Mutant p53 drives pancreatic cancer metastasis through cell-autonomous PDGF receptor β signaling

Abstract

Missense mutations in the p53 tumor suppressor inactivate its antiproliferative properties but can also promote metastasis through a gain-of-function activity. We show that sustained expression of mutant p53 is required to maintain the prometastatic phenotype of a murine model of pancreatic cancer, a highly metastatic disease that frequently displays p53 mutations. Transcriptional profiling and functional screening identified the platelet-derived growth factor receptor b (PDGFRb) as both necessary and sufficient to mediate these effects. Mutant p53 induced PDGFRb through a cell-autonomous mechanism involving inhibition of a p73/NF-Y complex that represses PDGFRb expression in p53-deficient, noninvasive cells. Blocking PDGFRb signaling by RNA interference or by small molecule inhibitors prevented pancreatic cancer cell invasion in vitro and metastasis formation in vivo. Finally, high PDGFRb expression correlates with poor disease-free survival in pancreatic, colon, and ovarian cancer patients, implicating PDGFRb as a prognostic marker and possible target for attenuating metastasis in p53 mutant tumors.

Figure 1. Depletion of Mutant p53 Abrogates Invasiveness of Pancreatic Cancer Cells

(A) Quantifications of wound distances in scratch-wound assays from 0, 3, 7, and 10 h after wounding of KPC cells stably expressing a nontargeting control shRNA (sh.Ctrl) or a shRNA targeting mutant p53 (sh.p53) or KP_flC cells stably expressing a sh.Ctrl (left panel). Data presented as mean ±SD. ***p < 0.001. Representative phase contrast images from live cell recordings of each condition are shown at 0 and 10 h (right panel). (B) KPC+sh.p53 and +sh.Ctrl as well as KP_flC cells expressing the GFP control and mutant p53 (175H and 273H) vector were allowed to invade into Collagen for 72 h before quantification as described in the Experimental Procedures (left panel). The average of invaded cells from 9 replicates ±SD is shown. A representative result of three repeated experiments is shown. **p < 0.01, ****p < 0.0001. Representative 3D reconstructions of each condition are shown (right panel). Cells were stained for F-actin (red) and DAPI (blue); dashed line indicates the approximate position of the Transwell membrane; the arrow indicates the direction of movement. (C) KPC+sh.p53 or +sh.Ctrl were orthotopically injected into the pancreata of athymic mice. When symptomatic, mice were euthanized and metastatic spread in lung and liver was quantified by counting GFP-positive macroscopic nodules (left panel). Data presented as mean ±SD. *p < 0.05, **p < 0.01. Panels show representative merged brightfield/GFP images from lung and liver (right panel).

Figure 2. Identification of PDGFRb as a Downstream Mediator of Mutant p53 in Regulating Cell Invasion

(A) Schematic workflow of RNA sequencing. (B) Ingenuity pathway analysis (Ingenuity Systems, www.ingenuity.com). Bars (p < 0.05) represent molecular and cellular functions that are significantly changed following mutant p53 depletion. (C) One-by-one invasion assay screen. Quantification of invaded KPC cells infected with individual shRNA-pools (∼3.6 shRNAs/gene) targeting the top 40 upregulated genes identified by RNAseq. Data presented as mean ±SD. (D) qRT-PCR for PDGFRb in KPC+sh.p53 (2 or 3) or +sh.Ctrl cells. Data presented as mean normalized PDGFRb expression ±SD of triplicate samples. A representative result of three repeated experiments is shown. (E) Western blotting analysis of PDGFRb, p53, and Actin in sh.p53- or sh.Ctrl-expressing KPC cells. The two bands of PDGFRb represent differentially glycosylated forms of the protein. See also Figure S1 and Table S1.

Figure 3. Depletion of Mutant p53 in Murine and Human Pancreatic Cancer Cells Decreases PDGFRb Expression Levels to Enhance Cell Invasion

(A) PDGFRa, PDGFRb and Actin levels of KPC cells infected with shRNAs targeting PDGFRa, PDGFRb or a nontargeting control (Ctrl) as determined by western blotting. (B) Quantification of the invasion into collagen of cell lines from (A), compared to KPC+sh.p53. Data presented as mean ±SD. **p < 0.005. (C) qRT-PCR for PDGFRb in 21 human pancreatic cancer cell lines of different p53 status. Data presented as mean normalized PDGFRb expression ±SD. (D) qRT-PCR for PDGFRb in the human pancreatic cancer cell lines Miapaca2, BXPC3, CFPAC, and A2.1 expressing sh.p53 or sh.Ctrl. Data presented as mean normalized PDGFRb expression ±SD. *p < 0.05, ***p < 0.001. p53 mutation of each cell line as indicated. p53 and actin levels were determined by western blotting (lower panel). (E) Quantification of invasion of human A2.1 cells infected with sh.PDGFRb, sh.p53 or sh.Ctrl (right panel). Data presented as mean ±SD. **p < 0.005, ***p < 0.001, ****p < 0.0001. Representative 3D reconstructions of invaded cells are shown (Left panel). See also Figure S2.

Figure 4. Mutant p53 Sequesters p73 to Impede the Repressive Function of the p73/NF-Y Complex on the PDGFRB Promoter

(A) KP_flC cells stably expressing a GFP-, p53^R175H-, or p53^R273H vector were transfected with HA.TAp73α. Either p53 or HA were immunoprecipitated and the expression of HAp73α, p53, or NF-YB was determined in both the input (10% of lysates) and immunoprecipitation. (B) Chromatin immunoprecipitation (ChIP) using NF-YB antibodies in KP_flC cells stably expressing sh.Ctrl, sh.NF-YB or HAp73α. Values are means ±SD. **p < 0.01. (C) KP_flC cells stably expressing a GFP-, HAp73α-, p53^R175H-, or p53^R273H vector or sh.p73 (1 or 2) were co-transfected with the PDGFRB-promoter-luciferase construct and renilla-luciferase vector. Firefly-luciferase activity of GFP-vector cells was set to 1. Values are relative Firefly-luciferase (Fluc) units normalized by renilla expression (Rluc) ±SD of quadruplicate samples. **p < 0.01, ***p < 0.001. A representative result of three repeated experiments is shown. (D) KP_flC+GFP cells as well as KP_flC+HAp73α superinfected with sh.Ctrl, or sh.NF-YB (1 or 2) were co-transfected with the PDGFRB-promoter-luciferase construct and renilla-luciferase vector. Luciferase activity was measured as described above. ***p < 0.001. (E) Quantification of invasion of the same cells as in (C). Data presented as mean ±SD. *p < 0.05, ***p < 0.001, ****p < 0.0001. (F) Quantification of invasion of same cells as in (D). Data presented as mean ±SD. **p < 0.01, ****p < 0.0001. (G) Scheme summarizing the mechanism of action of mutant p53 in promoting invasiveness. See also Figure S3.

Figure 5. PDGFRb Mediates Mutant p53 Pro-Metastatic Function in vivo

(A) Lung colonization assays after tail vein injection of KPC cells +sh.PDGFRa, +sh.PDGFRb (1 or 2), +sh.p53, or +sh.Ctrl. Total number of lung metastatic nodules in individual mice (n>6) was counted on serial histological sections (left panel). Data presented as mean ±SD. **p < 0.01, ****p < 0.0001. Representative merged brightfield/GFP images of whole lung from indicated mice (right panel). (B) MTS assay (E₄₉₀) of murine KPC and human A2.1 cells treated with crenolanib with various doses for 72 h. Normalized values presented as mean ±SD form quadruple replicates. (C) Immunoprecipitation of PDGFRb from KPC cells stimulated with 50 ng/ml PDGF-BB, after crenolanib or DMSO treatment for 4 h. Protein levels of PDGFRb, phospho-Tyrosine and Tubulin were determined by western blotting. (D) Quantification of invasion of murine KPC and human A2.1 treated with either DMSO or crenolanib at 300 nM (left panel). Data presented as mean ±SD. **p < 0.01, ***p < 0.001. Representative 3D reconstructions of invaded cells are shown (right panel). (E) Lung colonization assays after tail vein injection of crenolanib (300 nM)- or DMSO-treated KPC cells. Representative merged brightfield/GFP imaged of whole lung as well as H&E stains of pulmonary lobes are shown (left panel). Quantification of total number of lung metastatic nodules in individual mice (n > 6) (right panel). Data presented as mean ±SD. **p < 0.01. Scale bars represent 1000 μm. See also Figure S4.

Figure 6. Imatinib Reduces the Incidence of Metastasis in KPC Mice Through PDGFRb Inhibition

(A) Immunoprecipitation of PDGFRb from KPC cells stimulated with 50 ng/ml PDGF-BB, after imatinib or DMSO treatment for 4 h. Protein levels of PDGFRb, phospho-Tyrosine and Tubulin were determined by western blotting. (B) Lung colonization assays after tail vein injection of imatinib (3 μM)- or DMSO-treated KPC cells. Representative merged brightfield/GFP imaged of whole lung as well as H&E stains of pulmonary lobes are shown. Arrows indicate metastases (left panel). Quantification of total number of lung metastatic nodules in individual mice (n > 5) (right panel). Data presented as mean ±SD. **p < 0.01. Scale bars represent 100 μm. (C) Weight of pancreatic tumors of KPC mice treated with vehicle or imatinib at time of death. (D) Quantification of the number of mice with metastatic disease at the time of death. Values are percentages of the total number of mice in each cohort. Colored columns represent mice with metastases (METS) and white columns represent disease-free (DF) animals. (E) Quantification of the number of mice with lung, peritoneal (Peri.) or liver metastatic disease at the time of death. Values are percentages of the total number of mice in each cohort. (F) Representative H&E stains of harvested organs (primary tumor, lung, liver, peritoneal tissue) from vehicle and imatinib-treated animals. Scale bars represent 100 μm or 50 μm (insets). (G) Representative immunofluorescence images of pancreatic tumors of vehicle- or imatinib-treated KPC mice. DAPI, blue; CK8, red; and pPDGFRb, green. Scale bars represent 100 μm. See also Figure S5.

Figure 7. High PDGFRb Expression in Human Pancreatic and Colorectal Cancer Correlates with Reduced Metastases Free Survival

(A) Kaplan-Meier survival curves of 103 pancreatic cancer patients (clinical variable = DFS) as a function of PDGFRb-high versus PDGFRb-low expressing tumors. (B) Box plot of PDGFRb expression versus tumor grade of pancreatic tumors. (C) Clinicopathologic analysis of vascular space invasion of pancreatic cancer patients stratified by the expression levels of PDGFRb in the primary tumor. (D) Stratification of human PDAC samples (n = 961) based on high and low pPDGFRb and p53 expression levels. Chi-Square test was performed (p = 0.009). p53 and pPDGFRb levels were assessed by IHC and scored using a relative scale from 0 to 3. (E) Kaplan-Meier survival curves of colorectal cancer patients (clinical variable = DFS) as a function of PDGFRb-high versus PDGFRb-low expressing tumors. (F) Box plot of PDGFRb expression versus tumor grade of colon tumors. See also Figure S6.