Mouse Models for Mesothelioma Drug Discovery and Development

Abstract

INTRODUCTION

Mesothelioma is a rare, highly aggressive, incurable tumor arising from mesothelial linings in the body. Many drug therapies have been developed to treat this tumor including generalized chemotherapeutic agents, targeted molecular therapies, and immune system modulators. Mouse models including genetically engineered mice and patient-derived xenografts have been instrumental in the discovery and evaluation of such therapies for mesothelioma, but there is a need for improved understanding of the role of inflammation, tumor heterogeneity, mechanisms of carcinogenesis, and the tumor microenvironment.

AREAS COVERED

This review is a comprehensive analysis of available mouse models for mesothelioma drug discovery and development. Gaps in our current knowledge of mesothelioma are highlighted, and future directions for mouse model research is considered.

EXPERT OPINION

Soon, CRISPR-Cas gene-editing technology will improve our understanding of mesothelioma molecular mechanisms foundational to the discovery and testing of efficacious therapeutic agents. There are at least two areas of methodology development that are likely on this near horizon. One is concerned with better modeling of the role of inflammation associated with mesothelioma. The other is related to the availability of humanized immunocompromised mice harboring patient-derived xenografts. Combining human tumors in an environment with functional human immune cells will enable rapid innovation in immuno-oncology therapeutics.

1. Introduction

Mesothelioma is a rare, highly aggressive tumor arising from the mesothelial lining of the pleura, pericardium, peritoneum, and tunica vaginalis that is currently incurable. Malignant pleural mesothelioma (MPM) accounts for approximately 80% of cases with a median overall survival of about 12 months and 5-year overall survival of less than 10% [1]. Contrary to historical predictions, MPM incidence continues to increase worldwide [2–4]. So, much of this discussion will focus on MPM but, where relevant, other forms of mesothelioma will be highlighted.

The clinical approach to MPM is based on its histologic subtypes: epithelioid, biphasic (a mixture of histologic types), or sarcomatoid. The epithelioid subtype is most common, with the sarcomatoid subtype portending a poor prognosis [5]. Beyond these histologies, MPM tumors harbor diverse molecular heterogeneity and numerous chromosomal abnormalities (chromothripsis, chromoplexy, or both) [6]. The most frequently mutated tumor suppressor genes (TSGs) include BAP1, CDKN2A (encoding p16INK4A and p14ARF), and NF2, and to a lesser extent, include TP53, LATS2, and SETD2 [7]. Proto-oncogenes are rarely mutated in MPM; however, several pathways are consistently affected in MPM and, thus, the subject of targeted therapies. These include the PI3K-AKT-mTOR pathway, various metabolic vulnerabilities such as methylthioadenosine phosphorylase (MTAP) or argininosuccinate synthetase 1 (ASS1) loss, as well as targets of the tumor microenvironment such as cancer-associated fibroblasts [8].

The mainstay chemotherapeutic treatment of MPM consists of pemetrexed and cisplatin or carboplatin [9]. More recently, the randomized Phase III Mesothelioma Avastin Cisplatin Pemetrexed study (MAPS) showed that in the first-line treatment of unresectable MPM, bevacizumab (anti-vascular endothelial growth factor [VEGF] A) added to cisplatin-based chemotherapy improved overall survival (OS) to 18.8 months versus 16.1 months (without) [1]. Notably, because MAPS was not a registration trial, it did not receive Food and Drug Administration (FDA) approval. However, these positive findings prompted the National Comprehensive Cancer Network and the European Respiratory Society to recommend this regimen in unresectable MPM cases [10, 11]. Moreover, the latest development anticipated to become a new standard-of-care in MPM drug therapy (since being granted FDA approval in 2020) is the combination of nivolumab (anti-programmed cell death protein-1 [PD-1]) plus ipilimumab (anti-cytotoxic T-lymphocyte antigen 4 [CTLA-4]). This regimen was investigated in CHECKMATE-743 (NCT02899299), a randomized, open-label Phase III trial in patients with unresectable MPM. Patients received either nivolumab and ipilimumab for up to 2 years (N=303) or six cycles of combination chemotherapy with cisplatin or carboplatin plus pemetrexed (N=302). In a pre-specified interim analysis at a median of 29.7 months, the median OS was 18.1 months for the immunotherapy group versus 14.1 months for the chemotherapy group. Subgroup analysis showed a significant OS benefit of nivolumab–ipilimumab versus chemotherapy in patients with non-epithelioid histology (HR=0.46), but not epithelioid, and among those with a PD-L1 expression level of at least 1% (HR=0.69). The objective response rate was lower in the immunotherapy than the chemotherapy group, at 40% versus 43%, but the median duration of response was higher, at 11.0 versus 6.7 months.

Meanwhile, many other classes of anti-MPM therapies continue assessment in clinical trials such as 1) dendritic cell therapy [12]; 2) cancer vaccine to Wilm’s tumor 1 (WT1) Galinpepimut-S [13] combined with nivolumab (NCT04040231); 3) mesothelin-targeted immunotoxins SS1P [14] and LMB-100 (NCT03644550); 4) mesothelin-targeted bacterial strain vaccine CRS-207 [15]; 5) mesothelin plus fibroblast activating protein-directed chimeric antigen receptor (CAR) T-cell therapy [16]; 6) Enhancer of Zeste Homolog 2 (EZH2) inhibitor tazemetostat (NCT02860286); and 7) arginine-lowering agent pegylated arginine deiminase (ADI-PEG20) in ASS1-negative tumors [17]. Part of the reason for so many trials and therapeutic types designed to address MPM is that knowledge gaps remain surrounding this recalcitrant surface cancer [18].

Consequently, in vivo models of mesothelioma are necessary to dissect the precise pathologic processes driving disease development and provide more accurate preclinical environments for identifying novel treatments that more effectively translate to humans. Since mice are the most frequently utilized animals, this review will focus on several of those mouse models organized into three major groups based on how MPM is originated: asbestos induction, genetically engineered mouse (GEM) models, and graft models (Table 1). We conclude by providing our expert opinion pre-empting future directions for this challenging field.

Table 1:

Mouse Models of Malignant Pleural Mesothelioma

Model	Major Advantages and Disadvantages
Asbestos induction  Inhalation  Intraperitoneal injection	Advantages Representative of human exposure Same human carcinogen Ease of delivery Disadvantages Difficult to regulate asbestos inhalation Hazardous technique Long latency period Few mesothelioma tumors arise Difficult to measure tumors
Genetically Engineered Mice (GEM)  Asbestos induction – intraperitoneal injection  Conditional Knockouts/Knockins (CKO/CKI)	Advantages Rapid tumor growth Targets mutations seen in human disease Immunocompetent Can test tumor biology and treatment responses Reproducible Disadvantages Prolonged latency in many Many peritoneal vs. pleural Histologies often differ from humans, treatment responses may not translate Many other tumors develop Unable to reproduce chromosomal abnormalities seen in humans
Graft Models  Cell-line subcutaneous transplantation:   Syngeneic/Xenograft  Cell-line orthotopic transplantation   Intrapleural/Intraperitoneal   Syngeneic/Xenograft [57, 58]  Patient Derived Xenograft	Advantages Tumors grow rapidly and easily assessed Immunocompetent mice can be used Reproducible Patient-derived Xenograft (PDX) allows for study of heterogeneity and treatment response on human tumor cells Disadvantages Some models require immunodeficient mice Cell-lines may develop genomic instability Unable to evaluate tumor microenvironment with cell lines Fresh tumor samples rare for PDX PDX microenvironment replaced by mouse cells over passages Can take significant time to create the models
Future Models  Humanized Mouse Models   Not currently available  CRISPR-Cas models   Not currently available	Advantages Potential for microenvironment assessment Potential for study of role of inflammation Potential for immunotherapy evaluation Disadvantages Difficult to create these models Would not fully represent the human tumor environment or immunophenotype Significant cost Risk of graft versus host reaction Risk of incomplete differentiation of hematopoietic stem cells

2. Mouse Models of Mesothelioma

We will discuss the advantages and disadvantages unique to mouse models in use as preclinical in vivo testing environments to assess novel therapies. From a drug discovery perspective, recombinant genetics in small animal models has been a fundamental method to assess oncogenes and tumor suppressors related to any specific cancer. Because the gene-editing techniques were achieved first in the mouse instead of rats, this review will remain exclusive to mouse models related to MPM induction and therapeutic testing (Figure 1). Additionally, comprehensive reviews encompassing the entire range of preclinical mesothelioma models from in vitro to in vivo constructs are available [19–21].

Figure 1. Classification of preclinical models of mesothelioma for drug development.

Genetically Engineered Mice (GEM) with genomic modifications, commonly gene(s) knockout/knockin, are useful for investigating the molecular mechanisms underlying the development and progression of mesothelioma. Graft models involve the implantation of syngeneic or xenografted cell lines which harbor tumor-generating modifications within defined anatomical locales (intrapleural, intraperitoneal, or subcutaneously) for the evaluation of drug-based interventions. Patient-derived xenograft models are produced in a similar fashion to that of cell line-based models and offer the added benefit of more accurately reproducing the tumor microenvironment and mimicking the heterogeneic nature of human mesotheliomas. Each model is useful for developing systemic therapies (standard chemotherapeutic agents), targeted molecular therapies (disrupting select genes or signaling pathways), and immunotherapies to alter the function of immune cells to target tumor cells. KO/KI – Knockout/Knockin.

2.1. Asbestos Induction

Some of the earliest in vivo models of mesothelioma comprised immunocompetent mice strains exposed to asbestos through inhalation [22]. The advantages of this approach include using the same carcinogen known to cause disease in humans to replicate human pathomechanisms. However, with the inhalation model, there is inherent variability and technical difficulty in regulating the mice’s intake of asbestos. Airborne asbestos also posed an occupational hazard for the researchers. The Balb/c mice developed lung adenocarcinoma more frequently instead of tumors resembling MPM. Lastly, the long latency period of at least 12 months to develop tumors was too costly and a very time-intensive commitment of resources.

Alternatively, others made use of intraperitoneal injection of asbestos in similar kinds of mice [23]. This delivery route for the carcinogen was technically much easier (including in comparison to intrapleural) and quickly became the most common method of asbestos inoculation in mice. For this model, the mesothelioma’s peritoneal location was considered to reflect similar biology to MPM, but this notion has been challenged because there appear to be some biologic differences in pleural versus peritoneal mesothelial cells [24]. The impact of these differences on pre-clinical research models continue to be investigated. A second major caveat with intracavitary (intrapleural or intraperitoneal) asbestos inoculation is the bypassing of the body’s natural defenses so that mesothelial cells are subjected to doses of silicate fibers that would not be encountered under normal circumstances with the human case. Similarly, this model also demonstrated unacceptably long latency periods of 7 months or more, with less than 30% of mice developing tumors.

Generally, it was difficult to measure the tumor size in these in vivo inoculation models. Therefore, given the long latency period, inefficient rates of tumor development, and anatomic differences that impacted disease biology, these direct asbestos exposure models have not been widely adopted to test novel drug therapies.

2.2. Genetically Engineered Mice – Asbestos Induction

Genetically engineered mice (GEMs) are in vivo models permitting hetero- or homozygous deletions of TSGs implicated in MPM that represent a significant step forward in cancer study biology. A Tp53 knockout of p53^{+/−, −/−} 129/Sv mice with an intraperitoneal injection of asbestos demonstrated faster mesothelioma tumor incidence and growth than wild type mice [25]. However, not all of these mice developed mesothelioma, with only 13% of the homozygous and 76% of the heterozygous mice developing it. Many of these mice died of other tumors such as thymic lymphomas and hemangiosarcomas, and these models also suffered from long latency periods to tumor development (10 weeks for homozygous, 11 months for heterozygous mice). Since a TP53 mutation is found in a minority (~9%) of human MPM cases [7], the general applicability of this specific model’s mechanistic insights may be limited.

Targeting mutations more commonly seen in humans, Nf2^+/− knockout Nf2^KO3/+ FVB/N, Nf2^+/− 129Sv/Jae mice with an intraperitoneal injection of asbestos generated mesothelioma much more quickly than wild type mice [26, 27]. While 85% of these mice developed mesothelioma, a prolonged latency period of 12 months was required to do so.

A model of Cdkn2a with a knockout of exon 1α (loss of p16Ink4a), 1β (loss of p19Arf), or exon 2 (loss of both) (Ink4a mice (01XE4, FVB.129-Cdkn2a^tm2.1Rdp) and Ink4a;Arf mice (01XB2, FVB/N.129-Cdkn2a^tm1Rdp)) with an intraperitoneal injection of asbestos reliably generated mesothelioma, but still with a long latency of 30 weeks for Ink4a;Arf^+/−, 35 weeks for Ink4a^+/−, 38 weeks for Arf^+/−, and 49 weeks for wild type mice [28]. The accelerated tumor growth compared to wild type mice suggested that both Cdkn2a gene products suppress mesothelioma development. The majority of this model’s tumors were of sarcomatoid histology, unlike the more common epithelioid histology seen in humans.

Targeting both of these genes, Nf2^+/−;Cdkn2a^+/− knockout mice were developed (p16/p19^+/− mice (01XB2, FVB/N.129-Cdkn2a^tm1Rdp) crossed with Nf2^+/− mice), again with an intraperitoneal injection of asbestos [29]. These mice demonstrated accelerated mesothelioma onset, a progression of disease, and metastatic potential. The Nf2^+/−;Cdkn2a^+/− mice demonstrated a median latency of 24 weeks to tumor development, compared to the Nf2^+/− mice, which had a median latency of 38 weeks, while the wild-type mice had a median latency of 45 weeks. Additional findings with this model include: 1) tumor spread being associated with c-Met upregulation via p53/miR-34a, and 2) an increased cancer stem cell population within the tumors. A similar model involving Nf2^+/−;Cdkn2a^+/− knockout mice (Cdkn2a^+/− mice (01XB2, FVB/N.129-^Cdkn2atm1Rdp) crossed with Nf2^+/− mice) was developed and tested in the presence of the IL1β receptor (IL1R) antagonist anakinra to evaluate if there was a connection between inflammation-related IL1β/IL-1R signaling and tumor development [30]. These mice received an intraperitoneal asbestos injection. Mesothelioma development in anakinra treated mice was delayed (33 weeks versus 23 weeks in untreated mice), suggesting inflammation-related IL1β/IL-1R signaling is vital for asbestos mediated tumors. However, anakinra’s inhibitory effect was not absolute, whereby all Nf2^+/−;Cdkn2a^+/− mice with drug treatment eventually developed mesothelioma, implicating that other inflammatory pathways can become involved in mesothelioma development and indirectly that inflammatory circuits are redundantly complex.

Bap1^+/− knockout (constitutive Bap1^+/− in C57BL/6 background mice) tumor models based on an intraperitoneal injection of asbestos have been developed [31]. The GEM mice develop mesothelioma in only 36–60% of cases depending on asbestos dose, with the majority being sarcomatoid histology. Consistent with cancer predisposition as seen in human BAP1 mutants, mesothelioma occurs at lower doses of asbestos exposure than other mouse models.

Lastly, of historical note, a GEM model of mesothelioma driven by simian virus 40 (SV40) has been explored. These mice are created with the TAg open reading frame (ORF; SV40 strain 776) inserted into C57/BL 6J embryonic stem cells to create MexTAg mice and stimulated via intraperitoneal injection of asbestos [32]. In 100% of mice exposed to asbestos, mesothelioma appeared after 20–40 weeks, providing an opportunity for early therapeutic interventional studies. However, this model has many caveats compared to human MPM, as tumorigenesis could be altered by TAg expression, the subsequent tumor genetic landscape is dissimilar from human disease, and SV40 is no longer accepted as a causative agent in human MPM [33].

2.3. Genetically Engineered Mice – Conditional Knockout or Knockin, No Asbestos Induction

Mouse models involving conditional knockout (CKO) of genes have also been developed to determine the significance of those mutations without asbestos exposure. One example is a CKO of Nf2, Tp53, Rb, and Ink4a/Arf (in single and compound genetic lesions) mice using the Cre-LoxP system and luciferase reporter (LucR) Adeno-Cre virus injected intrapleurally [34]. This model spontaneously yielded MPM of all three histologies in as short as 9 weeks and at high frequency, whereby these tumors more closely resembled the human pathology. Tumor development was easy to measure and track with bioluminescence imaging. However, due to small amounts of adenovirus entering the circulation, multiple other tumor types in different anatomic locations formed.

Another CKO model with Bap1^f/f, Nf2^f/f, and Cdkn2a^f/f mice in a background of FBV/N was crossed to generate the cohorts and utilized the Adeno-Cre system [35]. This model demonstrated rapid onset of high-grade MPM in triple knockout mice (85%, with a median survival of 12 weeks), making this a practical-use model. The authors demonstrated a possible role of the polycomb repressive complex 2 (PRC2) in Bap1-mediated mesothelioma. PRC2 is a therapeutic strategy under investigation with tazemetostat targeting EZH2, a catalytic subunit of PRC2 (NCT02860286). As with other models, sarcomatoid histology arose, but the epithelioid subtype was absent, and other kinds of tumors developed in extrathoracic sites.

To investigate the role of inactivation of the entire Cdkn2ab locus in addition to Bap1 and Nf2 inactivation, CKO Bap1^f/f, Nf2^f/f, Cdkn2ab(Ink4a^−/−; Ink4b^−/−;Arf^f/f) and Cdkn2a(Ink4a^f/f;Arf^f/f) mice in a mixed background (FBV, Bl6, 129Ola) were crossed to generate cohorts, utilizing Adeno-CMV-Cre injection in the pleural space [36]. This model demonstrated the additional knockout of the Cdkn2b locus sped the MPM development, which matured in 8 to 14 weeks in 100% of mice. Tumor size could not be measured directly, so mice were sacrificed based on symptoms. All three histologic subtypes were formed and recapitulated signaling pathway activation of PI3K, MAPK, FAK, and Hippo pathways, as seen in human MPM. As these mice were immunocompetent, a similar immunophenotype to asbestos-induced human MPM could be appreciated. Treatment with cisplatin and pemetrexed demonstrated a response similar to human cohorts. Interestingly, elevations in p53 protein and elevated EZH2 expression after Bap1 loss were observed in this CKO context. This model is well suited for drug treatment study because of the short time required to generate MPM with defined mutations, an inflammatory response similar to that in humans, and a reactivity profile similar to human tumors treated with standard chemotherapy.

Most recently, the importance of the MAPK/ERK and PI3K/AKT/mTOR pathways in MM development has been investigated with a GEM model utilizing Pten^lox/lox;Trp53^lox/lox (mesothelial-specific) knockout mice and Pik3ca^wt/H1047R;Trp53^lox/lox knockin mice inoculated with Cre-recombinase coding adenoviruses into the bladder wall [60]. The Pten^lox/lox;Trp53^lox/lox mice developed an aggressive form of sarcomatoid mesothelioma in the peritoneum within 8–12 weeks, and the Pik3ca^wt/H1047R;Trp53^lox/lox knockin mice developed histologically similar tumors but had longer survival, less metastases, less proliferation, and less inflammation compared to the Pten^lox/lox;Trp53^lox/lox mice. Complementary mechanistic studies implicated Gαi2, a G-protein coupled receptor subunit, signaling mediated reduced pErk and pAkt levels in Pten^lox/lox;Trp53^lox/lox tumor cells. Given the poor prognosis associated with the sarcomatoid mesothelioma histology in humans owing to higher levels of chemoresistance, the authors evaluated a novel drug combination therapy by hypothesizing p110β and Mek1/2 cooperate in driving the sarcomatoid phenotype of Pten^lox/lox;Trp53^lox/lox tumors. They found reduced tumor growth and increased survival after combined p110β/MEK inhibition using selumetinib and AZD8186. This new GEM model demonstrates the flexibility to study any series of genetic lesions, and how robust integrative genomics analyses can identify rational new therapies.

2.4. Graft Models – Cell Line Transplantation

Several mesothelioma cell lines have been developed both of mouse origin (syngeneic) and human origin (xenograft) then injected subcutaneously or orthotopically to study mesothelioma, contributing insights on pathogenesis and to rational therapeutic targets. Genomic instability over time in the cell lines may adversely affect their clinical relevance, however. Additionally, these lines do not represent the three-dimensional nature of human disease and lack the heterogeneous extracellular components typical of MPM [37].

One study delivered mouse mesothelioma cell lines via subcutaneous injection into syngeneic mice, then treated the resulting tumor isografts with the immunosuppressant FTY720 [38]. Balb/c mice were injected in their flanks with AB1 murine mesothelioma cells and randomized to FTY720 or vehicle via intraperitoneal injection. This type of model is easily reproduced due to rapid tumor growth with efficient engraftment frequency. The assessment of treatment response is facilitated because of the subcutaneous tumor location that is highly accessible. Also, the study of immunotherapeutics is possible since these mice are immunocompetent. This model conveniently demonstrated the efficacy of a novel immunosuppressant FTY720 providing a rationale for in-human testing, although such effects may not easily translate when applied against human MPM.

Addressing this shortcoming of syngeneic tumor models, human mesothelioma cell lines can be delivered subcutaneously into immunodeficient mice. The resulting xenograft is treated with, for example, chimeric antigen receptor (CAR) T lymphocytes [39]. This model involved non-obese diabetic (NOD) severe combined immune deficient (SCID) IL2Rγ^−/− (NSG) mice injected subcutaneously with M108 human mesothelioma cells, and then tail vein injected with either activated T-cells, mesothelin-targeted (meso) CAR T-cells, or mesoCAR T-cells with transduced chemokine receptor CCR2b (human MPM secrete the chemokine CCL2). This study showed a more than twelve-fold increase in T-cell tumor infiltration by the mesoCAR T-cells with CCR2b compared to those without CCR2b, as well as enhanced anti-tumor activity. However, although the use of human mesothelioma cell lines may be more reflective of human treatment responses, they require immunodeficient mice, limiting a comprehensive evaluation (no immune cell activity) of response to therapy. The xenograft’s subcutaneous flank location lying outside of the human tumor’s familiar anatomic confines may not respond similarly to drug therapy as their in-situ counterparts. Specific to this model, CCR2b mesoCAR T-cells may accumulate elsewhere.

Orthotopic cell line transplantation models (both syngeneic and xenograft) have been developed to more closely mimic the anatomic locations of human disease. One such model involved BALB/c mice injected with the syngeneic murine mesothelioma cell line AB-12 intrapleurally. The authors then instilled two anti-angiogenic agents via adenovirus vectors expressing human pigment epithelium-derived factor (AdPEDF) and the human vascular endothelial growth factor receptor-1 (Adsflt-1) intrapleurally to evaluate their effect on tumor burden and survival [40]. They found these two agents together effectively inhibited tumor growth and improved survival. This is a valuable model in an anatomically relevant location, with an intact immune system. However, it suffers from the same clinical translation issues given the murine origin of mesothelioma cell line and treatment responses. Additionally, tumor size (total volume burden) was challenging to measure, leading to all models being sacrificed at empirically determined time points for evaluation.

Orthotopic transplantation of cancer cells is generally more challenging to establish due to the need for surgical skills to deliver cells to specific anatomic sites and the necessary infrastructure for mouse survival and maintenance. As an example of an orthotopic MPM model, human tumor cell lines are syringe-injected intrapleural into mice, then treated with mesothelin-targeted CAR T-cells. NSG mice were injected intrapleurally with MSTO-211H biphasic mesothelioma cells transduced with GFP/firefly luciferase fusion protein (for visualization and imaging of T-cells and tumors) and the human mesothelin (MSLN)-variant 1 to generate MSTO MSLN+ GFP-ffLuc+ cells [41]. With this model, the xenografts were situated in an accurate anatomic environment (pleural space) and utilized a human cell line that may demonstrate more meaningful drug responses. Interestingly, the intrathoracic injection of CAR T-cells showed superior treatment response compared to peripheral injection for this mesothelioma model. However, this model still suffers from the need for immunodeficient mice, limiting meaningful evaluation of response to some classes of therapies, especially immune-mediated drugs.

2.5. Graft Models – Patient-Derived Xenograft

Patient-derived xenograft (PDX) models can be valuable to study the histologic intra- and inter-tumor heterogeneity, unique genomic features, and the response to non-immune therapeutic agents. PDX models are created by implanting a small fragment of human tumor tissue subcutaneously or orthotopically into immunodeficient mice. These tumors then grow and are removed, sequenced, and analyzed, then re-implanted further in mice and expanded to test drug therapies [42]. One example is a model with surgically resected MPM specimens, each subcutaneously implanted into the flanks of immunodeficient (NOD/SCID) mice to create a unique PDX. Ten of these PDX models were treated with cisplatin and pemetrexed to evaluate treatment response [43]. The histologies of PDX remained largely concordant to their native patient tumors, as did to selected biomarkers. Also, seven of the 10 PDX models treated with chemotherapy demonstrated significant growth inhibition. This type of model highlights many of the disadvantages of PDX models, including 1) requiring significant time (months) to develop; 2) fresh tumor specimens are not readily available as MPM is a rare cancer; 3) immunodeficient mice are used, precluding evaluation of response to immune-based therapies; and 4) ectopic tumor location (subcutaneous) is generally less physiologically relevant. Additionally, the PDX engraftment rate was 40% overall and 32% for epithelioid histology. This rate was further decreased to only 29% for patients who received neoadjuvant radiation. Of the PDX that did engraft, this parameter correlated with poorer patient outcomes regardless of histology or specific therapy. Lastly, there are bona fide concerns that, with multiple passages, the intratumor microenvironment of the PDX will be progressively replaced by mouse-origin cells.

3. Mouse Models and their Suitability for Drug Discovery

Specific mouse models are better suited for drug discovery given the model’s characteristics and type of therapy under investigation (Table 2). Many new therapies are currently being studied using mouse models of mesothelioma (Table 3). Herein, we discuss mouse model suitability for drug discovery.

Table 2:

Suitable Mouse Models for Novel Drug Testing by Drug Class

Treatment Class	Appropriate Model
Generalized Therapies	All models
Targeted Molecular Therapies	GEMs  Asbestos induction Knockouts/ins  Conditional Knockouts/ins (CKO/CKI) Graft Models  Cell-line subcutaneous transplantation:   Syngeneic/Xenograft  Cell-line orthotopic transplantation:   Intrapleural/Intraperitoneal   Syngeneic/Xenograft Patient Derived Xenograft (PDX)
Immune System Modulators	GEMs  Asbestos induction Knockouts/ins  Conditional Knockouts/ins (CKO/CKI) Graft Models  Cell-line subcutaneous transplantation:   Syngeneic  Cell-line orthotopic transplantation:   Intrapleural/Intraperitoneal   Syngeneic Humanized Patient Derived Xenograft (PDX)  Not currently available

Table 3:

Recent Drugs Under Investigation for Mesothelioma Utilizing Mouse Models

Treatment class	Mechanism	Model
Generalized Therapies
Non-viral anti-thymidylate synthase RNAi embedded liposome (TS shRNA lipoplex) and pemetrexed [59]	Downregulation of thymidylate synthase	Xenograft – cell line
Targeted Molecular Therapies
Selumetinib and AZD8186 [60]	Mitogen-activated protein kinase (MEK) and p110β /PI3K inhibition	GEM
Galectin-9 mAb (Clone P4D2) [61]	Galectin-9 agonist activity	Syngeneic – cell line
Thiaproline [62]	Cell collagen production inhibition	Syngeneic – cell line
(−)-gossypol (AT-101) [63]	Apoptosis induction and cell cycle distribution modulation	Syngeneic – cell line
CFI-402257, cisplatin and Pemetrexed [64]	Thr/Tyr kinase (TTK)/monopolar spindle 1 kinase (Mps-1) inhibition	Syngeneic – cell line
FTY720 [65]	Phosphatase protein 2A (PP2A) inhibition	Syngeneic – cell line
α-difluoromethylornithine (DFMO) [66]	Ornithine decarboxylase (ODC) inhibition	Xenograft – cell line
gC1qR monoclonal antibody 60.11 [67]	gC1qR complement receptor blockade	Xenograft – cell line
Trametinib and 4-methylumbelliferone (4-MU) [68]	MEK and hyaluronic acid inhibition	Xenograft – cell line
Pevonedistat [69]	Cullin inhibition	Xenograft – cell line
Curcuminoids [70]	Cell cycle regulation	Xenograft – cell line
Anti-CD26 mAb (YS110) and triptolide (TR-1) antibody drug conjugate [71]	Immune cell recruitment, and inhibition of transcription of RNA polymerase II (Pol II)	Xenograft – cell line
microRNA (miR-215–5p) [72]	p53 activation	Xenograft – cell line
Tocotrienol (6-O-carboxypropyl-α-tocotrienol; T3E) [73]	Rat sarcoma (RAS) pathway modification, full mechanism unclear	Xenograft – cell line
GHRH antagonists MIA-602 and MIA-690 [74]	GHRH antagonism	Xenograft – cell line
Nintedanib [75]	Angiokinase inhibitor	Xenograft – cell line
Pamrevlumab (FG-3019) [76]	Connective tissue growth factor mAb antagonist	Xenograft – cell line
MT95–4 [77]	Anti-Aminopeptidase N/CD13 mAb	Xenograft – cell line
Sialic acid-binding lectin (cSBL) and pemetrexed [78]	RNA degradation	Xenograft – cell line
RG7112 and Tumor necrosis factor-related apoptosis-inducing ligand agonist (rhTRAIL) [79]	Murine Double Minute 2 (MDM2) inhibition, and TRAIL agonism	Xenograft – cell line
Sodium 4-carboxymethoxyimino-(4-hydroxyphenyl retinamide) [80]	Reactive oxygen species related apoptotic cascade activation as well as mitotic arrest	Xenograft – cell line
Ethyl pyruvate [81]	HMGB1 inhibition	Xenograft – cell line
LMB-100 (RG7787) and nab-Paclitaxel [82]	Anit-mesothelin immunotoxin	Xenograft – cell line
Pegylated arginase (PEG-BCT-100) [83]	Amino acid depletion	Xenograft – cell line
Amatuximab [84]	Anti-mesothelin mAb	Xenograft – cell line
HA15 [85]	Glucose-regulated protein 78 (GRP78/BiP) inhibition	PDX
OTX015/MK-8628 [86]	Bromodomain inhibition	PDX
VS-4718 [87]	Focal adhesion kinase (FAK) inhibition	PDX
Immune System Modulators
p-Tvax and OX40 agonist [88]	Immunogenic vaccine with T cell proliferation	Syngeneic – cell line
Pexidartinib (PLX3397) and dendritic cell vaccination [89]	CSF-1R kinase inhibition	Syngeneic – cell line
Anti-PD-1 antibody, cisplatin, pemetrexed [90]	PD-1 inhibition	Syngeneic – cell line
9H10 [91]	Anti-CTLA-4 monoclonal antibody	Syngeneic – cell line

GEM = Genetically Engineered Mice, PDX = Patient Derived Xenograft, mAb = mono-clonal antibody

Although the MPM models utilizing asbestos induction without genetic modification are immunocompetent with MPM induced by asbestos, the same carcinogen as in humans, they are not well suited for novel drug testing. Foremost is the prolonged latency period required for tumor development that is costly and impractical in modern labs [22, 23]. Additionally, tumors from murine models subjected to asbestos inoculation do not express the same common mutations seen in human mesothelioma [44]. Given these limitations, GEMs and graft models have become the preferred models for drug discovery.

GEMs have several advantages over the asbestos induction models for novel drug therapy investigation. The ability to knockout or conditionally knockout genes of interest permits interrogation of multiple genetic loci in a modestly short time period. These models also allow for bioluminescent imaging of tumors so that in vivo response can be tracked at convenient time points [34]. Therefore, models can be created targeting several genes from those commonly mutated in human MPM to those affecting various signaling pathways of relevance for targeted drug development testing [36]. These models are also immunocompetent, making them excellent candidates for studying the immunophenotype and response to various targeted novel drug therapies, including immune-modulating therapies. However, many of these models develop tumors in anatomically irrelevant locations to MPM, the histologies of the tumors developed can vary significantly from human tumors, and these models do not mimic the chromosomal abnormalities seen in human disease.

Graft models are useful for novel drug evaluation in MPM because the engraftment efficiency is high, and growth is relatively rapid. Immunocompetent mice injected subcutaneously and orthotopically with mouse mesothelioma cell lines have been used to study the effects of therapies, including immune-modulating therapies [38, 40]. Graft models can also study response to treating human mesothelioma cell lines and tumor tissues, although under an immunodeficient context [41]. PDX models offer much promise for new therapeutic drug developments given the unique ability to study human mesothelioma tumor and intra- and inter-tumor heterogeneity with ease of tumor evaluation, whether it is in ectopic or orthotopic locations [43]. Nevertheless, PDX models have drawbacks as engraftment rates can be low, patients and primary tumor availability are scarce, the immune environment is compromised, and murine cells may replace human cells with repeated graft passage.

4. Limitations of Mouse Models in Drug Discovery

Despite a varied collection of murine tumor models for drug testing applications based on a diversity of mouse strains, there are still pathophysiologic aspects (i.e., hallmarks of cancer) of human MPM that have yet to be fully understood and replicated in these in vivo models. In the context of drug testing and discovery, several common factors underpin the reasons why drugs that appear efficacious in mouse models are ineffective when applied to humans.

In general, while humans and mice share virtually the same set of genes, it is not always true that a drug targeting a mouse gene would exert an identical effect on the same gene in humans. The function of genes in different organisms (i.e., human versus mouse) may differ and be utilized in physiologic processes in entirely disparate ways, which would confound the interpretation of drug effects [45]. Specifically, regulation of p53 target genes diverged along species-specific pathways with dramatically different DNA binding landscapes between human and mouse [46]. Another category of differential drug response is in the spatio-temporal location of proteins impacting drug metabolism and pharmacodynamics. A well-documented example is that of fialuridine that worked against hepatitis B in mice but was toxic to humans because the protein transporter of the drug was also located in human mitochondria and not in mouse leading to human-specific mitochondria poisoning [47]. The genes of mouse and human for this drug transporter differed by three base pairs resulting in proteins with a radically different location and function. Lastly, in human diseases like cancer, the genetic landscape(s) are intrinsically complex and depend on the orchestration of gene pathways to drive biologic behavior, some of which are known and many of which remain obscure in MPM [7]. Incomplete knowledge of the entire genetic profile underlying the MPM tumor phenotype explains why the GEM mice, for example, do not manifest mesothelioma that genuinely mimics human cancer.

These general considerations highlight the importance of specifying specific objectives when testing novel therapies and then selecting a mouse model that is best suited to provide relevant and compelling insights to test further in clinical settings. Importantly, understanding the immunological variation inherent among diverse, inbred mouse strains, especially in genetically complex GEM models, is a prerequisite to the accurate interpretation of expected immunophenotypes. For instance, two commonly used mouse strains C57BL/6NCr (Th1 responder) and BALB/cAnNCr (Th2 responder) have been shown to exhibit significant strain differences in kinetics of tumor growth as well as baseline T-cell profile differences that would result in differential response to pre-clinical evaluation of immunotherapeutic drugs [48]. In a comprehensive screen of commonly used inbred mouse strains, significant differences in peripheral blood leukocyte sub-populations as well as differences in lymphocyte subpopulations that would greatly complicate interpretation of drug treatments between mouse groups of divergent genetic backgrounds have been observed consistently [49]. Each of the mouse models examined in this review does not similarly qualify for general purpose testing of all the various therapeutics classes. For example, the tumor microenvironment and tumor heterogeneity of human MPM are currently only able to be partially evaluated in PDX models during early graft passages [43], but cell line xenograft models cannot assess this facet of tumor response during drug therapy as only cancer cell clones comprise such tumor grafts. Human MPM, especially those cases associated with asbestos exposure, is heavily influenced by chronic inflammatory effects [50], yet the immunocompetent syngeneic mouse models are not equipped to reveal the intricacies of the inflammatory repertoire that promotes MPM, owing to species-specific cancer cells and a murine immune response that does not fully recapitulate adult human innate and adaptive immune processes. Notably, the steps of MPM carcinogenesis initiated by asbestos fibers remain incompletely described, such as not knowing the selective gene targets of NF-κB (regulator of inflammation) that are the central effectors of mesothelial cell survival facilitating cellular transformation and tumor initiation [18].

5. Conclusion

Mouse models of mesothelioma have played a significant role in improving our understanding of MPM mechanisms that point to rational therapeutics that are clinically relevant. However, more progress in this field is needed for this aggressive, heterogeneous tumor. Important work has been done with immunocompetent mice models such as with asbestos induction, GEMs with asbestos induction and with CKOs, and graft models with syngeneic cell line transplantation. Immunodeficient models using human tumor with xenograft cell line transplantation and PDX models have also greatly improved our understanding of MPM. To date, no one model has precisely represented the human condition, yet these models are critical in expanding our knowledge and provide an environment to test therapeutic principles. We look forward to future contributions to help direct new therapeutic strategies.

6. Expert Opinion

MPM remains a fatal disease in need of highly effective targeted therapeutic agents and both in vitro and in vivo disease models for drug testing. Many unique murine models exist in this regard, each with their advantages and disadvantages depending on the drug study’s objective. Better models are needed to continue to understand human MPM biology, including closer anatomic representation of the disease, further delineation of genetic alterations, the importance of immunologic involvement in disease, interactions with the tumor microenvironment, mechanisms of carcinogenesis (e.g., the precise contribution of chronic inflammation), targeted drug responses, and tumor inter- and intra-heterogeneity. The ability to alter the genomic backgrounds of mice utilizing CRISPR-Cas genome editing technologies [51] has contributed significantly to the production of useful models with which to explore disease progression and for testing interventional therapies. However, initial studies have focused on gene knockout models that mainly predispose mice to develop mesothelioma (there is no known oncogenic driver mutation in human mesothelioma, as discussed previously). To better understand the oncogenic mechanisms superseding these mutational events, one must develop models that explore these initiating conditions.

To address some current limitations in the information available from mouse models of MPM, we anticipate, shortly, the development of models that improve upon the role of inflammation and PDX methodology.

Inflammation has long been implicated as a crucial effector mechanism driving early but ill-defined events that lead to mesothelioma [52–54]. Proposed mouse models that might replicate this early state would involve driving inflammation in a mesothelial-specific manner. One method to do this would be to drive NF-κB signaling in a hypermorphic transgenic mouse model where inflammation initiating genetic elements are placed under the control of MSLN cis-regulatory transcriptional control elements (MSLN promoter). However, a more attractive model would involve CRISPR-Cas mediated introduction of these same inflammation-driving genetic elements directly into the MSLN locus in a single allele manner. This is permissible since there are no deleterious effects to having a non-functional MSLN allele [55]. Such a model would ensure that the complete repertoire of MSLN regulating genetic elements, such as epigenetic influences, are fully represented. Such a model could be used with or without introducing asbestos fibers to explore the contribution of inflammation to the oncogenic state.

Humanized PDX models represent a possible future avenue to understand MPM better in the setting of a humanized immune system; however, these models are not without significant disadvantages. These models involve using immunodeficient mice such as NSG, which are irradiated then engrafted with human leukocytes and CD34⁺ hematopoietic stem cells. These models are considered successful once confirmed to have more than 25% of CD45⁺ human cells in peripheral blood. A PDX can then be engrafted in these mice who now possess a human-derived immune system, allowing for preclinical assessment of immune-directed therapies. Such humanized PDX models have been developed for cancers such as lymphoma, but have yet to be reported for MPM [42, 56]. However, the relative disadvantages of a humanized PDX model include the difficulty in recruiting patients with a rare tumor (like with PDX) and technical challenges with reliable induction of a complete immune cell repertoire. Additional caveats about this model are 1) high cost and low throughput testing format, 2) risk of graft versus host reaction, and 3) incomplete differentiation of hematopoietic stem cells and weak immune reactivity.

An ideal mouse model for MPM would be easily reproducible and represent rapid tumor development in the thorax with the same genetic and microenvironment seen in human disease and allow for easy and precise tumor assessment. There would be an intact human immune system and reliable methodology for measuring response to various chemotherapeutic, targeted therapies, and immunotherapeutic drugs. This type of model hopefully awaits as technological innovations arise.

ARTICLE HIGHLIGHTS:

• Many drug therapies have been tested against mesothelioma with overall poor survival results necessitating the discovery of novel agents with improved efficacy

• Several mouse models are available: 1) asbestos induction, 2) genetically engineered models, and 3) graft models, including patient-derived xenografts

• No single model accurately represents human disease, and each model has its advantages and disadvantages

• Current gaps in our knowledge of mesothelioma include the role of inflammation, tumor heterogeneity, mechanisms of carcinogenesis, and the tumor microenvironment

• Novel mouse models are needed to better understand the pathophysiology of mesothelioma, with CRISPR genome editing and humanized patient-derived xenograft models as likely future avenues for innovative drug testing and development