Pleural mesothelioma is a cancer of the lining of the lungs that is predominantly caused by asbestos exposure. Most cases can be histologically classified as either epithelioid, sarcomatoid or biphasic ( Lokuhetty et al., 2021), with loss of tumor suppressors being a common genetic feature. For example, loss or inactivation of the nuclear deubiquitylase BRAC1-associated protein (BAP1) is found in 60-70% of cases and is most frequent in the epithelioid subtype ( Bott et al., 2011). Despite the recent approval of immunotherapy for untreated mesothelioma ( Baas et al., 2021), pleural mesothelioma remains an incurable disease with poor prognosis. There is, therefore, still a pressing need for new therapies, and consequently the requirement for drug screening and preclinical testing. Typically, rodent models would be used for such studies; however, these models are costly, labor-intensive to maintain, take months to produce results ( Reyal et al., 2012), and there is a growing demand to reduce the use of protected animals in research. Despite a declining trend in the use of mice since 2007, they remain a commonly used model with 60% of all animal experiments in 2023 being conducted in this species, which equates to around 0.9 million mice in the UK alone ( Baxter, 2024). Around one-fifth of these animals were used for either basic or translational research of human cancer and 40% of all mouse experiments involved non-genetically modified models ( Baxter, 2024), such as xenografts. If a fraction of these were replaced with non-mammalian models it would save thousands of mice per year.

The extra-embryonic chorioallantoic membrane (CAM) of fertilized hens’ eggs can be used to generate cancer xenografts for the translation of in vitro cancer studies. The CAM model has multiple benefits; being cheaper and having fewer husbandry requirements compared to mammalian models. It is also an attractive option as avian embryos are non-protected up to two thirds gestation, with pain perception only becoming functional from E15 onwards in chickens ( Süß et al., 2023), and so they do not fall under Home Office regulations up to embryonic day 14 (E14) in the UK. Thus, the CAM is both more ethical and simpler for researchers to adopt as an in vivo model than mice. This 3Rs-compliant model has been successfully employed to study the key hallmarks of cancer, and for evaluation of new therapeutics, and thus can significantly contribute to reducing or replacing use of rodent models in cancer research ( Schneider-Stock and Ribatti, 2021).

We have previously shown the utility of the CAM model for the generation of mesothelioma cell line xenografts ( Barnett et al., 2022). While cell lines are a tractable option and are useful for optimizing protocols and experimental workflows, patient-derived models are more representative. The recent development of a breadth of in vitro patient-derived models for mesothelioma, such as organoids ( Shamseddin et al., 2021) and explants ( Powley et al., 2020), is important for driving forward mesothelioma research. However, the choice of model should be carefully considered. For instance, patient-derived organoids (PDO) lose the original tissue architecture and cellular composition, whereas both patient-derived explants (PDE) and patient-derived xenografts (PDX) retain these features. Since mesothelioma is a highly fibrotic cancer, with stromal and immune cells influencing tumor progression, maintaining the tumor microenvironment is particularly important to ensure a more representative treatment response ( Chrisochoidou et al., 2023). Furthermore, a key benefit of CAM xenografts over both explants and organoids is the vascularization of tumor nodules, which maintains viability, whereas explants frequently become necrotic. This vascularization also enables systemic delivery of therapies and reagents through chick-derived blood vessels, via intravenous or yolk-sac injection, thus offering a more physiologically relevant means of treatment administration.

Although, several mouse models of mesothelioma have been developed including asbestos-induced ( Grosso et al., 2021), conditional genetic knockouts ( Kukuyan et al., 2019) and, of most relevance here, xenograft or PDX models ( Wu et al., 2017), these can manifest severe phenotypes. As well as ethical considerations, establishing PDXs in mice can be difficult, with a take rate of only 40-50% in some cases ( Boughey et al., 2021) and mesothelioma is no exception ( Wu et al., 2017). Furthermore, generating mouse PDX models can take several months, with reports of 2-8 months for first generation engraftment in breast cancer for example ( Reyal et al., 2012). Thus, for applications such as patient avatars, this may be futile in mesothelioma where median survival is 12-21 months, and only 7 months for the sarcomatoid subtype ( Amin et al., 2018). The CAM xenograft assay can be considered as a direct replacement for mouse xenograft flank models and has been tested as a PDX platform for a variety of other cancer types such as glioblastoma ( Shoin et al., 1991), clear cell renal cell carcinoma ( Fergelot et al., 2013), as well as one publication reporting use of mesothelioma tissue ( Mîndrilă et al., 2017). Published studies report high engraftment rates (70-80%), histological concordance with tumor type and a strong correlation between CAM-PDX assay results and patient outcome ( Shoin et al., 1991, DeBord et al., 2018, Charbonneau et al., 2023), supporting the CAM model as an alternative method for generating physiologically applicable PDXs that can be utilized for drug screening and personalized medicine.

PET/CT is routinely used in the clinic for detection and staging of cancer, mesothelioma included, as well as monitoring of treatment response in patients ( Kruse et al., 2013). Moreover, since mesothelioma has a large stromal compartment, an imaging technique that selectively reports on cancer cell metabolism as a surrogate marker for viability may be more informative of treatment response than standard readouts such as tumor weight. Additionally, as others have previously demonstrated the feasibility of applying PET/CT imaging to cell line-derived CAM xenografts this was our modality of choice ( Warnock et al., 2013, Smith et al., 2023, Schulze et al., 2023a). Hence, we developed a workflow to combine CAM-PDX and PET/CT imaging to facilitate assessment of treatment response in this model.

To make this approach accessible to other laboratories as a replacement for mouse PDXs, our aim was to develop robust protocols that describe in detail how to generate CAM-PDX models of mesothelioma and monitor tumor cell composition and health to assess therapeutic responses. Our specific objectives were to: (1) Optimize a pipeline for processing, characterization and cryopreservation of biopsy samples to ensure availability of high-quality tissue for engraftment; (2) Demonstrate robust engraftment of tissue, and preservation of architecture and cellular composition in PDX nodules; and (3) Apply in ovo [ ¹⁸F]-FDG PET/CT imaging to measure cancer cell metabolism within PDX nodules. These CAM-PDX methods can be extrapolated to other tumor types, and the tissue processing method could be applied to the generation of PDEs, or even mouse PDX models, to increase experimental efficiency thus having a wider impact across models.

Supplier information and catalogue numbers for reagents, consumables and equipment are detailed in Table 1.

As initial proof of principle that freshly collected mesothelioma tissue can engraft onto the CAM, we scheduled initiation of embryo development and windowing so that eggs were ready for engraftment at E7 on the day of the biopsy surgery. We implanted 1 mm ³ fragments of fresh sarcomatoid tissue at E7 and, by E14, we observed that 93% (n = 15) of nodules were vascularized based on evidence of remodeling of the CAM vasculature and feeder vessels ( Figures 3 and 7A; Table 11). Immunohistochemical assessment of the PDXs showed the presence of calretinin (CalB2)-positive mesothelioma cells and expert pathological assessment of H&E staining found localized spindle cell proliferation showing atypical cells with hyperchromatic nuclei favoring sarcomatoid mesothelioma cells, which was consistent with the patient’s diagnosis. The mesothelioma cells were BAP1-positive, in line with expectations for the sarcomatoid subtype since only 20% are BAP1 negative ( Righi et al., 2016). Additionally, Ki67-positive proliferating cells were present (black arrowheads, Figure 7B), with low levels of apoptosis as determined by caspase-3 (cCasp3) staining ( Figure 7B). We also observed the presence of host and/or patient-derived alpha smooth muscle actin (α-SMA)-positive fibroblasts, as well as preservation of patient-derived CD8+ and CD4+ immune cells ( Figure 7B).

Since pleural mesothelioma is a heterogenous cancer, where tumor cell content and distribution can differ greatly between, and within patient samples ( Figure 8A), a workflow to facilitate pre-screening of the tissue samples ( Figure 1A) was employed to optimize selection for implanting and ensure regions with a greater proportion of tumor cells are prioritized for generating the CAM-PDXs. Moreover, surplus biopsy material was obtained immediately after procedure and only 50% of samples collected were later determined to be mesothelioma by the clinical diagnostic service ( Table 10). Therefore, a freezing protocol was utilized to preserve the tissue whilst maintaining histological concordance with the fresh counterpart ( Figure 8B), to allow time for histopathological confirmation of mesothelioma and, in line with 3Rs principles, prevent egg development being initiated unnecessarily should surgery not go ahead as planned, or implanted samples later be determined as not mesothelioma.

Next, we sought to assess two commonly used freezing media, termed “High FBS” or “Low FBS” ( Table 2), to determine which best preserves the tissue architecture and viability of the collected mesothelioma biopsy material. Adjacent fragments of tissue were slow-frozen in either high or low FBS freezing media with one piece of fresh tissue immediately fixed (“fresh-fixed”) for histology ( Figure 1). After a period (>1 month) in liquid nitrogen storage, exemplar fragments were thawed and fixed for histology. H&E staining revealed standard artifacts from freezing in both cryopreservation media, however viable tumor cells were present as confirmed by a pathologist ( Figure 9). Immunohistochemical staining confirmed the presence of tumor cells (panCK-positive) that histologically matched the patient diagnosis of BAP1-positive epithelioid mesothelioma, across all three conditions ( Figure 9). Further characterization revealed the presence of proliferating cells and no visible apoptosis. Staining for α-SMA showed similar fibroblast composition, whilst CD4+ and CD8+ immune cells were detected in all conditions. Overall, these data show that both freezing methods were successful in preserving tissue architecture whilst maintaining the cellular composition of the sample.

To assess reanimation of viable tissue, we thawed fragments slow-frozen in high or low FBS freezing media (from the same epithelioid biopsy sample region shown in Figures 8 and 9) and implanted them onto the CAM. Visual scoring for remodeling of the CAM vasculature and feeder vessels showed that 82% (n = 27) versus 74% (n = 26) of PDX generated from tissue fragments stored in high or low FBS freezing media, respectively, had formed well vascularized nodules after 7 days ( Figures 10A and 11A, Table 11). On E14, the resulting CAM-PDXs were dissected and processed for histological assessment. We observed retention of mesothelioma cells and BAP1 status, along with presence of proliferating Ki67 cells (black arrowheads, Figure 11B) and little evidence of apoptosis ( Figures 10B and 11B). There was also evidence on H&E of nucleated red blood cells inside cross sections of vessel-like (α-SMA positive) structures in the CAM-PDXs (white arrows, Figure 11B), indicating perfusion of chick blood within pre-existing or new vasculature within the engrafted tissue. The only discernible difference observed was that the frequency of CD4+ and CD8+ T-cells was lower in the CAM-PDXs derived from tissue frozen in the low FBS media ( Figure 11B). Overall, these data show successful engraftment on the CAM following reanimation after freezing and it should be noted that despite the sparse tumor cell content of the example CAM-PDX shown in Figure 10, it offers proof of principle of the methodology irrespective of how challenging the sample is. Finally, as it is common to serially passage mouse PDX ( Wu et al., 2017), we assessed the feasibility of this for the CAM-PDX model. Briefly, CAM-PDX generated from frozen epithelioid fragments were dissected on E14 (P0), cut in half and transplanted to new E7 eggs (P1). We observed vascularization of the nodules following passaging irrespective of the freezing method ( Figure 12). However, expansion of the tissue was not evident over the 7-day engraftment period, likely reflecting the diffuse tumor cellularity and biology of mesothelioma.

To visualize viable mesothelioma cells within the CAM-PDXs in ovo, we required an imaging modality that is suitable for unlabeled tissue and we therefore utilized [ ¹⁸F]-FDG PET/CT. [ ¹⁸F]-FDG is a non-metabolizable glucose analogue widely used as a radiotracer in oncology, including mesothelioma ( Szlosarek et al., 2013), as elevated glucose uptake is commonly observed in cancer cells due to the Warburg effect ( Tekade and Sun, 2017). PET imaging after intravenous administration of [ ¹⁸F]-FDG revealed tracer uptake ( Figure 13A) in the CAM-PDX which were established from the same cryopreserved epithelioid tissue shown in Figures 8- 11. [ ¹⁸F]-FDG uptake indicates that the CAM-PDXs were sufficiently well vascularized to accumulate administered substances, such as the radiotracer, from circulating blood, and that they are metabolically active. Quantification of the PET signal, reported as total accumulated SUV across the entire CAM-PDX (SUVacc), showed similar uptake in CAM-PDXs (Two Sample t-test, p = 0.76), irrespective of the cryopreservation media used ( Figure 13B), suggesting that both freezing methods preserve viable tissue that can be reanimated for engraftment. Despite this, we recommend high FBS freezing media, as a higher percentage of nodules became vascularized and there is evidence of better resident immune cell preservation.

The CAM assay is a cost-effective, higher throughput preclinical model that can contribute to accelerating translation of in vitro studies and the treatment discovery pipeline. With increased uptake of the CAM model in the research community, it will become more widely accepted as a direct replacement for the ‘gold standard’ mouse model, therefore having a greater contribution to reducing the use of mammalian models. We propose that as the model is moderate throughput, drug evaluation can be conducted on first generation (P0) CAM-PDXs, potentially further reducing numbers, as it would remove the standard practice of serial passaging in mice. If one successful mouse PDX is typically passaged 5-times, using 5 mice each time, then the CAM-PDX model could theoretically reduce the number of animals used per PDX by 25. We expect the CAM-PDX model to be utilized for applications such testing combination therapies, formulations, drug delivery systems and novel treatment strategies. To ensure experimental rigor, we recommend that the study design should incorporate the appropriate controls such as blinding to any information relating to the sample and treatment condition, for all stages of experimental intervention, imaging and analysis.

We have shown that viable, vascularized mesothelioma CAM-PDXs can be generated from both fresh and frozen fragments of biopsy tissue, and that these are amenable to viability assessment using PET/CT imaging. Importantly, morphological and immunological features of the primary sample are retained, the latter being a key advantage over cell line-derived CAM xenografts and methods that involve disaggregation of patient derived material, particularly with the growing importance of immunotherapies. The inclusion of PET/CT readout is highly translational and relevant, since it is routinely used in the clinic to detect and stage cancer as well as monitor treatment response. Together the methods we describe will allow assessment of response to a broad range of therapies in an ethical model that provides an excellent recapitulation of human disease and allows systemic administration of drugs.

We have established the feasibility of cryopreserving tissue prior to PDX generation bringing together several practical and 3Rs benefits, making these models more readily adoptable. For instance, when utilizing surplus material from diagnostic biopsies not all will be confirmed as cancer, indeed in our case only half of samples we collected were confirmed as mesothelioma. Careful tissue processing and cryopreservation enables histopathological confirmation of the cancer type prior to starting an experiment. Furthermore due to the heterogeneous nature of pleural mesothelioma, immunohistochemical pre-screening of the tissue to assess tumor cell content increases the chances of using tumor-cell rich regions. Although CAM xenograft models are non-protected up to embryonic day 14 ( The Animals (Scientific Procedures) Act 1986 Amendment Regulations, 2012) and more ethical than mouse xenograft models ( Ribatti and Annese, 2023), they are not truly animal-free. Therefore, the methodology adaptations we describe here are important to ensure that embryo development is only initiated when good quality and well characterized banked tissue is available for PDX engraftment.

Our work is also applicable across other mesothelioma models that utilize patient tissue. In an era when therapeutic surgery is increasingly rare following publication of the MARS2 study ( Lim et al., 2024), the tissue pre-screening workflow described here can be utilized for efficient generation of any patient-derived models from biopsy tissue, including in vitro PDEs. In addition, where mouse PDX models are necessary, animal numbers can be reduced by ensuring only tissue confirmed as the cancer of interest, as well as tumor cell-rich regions are utilized. Freezing samples prior to engraftment also has the benefit of widening the availability of tissue, and therefore the appeal of the CAM-PDX model, to researchers who may be unable to readily obtain fresh tissue, for example due to a lack of local specialist hospitals or clinical collaborators.

The CAM-PDX model sits between PDEs and mouse PDXs, offering a middle-ground between tractability, time scale and physiological relevance. A particular advantage over in vitro methods is host-derived vascularization that is only achievable with in vivo models. Moreover, the short experimental window of the CAM-PDX model means results can be available in weeks compared to the months required to obtain meaningful data from mouse PDX models. However, since there is only a 7-day window from engraftment to termination of the experiment using the CAM, animal models may still be required for longer term studies, for example if therapies require a prolonged treatment regime. Although the ability to maintain mouse PDX models for long periods does have drawbacks, for instance it allows more time for the human tissue to be infiltrated by host-derived cells. It is also common practice to passage mouse xenografts to extend study time and expand material ( Wu et al., 2017). However, serial passaging can result in loss of human-derived stroma ( Julien et al., 2012), immune cells and contribute to further colonization by mouse stromal components leading to a far less representative model over time ( Rosfjord et al., 2014). Although we have shown transplanting CAM-PDXs is possible, there was no evidence for expansion of the material and given the limitations observed with passaging PDX in mice, researchers should carefully consider whether this is necessary.

Here we describe in detail how to perform PET/CT to facilitate in ovo assessment of unlabeled xenografts. Whilst PET allows us to visualize the xenografts, CT scans enable the precise localization of the region of interest. It also serves to apply attenuation correction in the PET images, which increases the accuracy of the quantification of radiotracer uptake. Although dynamic scanning allows detailed analysis of physiological processes, it is time-consuming and requires complex data processing. Combining static PET imaging with an efficient injection technique means up to 24 eggs can be scanned per day (8 working hours), only limited by the activity of the decaying radiotracer, making it a high-throughput preclinical screening tool.

The experimental pipeline we describe has the potential to significantly accelerate the advancement of personalized medicine, as it could be utilized as a rapid avatar for treatment selection or establishing patient suitability for clinical trials ( Izumchenko et al., 2017), as opposed to the slower, more costly and less ethical murine PDX models. The option to test treatments in a rapid manner using the CAM-PDX platform, where turnaround time can be weeks instead of months, is highly desirable for cancers with poor prognosis such as mesothelioma, where patients may only survive a few months after their diagnosis.

We believe that these combined protocols for efficient tissue processing and the generation of CAM xenografts from mesothelioma tissue can significantly reduce requirements for mouse PDX models. Furthermore, due to the success rate of generating CAM-PDX from a range of cancers ( DeBord et al., 2018) we envisage this model accommodating most solid cancer types, bringing wider impact in the reduction and replacement of murine models in cancer research. Additionally, due to the minimal requirement for expensive reagents or equipment, the CAM model can be easily adopted in standard research labs. As well as expanding to other cancer types, there is also scope for combining with other preclinical imaging such as magnetic resonance imaging ( Barnett et al., 2022), adding alternative read-outs such as spatial profiling, or volatile organic compound analysis ( Little et al., 2024), and investigating novel therapies ( Schulze et al., 2023b) and delivery vehicles ( Butler et al., 2022). Overall, the need for more accessible and physiological models of cancer, faster translation of therapies and the desire to reduce rodent model usage, makes the CAM-PDX model an attractive option.

The collection and use of human tissue samples for this study was covered by the generic ethical approval granted to Mesobank – a Research Tissue Bank, approved by East of England - Cambridge Central Research Ethics Committee (approved 9 ^th August 2019, 18/EE/0161; renewed 30 ^th August 2023, 23/EE/0139). Therefore, a separate REC approval for this project was not required. Patient samples were donated under written informed consent and collected between March 2021 and November 2024. All experiments using human tissue were conducted in accordance with the Declaration of Helsinki and in compliance with all local policies and standard operating procedures for working with human material. Samples stored and research performed at the University of Liverpool under licensing no. 12020 granted under section 16(2)(e)(ii) of the Human Tissue Act 2003.

All experiments utilizing fertilized Hens’ eggs were terminated by embryonic day 14 (two thirds of the gestation period) meaning the model is classified as non-protected under The Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 in the UK and therefore no Home Office approval was required. Standard operating procedures were reviewed and approved by the University of Liverpool Animal Welfare and Ethical Review Body.