Patient-Derived Urothelial Cancer Xenograft Models: A Systematic Review and Future Perspectives

Search results of literature

The systematic review process is represented in Fig. 1. A total of 971 articles were identified through PubMed and Web of Science databases, and manual searching on the reference lists of the included studies increased the total references by an additional 4 articles. Of those 975 references, 853 articles were excluded after title and abstract screening and 73 were further excluded after full text analysis. The resulting 49 articles were therefore analyzed for this systematic review.

OPEN IN VIEWER

Fig.1

Flow diagram of the systematic review process.

These 49 reports were divided into two groups: one included 29 papers focused on methods to establish PDX models and the other included 37 papers (partly overlapping with the former) focused on the use of PDX models. Because the use of severe combined immunodeficiency (SCID) mice and its derivatives for PDX generation did not become widespread until the year 2000, we decided to limit our review of the literature to after the year 2000, resulting in a detailed analysis of 41 reports.

The conventional method of PDX establishment

A variety of methods to establish UC PDX establishment methods have been reported including the origin of specimen (i.e. TURBT sample or metastatic biopsy), the method of implantation, the host mice and the use of Matrigel, etc. (Table 1). Unfortunately, it is difficult to assess the superiority or inferiority of the differing methods because of the small sample sizes per study as well as a lack of reports that directly compare the different methods. That said, the overall engraftment rate, which represents the number of successful tumor growth per patient sample, in reported series with 10 or more clinical specimens ranged from 15 to 60%. In general, immediately after specimen collection and before transplantation, the tumors were maintained in medium containing antibiotics or sterile salt solution. The samples were then cut into 1–3 mm cubes to implant subcutaneously through small dorsal incisions in immunocompromised mice. Implanted tumor fragments often engrafted within about 1 to 3 months, and engraftment failure was judged after 5-6 months of follow-up [13–15]. These grown fragments can be passaged by dividing the excised tumor into smaller pieces and re-implanting them into additional host mice. Efficient cryopreservation protocols for PDX, established in prostate, ovarian and pancreas cancer [16–18], but have not yet been reported for UC. It is necessary to examine whether these methods can be applied to UC PDX.

Immunodeficient mice commonly reported in the literature for UC PDX include nude mice, severe combined immunodeficiency (SCID) mice, nonobese diabetic scid (NOD/SCID) mice and NOD.Cg-Prkdc^scidIl2rgt^m1Sug (NOG) mice as well as NOD.Cg-Prdkc^scid/l2rgt^m1Wjl (NSG) mice. SCID mice are deficient in mature T/B cells. The NOD/SCID strain not only lack mature T and B cells but also are deficient in macrophages and dendritic cells. NOG and NSG strains carry the NOD/SCID background combined with the IL2 receptor common gamma-chain knockout background and thus also lack NK cells [19].

While never formally directly compared, relative to nude mice, the other strains (SCID and NOD/SCID as well as NOG and NSG) of mice appear to have higher engraftment rates (nude: 15% (10/65 tumors) [20], SCID; 50% (15/30) [13], NOD/SCID; 58% (51/87) [15, 22], NOG and NSG; 40% (27/68) [14, 23]. Although the number of samples was small and no direct comparison was made, the use of Matrigel^® did not appear to positively affect the engraftment rates (use; 32% (8/25 tumors), no use; 53% (85/160) in SCID, NOD/SCID and NSG mice) nor was there a difference in the engraftment rate between placing a tumor fragment under the skin using an incision versus injecting the shredded graft with the needle. It is notable however, that tumors did not establish in any of the three cases where the source of xenografted material was a needle aspiration biopsy sample [22]. Interestingly, transplantation under the renal capsule showed a higher establishment rate than other methods [21]. Muscle-invasive tumors demonstrated higher success rate of establishment than non-muscle-invasive tumors (invasive; 63% (43/74 tumors), non-invasive; 42% (10/24) [13, 15]). Skowron et al. reported that success of PDX engraftment predicted poor survival in locally advanced bladder cancer [15]. On the other hand, it has been reported that the TP53 mutation status, one of the most frequently mutated genes in bladder cancer, did not correlate with differences in the tumor take rate [13].

It should be noted with caution that subcutaneous tumors differ from orthotopic tumors in terms of intrinsic gene expression profiles and extrinsic pharmacokinetics, as reported in other cancer types [24, 25]. In this regard, however, it is a tough challenge to establish orthotopic PDX from a bladder tumor. Recently, two reports showed that bladder instillation of dissociated cells or cultured cells from once subcutaneously established PDX tumors enabled to create orthotopic PDX models [26, 27]. These models recapitulated invasion from the bladder lumen and metastasis to the lung as well as the architectural and molecular expression characteristics of patient primary tumors. However, further assessment may be needed to determine whether the cell dissociation process necessary for bladder instillation can retain the intratumoral diversity that is an advantage of PDX.

Histologic and Genomic Characterization of PDX models

Based on hematoxylin and eosin (H&E) staining with or without immunohistochemical staining, most papers claim that PDX maintain their original tumor morphology. This appears to be true in the case of bladder cancer as well [28]. However, it is imperative that bladder cancer histology be confirmed in UC PDX since Abe et al. reported a number of cases in which engrafted tumors were morphologically different from the original tumors. While the pathologic diagnosis from the patient’s tumor was UC, the corresponding PDX turned out to be Epstein-Barr virus (EBV)-associated lymphoproliferative disease [13]. In this case, the authors retrospectively observed EBV-encoded small RNA positive infiltrating lymphocytes in the original tumor, suggesting that the infected lymphocytes were transplanted alongside with the UC cells were able to proliferate neoplasticly in the immunodeficient mice. These PDX lymphocytic tumors have also been reported in PDX models of other cancer types such as breast cancer and pancreatic cancer [29]. Attention should always be paid to the cell morphology of PDX, and if morphological changes are observed, it should be confirmed by immunostaining of lymphocytic markers.

In addition to histology, groups have also examined the genomic fidelity of UC PDX to their initial tumor. Several studies analyzing various genomic parameters showed that genetic identity is preserved between UC PDXs and their original tumor samples, based on karyotyping [28, 30–32], flow cytometric quantification of cellular DNA content [31, 33], short tandem repeat (STR) polymorphism assay [20], targeted mutational analysis [13, 21], array-based comparative genomic hybridization (aCGH) [20, 34], and next-generation sequencing [11, 35]. These studies revealed that the genetic profiles of UC PDXs and their original tumors while not perfectly matched, share very high homology. Jager et al. performed aCGH on 6 UC samples and revealed striking concordance in the copy number profiles between patient tumors and their matched PDX [21]. Interestingly, copy number changes in the PDX models were more prominent, presumably due to reduction in the contribution of human stroma cells. Pan et al. compared whole exome sequencing data of two UC PDXs and original patient tumors to show that they shared 92–97% of genetic aberrations, including multiple druggable targets [14]. Wei et al. also showed that in two cases, the majority (91% and 82%) of all mutations were shared by primary and UC PDX samples [35]. Although both comparisons are limited to a small number of cases, these reports prove that UC PDX retains the genetic identity of the primary tumor as reported in other cancer types [7]. In recent years, molecular subtype classification of bladder cancer has attracted attention, however at this time there are no definitive studies addressing whether RNA molecular subtype is conserved in UC PDX tumors.

PDX establishment of rare histologic variants

For rarer bladder cancer histologic variants that lack other research models (i.e. cell lines or genetically engineered murine (GEM) models), UC PDX models can be very useful tool. Some reports on the establishment of PDX from clinical specimens with rare genotypes or phenotypes have been published. For example, Hofner et al. established neuroendocrine bladder cancer (NEBC) xenograft models from two patient’s tumors and identified MET as a potential new treatment target for NEBC [36]. Furthermore, Funada et al. succeeded in establishing a PDX model of IGF-II producing giant cell urothelial carcinoma an aggressive and rare variant of high-grade UC. Interestingly, the tumor caused non-islet cell hypoglycemia that was recapitulated in the UC PDX mice [37]. There was no apparent therapeutic effect of treatment with a neutralizing IGF II antibody in the UC PDX model suggesting that the IGF-II production was a paraneoplastic phenomenon and not growth promoting [38].

Application of UC PDX models

There are a variety of potential applications for the use of PDX in bladder cancer research. For example, PDX have been studied as a tool for precision medicine. While PDX models have been proposed as avatars to predict or personalize therapeutic response in other tumor types, careful studies (i.e. co-clinical trials) have not yet been carried out in UC PDX to support this approach. Perhaps the closest to demonstrating the promise of UC PDX for precision oncology was Pan et al. who showed that their original 22 lines of PDX model morphologically and genetically preserved the trait of the original specimen and performed molecularly targeted therapy with small molecule inhibitors targeting kinases such as HER2 or FGFR3 [14]. Nonetheless, whether the response pattern seen would be reproduced in the patient counterparts is unknown.

Another promising application is to analyze the established PDX models to better understand bladder cancer biology and identify new drug targets. For example, Skowron et al. found that the PDX engraftment allowed amplification of rare basal cell populations with the expression of cell surface markers (CD90 + /CD44 + /CD49f+), which have tumor-initiating capacity and correlate with prognosis, at similar frequencies when compared to the original donor tumor, while maintaining a heterogenous tumor cell population. They analyzed transcriptome of the sorted basal cells and other differentiated cells which were expanded in the PDX through the process from establishment to serial passages and found that CDC25C was a prognostic marker in bladder cancer patients [15]. Lee et al. focused on collagen I, which is highly expressed in invasive bladder cancer, and examined a rare bladder cancer PDX model that spontaneously developed lung metastasis to find that bladder tumor cells expressing the collagen receptor, CD167a, responded to collagen I stimulation at the primary tumor to promote local invasion and utilized the same receptor to preferentially colonize at airway smooth muscle cells [39]. This result was confirmed in T24 tail vein injection models.

Strategies using complementarity between PDX and 3D culture

Due to the recent development of 3D culture techniques including organoids and spheroids for bladder cancer, there are a growing number of papers combining 3D culture models and PDX models for complementary purposes. Lee et al. were the first to report a comprehensive panel of UC organoid lines in 2018. They succeeded in establishing organoid lines from transurethral resection (TUR) specimens with a high probability (70%) as well as conventional PDX tumors, which recapitulated the histopathological and molecular diversity of human bladder cancer and often retained parental tumor heterogeneity [11]. These organoids could be orthotopically transplanted into NSG mice and the in vivo model recapitulated the drug responses observed in organoid culture. Interestingly, it was also noted that a subset of mutations was either lost or gained, or that molecular subtypes were changed, during serial passaging in culture and/or during grafting or reestablishment of organoids from grafts. This finding implies that clonal selection and cell plasticity may depend on the culture environment and promises that organoids will be a tool to accelerate the elucidation of bladder cancer biology.

Spheroid culture is also useful for drug screening and drug target discovery, although it is difficult for long-term subculture. We have recently demonstrated that drug susceptibilities in PDX models were paralleled with corresponding cancer tissue originated spheroid (CTOS) models, established using a 3D culture method by Yoshida et al. [40]. Using multiple clinical specimens, we successfully identified cases in which a combination of disulfiram, an FDA approved drug for alcoholism, and cisplatin was predicted to produce synergy [41]. In another example of the potential superiority of 3D culture, Namekawa et al. showed that ALDH1A1 expression was increased in T24 cells cultured in 3D compared to 2D cultures and that ALDH1A1 inhibition in patient-derived spheroids and mouse models suppressed the tumor growth [42].

In addition to 3D culture, another recent study described a technology termed conditional reprogramming, which allows efficient establishment of patient tissue-derived 2D cancer cell cultures in the presence of RHO kinase inhibitor and a fibroblast feeder layer [43]. Mondal et al. showed that these conditionally reprogrammed cancer cells established from early passage PDX tumors retain the same STR, genetic alterations and drug sensitivity/resistance profiles. Their proposed strategy for using UC PDX models in combination with the ex vivo culture system should facilitate biobanking and reimplantation as well as high-throughput drug screening and genetic manipulation, which are difficult with UC PDX models alone [44].

Development of humanized mice

As cancer immunotherapy has come to the center of cancer therapy, PDX models using conventional immunodeficient mice have a serious limitation that the local immune response is not reproduced. However, Wang et al. recently reported that programmed cell death protein 1 (PD-1) inhibitors could be evaluated with hematologically humanized mice [9]. Humanized NSG (HuNSG) mice were generated by transplanting human (h)CD34⁺ hematopoietic progenitor and stem cells (HSPC) into irradiated NSG mice. The engraftment levels of hCD45+ cells were determined 12 weeks post-HPSC transplantation by flow cytometric quantification of peripheral blood hCD45+ cells. HuNSG mice that had over 25% hCD45+ cells in the peripheral blood were considered as engrafted and humanized. Treatment with pembrolizumab, which targets PD-1, showed significant growth inhibition in non-small cell lung cancer PDX tumor in HuNSG but not in hematologically unmanipulated NSG mice. Bladder cancer PDX have also been shown to grow in HuNSG mice with partially HLA-matched allogenic human immune systems and showed tumor responses when treated with pembrolizumab. However, in mice that were transplanted HSPC from different donors, pembrolizumab was no longer effective. It is not yet clear what caused it, and further investigation is needed. Blinoba et al. also created NOG / SCID mice transplanted with UC PDX and human lymphocytes from healthy donors at the same time and verified the effect of anti-PD-L1 antibody [10].