KRAS signaling in malignant pleural mesothelioma

Abstract

Malignant pleural mesothelioma (MPM) arises from mesothelial cells lining the pleural cavity of asbestos-exposed individuals and rapidly leads to death. MPM harbors loss-of-function mutations in BAP1, NF2, CDKN2A, and TP53, but isolated deletion of these genes alone in mice does not cause MPM and mouse models of the disease are sparse. Here, we show that a proportion of human MPM harbor point mutations, copy number alterations, and overexpression of KRAS with or without TP53 changes. These are likely pathogenic, since ectopic expression of mutant KRAS^G12D in the pleural mesothelium of conditional mice causes epithelioid MPM and cooperates with TP53 deletion to drive a more aggressive disease form with biphasic features and pleural effusions. Murine MPM cell lines derived from these tumors carry the initiating KRAS^G12D lesions, secondary Bap1 alterations, and human MPM-like gene expression profiles. Moreover, they are transplantable and actionable by KRAS inhibition. Our results indicate that KRAS alterations alone or in accomplice with TP53 alterations likely play an important and underestimated role in a proportion of patients with MPM, which warrants further exploration.

Keywords: BAP1; KRAS; NF2; TP53; asbestos.

Figure 1. KRAS alterations in human MPM from published datasets and the cancer genome atlas (TCGA) pan‐cancer MPM cohort

1. A

   KRAS and TP53 mutation frequencies in MPM from the catalogue of somatic mutations in cancer (COSMIC) stratified by histologic subtype (n = 775 patients).

2. B

   Top 25 mutated genes from 10 molecular studies of human MPM (n = 838 patients).

3. C–E

   KRAS and TP53 alterations in the cancer genome atlas (TCGA) pan‐cancer MPM dataset (n = 86 patients). Shown are clinical and molecular data plot with alteration frequencies (C) and patients reclassified as KRAS‐ or TP53‐altered (asterisks), copy number variation data summary (D), and segments of the KRAS and TP53 loci (E).

Data information: In (A), data are presented as cumulative percentages of patients tested mutant respective to patients tested for every gene. P, overall probability, two‐way ANOVA. In (B), data are presented as cumulative numbers (n; numbers above bars) and percentages (%) of patients with KRAS (red bar), TP53 (blue bar), and other (gray bars) mutations. In (C), each column represents one patient and each row one clinical or molecular feature. Asterisks indicate KRAS and TP53 alterations not identified by the TCGA, but reclassified as altered in this study due to 12p gain, 17p loss, KRAS locus gain (z > 0.3), and/or TP53 locus loss (z < −0.3). In (D), data are presented as raw data points (circles), rotated kernel density distributions (violins), and patient numbers (n) between thresholds of normal (solid black line at z = 0), low amplification (dotted red line at z = 0.1), low loss (dotted blue line at z = −0.1), high amplification (solid red line at z = 0.3), and deep loss (solid blue line at z = −0.3). P, probability, paired Wilcoxon rank sum test. In (E), KRAS (red line) and TP53 (blue line) loci segments of all 87 patients are shown. Each horizontal segment represents one patient. White and shades of red and blue indicate no change and magnitude of gain and loss, respectively. Source data are available online for this figure.

Figure 2. KRAS pathway activation in MPM from the cancer genome atlas (TCGA) pan‐cancer MPM dataset

1. A–F

   Molecular and clinical features of the cancer genome atlas (TCGA) pan‐cancer MPM patients (n = 87) stratified by the presence of KRAS standalone (n = 10) and combined KRAS/TP53 (n = 7) alterations. Shown are unsupervised hierarchical clustering of n = 86 patients (gene expression data were not available for one patient) by 40 genes significantly overexpressed in KRAS/TP53‐altered over KRAS‐altered over KRAS/TP53‐normal patients (A) and data summaries of mononucleotide change signatures (B), of indices of genomic instability and mutation burden (C), of clinical features and KRAS/TP53/NF2 co‐mutation frequency (D, E), and of overall survival (F).

2. G

   KRAS/TP53 pathway adapted from Matallanas et al (2011) and Tikoo et al (1994). Color‐coded genes were identified by TCGA and PANTHER pathway analyses.

3. H, I

   PANTHER and Reactome KRAS and TP53 pathways significantly altered in the cancer genome atlas (TCGA) pan‐cancer MPM patients. Shown are volcano plot of fold‐enrichment versus −log₁₀(probability) (H), and data summary of fold‐enrichment of KRAS and TP53 versus all other pathways with fold‐enrichment > 0.5 (I).

Data information: In (A), data are presented as heatmap of 40 differentially expressed genes (rows) in 86 individual patients (columns), color code of unsupervised hierarchical clusters, KRAS/TP53 status, and heatmap (legend), and probabilities (P) for enrichment of KRAS‐ and KRAS/TP53‐altered patients in cluster 1. The scale bar represents the color‐coded z‐scores. In (B), data are presented as heatmap of six different possible mononucleotide changes (rows) in patients grouped by KRAS/TP53 status (columns) and color code of mean mutation number (legend). ****, FDR q < 2 × 10⁻⁷ compared with all other mononucleotide changes, 2‐way ANOVA with Benjamini, Krieger, and Yekutieli two‐stage linear step‐up procedure. In (C) and (I), data are presented as raw data points (circles), rotated kernel density distributions (violins), medians (solid lines), and quartiles (dotted lines). P, overall probability, Kruskal–Wallis test. (C): * and **: P < 0.05 and P < 0.01, respectively, compared with KRAS/TP53‐normal patients, Dunn's post‐tests. (I): ** and ****: P < 0.01 and P < 0.0001, respectively, compared with other pathways, Dunn's post‐tests. In (D) and (E), data are presented as patient numbers (n) and overall probability (P) by χ ² or Kruskal–Wallis tests (D) or hypergeometric test for enrichment of KRAS mutations in TP53‐altered or biphasic MPM (E). In (F), data are presented as sample size (n), Kaplan–Meier survival estimates (lines), censored observations (line marks), log‐rank P value, and hazard ratio (HR) with 95% confidence interval (95% CI). In (H), data are presented as color‐coded individual pathways (circles), threshold of significance (horizontal dotted line), no enrichment baseline reference (vertical dotted line), and selected pathway names and codes. P and R initials in pathway codes denote PANTHER and Reactome pathways, respectively. n, sample size; FDR q, probability, false discovery rate; ΔGE, differential gene expression. Source data are available online for this figure.

Figure 3. KRAS and TP53 alterations in human MPM from Germany and human MPM cell lines from France

1. A–D

   Pleural fluid cell pellets and supernatants from 45 patients (called ASK #) with pleural effusion from Munich, Germany (Klotz et al, 2019a, 2019b), were subjected to digital droplet polymerase chain reaction (ddPCR) for the detection of mutant (^MUT) copies of KRAS codon 12/13 (KRAS ^G12/13) and KRAS codon 61 (KRAS ^Q61), as well as copies of TP53 and TERT. Diagnoses were benign pleural effusion (n = 5), lung adenocarcinoma (LUAD; n = 16), MPM (n = 12), and other extrathoracic cancers (n = 12). The assays were designed for detection of down to 1:20,000 copies using EKVX (KRAS ^WT TP53 ^G610T), A549 (KRAS ^G12S TP53 ^WT), NCI‐H460 (KRAS ^Q61H TP53 ^WT), NCI‐H3122 (KRAS ^WT TP53 ^E285V), and NCI‐H3255 (KRAS ^WT TP53 ^G560‐1A) human LUAD cells as controls. Shown are individual patient (KRAS plot) and individual sample (TP53 plot) allelic frequencies with color code and limits of normal TP53 allelic frequency as vertical dashed lines in the TP53 plot (A), representative gated dotplots of codon 12/13 KRAS ddPCR (B) and TP53/TERT (C), and results summary table (D). Any number of KRAS‐mutant droplets detected in any sample (KRAS plot in A) and any patient that failed to achieve normal TP53 ploidy by any sample (TP53 plot in A) was deemed altered.

2. E–G

   Results summary (E), representative KRAS CNA segments (F), and data summary of individual cell line CNA z‐score (G) from Affymetrix CytoScanHD Arrays of 33 primary MPM cell lines (called MESO #) from Nantes, France (GEO dataset GSE134349). Red lines denote the KRAS locus on chromosome 12p12.1.

3. H

   Data summary of mutant allelic frequency of KRAS compared with NF2 and BAP1 in all mutated samples from (A–G).

Data information: In (A), data are presented as data summary of the highest mutant copy percentage detected per individual sample (KRAS plot) or of all individual samples assessed (TP53 plot). In (D), data are presented as number of patients (n). P, probability, hypergeometric test for enrichment of KRAS mutations in MPM versus other tumors. In (E), data are presented as individual cell lines (columns), genes (rows), legend, and number of patients (n in table). P, probability, hypergeometric test for enrichment of KRAS mutations in TP53‐mutant MPM. In (G), data are presented as raw data points (circles), rotated kernel density distribution (violins), and cell line numbers (n) outside thresholds of amplification (dotted red line at 2.3) and loss (solid blue line at 1.7). P, probability, paired Wilcoxon rank sum test. In (H), data are presented as raw data points (circles), rotated kernel density distributions (violins), and medians (lines). P, overall probability, one‐way ANOVA. * and **: P < 0.05 and P < 0.01, respectively, compared with KRAS, Tukey's post‐test. Source data are available online for this figure.

Figure EV1. KRAS and TP53 alterations in a patient with malignant pleural mesothelioma from the MAPED study (NCT03319472)

1. Lollipop plot showing the four different missense KRAS mutations found (D57H, A134T, R151G, and E153D).

2. 3D rendering of KRAS protein showing the three KRAS mutations predicted by OncoKB to be non‐functional (green color) and the E153D mutation predicted by OncoKB to be oncogenic (yellow color).

3. Representative Sanger sequencing traces. Arrows indicate point mutations.

4. TP53 RT–PCR in comparison to cancer cell lines and other patients with malignant pleural effusion. Note the decreased band intensity in the patient with MPM.

5. TP53 qPCR in comparison to cancer cell lines and other patients with malignant pleural effusion. Note the markedly increased TP53 transcript abundance in the patient with MPM that, together with (D), indicates a TP53 mutation.

Data information: In (A), the likely oncogenic E153D mutation is shown enlarged compared with the other three mutations. In (D), arrows at 550 base pairs (bp) indicate amplicon size. In (E), data are presented as raw data points (circles), rotated kernel density distributions (violins), and medians (lines). P, overall probability, one‐way ANOVA. *, ***, and ****: P < 0.05, P < 0.001, and P < 0.0001 compared with A549 cells, Bonferroni post‐tests. In (D) and (E), abbreviations are as follows: BPE, benign pleural effusion; MPM, malignant pleural mesothelioma; STAD, stomach adenocarcinoma; BRCA, breast cancer; LUAD, lung adenocarcinoma; CUP, cancer of unknown primary; sh, cell line stably expressing anti‐TP53 short hairpin RNA. The TP53 status of the EKVX and A549 cell lines is indicated in parentheses. Source data are available online for this figure.

Figure 4. KRAS and TP53 alterations in MPM patients from France and Turkey

1. A, B

   Pleural fluid cell pellets and supernatants from 10 patients (called CRCINA #) with pleural effusion from Nantes, France (Gueugnon et al, ; Smeele et al, 2018), and pleural tumor samples from 17 patients (called TR#) with MPM from Istanbul, Turkey, were subjected to digital droplet polymerase chain reaction (ddPCR) for the detection of mutant (^MUT) copies of KRAS codon 12/13 (KRAS ^G12/13) and KRAS codon 61 (KRAS ^Q61), as well as copies of TP53 and TERT. Diagnoses were lung adenocarcinoma (LUAD; n = 4) and MPM (n = 23). The assays were designed for detection of down to 1:20,000 copies using EKVX (KRAS ^WT TP53 ^G610T), A549 (KRAS ^G12S TP53 ^WT), NCI‐H460 (KRAS ^Q61H TP53 ^WT), NCI‐H3122 (KRAS ^WT TP53 ^E285V), and NCI‐H3255 (KRAS ^WT TP53 ^G560‐1A) human LUAD cells as controls. Shown are individual patient (KRAS plot) and individual sample (TP53 plot) allelic frequencies with color code and limits of normal TP53 allelic frequency as vertical dashed lines in the TP53 plot (A) and results summary table (B). Any number of KRAS‐mutant droplets detected in any sample (KRAS plot in A) and any patient that failed to achieve normal TP53 ploidy by any sample (TP53 plot in A) was deemed altered.

Data information: In (A), data are presented as data summary of the highest mutant copy percentage detected per individual sample (KRAS plot) or of all individual samples assessed (TP53 plot). In (B), data are presented as number of patients (n). P, probability, χ ² test. Source data are available online for this figure.

Figure EV2. Adenoviral‐mediated mesothelial recombination

Dual‐fluorescent mT/mG CRE‐reporter mice (C57BL/6 background) received 5 × 10⁸ PFU intrapleural (A–E) or intraperitoneal (F) Ad‐Luc or Ad‐Cre and were serially imaged for bioluminescence.

1. Data summary of chest light emission (top; n = 5 mice/group) and representative bioluminescence images (bottom). Note cessation of transient Ad‐Luc expression by day 14.

2. Data summary of mG⁺ and mT⁺ mesothelial, lung, and chest wall cell percentage (n = 10 mice/group; B), representative macroscopic (top) and microscopic (middle, bottom) fluorescent images (C), optical frontal sections of stripped parietal pleura placed apical side up on glass slides (D), and deep lung sections (E, top) and fluorescent image of pleural lavage cells (E, bottom). z, focal plane distance from slide. a, alveoli; b, bronchi; ps, pleural space; arrows, recombined mesothelium.

3. Data summary of mG⁺ and mT⁺ mesothelial and deeper located (other) abdominal cell percentage (n = 10 mice/group) and representative merged microscopic fluorescent image of peritoneal surface mesothelium showing Cre‐recombined mesothelium (arrows).

Data information: In (A), data are presented as mean ± 95% confidence interval. P, overall probability, two‐way ANOVA. * and ****: P < 0.05 and P < 0.0001 for comparison between groups at the indicated time points, Bonferroni post‐tests. In (B) and (F), data are presented as raw data points (circles), rotated kernel density distribution (violins), and medians (lines). P, overall probability, two‐way ANOVA. ****: P < 0.0001 for the indicated comparisons, Bonferroni post‐tests. Ad, adenovirus; PFU, plaque‐forming units; Luc, luciferase gene; Cre, CRE recombinase gene; mT, membranous tomato red; mG, membranous green fluorescent protein. Source data are available online for this figure.

Figure EV3. Pleural mesothelial KRAS ^G12D expression causes inflammation

Wild‐type (WT) and KRAS ^G12D mice were lethally irradiated (1,100 Rad) and received same‐day bone marrow transfer of 10 million bone marrow cells from ubiquitously luminescent CAG.Luc.eGFP donors (all on the C57BL/6 strain). After 1 month required for bone marrow reconstitution, chimeras received 5 × 10⁸ PFU intrapleural Ad vectors, were longitudinally imaged for bioluminescence, and were sacrificed for pleural lavage cell analysis.

1. Representative chest bioluminescence images taken 2 weeks post‐Ad (top), pleural lavage cytocentrifugal specimens stained with May–Gruenwald–Giemsa (middle), and dotplots of CD11b and Gr1 expression by flow cytometry (bottom). Dotted lines in top panels denote the chest. Arrows in middle and bottom panels indicate increased mononuclear cells.

2. Summary of longitudinal chest light emission and total pleural cell number (dotplots), legend to dotplots, as well as of CD11b⁺Gr1⁺ pleural cells at day 14 post‐Ad (violin plot).

Data information: In (B), data are presented as mean ± 95% confidence interval (dotplots; n = 5–6 mice/data‐point) or as raw data points (circles), rotated kernel density distribution (violins), and medians (lines). P, overall probability, one‐way (violin plot) or two‐way (dotplots) ANOVA. ** and ****: P < 0.01 and P < 0.0001, respectively, for Ad‐Cre‐treated KRAS ^G12D mice compared with all other groups, Bonferroni post‐tests. WT, wild‐type; KRAS ^G12D, Lox‐STOP‐Lox.KRAS ^G12D; CAG. Luc.eGFP, ubiquitously luminescent mice; Ad, adenovirus type 5; PFU, plaque‐forming units; Cre, CRE recombinase gene; GFP, green fluorescent protein; ANOVA, analysis of variance. Source data are available online for this figure.

Figure 5. Human‐like malignant pleural mesotheliomas and effusions of mice with pleural mesothelial‐targeted oncogenic KRAS ^G12D and/or Trp53 deletion

Wild‐type (Wt), KRAS ^G12D, and Trp53f/f mice (all C57BL/6) were intercrossed and all possible offspring genotypes received 5 × 10⁸ PFU intrapleural Ad‐Cre (n is given in survival table in [C]).

1. Representative photographs of the thorax before (top) and after (bottom) chest opening (t, tumors; l, lungs; cw, chest wall; h, heart; dashed lines, effusion; ppt, parietal pleural tumors).

2. Kaplan–Meier survival plot.

3. Survival table.

4. Data summary of pleural effusion volume and nucleated cells (n is given in table in [C]).

5. Incidence of pleural tumors and effusions.

6. Representative May–Gruenwald–Giemsa‐stained pleural fluid cytocentrifugal specimen from a KRAS ^G12D;Trp53f/f mouse showing macrophages (MΦ, black arrow), lymphocytes (LΦ, purple arrow), and neutrophils (NΦ, green arrow) and summary of cellular and biochemical features of effusions of KRAS ^G12D;Trp53f/f mice (n = 10).

7. Gross macroscopic and microscopic images of visceral and parietal tumors stained with hematoxylin and eosin or PCNA (n is given in table in [E]).

Data information: In (B) and (C), data are presented as Kaplan–Meier survival estimates (lines), censored observations (line marks) 95% confidence interval (shaded areas) and number of mice at risk. P, overall probability, log‐rank test. ** and ***: P < 0.01 and P < 0.001, respectively, for the comparisons indicated, log‐rank test. In (D), data are presented as raw data points (circles), rotated kernel density distribution (violins), and medians (lines). P, overall probability, one‐way ANOVA. ****: P < 0.0001, for comparison with all other groups, Bonferroni post‐tests. In (E), data are presented as number of mice (n). P, probability for comparison with the top‐three groups, Fischer's exact test. In (F), data are presented as mean ± 95% confidence interval. Wt, wild‐type; KRAS ^G12D, Lox‐STOP‐Lox.KRAS ^G12D; Trp53f/f, conditional Trp53‐deleted; Ad, adenovirus type 5; PFU, plaque‐forming units; Cre, CRE recombinase gene; PCNA, proliferating cell nuclear antigen; LDH, lactate dehydrogenase; ANOVA, analysis of variance; VEGF, vascular endothelial growth factor. Source data are available online for this figure.

Figure EV4. Malignant peritoneal mesothelioma of KRAS ^G12D ;Trp53f/f mice

Wild‐type (Wt), Trp53f/f, and KRAS ^G12D ;Trp53f/f mice (all C57BL/6) received 5 × 10⁸ PFU intraperitoneal Ad‐Cre and were harvested when moribund.

1. Kaplan–Meier survival plot and survival table.

2. Tumor and ascites incidence table.

3. Representative macroscopic images of peritoneal tumors (dashed outlines).

4. Representative hematoxylin‐and‐eosin‐stained tissue sections of peritoneal tumors.

Data information: In (A), data are presented as Kaplan–Meier survival estimates (lines), 95% confidence intervals (shaded areas) and numbers of mice at risk. P, probability, log‐rank test. In (B), data are presented as number of mice (n). P, probabilities, Fischer's exact tests. Wt, wild‐type; KRAS ^G12D, Lox‐STOP‐Lox.KRAS ^G12D; Trp53f/f, conditional Trp53‐deleted; Ad, adenovirus type 5; PFU, plaque‐forming units; Cre, CRE recombinase gene. Source data are available online for this figure.

Figure 6. Molecular phenotyping of murine mesothelioma

Wild‐type (Wt), KRAS ^G12D, and Trp53f/f mice were intercrossed, and all possible offspring genotypes received 5 × 10⁸ PFU intrapleural or intratracheal Ad‐Cre and were sacrificed when moribund. In parallel, C57BL/6 mice received 10 consecutive weekly intraperitoneal injections of 1 g/kg urethane and were sacrificed after 6 months. Data summary (heatmap) and representative images of immunoreactivity of tissue sections of pleural and pulmonary tissues and tumors from these mice for different markers of human malignant pleural mesothelioma (MPM) and lung adenocarcinoma (LUAD). n = 10 mice/group were analyzed for each marker. Brown color indicates immunoreactivity and blue color nuclear hematoxylin counterstaining. Note the ubiquitous strong expression of Wilms' tumor 1 (WT1), patchy moderate expression of vimentin (VIM), ubiquitous moderate expression of mesothelin (MSLN), ubiquitous strong expression of calretinin (CALB2), podoplanin (PDPN), and osteopontin (SPP1), patchy moderate expression of cytokeratin 5/6 (CK5/6), and the absence of expression of surfactant protein C (SFTPC) in murine KRAS‐driven mesotheliomas. Note also the ubiquitous strong expression of WT1, the patchy moderate expression of VIM, the ubiquitous low‐level expression of MSLN, the ubiquitous strong expression of CALB2 and SPP1, the ubiquitous low‐level expression of PDPN, the variable moderate expression of CK5/6, and the ubiquitous moderate expression of SFTPC in murine KRAS ^G12D‐driven and urethane‐induced LUAD.

Figure 7. Transplantable KRAS/TP53‐mutant murine mesothelioma (KPM) cell lines

KRAS ^G12D ;Trp53f/f pleural mesothelioma (KPM), pleural mesothelial (PMC), and asbestos‐induced AE17 mesothelioma cells (all from C57BL/6 mice) were analyzed.

1. A

   KPM cell culture showing anoikis (white arrows) and spindle‐shaped morphology (black arrows).

2. B

   Representative colonies of KPM1 cells (7.5 × 10³ cells/vessel) seeded on a soft agar‐containing 60‐mm petri dish and stained with crystal violet after a month (n = 3/group).

3. C

   Data summaries from in vitro MTT reduction (top; 2 × 10⁴ cells/well; n = 6 independent experiments) and in vivo subcutaneous tumor growth after injection of 10⁶ cells per C57BL/6 mouse (bottom; n = 5/group).

4. D

   KRAS/Kras mRNA Sanger sequencing shows wild‐type Kras (Kras ^WT) of PMC and mutant murine Kras/human KRAS alleles (KRAS ^G12D and Kras ^G12C) of KPM and AE17 cells (arrows).

5. E, F

   RT–PCR (E) and qPCR (F) of KPM cells and parental tumors show Trp53f/f allele deletion (Δ) and Bap1 and Cdkn2a overexpression compared with PMC.

Data information: In (C), data are presented as mean (circles) and 95% confidence interval (bars). P, overall probability, two‐way ANOVA. ****: P < 0.0001 for AE17 cells (top) or for KPM cells (bottom) compared with all other groups, Bonferroni post‐tests. In (F), data are presented as raw data points (circles), rotated kernel density distribution (violins), and medians (lines). P, overall probability, two‐way ANOVA. *, **, and ****: P < 0.05, P < 0.01, and P < 0.0001, respectively, for comparison with PMC, Bonferroni post‐tests. Source data are available online for this figure.

Figure EV5. Bap1 mutations of KPM cells

KRAS ^G12D ;Trp53f/f pleural mesothelioma (KPM) and pleural mesothelial cells (PMC) were analyzed by RNA sequencing (GEO dataset GSE94415), Sanger sequencing for Bap1, and immunohistochemistry for BAP1 protein expression.

1. Coverage and alignment plot from RNA sequencing. Alignments are represented as gray polygons with reads mismatching the reference indicated by color. Loci with a large percentage of mismatches relative to the reference are flagged in the coverage plot as color‐coded bars. Alignments with inferred small insertions or deletions are represented with vertical or horizontal bars, respectively.

2. Bap1 mRNA Sanger sequencing shows a G>A transition (arrow) at c.829 that generates a missense mutation in codon E278K (top), as well as a single nucleotide insertion in position c.831 with a consequent frameshift mutation in codon S279insA and a single nucleotide insertion resulting to a frameshift mutation in codon K340insA at c.1072 (bottom).

3. Representative immunohistochemical images of BAP1 immunoreactivity (brown) of lungs with normal PMC and mouse tumors caused by transplanted KPM cells counterstained with hematoxylin (blue). Arrows indicate nuclear BAP1 staining.

4. Lollipop plot for each KPM cell line visualizing all Bap1 mutations detected. Red and blue lollipops with their numbers represent, respectively, missense mutations and insertions causing frameshift with their positions after the ATG start codon.

Figure 8. Transplantable and actionable murine mesothelioma models using KPM cells

C57BL/6 mice received 2 × 10⁵ intrapleural KRAS ^G12D ;Trp53f/f pleural mesothelioma cells (KPM), pleural mesothelial cells (PMC), or asbestos‐induced AE17 MPM cells.

1. A

   Kaplan–Meier survival plot with survival table.

2. B

   Data summary of pleural effusion volume and total cells (n = 10, 12, 10, 9, and 9 mice/group, respectively, from left to right).

3. C

   Images of the chest before and after opening, showing effusion (dashed lines), visceral (vpt), and parietal (ppt) pleural tumors on the costophrenic angle (ca), the diaphragm (d), and the chest wall (cw, arrows). t, tumors; l, lungs; h, heart.

4. D

   May–Gruenwald–Giemsa‐stained pleural cells (macrophages, MΦ: black arrow; lymphocytes, LΦ: purple arrow; neutrophils, NΦ: green arrow; eosinophils, EΦ: orange arrow).

5. E

   Effusion cytology and biochemistry data summary (total n = 10 mice; n = 4, 3, and 3 effusions from mice injected with KPM1, KPM2, and KPM3 cells, respectively, were analyzed and shown are pooled data).

6. F, G

   C57BL/6 mice received pleural KPM1 cells followed by a single intrapleural injection of liposomes containing 1% DMSO or 15 mg/kg deltarasin in 1% DMSO at day 9 post‐tumor cells. Shown are data summaries of MPE volume (n = 8 and 7 DMSO and deltarasin‐treated mice/group, respectively) and pleural fluid nucleated cells at day 19 post‐KPM1 cells (F), as well as representative images of pleural effusions (dashed lines) and tumors (t in [G]).

Data information: In (A), data are presented as Kaplan–Meier survival estimates (lines), 95% confidence interval (shaded areas), and number of mice at risk (n). P, probability of overall comparison and of any comparison to PMC, log‐rank test. In (B) and (F), data are presented as raw data points (circles), rotated kernel density distribution (violins), and medians (lines). Numbers in red font and arrows in (F) indicate end‐point reduction by deltarasin effect. P, probability, one‐way ANOVA (B) or Student's t‐test (F). *, **, ***, and ****: P < 0.05, P < 0.01, P < 0.001, and P < 0.0001, respectively, for comparison with PMC, Bonferroni post‐tests. In (E), data are presented as mean ± 95% confidence interval. LDH, lactate dehydrogenase. Source data are available online for this figure.

Figure 9. The molecular signature of KPM cells is enriched in human mesothelioma

RNA sequencing results (GEO dataset GSE94415) of KRAS ^G12D ;Trp53f/f mesothelioma (KPM) cells (n = 3) compared with pleural mesothelial cells (PMC; n = 1 pooled triplicate). n denotes biological replicates, since pooled triplicate technical replicates from each cell line were sequenced.

1. Unsupervised hierarchical clustering shows distinctive gene expression of KPM versus PMC.

2. Volcano plot showing some top KPM versus PMC differentially expressed genes.

3. KPM and PMC expression of classic mesothelioma markers (top) and top KPM versus PMC overexpressed genes (bottom).

4. Gene set enrichment analysis, including enrichment score and nominal probability value of the 150 gene‐signature specifically over‐represented in human mesothelioma compared with other thoracic malignancies derived from 113 patients (GSE42977) within the transcriptome of KPM cells versus PMC shows significant enrichment of the human mesothelioma signature in KPM cells.

Data information: In (C), data are presented as mean (columns) and 95% confidence interval (bars). P: probability, two‐way ANOVA. ns, *, **, and ***: P > 0.05, P < 0.05, P < 0.01, and P < 0.001, respectively, compared with PMC, Bonferroni post‐tests. Source data are available online for this figure.