with Tsc2 knockdown greatly alleviating cell-cycle arrest upon p53^53,54 expression (Fig. 4G and H), we found that Tsc2 knockdown diminished senescence dramatically, as assessed by p21 expression and SA-β-gal staining after p53^53,54 restoration (Fig. 5C–E). Analyses of the expression of canonical p53 target genes, including Mdm2, p21, Bax, Noxa, and Zmat3, revealed that Tsc2 knockdown attenuates their expression, suggesting that the p53 pathway is dampened with Tsc2 knockdown (Supplementary Fig. S9), consistent with previous findings that mTORC1 activation stimulates MDM2 activity (40). We found further that Tfeb knockdown significantly reduced senescence driven by p53^53,54 restoration, again measured by BrdU incorporation, p21 expression, and SA-β-gal staining (Figs. 4G, H, and 5C–E). Thus, p53^53,54 induces Tsc2, which inhibits mTORC1, leading to TFEB translocation and senescence.

To interrogate the importance of autophagy in promoting p53-driven senescence, we performed Atg5 knockdown experiments. Using siRNA directed against Atg5 and then reactivating p53^53,54 by Ad-Cre virus infection of lung adenocarcinoma cells, we observed efficient Atg5 knockdown and inhibition of autophagy, indicated by p62 upregulation (Fig. 5F; ref. 41). We found further that Atg5 knockdown significantly reduced senescence in cells upon p53^53,54 restoration, as indicated by diminished p21 expression and reduced SA-β-gal staining, relative to siRNA control-treated cells (Fig. 5F–H). Together, our studies of p53^53,54 have revealed a TSC2–mTORC1–TFEB axis required for p53-driven senescence.

MDM2 Antagonist Treatment of Mouse and Human p53^wt Cells Mimics p53^53,54 Action

Although analysis of p53^53,54 helped delineate the sequelae underlying p53-driven lung adenocarcinoma regression, the ultimate therapeutic goal is to discover a strategy to convert p53^wt to a more active p53^53,54-like state. As our previous work in fibroblasts revealed that p53^53,54 protein accumulates to higher levels than p53^wt and that it binds MDM2 less avidly than p53^wt, an observation we confirmed in lung adenocarcinoma cells (Supplementary Fig. S10A), we investigated whether the MDM2 inhibitor (MDM2i) navtemadlin (42) could augment p53^wt activity to mimic p53^53,54 action. Indeed, IF and Western blot analyses of mouse lung adenocarcinoma cells showed that navtemadlin treatment increased p53^wt accumulation as well as promoting Tsc2 induction, mTORC1 inhibition, TFEB nuclear translocation, LAMP1 upregulation, and cellular senescence to levels similar to those observed with p53^53,54 (Fig. 6A–C; Supplementary Fig. S10B). We also utilized human A549 lung adenocarcinoma cells, which retain wt TP53, and isogenic A549/TP53^KO cells that we generated using CRISPR/Cas9-mediated genome editing. As in mouse cells, we found that navtemadlin treatment triggered TSC2, LAMP1, and LC3-II protein induction, as well as cellular senescence, only in the presence of p53 (Fig. 6D–F). Finally, we mined expression data from a lung adenocarcinoma cell line with intact TP53 in which RNA-seq was used to compare gene expression profiles in the presence or absence of the MDM2i milademetan (43). Gene set enrichment analysis revealed that p53 and senescence signatures were induced and that mTOR signaling was repressed upon milademetan treatment (Supplementary Fig. S10C). Taken together, these findings suggest that p53^wt activation through MDM2 antagonism can phenocopy the robust antiproliferative program driven by p53^53,54, both in mouse and human contexts.

Figure 6.

Treatment of mouse and human p53^wt lung adenocarcinoma cells with an MDM2 antagonist phenocopies p53^53,54 responses. A, (Left) Representative IF images of p53 and TFEB in indicated lung adenocarcinoma cell lines after infection with Ad-Empty or Ad-Cre for 48 hours. For MDM2i-treated samples, this includes a 16-hour treatment with 5 μmol/L navtemadlin. DAPI marks nuclei. Scale bar, 50 μm. (Right) Representative images of SA-β-gal staining for p53^null, p53^wt-restored, and p53^wt-restored lung adenocarcinoma cells after infection with Ad-Empty or Ad-Cre for 120 hours. For MDM2i-treated samples, this includes a 24-hour treatment with 5 μmol/L navtemadlin (MDM2i). B, Quantification of fluorescence intensity of TFEB from A in the nucleus and in the cytoplasm in indicated cell lines per 200× field (n = 10 fields). C, Quantification of analysis from A showing the percentage of SA-β-gal–positive cells among all cells per 100× field (n = 10 fields). D, Representative immunoblots of A549/TP53^wt and A549/TP53^KO cells after 5 μmol/L navtemadlin (MDM2i) treatment for the indicated time length (n = 2 experiments). GAPDH serves as a loading control. E, Representative images of SA-β-gal staining for A549/TP53^wt and A549/TP53^KO cells with or without 5 μmol/L navtemadlin (MDM2i) treatment for 96 hours. F, Quantification of analysis from E showing the percentage of SA-β-gal–positive cells among all cells per 100× field (n = 6 fields). G, Lung cancer study schematic. Indicated mice were infected with Ad-CMV-FLPo virus to initiate lung adenocarcinoma. After 12 weeks, mice were treated with tamoxifen for 3 consecutive days per week for 2 weeks (days 1, 2, and 3 and 8, 9, and 10) and received indicated drug treatments for 10 days (vehicle or 50 mg/kg milademetan on days 1–10). Lungs were collected 11 days after the first dose of tamoxifen (day 12). H, Quantification of lung tumor burden (percentage of lung tumor area relative to total lung area) for each mouse cohort (n = 7 for p53^null + vehicle, n = 6 for p53^wt-restored + vehicle, n = 7 for p53^wt-restored + MDM2i). P values were calculated by one-way ANOVA with Dunnett’s post hoc testing for pairwise comparisons.

We next sought to examine the therapeutic potential of MDM2 antagonism for lung adenocarcinoma treatment in vivo. We induced lung adenocarcinoma in Kras^FSF-G12D/+;Trp53^LSL-wt/Frt and Kras^FSF-G12D/+;Trp53^LSL-wt/Frt;Rosa26^CreER mice with Ad-FLPo. After 12 weeks, a time of significant tumor burden, p53^wt was reactivated in Kras^FSF-G12D/+;Trp53^LSL-wt/Frt;Rosa26^CreER mice by tamoxifen combined with either milademetan or vehicle (Fig. 6G). Kras^FSF-G12D/+;Trp53^LSL-wt/Frt mice treated with vehicle provided a baseline for tumor burden in the complete absence of p53. Analysis of lung adenocarcinoma burden revealed that combined p53^wt expression and milademetan led to significantly reduced tumor burden relative to p53^wt expression and vehicle, reminiscent of the p53^53,54 mutant (Fig. 6H). Monitoring of mouse weight suggested limited toxicity from the drug treatment (Supplementary Fig. S10D). MDM2 antagonists thus provide a proof of concept for a pharmacologic approach to mimic the p53^53,54 super–tumor suppressor and suggest a therapeutic approach for treating p53^wt-expressing tumors. Notably, IHC analysis of a soft tissue sarcoma from a patient who was treated with milademetan revealed robust p53 expression accompanied by induction of the lysosomal marker LAMP1 and the senescence marker p16 (Supplementary Fig. S10E). These findings support the idea that MDM2i’s are triggering the same pathway during the treatment of human patients as we observed with p53 reactivation in mouse models.

p53 Drives Lung Adenocarcinoma Regression through Macrophage Activation

We next set out to understand how p53 reactivation and induction of senescence promote lung adenocarcinoma regression. Senescent cells in vivo have been shown in certain contexts to trigger the recruitment of immune cells, such as macrophages, to induce tumor regression (44–46). Indeed, IHC analysis for CD68, a macrophage marker, showed that both restoration of p53^53,54 and p53^wt along with milademetan treatment significantly increased the number of macrophages in senescent tumors in vivo, suggesting a role for macrophages in senescent cancer cell phagocytosis, thus driving tumor regression (Fig. 7A and B). Consistent with this notion, we found that p53 reactivation could stimulate a senescence-associated secretory program (SASP), which is known to promote the recruitment of immune cells, including macrophages, to tumors (47). Specifically, upon performing qRT-PCR to examine the expression of cytokines known to attract macrophages, Csf1 and Mcp1, we found that these cytokines are upregulated upon p53 reactivation and that MDM2i treatment or p53^53,54 mutation potentiated this upregulation (Fig. 7C). We also performed cytokine arrays to analyze cytokines secreted by lung adenocarcinoma cells upon reactivation of p53^53,54. Among 111 cytokines that we tested, CCL2/MCP1 (known to attract monocytes) and CCL3/CCL4/MIP1 (known to attract macrophages) were significantly upregulated with p53^53,54 reactivation compared with p53^null cells (Fig. 7D and E). These signaling profiles support the idea that robust p53 activity drives an SASP to promote macrophage recruitment, tumor cell phagocytosis, and tumor regression. To further test this hypothesis, we performed phagocytosis assays using mCherry-labeled macrophages and carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled lung adenocarcinoma cells (Fig. 7F). We cocultured macrophages with lung adenocarcinoma cells for 4 hours after Ad-Empty or Ad-Cre transduction. Using FACS, we assessed levels of phagocytosis by determining the percentage of double-positive cells relative to total macrophages. In this assay, we observed dramatically enhanced phagocytosis when macrophages were cocultured with p53^53,54 lung adenocarcinoma cells or with p53^wt lung adenocarcinoma cells + MDM2i compared with p53^null cells (Fig. 7F and G; Supplementary Fig. S11A). Finally, to investigate the requirement for macrophages in driving tumor regression, we depleted macrophages in mice with clodronate and queried whether macrophage depletion would compromise tumor regression induced by the MDM2i. Vehicle or clodronate was delivered to mice 2 days before p53 reactivation (Fig. 7H). Immunostaining for CD68 confirmed that macrophages were significantly depleted 2 and 7 days after the clodronate treatment (Supplementary Fig. S11B and S11C). Tumor burden quantification indicated that depletion of macrophages significantly reverted tumor regression induced by p53 reactivation (Fig. 7I). These observations are consistent with the notion that the senescence program driven by augmented p53 triggers a SASP to stimulate macrophage activity, phagocytosis, and lung adenocarcinoma regression.

Figure 7.

p53 drives lung adenocarcinoma regression through macrophage activation. A, Representative images of SA-β-gal staining and CD68 immunostaining in consecutive sections of lung adenocarcinomas from each cohort in Fig. 6G. B, Quantification of analysis in A (bottom), showing the number of CD68-positive cells per field (200 × 200 pixel region from a 200× field; n = 6 fields). C, qRT-PCR of cytokines in indicated lung adenocarcinoma cells, normalized to β-actin, and graphed relative to levels in p53^null cells. D, Cytokine arrays on supernatant from lung adenocarcinoma cells that were either transduced with Ad-Empty to retain p53^null status or Ad-Cre to restore p53^53,54 expression for 5 days. E, Intensity quantification of C17/18 (CCL2/JE1/MCP1) and 19/20 (CCL3/CCL4/MIP1α/β) replicate spots from cytokine arrays using ImageJ. F, Schematic for phagocytosis assays in which mCherry-positive J774.1 macrophage cells were cocultured with CFSE-stained cancer cells for 4 hours. Phagocytosis was determined as the percentage of mCherry/CFSE double-positive cells relative to total mCherry-positive cells. G, Quantification of FACS analyses from F (n = 3). P values were calculated by one-way ANOVA. H, Lung cancer study schematic. Indicated mice were infected with Ad-CMV-FLPo virus to initiate lung adenocarcinoma. After 12 weeks minus 2 days, clodronate (in a liposome suspension) was introduced into 8 mice by tail vein injection (200 μL) and intratracheal injection (50 μL) to deplete the macrophages in the lung, and vehicle liposomes were introduced into six mice in the same fashion. All mice were then treated with tamoxifen combined with the MDM2 antagonist milademetan 2 days after clodronate/liposome injection. Tamoxifen was delivered on days 1, 2, 3, 8, 9, and 10, whereas milademetan was delivered daily from day 1 to day 10. Tumors were collected on day 12. I, Quantification of lung tumor burden (percentage of lung tumor area relative to total lung area) for each mouse cohort (n = 6 for vehicle; n = 8 for clodronate treatment). P value was calculated by Student t test. J, Model for how p53 restoration in lung adenocarcinoma drives tumor regression. p53 transcriptionally activates its direct target gene Tsc2, leading to mTORC1 inhibition and TFEB nuclear translocation, followed by transcriptional induction of a lysosomal/autophagy gene expression program. The induction of autophagy contributes to the activation of cellular senescence, which in turn leads to the secretion of factors, macrophage recruitment, tumor cell phagocytosis, and tumor regression.

A Framework for p53-Based Therapy

Together, our findings based on the restoration of p53^53,54 or p53^wt concomitant with MDM2 antagonists suggest that p53 induces lung adenocarcinoma regression by activating Tsc2 to limit mTORC1, which in turn triggers robust TFEB nuclear accumulation and activation of lysosomal/autophagy gene expression programs, followed by extensive autophagy and then increased senescence. Senescent cancer cells, in turn, release secreted factors that stimulate their clearance by macrophages (Fig. 7J).

p53-based therapy approaches have been limited by a lack of efficacy and toxicity, but the framework we have deconstructed provides a first step for suggesting new combination therapy approaches that could augment therapeutic efficacy while preserving normal tissue integrity. For example, low-level p53 activation combined with modulation of other steps in the p53 therapeutic response pathway (Fig. 7J) could provide an improved therapeutic strategy for lung adenocarcinoma. To test this idea, we assessed the potential of combined p53 activation with mTOR inhibition using Torin1. Treatment of p53^null lung adenocarcinoma cells with navtemadlin had little effect on cell viability, as expected, and Torin mildly reduced viability, which was not enhanced by combined treatment with navtemadlin. In stark contrast, treatment of p53^wt-restored cells with either Torin or navtemadlin reduced viability, but together, they showed cooperative effects on viability (Supplementary Fig. S12). These observations illustrate the importance of our pathway deconstruction as a framework for identifying targetable nodes, which, through the development of combination therapies, could lead to improved p53-based therapeutics.

Discussion

Despite the demonstration that p53 restoration can promote lung adenocarcinoma regression over a decade ago (8, 9), the precise mechanisms by which it acts have not been fully elucidated. Here, we leveraged a panel of mouse and human cellular models, combined with genomic approaches, to illuminate these mechanisms. Critical for our analyses was a hyperactive TAD mutant, p53^53,54, which we previously described as a “super–tumor suppressor” that shows enhanced tumor suppressor activity relative to p53^wt in both lung adenocarcinoma and PDAC (20, 26). p53^53,54 induced tumor regression better than p53^wt, associated with significant proliferative arrest, AT1 cell differentiation, and, ultimately, senescence. We deconstructed the pathway by which p53^53,54 restoration induces tumor regression: p53^53,54 transcriptionally activates the Tsc2 gene, leading to mTORC1 inhibition and accumulation of the lysosomal gene transcription factor TFEB in the nucleus, in which it induces lysosomal and autophagy gene expression programs that trigger autophagy, senescence, and tumor cell phagocytosis. p53^53,54 bolsters the expression of the lysosomal/autophagy program by directly inducing some of these target genes. Although the enhanced activity of the p53^53,54 mutant facilitated the discovery of these pathways, all phenotypes were observed in an attenuated fashion with p53^wt. Moreover, treatment of p53^wt-expressing cells with an MDM2 antagonist activated p53 to an extent similar to p53^53,54 to drive responses associated with tumor regression. Hence, p53^53,54 served as a powerful tool to uncover a new pathway of p53^wt action in tumor regression.

Notably, our findings suggest that p53^53,54 does not simply induce cytostasis but rather tumor regression. Whether p53 reactivation induces lung adenocarcinoma regression via apoptosis has been controversial: Although apoptosis was observed upon p53 reactivation in mice with Kras-driven lung adenocarcinomas in one study (9), other studies have suggested that restoration of p53 does not induce apoptosis in lung adenocarcinoma, either based on analysis of lung adenocarcinoma genetically engineered mouse model (GEMM) samples or analysis of lung adenocarcinoma cells, which display low levels of mitochondrial apoptotic priming (8, 48). We similarly found that restoration of either p53^wt or p53^53,54 induced minimal apoptosis in vivo. Our findings suggest instead that the lung adenocarcinoma regression we observed in mice is likely triggered by p53-induced senescence and the recruitment of macrophages capable of phagocytosing tumor cells. Consistent with this notion, either restoration of p53^53,54 or restoration of p53^wt concomitant with milademetan treatment significantly increased the number of macrophages in senescent tumors in vivo. Moreover, factors involved in macrophage recruitment were upregulated in cultured lung adenocarcinoma cells upon p53 reactivation, and p53 activity in lung adenocarcinoma cells promoted their engulfment by macrophages in vitro. Finally, we showed that depletion of macrophages with clodronate compromised the reduced tumor burden seen with p53 restoration, suggesting that macrophages are important mediators of p53-driven lung adenocarcinoma regression. These findings are consistent with previous findings that senescent HCC cells are eliminated by immune-mediated clearance involving macrophages following p53 reactivation (49, 50) and expand our understanding of the mechanisms of p53-based therapeutic responses to lung adenocarcinoma.

The transcriptional programs and molecular mechanisms underlying p53-driven tumor regression have not been well characterized. It also remains unclear whether the mechanisms responsible for tumor regression upon p53 reactivation in p53-deficient tumors are the same as those underlying p53 action in the suppression of emerging tumors. Although p53 activates a plethora of diverse target genes, recent work using unbiased screens has illuminated just a few of these genes with prominent roles in p53-mediated tumor suppression, such as Zmat3 and Arrdc3 (51–53). Here, we similarly identified a significant target gene through which p53 acts to drive a tumor-regressive senescence program, Tsc2, and we show that TSC2 acts via inhibition of mTORC1. Although it has been suggested that p53 represses mTORC1 and that this response contributes to tumor suppression (54, 55), neither TSC2 nor p53-mediated mTORC1 repression has been implicated previously in active tumor regression. Of note, some p53 targets, including Tsc2 (56), are activated by p53 in a tissue-specific manner; therefore, pathways of p53-driven tumor suppression and tumor regression may vary by context.

Pharmacologic restoration of p53 in cancers, either by driving mutant p53 to a wt conformation or by activating p53^wt through MDM2/4 antagonism, is an attractive strategy. However, little attention has been directed to this approach in lung adenocarcinoma treatment or in epithelial cancers (reviewed in refs. 16, 57). Half of lung adenocarcinomas express mutant versions of p53 with the potential for reactivation once small molecules with those capabilities are discovered. As of now, the most promising approach is to reactivate the p53 Y220C mutant, which has a pocket that will accommodate a small-molecule stabilizer (15). In addition, half of lung adenocarcinomas in humans retain a wt TP53 gene, and our study provides a strong rationale for the value of p53 reactivation as a therapeutic strategy for lung adenocarcinoma with an intact TP53 gene. Strategies that increase p53^wt protein stability and stimulate its activity, such as through the use of MDM2i’s, could drive tumor regression, as we showed in mouse models. MDM2 antagonists have been used in clinical trials for multiple types of cancers over the years (reviewed in ref. 58), and recent phase I studies of milademetan have generated promising results for patients with advanced liposarcoma, solid tumors, and lymphomas (59). However, a major obstacle for most MDM2i’s is the normal tissue toxicity caused by p53 activation throughout the body. Efforts are underway to determine whether manipulating the dosing schedule for each MDM2i can achieve the goal of robust p53 activation in tumors but with manageable toxicity (reviewed in ref. 60). Another significant obstacle to MDM2i use is the ability of tumors to develop therapeutic resistance. In our GEMMs carrying LSL-p53 alleles, we could not address the durability of therapeutic responses and the emergence of resistance, as we reproducibly detected p53^null escaper cells failing to recombine the LSL-p53 allele (25) and ultimately fueling tumor growth. Future studies should evaluate the durability of therapeutic responses and investigate the potential resistant mechanisms in GEMMs with a non-LSL p53^wt allele.

Gaining a detailed understanding of the pathway by which p53 induces tumor regression, as presented here, provides another critical strategy for improving p53-based cancer therapies by identifying multiple points of intervention that could enhance therapeutic efficacy, minimize toxicities, and avert resistance. p53 reactivation, coupled with mTOR inhibition, TFEB agonism, or macrophage activation, would be predicted to cooperatively promote tumor regression and may allow reduced doses of p53-activating drugs to be used, thereby minimizing normal tissue toxicity. Indeed, our demonstration of cooperative inhibition of lung adenocarcinoma cell viability in vitro with p53 activation and mTOR inhibition provides an important proof of concept for this strategy. It would be interesting to expand this observation by combining MDM2i’s and mTORC1-specific inhibitors in vivo to determine whether this regimen would allow reduced doses of MDM2i’s to be utilized, thereby mitigating MDM2i-associated toxicities while still driving tumor regression. Additional studies will be needed to address whether combination therapies suggested by our model will indeed lead to a new strategy for p53-based therapeutics. A proposed scheme for identifying therapeutics that cooperate with RAS inhibitors provides an important conceptual framework for a similar approach with p53 (23, 61). Specifically, deeper inhibition of the pathway, cotargeting a critical downstream node, inhibition of protective adaptive responses, and exploiting other vulnerabilities are all potential approaches to improve therapeutic success (23). Collectively, our work not only reveals a fully elaborated mechanism by which p53 restoration induces lung adenocarcinoma regression but also provides an important framework to best advance strategies for lung adenocarcinoma treatment.

Methods

Mouse Lung Adenocarcinoma Studies

All mouse work was approved and performed in compliance with Stanford University’s Institutional Animal Care and Use Committee, known as the administrative panel on laboratory animal care. Kras^FSF-G12D/+;p53^LSL-wt/Frt, Kras^FSF-G12D/+;p53^LSL-wt/Frt;Rosa26^CreER, Kras^FSF-G12D/+;p53^{LSL-53,54/Frt};Rosa26^CreER, Kras^FSF-G12D/+; p53^{LSL-25,26/Frt};Rosa26^CreER, and Kras^FSF-G12D/+;p53^{LSL-25,26,53,54/Frt};Rosa26^CreER mice (mixed 129/Sv-C57BL/6) were used (24, 62, 63). Lung tumors were induced as previously described (64). Specifically, 8- to 12-week-old male and female mice were anesthetized by i.p. injection of avertin (2,2,2-tribromoethanol) and given a dose of Ad5-CMV-FLPo intratracheally. Viral particles (1 × 10⁹) of Ad5-CMV-Cre (Ad5mCMV-FLPo from the University of Iowa Viral Vector Core, cat. #VVC-U of Iowa-530) were diluted in minimum essential medium, precipitated with CaCl₂, and delivered to mice after a 20-minute incubation. Sixteen weeks after infection, reactivation of p53^wt or p53 TAD mutants was achieved by delivering tamoxifen (on days 1,2,3,8,9,10) by oral gavage to the mice, and lungs were collected 2 weeks later. Mice were intraperitoneally injected with 200 μL of 10 mg/mL BrdU 24 hours before lung collection. Lungs were inflated/fixed with formalin for 24 hours before processing and paraffin embedding. For the drug treatment cohorts, mice were infected with Ad-CMV-FLPo virus to initiate lung adenocarcinoma. After 12 weeks, mice were treated with tamoxifen for 3 consecutive days per week for 2 weeks. Mice were also treated with vehicle or 50 mg/kg milademetan by oral gavage for 10 days. For the macrophage-depletion experiment, clodronate/liposomes were introduced into mice by both tail vein injection (200 μL) and intratracheal instillation (50 μL) 2 days before tamoxifen treatment. Lungs were collected on day 12 relative to day 1 of tamoxifen treatment. Mice that were dead for a random reason (other than lung tumors) were excluded from the study. Mice of each genotype were randomly assigned to each group. Mice grouping and data analyses were performed in a double-blind manner. The numbers of mice used were determined by power calculations based on published data as well as data collected in the pilot study.

Analysis of Mouse Lung Adenocarcinomas

Formalin-fixed, paraffin-embedded, hematoxylin and eosin (H&E)–stained slides were scanned using a NanoZoomer 2.0-RS slide scanner (Hamamatsu) and scored using NDP.view2 (Hamamatsu). Briefly, all lung lobes and lesions were outlined to quantify overall lung size and tumor sizes. Tumor burden was calculated as the percentage of lesion area compared with whole lobe area. For tumor quantifications, formalin-fixed, paraffin-embedded slides were scanned and processed using ImageJ.

GLASS-AI

All the formalin-fixed, paraffin-embedded, H&E-stained slides were analyzed using the GLASS-AI pipeline (31). Grading guidelines used by GLASS-AI were based on the system proposed by the laboratory of Dr. Tyler Jacks (65) and are described as follows: Grade 1 tumors have uniform nuclei showing no nuclear atypia. Grade 2 tumors contain cells with uniform but slightly enlarged nuclei that exhibit prominent nucleoli. Grade 3 tumors have cells with enlarged, pleomorphic nuclei showing prominent nucleoli and nuclear molding. Grade 4 tumor cells have very large, pleomorphic nuclei exhibiting a high degree of nuclear atypia, including abnormal mitoses and hyperchromatism, and contain multinucleate giant cells.

Lung Adenocarcinoma Cell Generation

Kras ^LA2/+ ;p53 ^{LSL-wt/LSL-wt} or Kras^LA2/+;p53^{LSL-mut/LSL-mut} (where mut = 25,26; 53,54; or 25,26,53,54) mice were used to generate the lung adenocarcinoma cell lines. Briefly, 11-week-old mice were sacrificed, and tumors were microdissected from the lungs. Tumors were dissociated using collagenase/dispase and DNase for 2 hours at 37°C. Cells were grown in N5 media [Gibco DMEM/F12 (+HEPES/glutamine), 5% FBS, 35 μg/mL Bovine Pituitary Extract, N2, antimitotic] for at least 3 days. Once cells began to proliferate, they were sorted based on Epithelial Cell Adhesion Molecule (EpCAM) positivity with a mouse anti-EpCAM antibody (BioLegend, cat. #118213, RRID: AB_1134105; 0.2 μg per 1 × 10⁶ cells) and kept in N5 media until proliferating. Cells were subsequently cultured in Dulbecco’s Modified Eagle Medium (DMEM) + 10% FBS. In some experiments, cell lines were treated with BrdU for 4 hours to assess proliferation. Cell lines were authenticated and checked for Mycoplasma contamination monthly.

Adenovirus Infections

To reactivate p53 or p53 mutants, cell lines were infected with Ad5-CMV-Cre (University of Iowa Viral Vector Core, cat. #VVC-U of Iowa-5), or cell lines were infected with Ad5-CMV-Empty (University of Iowa Viral Vector Core, cat. #VVC-U of Iowa-272) to retain p53^null status. Cell lines were infected at a multiplicity of infection (MOI) of 100.

Immunostaining and Microscopy

H&E staining, IHC, and IF staining were performed on paraffin-embedded lungs using standard protocols. Immunostaining was performed using the following primary antibodies: goat anti-AGER (R&D Systems, cat. #AF1145, RRID: AB_354628; 1:200), rat anti-AGER (R&D Systems, cat. #MAB1179, RRID: AB_2289349; 1:100), rabbit HOPX (Proteintech, cat. #11419-1-AP, RRID: AB_10693525; 1:100), rabbit anti-SPC (Millipore, cat. #AB3786, RRID: AB_91588; 1:100), rabbit anti-p44/42 MAPK (Erk1/2; Cell Signaling Technology, cat. #4370, RRID: AB_2315112; 1:100), rabbit anti-CC3 (Cell Signaling Technology, cat. #9661, RRID: AB_2341188; 1:100), rabbit anti-TFEB (Bethyl Laboratories, cat. #A303-673A, RRID: AB_11204751; 1:100), rat anti-LAMP1 (Developmental Studies Hybridoma Bank (DSHB), cat. #1d4b, RRID: AB_2134500; 1:100), rabbit anti-p53 (Leica Biosystems, cat. #P53-CM5P, RRID: AB_2744683; 1:100), mouse anti-p53 (Cell Signaling Technology, cat. #2524, RRID: AB_331743; 1:100), and mouse anti-BrdU (BD Biosciences, cat. #555627, RRID: AB_395993; 1:50). Secondary antibodies used were as follows: goat anti–rat-AF488 (Thermo Fisher Scientific, cat. #A-11006, RRID: AB_2534074), goat anti–rabbit-FITC (Vector Laboratories, cat. #FI-1000, RRID: AB_2336197), horse anti–mouse-FITC (Vector Laboratories, cat. #FI-2000, RRID: AB_2336176), goat anti–mouse-AF546 (Thermo Fisher Scientific, cat. #A-11030, RRID: AB_2737024), goat anti–rabbit-546 (Thermo Fisher Scientific, cat. #A-11035, RRID: AB_2534093), biotinylated goat anti-mouse (Vector Laboratories, cat. #BA-9200, RRID:AB_2336171), biotinylated goat anti-rabbit (Vector Laboratories, cat. #BA-1000, RRID: AB_2313606), and biotinylated horse anti-goat (Vector Laboratories, cat. #BA-9500, RRID: AB_2336123). For IHC experiments, paraffin sections were deparaffinized, rehydrated, and unmasked in 10 mmol/L sodium citrate buffer with 0.05% Tween-20 in a pressure cooker for 10 minutes, quenched for 20 minutes in 3% H₂O₂, permeabilized for 10 minutes in TBS with 0.3% Triton X-100, and blocked for 30 minutes in a solution of TBS + 0.3% Triton X-100 + 10% serum + 10% bovine serum albumin (BSA). Slides were incubated overnight at 4°C with the primary antibody diluted in blocking solution and subsequently incubated for 1 hour at 37°C with a biotinylated secondary antibody compatible with the primary antibody (1:1,000, Vector Laboratories). Slides were washed between steps with TBS. Slides were then incubated with the VECTASTAIN Elite ABC-HRP Kit (Vector Laboratories, cat. #PK-6100) according to the manufacturer’s instructions. DAB peroxidase kit (Vector Laboratories, cat. #SK-4100) was used for staining, and Gill’s hematoxylin was used for counterstaining, after which slides were dehydrated and mounted with Permount. A NanoZoomer 2.0-RS slide scanner (Hamamatsu) was used for imaging. For IF experiments, samples were unmasked in the pressure cooker for 10 minutes. All slides were then permeabilized for 10 minutes in 3% Triton X-100 in TBS, followed by blocking in a solution of TBS + 3% Triton X-100 + 10% serum + 10% BSA, and then incubated overnight at 4°C with primary antibody diluted in the blocking solution. Subsequently, slides were incubated for 1 hour at 37°C with an Alexa Fluor–conjugated secondary antibody compatible with the primary antibody (1:200, Thermo Fisher Scientific). Slides were washed with TBS between steps. Slides were mounted in ProLong Gold Antifade Mountant with DAPI. Images were acquired with a Leica DM4B microscope (Leica Microsystems) and analyzed with LAS X software (Leica Microsystems).

qRT-PCR

TRIzol reagent (Invitrogen) was used for RNA preparation using standard protocols (https://assets.thermofisher.com/TFS-Assets/LSG/manuals/trizol_reagent.pdf). MMLV reverse transcriptase (Invitrogen) was used to perform reverse transcription reactions. qPCR was performed in triplicate using gene-specific primers and SYBR Green master mix (Life Technologies) in a 7900HT Fast Real-Time PCR machine (Applied Biosystems). Changes in transcript abundance were calculated using a standard curve.

Immunoprecipitation

Immunoprecipitation experiments were performed according to standard protocols (https://www.abcam.com/en-us/technical-resources/protocols/immunoprecipitation). Cells were lysed with NP-40 buffer (51) to generate protein lysates. Lysates were incubated with rabbit anti-p53 antibody (CM5) overnight at 4°C. The next day, Protein A Sepharose beads (GE HealthCare Life Sciences, cat. #17012001) were washed, blocked, and added to the lysates for 4 hours at 4°C. The beads were washed, and protein was eluted in sample buffer for subsequent Western blot analyses.

ChIP

E1A;HRas ^G12V MEFs (25) were cultured in DMEM (Gibco), supplemented with 10% fetal calf serum (FCS), 1% penicillin–streptomycin, and 50 μg/mL gentamicin. Mouse ES cells were cultured according to the previously established protocols (66). Cells were maintained at 37°C in a humidified incubator with 5% CO₂. Following exposure to 0.2 μg/mL doxorubicin (Sigma-Aldrich, cat. #D1515) for 6 hours, cells were harvested and processed for chromatin preparation. Immunoprecipitation was performed using an anti-p53 polyclonal antibody (CM5, Leica Novocastra), with normal IgG included as a negative control. Chromatin preparation, immunoprecipitation, and ChIP–qPCR were performed as previously described (67).

siRNA Transfection

All the siRNAs were purchased from Horizon Discovery: TFEB siRNA SMARTpool (E-050607-00-0005), TSC2 siRNA SMARTpool (E-047050-0005), ATG5 siRNA SMARTpool (E-064838-00-0005), and control siRNA pool (D-001910-10-05). siRNAs were transfected into cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. After 24 hours of transfection, the cells were infected at an MOI of 100 with either Ad5-CMV-Cre to reactivate p53^wt or p53^53,54 or Ad5-CMV-empty to retain p53^null status. Protein lysates were collected 48 hours after viral infection. SA-β-gal staining was performed 120 hours after viral infection.

Western Blot Analysis

Western blots were performed according to the standard protocols (https://www.abcam.com/en-us/technical-resources/protocols/western-blot). Cells were lysed with NP-40 buffer, and protein was extracted (51). Protein extracts were run on SDS-PAGE gels and then transferred to a PVDF membrane (Immobilon, Millipore). Membranes were blocked with 5% milk in TBS-Tween and probed with the following primary antibodies: rabbit anti-p53 (Leica Biosystems, cat. #NCL-L-p53-CM5p, RRID: AB_2895247; 1:1,000), rat anti-LAMP1 (DSHB, cat. #1d4b, RRID:AB_2134500; 1:100), rabbit anti-TFEB (Bethyl Laboratories, cat. #A303-673A, RRID: AB_11204751; 1:1,000), rabbit anti-LC3II (Cell Signaling Technology, cat. #3868, RRID: AB_2137707; 1:1,000), mouse anti-GAPDH (Biosynth, cat. #10R-G109A; 1:20,000), rabbit anti-ATG5 (Cell Signaling Technology, cat. #12994, RRID: AB_2630393; 1:1,000), rabbit anti-p62 (Enzo Life Sciences, cat. #BML-PW9860, RRID: AB_2196009; 1:1,000), rabbit anti–4E-BP1 (Cell Signaling Technology, cat. #9644, RRID: AB_2097841; 1:1,000), rabbit anti–phospho-4E-BP1 (Thr70; Cell Signaling Technology, cat. #9455, RRID: AB_330949; 1:1,000), rabbit anti-p70 S6 kinase (Cell Signaling Technology, cat. #2708, RRID: AB_390722; 1:1,000), rabbit anti–phospho-p70 S6 kinase (Thr389; Cell Signaling Technology, cat. #9234, RRID: AB_2269803; 1:1,000), and rabbit anti-TSC2 (Cell Signaling Technology, cat. #4308, RRID: AB_10547134; 1:1,000). Blots were developed with ECL Prime Western Blotting Detection Reagent (Amersham) and imaged using a ChemiDoc XRS (Bio-Rad).

Subcellular Fractionation

Subcellular fractionation was performed according to a previously published protocol (68).

RNA-seq

For RNA-seq, lung adenocarcinoma cells were transduced with Ad-Empty or Ad-Cre for 48 hours and then harvested, pelleted, and flash-frozen in liquid nitrogen. Once all samples were collected, RNA was harvested for sequencing on a HiSeq 4000 (Illumina). High-performance computation was performed on the Stanford SCG Informatics Cluster, and subsequent bioinformatics analyses were performed in R (version 4.0.3) using various R packages unless otherwise noted. RNA-seq reads were aligned to the mouse genome (mm10) using HISAT2 (version 2.0.5, https://github.com/DaehwanKimLab/hisat2), sorted based on genomic location, indexed using Samtools (version 1.3.1, https://github.com/samtools/samtools), and counted and mapped to each gene using HTSeq-count (version 0.6.1, https://github.com/simon-anders/htseq). Differentially expressed genes (DEG) were identified using DESeq2 with a cutoff of a P-adjusted value of <0.05 (version 1.24.0, https://github.com/mikelove/DESeq2). PCA was performed using DESeq2. Unsupervised hierarchical clustering and heatmap visualization of DEGs were performed using the “pheatmap” package in R (version 1.0.12, https://github.com/raivokolde/pheatmap). Metascape (https://metascape.org) was used for functional enrichment analysis (69, 70).

Senescence Staining

SA-β-gal staining was performed with a kit (Cell Signaling Technology) according to the manufacturer’s protocol (https://www.cellsignal.com/products/cellular-assay-kits/senescence-b-galactosidase-staining-kit/9860).

Construction of A549/p53^ko Cells

A549 cells were purchased from the ATCC (RRID: CVCL_0023). The pX459 plasmid (Addgene, plasmid #62988) was cut with the BbsI restriction enzyme and ligated to complementary annealed oligos containing the BbsI overhang site and a TP53 sgRNA sequence. The TP53-targeted sgRNA was designed using the designing tool at https://chopchop.cbu.uib.no/ with the following sequences: 5′-CACCGCGACGCTAGGATCTGACTG-3′ and 5′-AAACCAGTCAGATCCTAGCGTCGC-3′. To construct the A549/p53^ko cell line, A549 cells were transfected with pX459-sgTP53 using Lipofectamine 2000 (Invitrogen, cat. #11668027). The transfected cells were then passaged and cultured in medium containing 2 μg/mL puromycin 48 hours after transfection for 3 days. Cells were then changed to medium without puromycin and grown for an additional 10 days to form single-cell colonies. Individual cell colonies were picked and expanded. The knockout efficiency of p53 in each colony was validated using immunoblotting with an anti-p53 antibody (Cell Signaling Technology, cat. #2533, RRID: AB_10547134; 1:1,000).

Tissue Collection

All specimens and clinical data were collected and analyzed in accordance with an Institutional Review Board–approved protocol, to which patients provided written informed consent, and all studies were conducted in accordance with the Declaration of Helsinki. A standard hospital biopsy consent form was also signed by the patient, as per institutional policies, when undergoing fresh clinical biopsies. Formalin-fixed, paraffin-embedded tissue samples were collected, processed, and archived per institutional clinical guidelines.

Cytokine Array

The cytokine array analysis was performed using the Proteome Profiler Mouse XL Cytokine Array (Bio-Techne, cat. #ARY028) according to the manufacturer’s protocol (https://resources.rndsystems.com/pdfs/datasheets/ary028.pdf).

Phagocytosis Assay

The J774.1 macrophage cell line was infected with the pMCB320 vector (Addgene, plasmid #89359) to permanently express mCherry. Cancer cells were stained with CFSE (Thermo Fisher Scientific, cat. #C34570) prior to coculture with macrophages. Macrophages (100,000) and cancer cells (50,000) were cocultured at 37°C for 4 hours, and then phagocytosis events were analyzed by FACS analysis. Cancer cell lines were either transduced with Ad-Empty (to retain p53^null status) or Ad-Cre (to restore p53^wt or p53^53,54 expression) for 7 days. One p53^wt-restored cell line was treated with 5 μmol/L navtemadlin for 24 hours.

Statistical Analysis

Data analysis and statistical tests were performed using GraphPad Prism software (version 9.0.1). All experiments were performed on at least three biological replicates. All measurements were taken on discrete samples. Student t tests were two-sided. All statistical tests are denoted in the figure legends, and data are presented as the mean ± SD unless otherwise noted.

Supplementary Material

Supplementary Table 1

Gene lists of the RNA-seq analyses of p53null, p53wt-restored, p5353,54-restored, p5325,26-restored, and p5325,26,53,54-restored LUAD cells.

Supplementary Table 2

List of genes upregulated with restoration of p53wt, relative to p53null cells.

Supplementary Table 3

List of genes upregulated with restoration of p5353,54, relative to p53null cells.

Supplementary Table 4

List of genes upregulated with restoration of p5325,26, relative to p53null cells.

Supplementary Table 5

List of genes upregulated with restoration of p5325,26,53,54, relative to p53null cells.

Supplementary Table 6

List of differentially expressed genes with restoration of p53wt that overlap with p53-bound genes.

Supplementary Table 7

Gene list of upregulated genes with p53null in comparison with restoration of p53wt.

Supplementary Table 8

List of differentially expressed genes with restoration of p5353,54 that overlap with p53-bound genes.

Supplementary Figure S1

Supplementary Figure S1 shows that p5353,54 restoration induces LUAD regression.

Supplementary Figure S2

Supplementary Figure S2 shows that restoration of p53wt or p5353,54 induces LUAD regression and mild apoptosis in vivo.

Supplementary Figure S3

Supplementary Figure S3 shows that restoration of p5353,54 in LUAD induces cell differentiation.

Supplementary Figure S4

Supplementary Figure S4 shows that p5353,54 restoration induces potent arrest.

Supplementary Figure S5

Supplementary Figure S5 assesses transcriptional activities of p53 TAD mutants using RNA-seq.

Supplementary Figure S6

Supplementary Figure S6 shows that p5353,54 restoration hyperactivates autophagy and lysosomal gene expression programs.

Supplementary Figure S7

Supplementary Figure S7 analysize the autophagy and lysosomal genes that are regulated by TFEB.

Supplementary Figure S8

Supplementary Figure S8 shows that p5353,54 restoration induces TFEB nuclear translocation.

Supplementary Figure S9

Supplementary Figure S9 shows that knockdown of TSC2 dampens the p53 pathway.

Supplementary Figure S10

Supplementary Figure S10 shows that treatment of wild-type p53-expressing cells with an MDM2 antagonist across contexts.

Supplementary Figure S11

Supplementary Figure S11 shows that p53 drives LUAD regression through macrophage activation.

Supplementary Figure S12

Supplementary Figure S12 shows a p53-based therapy apporoach.

Data Availability

RNA-seq data from the in vitro mouse lung adenocarcinoma cell line experiments are available from the BioProject database under the accession number PRJNA1310623 and the Gene Expression Omnibus database under the accession number GSE231681. Human lung adenocarcinoma RNA-seq data were previously published (43).

Authors’ Disclosures

K.T. Bieging-Rolett reports personal fees from HiberCell, Inc. outside the submitted work. C.A. Brady reports personal fees from Relay Therapeutics and Nuvation Bio outside the submitted work. A. Elkrief reports grants from Kanvas Biosciences, GMT Science, AstraZeneca, Bristol Myers Squibb, and Merck outside the submitted work. C.T. Chen reports grants from Rain Oncology, ADC Therapeutics, ORIC Pharmaceuticals, Takeda Pharmaceuticals, Palleon Pharmaceuticals, Pionyr Immunotherapeutics, Kinnate Biopharma, Gilead Sciences, Mersana Therapeutics, Genentech/Roche, BioHaven Pharmaceuticals, D3 Bio, Merus, Revolution Medicines, Riboscience, and Tango Therapeutics and personal fees from Boxer Capital, Globepoint Global Advisors, Johnson & Johnson, and Mubadala Capital outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

M. Wang: Conceptualization, data curation, software, formal analysis, validation, investigation, writing–original draft, project administration. K.T. Bieging-Rolett: Investigation. A.M. Kaiser: Investigation. C.A. Brady: Investigation. J.H. Lockhart: Investigation. S. Ferreira: Investigation. K.T. Nguyen: Investigation. A. Rajeevan: Software, investigation. S.A. Evans: Investigation. T. Zhao: Investigation. N. Raj: Investigation. A. Elkrief: Resources. S.E. Tischfield: Resources. M. Ladanyi: Resources. M.G. Ozawa: Investigation. N.Q. Bui: Resources. C.T. Chen: Resources. E.R. Flores: Software. L.D. Attardi: Conceptualization, funding acquisition, writing–review and editing.