Valosin-Containing Protein Stabilizes Mutant p53 to Promote Pancreatic Cancer Growth

Pancreatic ductal adenocarcinoma (PDAC) accounts for more than 90% of all pancreatic tumors and is one of the leading causes of cancer mortality and one of the most lethal malignant neoplasms across the world. The incidence and the death rate rose over the past decade. The general 5-year survival rate for patients with PDACs is only 9% (1, 2). The possible reason for the overwhelmingly mortality is that most patients with PDACs are asymptomatic at the early stage. This often accounts for their late diagnosis at phases III and IV, at which, surgical removal of primary tumors becomes impossible as they have metastasized to other organs (3, 4). However, the causes of PDACs are still largely unknown, although certain risk factors have been identified, such as cigarette smoking, dietary factors, alcohol abuse, obesity, diabetes mellitus, genetic factors, chronic pancreatitis, etc. (2, 4).

Approximately 97% of PDACs have gene alterations, such as point mutations, amplifications, deletions, translocations, and inversions (5). There are four frequently altered driver genes for PDACs: the oncogene KRAS and the tumor suppressor genes, TP53, CDKN2A, and SMAD4 (4–7). TP53 is mutated in 70% to 85% of PDACs, among which, missense mutations are most common (2, 8, 9). These mutations not only disable wild-type p53′s normal tumor-suppressive functions, but also contribute to their oncogenic gain of function (GOF), such as enhancing tumorigenesis, invasion, metastasis, and drug resistance (9–11). There are six “hotspot” codons within the DNA binding domain (DBD) of p53 and disrupt its transcriptional activity: Arg-175, Gly-245, Arg-248, Arg-249, Arg-273, and Arg-282 (10, 11). The R273H and R248Q hotspot mutations are classified into “DNA-contact mutations,” which are different from the “conformational mutations,” such as R175H and R249S. “DNA-contact mutations” lose their ability to specifically bind to DNA, whereas the “conformational mutations” cause unfolded structure or altered conformation, indirectly affecting their DNA-binding capacity (11–13).

Among the p53 “hotspot” mutations, the most frequently mutated codon in PDACs is codon 273 with the most common histidine for arginine (R273H) substitution (https://p53.iarc.fr/TP53SomaticMutations.aspx; ref. 9). GOF of the mouse p53^R270H mutant (equivalent to human R273H) has been observed in genetically engineered mice, as they developed tumors with increased aggressiveness, metastatic, and chemoresistant potential at the organismal level (11, 14). Mutant p53^R270H also synergizes with oncogenic Kras to promote PDACs (9). This mutant drives altered metabolism, promotes epithelial–mesenchymal transition (EMT) and cancer cell invasion to characterize PDACs, and mediates its malignant potential, becoming an attractive target for cancer therapy (11, 15, 16). However, the mechanisms underlying regulation of mutant p53-R273H GOF in PDACs remain incompletely understood. Growing evidence suggests that the GOF of the mutant p53 is often acquired through protein–protein interactomes and regulation of specific transcriptional targets (17, 18). Hence, dissecting the underlying mechanisms for this regulation of its GOF in PDACs might provide valuable information for developing a promising therapeutic strategy against this type of highly aggressive cancers.

In our attempt to identify possible regulators of p53-R273H in PDACs, we uncovered valosin-containing protein (VCP/p97/Cdc48/Ter94) as a novel interacting protein of this mutant p53. VCP is an abundant hexameric AAA+-type ATPase and involved in several cellular events, including protein quality control mechanisms and chromatin regulation with thousands of client proteins (19, 20). Expression of VCP is markedly elevated in multiple cancers and well correlated with tumor progress, poor prognosis (21–23), and tumor metastasis (7, 24), including PDACs. On the basis of its critical role in protein homeostasis, inhibition of VCP functions would result in increased proteotoxic stress and subsequent cell death in cancer. Thus, it might also serve as an attractive anticancer drug target (20, 25). In this study, we found that VCP interacts with MDM2 and p53-R273H and disturbs their binding. By doing so, VCP protected the mutant p53 from degradation and ultimately enhanced its GOF in cancer growth. In line with these results, suppression or ablation of VCP retarded the growth of p53-R273H PDAC cells in vitro and in vivo. Hence, our results unveil a novel role of VCP in boosting the GOF of this hot spot mutant p53 by protecting it from MDM2-mediated degradation and consequently promoting the growth of PDAC cells and tumors. These findings further consolidate the idea that VCP could serve as a potential therapeutic target against PDACs.

Detailed methods are provided in the Supplementary Information section.

Human cancer cell lines, including PANC-1 (RRID:CVCL_0480), HT29 (RRID:CVCL_0320), H1299 (RRID:CVCL_0060), and HCT-116^p53−/− cells, were cultured in DMEM supplemented with 10% FBS (GIBCO), 1% penicillin and streptomycin (GIBCO). AsPC-1 cells (RRID:CVCL_0152) were maintained in RPMI1640 medium with the same supplements. PANC-1, HT29, H1299, and AsPC-1 cells were obtained from the ATCC. HCT-116^p53−/− cells were generous gifts from Dr. Bert Vogelstein at The Ludwig Center at John Hopkins Medical School, Baltimore, MD. All these cells were grown at 37°C under a humidified 95:5 (%; v/v) mixture of air and CO₂. STR profiling was performed to ensure cell identity. No mycoplasma contamination was found. Cell lines were not passaged more than 20 times.

PANC-1 cells with stable p53 knockdown were infected with lentiviral particles containing pLKO.1_Puro shp53, and then selected with puromycin (3 μg/mL). AsPC-1 cells that stably expressed vector or p53-R273H were infected by lenti-virus produced from the pLenti6/V5 vector or pLenti6/V5-p53-R273H and selected with Blasticidin (5 μg/mL).

To determine potential mutant p53-binding proteins, 4 mg of total protein from PANC-1 cell lysate was used for coimmunoprecipitation (co-IP) by using anti-p53 (DO-1) or IgG antibody followed by protein A/G Agarose incubation. Mouse IgG was used as a control for nonspecific antibody interaction. The p53 protein complex was then eluted and analyzed by LC/MS-MS.

For generation of the VCP knockout cell line using the CRISPR/Cas9, the sgRNAs (single-guide RNA) for VCP was designed at http://crispr.mit.edu/. The lentiCRISPR v2 plasmid (Addgene Plasmid #52961) containing VCP sgRNA or sg-Ctrl was packaged within the lentivirus and infected PANC-1 cells. Single-cell colonies were selected, and knockout (KO) status was validated by Western blot (WB) analysis. sgRNAs sequences are provided in Supplementary Table S1.

The xenograft mouse model was generated as described previously (26). All treatments were administered according to the guidelines of Institution Animal Care and Use Committee (IACUC), and all the protocols were approved by Shanghai Medical College of Fudan University. A total of 5 × 10⁶ PANC-1 stable cell lines were suspended in basal medium and injected subcutaneously into both flank of 5- to 6-week-old male BALB/cA nude mice (left flank: PANC-1 sg-Ctrl cells; right flank: PANC-1 sg-VCP cells). Tumor size was measured by calipers every other day. Volumes were determined by the formula: volume = Length × Width² × 0.52 (0.52 is the constant used to calculate volume for an ellipsoid). Tumor tissues were then excised, weighed, and snap-frozen in liquid nitrogen for WB analysis or fixed for analyses. A complete list of all primary and secondary antibodies is provided in (Supplementary Table S2).

The data are presented as the mean ± SD unless otherwise stated. Statistical tests were performed using Microsoft Excel and GraphPad Prism Software. For comparisons of two groups, a two-tailed unpaired t test was used. For comparisons of multiple groups, one-way ANOVA was performed. Pearson correlation was performed to analyze the correlation of the gene expression profiling. A log-rank test was performed for survival curves. The levels of significance were set at P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***). The specific tests applied are included in the figure legends.

According to the International Agency for Research on Cancer (IARC) TP53 Mutation Database (https://p53.iarc.fr/TP53SomaticMutations.aspx), the highest frequency of p53 missense hotspot mutation occurred in codon 273, accounts for around 8% of all TP53 mutations in PDACs that harbor 70% to 85% mutated TP53 (Fig. 1A). To gain molecular insight into the mechanism underlying the mutant p53′s GOF and identify potential targets for mutant p53 co-targeted therapies in PDACs, we screened for proteins interacting with mutant p53 using co-IP with an anti-p53 antibody (DO-1) followed by LC/MS-MS analysis in a PDAC cell line (PANC-1) that harbors p53-R273H. The proteins observed in the IgG group (as a negative control) were treated as nonspecific binding proteins to eliminate false positives. As a result, LC/MS-MS analysis of the co-IPed proteins revealed 77 potential p53-R273H-binding proteins (Fig. 1B; Supplementary Table S3). Top candidate proteins were identified by applying a peptide filter threshold of 2 (unique peptides ≥2) as a criterion (Fig. 1B, table). Several known p53-binding proteins, including ribosomal proteins, heat shock proteins, and 14–3-3 protein (27–29), were among the listed proteins identified in PANC-1cells, which confirmed the effectiveness of our approach. However, we did not detect the well-known p53 binding protein MDM2 in this analysis because MDM2 expression was quite low due to the p53 mutation in the cancer cells, and also highly expressed VCP likely competed off MDM2-binding to the mutant p53 in the cells as divulged later in Fig. 5. Interestingly, VCP was one of the top p53-R273H-binding candidate proteins (Fig. 1B; Supplementary Table S3). Because VCP is primarily amplified in PDACs based on The Cancer Genome Atlas (TCGA) database (Supplementary Fig. S1), and no study has been reported thus far for this VCP-mutant p53 relationship, we decided to explore this further.

First, the mutant p53–VCP interaction was confirmed by co-IP, followed by WB assays in human p53-null lung cancer H1299 cells that were transfected with the VCP and p53-R273H plasmids with an empty vector as a negative control (Fig. 1C). This interaction was also validated with endogenous mutant p53 and VCP in two human p53-R273H-containing cancer cells, including pancreatic PANC-1 and colorectal HT29 (R273H) cells, by co-IP-WB analysis (Fig. 1D). These results demonstrate that VCP is a novel mutant p53-binding protein in PDACs as further verified below.

To map the direct binding domains of p53-R273H and of VCP, we expressed and purified recombinant glutathione S-transferase (GST)–p53 fusion proteins with different fragments and His-VCP from bacteria and performed GST-fusion protein–protein interaction assays. As shown in Fig. 2A, the full length of His-VCP proteins were pulled down by the full length and amino acids (aa) 100–290 fragment of GST-p53-R273H only, encompassing the central DBD and the mutated R273 residue. VCP contains an N-terminal domain and two highly conserved ATPase domains (D1 and D2; Fig. 2C, right; ref. 30). The mapping of the p53-binding domain(s) with VCP by employing the same GST-pull down analysis with different His-VCP fragments revealed that the aa 1–479 fragment, encompassing both the N-terminal and the central D1 domains, of VCP is critical for binding to the mutant p53, as much less complexes were detected between the two proteins when either of these domains was deleted (Fig. 2B). These result suggest that VCP may need both of its 1–199 and 200–479 fragments to form a binding site for the mutant p53. Of note, this 1–479 fragment appeared to form more complexes with the mutant p53 than did the full length VCP (Fig. 2B), suggesting that its C-terminus might play a role in regulating the VCP-p53 binding. Nevertheless, our results demonstrate that VCP directly binds to the DBD domain of mtp53s via its N-terminal (1–479) domain (Fig. 2C).

One of the VCP's most critical functions is to act as a key regulator of protein homeostasis (20, 31). To determine the impact of VCP on p53-R273H homeostasis, we knocked down VCP using two independent siRNAs in PANC-1 cells and found that p53-R273H levels are dramatically decreased by depriving endogenous VCP, and this decrease was more pronounced when combining the two siRNAs (Fig. 3A). However, knockdown of the mutant p53 in PANC-1 cells did not affect the protein level of VCP (Supplementary Fig. S2B), whereas ectopic expression of VCP moderately induced the mtp53 protein level (Supplementary Fig. S2C). These results suggest that VCP might act as a new upstream regulator of mutant p53, whereas the mutant p53 has no effect on the VCP protein level. These results also suggest that VCP might be required for the mutant p53-dependent survival of pancreatic cancer cells.

Next, we tested if the decrease of p53-R273H levels due to the depletion of VCP might affect cancer cell proliferation and survival by conducting cell survival CCK-8, apoptosis, and cell-cycle analyses. Interestingly, depletion of VCP in the pancreatic cancer cells led to decline of pancreatic cancer cell survival (Supplementary Fig. S2A) as well as drastic induction of apoptosis (Fig. 3B) and G2–M phase cell-cycle arrest (Fig. 3C). To validate if VCP activity is required for the survival of pancreatic cancer cells, we challenged PANC-1 cells with a first-in-class and orally bioavailable VCP inhibitor CB-5083 (20, 32). In line with the results of VCP depletion, CB-5083 induced cell apoptosis in a dose-dependent manner (Fig. 3D; Supplementary Fig. S3A) and caused G2–M phase cell-cycle arrest (Fig. 3E).

To further verify the role of VCP in cell survival is mutant p53-dependent, we knocked down p53-R273H in PANC-1 cells and then treated them with CB-5083. As expected, mutant p53 depletion background induced more cell apoptosis but conferred resistance of PANC-1 cells to the VCP inhibitor with reduced induction of apoptosis after the treatment (Fig. 3F and Supplementary Fig. S3B). The IC₅₀ (half-maximal inhibitory concentration) of CB-5083 in PANC-1 shp53 cells increased by more than two-fold (0.85 μmol/L vs. 0.34 μmol/L) compared with the pLKO control (Fig. 3G). Conversely, overexpression of p53-R273H rendered p53 null AsPC-1 cells more sensitive to CB-5083, as the IC₅₀ value significantly declined in the presence of p53-R273H (Fig. 3H).

We also examined whether VCP could affect the transcriptional activity of p53-R273H by performing a set of RT-qPCR analyses of previously identified target genes for this mutant p53 (33). Indeed, knockdown of either endogenous VCP or mutant p53 significantly reduced the RNA expression of its targets, CDK1, MAP2K3, and NFKB2 (Supplementary Fig. S4A), which were required for mutant p53-dependent cancer cell growth and survival (33). Of note, knockdown of VCP also led to the decrease of the p53 mRNA level. This effect might be through an unknown but indirect mechanism, because overexpression of VCP did not change the p53 mRNA level at all, while still induced the protein level of p53-R273H (Supplementary Fig. S2C) and the expression of some of its known target genes (Supplementary Fig. S4B) to some degrees.

Taken together, these results demonstrate that VCP is required for the survival of pancreatic cancer cells that harbor mutant p53 in a p53-dependent manner probably by directly binding to the mutant p53.

Although depletion of VCP could affect both the protein and mRNA levels of p53-R273H, VCP overexpression appeared to elevate the level and activity of the mutant p53 to some degrees without altering the p53 mRNA level. Thus, we decided to determine if VCP might affect the homeostasis of the mutant p53 protein. Mutant p53 proteins are often more stable and accumulated to high levels in tumor cells likely due to lack of sufficient amounts of MDM2, as MDM2 can ubiquitinate and degrade mutant p53s as well (34, 35). To investigate whether VCP might influence the MDM2 regulation of p53-R273H stability, we first tested if VCP could affect ubiquitination of p53-R273H by performing a set of cellular ubiquitination assays. As shown in Fig. 4A, ectopic VCP reduced the amount of polyubiquitinated mutp53 in HCT-116^p53−/− cells. Conversely, knockdown of endogenous VCP by siRNA increased the polyubiquitination (Fig. 4B). Consistent with these results, the VCP inhibitor CB-5083 also led to a dramatic increase of polyubiquitinated mutp53 in a dose-dependent manner in the presence of ectopic MDM2 (Fig. 4C). Of note, the effect of this inhibitor on the basal level of p53 ubiquitination without ectopic MDM2 was not detected. This was perhaps because the basal p53 ubiquitination in the cells might be primarily attributed to other yet unknown E3 ligases, as the level of endogenous MDM2 is quite low in the cells. Consistently, ectopic VCP extended the half-life of mtp53 protein as revealed by the cycloheximide-chase experiment (Fig. 4D).

Remarkably, CB-5083 also reduced the level of the VCP-mutp53 complex in a dose-dependent fashion both in cells (Fig. 4E) and in vitro (Fig. 4F), whereas treatment of PANC-1 cell with ATP conversely increased the VCP-mutp53 complex level to some degrees (Figs. 4G and H). These results demonstrate that VCP can inhibit MDM2-mediated polyubiquitination of p53, and this inhibition requires its ATPase activity. These results also suggest that direct binding to the mutant p53 must be required for VCP's inhibitory effect on its ubiquitination.

To elucidate how VCP suppresses p53 polyubiquitination by MDM2, we first tested if VCP might affect the binding of MDM2 to p53-R273H by performing an in vitro pull-down assay with purified proteins (Supplementary Fig. S5). Increasing levels of VCP led to the decreasing levels of the p53–MDM2 complex as pulled down with flag-p53 beads (Fig. 5A). Conversely, increasing levels of MDM2 led to the decrease of the p53–VCP complex (Fig. 5B). Moreover, ectopic VCP reduced the formation of endogenous MDM2–p53-R273H complexes to a certain degree (Fig. 5C), whereas depletion of VCP increased the complex formation (Fig. 5D). These results suggest that VCP might compete with MDM2 for binding to the mutant p53. Furthermore, the negative effect of VCP depletion on p53 protein levels was partially rescued by MDM2 knockdown (Fig. 5E; of note, because of the lower level of endogenous MDM2 in the cells, we needed to expose the MDM2 blot for a bit longer time period). Taken together, these results indicate that VCP can inhibit the binding of MDM2 to p53-R273H, which may explain why VCP can inhibit MDM2-mediated ubiquitination of p53. These also suggest that VCP might bind to MDM2.

To test this idea, we first confirmed the interaction between ectopic VCP and ectopic MDM2 by introducing their expression plasmids into HCT-116^p53−/− cells followed by a co-IP-WB assay (Fig. 5F). We then mapped their binding domains by conducting in vitro GST-pull down assays. As shown in Fig. 5G, VCP appeared to more preferentially bind to both the N-terminal and C-terminal fragments of MDM2. Interestingly, the N-terminal and central D1 domains of VCP were also critical for binding to MDM2, the same domain that is required for binding to mutant p53 (Figs. 5H and 2B). Also, knockdown of endogenous VCP in PANC-1 cells moderately induced the protein level of endogenous MDM2 (Supplementary Fig. S6). Together, our results demonstrate that VCP can prevent MDM2 from binding to mutant p53 and thus stabilize it. Because both of the N- and C-terminal domains of MDM2 have been shown to bind to p53 and mediate its ubiquitination and degradation (36, 37), the same domains that VCP binds to (Fig. 5G), VCP can suppress MDM2 activity toward mutant p53 by competing with MDM2 for binding to p53. Again as mentioned earlier, this also explains why MDM2 was not pulled down in our initial co-IP-MS analysis of the cancer cells with the anti-p53 antibody (Fig. 1B).

Given the functional interaction of VCP with p53-R273H in PDACs in vitro, we next investigated whether VCP depletion could affect this mutant p53-associated tumorigenesis in vivo. To this end, we employed the four high scored sgRNAs (single guide RNAs)-guided CRISPR/Cas9 system (38, 39) to deprive endogenous VCP. As shown in Fig. 6A, the CRISPR-mediated deprivation of endogenous VCP led to more dramatic decrease of endogenous mutant p53 in several individual sgVCP colonies compared with siRNA knockdown (Fig. 3A). In line with the results of siRNA-mediated VCP knockdown, deprivation of endogenous VCP in the pancreatic cancer cells led to drastic decline of pancreatic cancer cell survival (Fig. 6B) and marked reduction of colony formation (Fig. 6C; Supplementary Fig. S7). We used one pair of these single-cell-derived colonies to establish a pancreatic tumor xenograft model by subcutaneously implanting these stable cell lines: CRISPR empty vector control cells (PANC-1 sg-Ctrl; Fig. 6D, left flank) and VCP knockout cells (PANC-1 sgVCP, clone 10–12, Fig. 6D; right flank). Pancreatic tumor growth was monitored every other day for 25 days. As shown in Figs. 6D and E, compared with the empty vector control cells, depletion of VCP markedly reduced the size of the implanted tumors. The growth of VCP knockout tumors was significantly retarded compared with the empty vector control group, starting from day 9 post-injection as determined by tumor volumes (9 days P < 0.05; 18 days P < 0.01; 25 days P < 0.001; Fig. 6F). Accordingly, the tumor weights were remarkably reduced upon VCP depletion (mean value: 0.22 g vs. 0.7g, P < 0.001; Fig. 6G). The inhibitory effect of VCP depletion on xenograft tumor growth was confirmed by IHC staining of tumor tissues for Ki-67, a cell proliferation marker, as the Ki-67 signals were drastically reduced in VCP-stably knockout xenograft tumors (Fig. 6H). Also, depletion of endogenous VCP markedly decreased the protein level of p53-R273H in the pancreatic tumors as determined by WB analysis (Fig. 6I), which is consistent with the results obtained from cultured cells. These results demonstrate that VCP is essential for the growth of pancreatic cancers that harbor p53-R273H.

To correlate the above discovery of the new VCP-p53-R273H pathway in pancreatic cancer cell growth in vitro and ex vivo with more clinical setting, we analyzed publicly available online gene expression data sets. Strikingly, we observed that the VCP mRNA level is significantly higher in PDACs than that in adjacent noncancerous tissues from different oncomine databases (Fig. 7A). This was in line with its aberrant amplification in PDACs based on TCGA genome database (Supplementary Fig. S1). Next, we analyzed the correlation between VCP and p53 in TCGA datasets. Consistent with the protection role of VCP in regulation of mutant p53 stability (Figs. 3 and 4), VCP expression was proportionally correlated with the level of TP53mRNA (Fig. 7B, r = 0.221, P < 0.01). Although protein levels of mutant p53s in the cancer samples remain to be examined, they would highly likely increase in PDACs that harbor these p53 mutants (4, 5). We further investigated the clinical significance of VCP expression in PDACs. Although Kaplan–Meier plotter analysis in overall patients with PDACs showed no significant prognostic value of VCP (Supplementary Fig. S8), when restricting our analysis to data sets with information on the p53 status, we found that the VCP gene is predictive of poor survival of patients with PDAC with mutant p53 (Fig. 7C), as higher levels of VCP were correlated with lower survival rates for these patients. Therefore, these findings suggest that VCP could serve as a potential biomarker for clinical treatment and prognosis of PDACs, as 70% to 85% of PDACs harbor mutant p53, even though analysis of more PDAC samples is necessary to consolidate this clinical significance.

Mutations of the tumor suppressor p53 are frequently found in PDACs, especially in advanced pancreatic neoplasia. Among them, R273H is the most frequent mutation in PDACs. Although GOF of this mutant p53 has been reported for other types of cancers (40, 41), few studies have been reported about its role in pancreatic cancer, not mentioning its regulation by other proteins in this type of cancer. Therefore, elucidating the mechanisms underlying regulation of p53-R273H's GOF in PDACs become crucially important and would be valuable for identifying new molecule targets for the development of a potential therapy or strategy against this type of cancer, as current regimens for PDACs are largely limited and ineffective (42, 43). In addressing this outstanding issue, we screened for mutp53 binding proteins in PDACs cells harboring hotspot mutation (PANC-1, R273H) by co-IP with anti-p53 antibodies followed by a LC/MS-MS analysis and identified VCP as a novel mutant p53-binding protein. VCP is an intracellular protein localized in both of the cytoplasm and the nucleus. It has been shown to play a broader role via working with other proteins in cellular processes and physiology, including proteasomal protein degradation, chromatin remodeling, autophagosome maturation, immune signaling, cell-cycle events, endoplasmic reticulum (ER) membrane fusion, and assembly of Golgi membranes (44). Our further studies demonstrate a new role for VCP in regulation of p53-R273H stability and activity. In doing so, VCP can enhance the oncogenic function of this mutant p53 by promoting the growth of pancreatic cancer cells and tumors (Fig. 7D).

First, VCP via its N-terminal and D1 domains directly bound to p53-R273H at its central DBD domain in vitro and in cells (Figs. 1 and 2). Also, knockdown of VCP via siRNA or CRISPR-CAS9 editing in pancreatic cancer cells led to the drastic decrease of endogenous p53-R273Hs. Correspondingly, knockdown of VCP also resulted in marked reduction of pancreatic cancer cell survival and colony formation, and induction of apoptosis and cell-cycle arrest (Figs. 3B and C and 6B and C; Supplementary Figs. S2A and S7). This reduction was partially dependent on mutant p53, as this cellular phenotype was more pronounced when VCP was inhibited in mutant p53-containing, but not p53-null, pancreatic cancer cells (Figs. 3F–H; Supplementary Fig. S3B). Previously, VCP was reported to promote TP53BP1 recruitment to DNA damage sites by promoting ubiquitination and removal of L3MBTL1 in response to DNA damage signals without showing their direct binding (45). It would be tempting to find out if mutant p53 might play a role in this response via VCP.

Interestingly, the decrease of p53-R273H levels upon knockdown of VCP in the cells must be due to the degradation of the mutant p53 by MDM2, as VCP suppressed MDM2-mediated mutant p53 ubiquitination (Fig. 4A) whereas either knockdown of VCP in the pancreatic cancer cells or treatment of the cells with a VCP inhibitor CB-5083 that can inhibit the ATPase activity of this protein elevated mutant p53 ubiquitination by MDM2 (Figs. 4B and C). Consistently the inhibitor CB5083 also reduced the formation of the VCP-mutant p53 complex in cells and in vitro, whereas treatment of the cells with ATP increased the formation of the complex (Figs. 4E–H). As a member of the AAA (ATPases Associated with diverse cellular Activities) protein family, VCP uses energy derived from ATP hydrolysis to structurally remodel client molecules (19, 44). These suggest that the ATPase active site of VCP is crucial for its interacting with and protecting mutant p53, although it remains to be investigated how exactly this activity is involved in regulation of p53 stability by blocking the MDM2 activity and whether ATP might remodel VCP and consequently enhance its interaction with the mutant p53. A previous study reported that mutp53 requires and binds with HSP90 for its own protection from MDM2 degradation (29). Interestingly, our LC/MS-MS data (Fig. 1B) also revealed HSPA1B, one of the HSP90 chaperone machinery, as a mutant p53 binding protein in PANC-1 cells. Because VCP also protects mutant p53 from MDM2-mediated degradation (Figs. 2–5), it would be enticing to determine if VCP might protect the mutant p53 by partnering with HSP70. Another study showed that the MDM2-mediated degradation of p53 is inhibited by reduced phosphorylation of Thr 155 due to O-GlcNAcylation on Ser 149 of p53 (46). Because VCP binds to the central DBD domain of p53-R273H (Fig. 2A), it will be appealing to check if VCP might regulate p53 stability by influencing these posttranslational modifications of p53 or vice versa in the future.

Further supporting the notion that VCP protects mutant p53 by suppressing MDM2 E3 ubiquitin ligase activity and competing with MDM2 for mutant p53 binding are the following pieces of evidence. VCP bounds to MDM2 in cells and in vitro (Figs. 5F–H). This binding is via MDM2′s N- and C-terminal domains that are also the contact regions for p53. These result suggest that VCP might prevent p53 from binding to MDM2 by binding to the same domains of the latter. Indeed, this was the case as VCP reduced the formation of the MDM2–p53 complex in cells and in vitro in a dose-dependent manner (Figs. 5A–C), whereas depletion of VCP increased the formation of endogenous MDM2–p53 complexes (Fig. 5D). Conversely, increasing levels of MDM2 reduced the p53–VCP complex (Fig. 5B), and the effect of VCP depletion on p53 was in part rescued by MDM2 knockdown (Fig. 5E). Taken together, these results demonstrate that VCP can protect p53-R273H from MDM2-mediated proteasomal degradation by directly binding to both of the proteins and particularly preventing the binding of MDM2 to the p53 protein, consequently boosting the oncogenic activity of this mutant p53 (Fig. 7D).

The above discovery of the role of VCP in supporting the oncogenic function of p53-R273H can be translated into biological significance in mouse and human PDACs. Our further study using the xenograft tumor model system (Fig. 6D) showed that VCP is crucial for the growth of p53-R273H-containing pancreatic xenograft tumors, as its knockdown markedly retarded the growth of the tumors (Figs. 6D–G), which was closely associated with drastic reduction of endogenous p53-R273H (Fig. 6I). This animal result is well in line with the clinical database. Our bioinformatic analysis of the online databases on clinical samples revealed that VCP is highly expressed in PDACs compared with their adjacent normal tissues (Fig. 7A). Also, the expression of VCP was proportional to that of p53 in PDACs (Fig. 7B). Although proteomic analysis of mutant p53 and VCP proteins in PDACs is necessary to further support this claim, hot spot mutant p53s are generally quite stable in PDACs. Significantly, the higher level of VCP mRNAs is closely associated with the lower survival rate of p53 mutant patients with PDAC (Fig. 7C). Intriguingly, the depletion of VCP also led to the drastic decrease of the mRNA level of mutant p53 (Supplementary Fig. S4A). This observation suggests that VCP might be involved in the regulation of either transcription or stability of p53 mRNAs. However, it appears to be less likely because overexpression of VCP did not change the mRNA level of mutant p53 (Supplementary Fig. S4B). Also, knockdown of VCP induced the protein level of wild-type p53 (Supplementary Figs. S9A and S9B). Therefore, the downregulation of mutant p53 RNA levels by depleting VCP must be through a yet unknown and indirect mechanism(s) that remains to be further explored. Regardless of this unaddressed question, the ample experimental and clinical bioinformatic results as presented here demonstrate that VCP plays a crucial role in boosting the oncogenic activity of mutant p53 in pancreatic cancer.

Our finding of the VCP-p53-R273H interplay in the development of PDACs has at least three implications. First, this could explain in part why mutant p53s are often more stable in addition to the lack of the feedback regulator MDM2 (35). Also, VCP might regulate other hot spot mutant p53s as suggested by the result in Supplementary Figs. S10A and S10B, which shows that VCP also interacts with other hot-spot p53 mutants, although more studies are needed to address this question. Furthermore, our identification of VCP as an important player in the development of PDACs suggests that VCP might serve as a potential target for future development of an anticancer therapy for PDACs that harbor mutant p53s. In particular, as suggested in Fig. 4G and H, targeting the ATPase activity of VCP could serve as one alternative therapy or combined with other chemotherapies against this highly aggressive malignancy. Finally, the well correlation of high levels of VCP with more PDAC incidents and their poorer survival rate suggests that VCP could serve as a biomarker for this type of cancers that harbor mutant p53s and their prognosis. One remaining outstanding question is whether VCP could regulate the stability and tumor suppressive activity of wild-type p53 directly or indirectly.

H. Lu reports other support from Tulane University during the conduct of the study. No disclosures were reported by the other authors.

J. Wang: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y. Chen: Data curation, formal analysis, validation, methodology. C. Huang: Data curation, validation, investigation. Q. Hao: Data curation, validation, investigation. S.X. Zeng: Supervision, investigation, methodology. S. Omari: Data curation. Y. Zhang: Supervision. X. Zhou: Supervision, validation, investigation. H. Lu: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The authors thank the members of the Lu lab for active discussion and Tim Z. Lu for modifying the model figure. H. Lu. was supported in part by the Reynolds and Ryan Families Chair Fund of Translational Cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.