Preclinical mouse models for ovarian cancer that faithfully recapitulate human disease on both histopathological and molecular levels are crucial for advancing novel interventions into the clinic. We developed three genetically engineered mouse (GEM) models for fallopian tube-originating ovarian cancer with the loss of Brca1, Trp53, and Rb expression driven by Pax8 or Ovgp1 promoters or virally induced directly in the oviductal epithelium. We profiled the tumors by histology and gene expression and compared them to a previously described ovarian cancer model derived from ovarian surface epithelium. Expression profiles from the oviductal and ovarian epithelium tumors fall within the four subtypes of human high-grade serous carcinoma (HGSC) and represent different patient subpopulations. Allograft tumor models derived from the GEMs are amenable to preclinical intervention studies and respond to standard of care therapies. These well-defined, tractable models present a valuable resource for assessing novel drugs and immunotherapies for patients with HGSC.

Legends for the three Supplementary Figures.

Supplemental Figure 1. TPO171 transgenic thyroid tissues are stained with CD3-specific and cleft caspase 3 (ClC3)-specific antibodies following treatment with active LMB100 itox (left set of images) vs. LMB100 inactive mutant variant (right set of images).

Supplemental Figure 3. A) Immunostaining with anti-hMSLN antibody in orthotopic KPCdelMsln+hMSLN tumors following LMB-100 treatment. Scale bar = 800 µm. B) IP and orthotopic KPCdelMsln+hMSLN tumors were stained with Maisson Trichrome and percent positive pixels was quantified by image analysis. Each point represents quantification from a separate tumor/animal.

Supplemental Figure 2. A) KPCdelMsln+hMSLN cells were treated with increasing concentrations of LMB-100 for 72 hours then viability was measured by colorimetric assay. B) IHC characterization of tumors that form following orthotopic implantation of KPCdelMsln and KPCdelMsln+hMSLN tumor cells into C57Bl/6 syngeneic mice. In each set of three images, the scale bars are: 7mm (whole tumor images for top two sets and bottom left sets) and 8mm (a whole tumor image for bottom right set); 600 ï­m (intermediate magnification insets in all four sets); 100 ï­m (high magnification insets in all four sets). C) KPCdelMsln+hMSLN cells were inoculated IP into Msl mice. Mice were euthanized at various time points and all tumor tissue was dissected from the peritoneal cavity and weighed to determine total tumor burden. Blood collected at time of euthanasia was assessed for serum hMSLN by ELISA. Correlation was assessed by Pearson method.

Abstract Antibody-based therapies designed for human use frequently fail to cross-react with the murine isoform of their target. Because of this problem, preclinical studies of antibody-based mesothelin (Msl)-targeted therapeutics in immunocompetent systems have been limited by the lack of suitable mouse models. Here, we describe two immunocompetent humanized mesothelin transgenic mouse lines that can act as tolerant hosts for C57Bl/6-syngeneic cell lines expressing the human isoform of mesothelin. Thyroid peroxidase (TPO) mice have thyroid-restricted human mesothelin expression. Mesothelin (Msl) mice express human mesothelin in the typical serosal membrane distribution and can additionally be utilized to assess on-target, off-tumor toxicity of human mesothelin–targeted therapeutics. Both transgenic strains shed human mesothelin into the serum like human mesothelioma and patients with ovarian cancer, and serum human mesothelin can be used as a blood-based surrogate of tumor burden. Using these models, we examined the on-target toxicity and antitumor activity of human mesothelin–targeted recombinant immunotoxins. We found that immunotoxin treatment causes acute and chronic histologic changes to serosal membranes in Msl mice, while human mesothelin-expressing thyroid follicular cells in TPO mice are resistant to immunotoxin despite excellent drug delivery. Furthermore, poor delivery of immunotoxin to syngeneic orthotopic human mesothelin-expressing pancreatic adenocarcinoma limits antitumor activity both alone and in combination with immune checkpoint inhibition. In summary, we have developed two high-fidelity, immunocompetent murine models for human cancer that allow for rigorous preclinical evaluation of human mesothelin–targeted therapeutics.

LMB-100 is a mesothelin-targeted recombinant immunotoxin (iTox) that carries a modified Pseuodomonas exotoxin A (PE) payload. PE kills cells by inhibiting synthesis of new proteins. We found that treatment of pancreatic cancer cells with LMB-100 for 24⁻48 h did not change total protein level despite inducing protein synthesis inhibition (PSI). Further, increased levels of ubiquitinated proteins were detected, indicating that cells may have limited ability to compensate for PSI by reducing protein degradation. Together, these data suggest that PE depletes concentrations of a minority of cellular proteins. We used reverse phase protein array and Luminex assay to characterize this subset. LMB-100 decreased the abundance of 24 of 32 cancer-related proteins (including Bcl-x, Her2, Her3 and MUC16) without compensatory increases in other analytes. Further, cancer cells failed to maintain extracellular concentrations of cancer cell secreted growth factors (CCSGFs), including Vascular Endothelial Growth Factor (VEGF) following treatment with cytostatic LMB-100 doses both in culture and in mouse tumors. Decreased VEGF concentration did not change tumor vasculature density, however, LMB-100 caused tissue-specific changes in concentrations of secreted factors made by non-cancer cells. In summary, our data indicate that PSI caused by cytostatic LMB-100 doses preferentially depletes short-lived proteins such as oncogenic signaling molecules and CCSGFs.

p53 is a well-characterized transcription factor that is mutated in about 50% of human cancers. The majority of p53 genomic alterations are missense mutations which result in expression of p53 protein isoforms with deficient functionality. Such missense mutations cluster in several hotspots, with the most frequent alterations located at amino acids 175, 248 and 273 of human p53 (mouse amino acids 172, 245 and 270). p53 missense mutants disrupt the function of p53 tetramers, thereby behaving similarly to loss of p53. However, there is evidence that p53 missense mutants have additional tumor-promoting functions that differ from p53 deletions: p53 missense mutants can transform p53 null cells in vitro and in mouse models, and the p53 missense mutants induce a different tumor spectrum from that induced by p53 loss or other p53 missense mutants. The information obtained from the engineered murine models featuring p53 missense mutations has greatly elucidated the role of p53 mutagenesis in cancer. However, the currently-available p53 models bear several drawbacks that make them sub-optimal for preclinical studies. These models contain a global p53 haploinsufficiency, potentially impacting the biology of tumor-modifying structures like stroma, the immune system, and the vasculature. Furthermore, activation of the p53 missense mutant alleles results in a transition from a single functional wild-type allele to a combination of a wild-type and a mutant allele. These features substantially deviate from those of observed in human tumorigenesis, where somatic p53 mutagenesis in cells undergoing transformation results in a transition from two wild-type alleles to a combination of a wild-type and a mutant alleles, while tumor-modifying structures retain an unaltered p53 biallelic configuration. Here we describe the molecular and phenotypic characterization of a new allelic series of conditional p53 missense mutant mouse lines in which Cre-mediated recombination converts p53 from a wild-type to a missense mutant (R172H, R270H or R270C) configuration. To characterize the gradual molecular changes induced by expression of R172H and R270H mutants, we derived MEF lines harboring the alleles in a heterozygous conditional (p53-R172H fl/+ or p53-R270H fl/+) configuration. We observed that the wild-type p53 mRNA was lost but p53 protein abundance was increased after Cre-mediated activation of the mutant allele. Interestingly, we found that the R270H mutant allele provided a greater proliferative advantage, a distinct growth pattern and a greater ability to grow in vitro under starvation conditions than a similarly engineered R172H mutant allele. Consistent with published evidence, these observations suggest that the acquired functional changes are dissimilar between different p53 missense mutants. We further characterized the transcriptome profiles in MEF lines harboring R172H and R270H alleles and identified a number of unique and common transcriptional changes that could be causal for gain-of-function phenotypes and informative for uncovering additional mechanisms for carcinogenesis driven by p53 missense mutants. We used the newly-derived p53 mutant alleles to establish improved autochthonous and orthotopic pancreatic cancer models suitable for preclinical efficacy studies and we are currently developing lung cancer, ovarian cancer, and medullary thyroid carcinoma models. Citation Format: Tomas Vilimas, Keith Collins, Theresa Guerin, Roackie Awasthi, Lionel Feigenbaum, Guillermina Lozano, Terry Van Dyke, Serguei Kozlov. p53 missense mutants R172H and R270H exhibit differential effects on tumorigenesis. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr A17.