Patient-derived xenograft (PDX) models, applications and challenges in cancer research

Generally, these models can be created as orthotopic, heterotopic, and metastatic. In heterotopic models, at the first step, the tumor tissue specimen should be prepared in a cold solution containing Fetal Bovine Serum (FBS), penicillin/streptomycin to increase the success rate of engrafting. The tumor tissue can be transplanted in three approaches; the tumor tissue can be cut into pieces and several 3–5 mm³ pieces, a large amount of tumor or injection of minced tumors can be used for transplantation. The tumor samples are washed three times with the above-mentioned solution before transplantation. Various parts of the mouse can be used as tumor tissue transplantation sites to create PDX models. The tissues can be engrafted heterotopically into the intracapsular fat pad [17, 18], the anterior compartment of the eye [19], under the renal capsule [20], subcutaneously, and orthotopically into the origin of cancerous tissue [21]. Orthotopic transplantation is usually preferred because it has a closer microenvironment to human cancer. But in general, both subcutaneous and orthotopic sites are most commonly used [19].

It should be noted that depending on the type of tumor and the sample area, the concentration and type of antibiotics may change. Small incisions are created on the lower back for subcutaneous transplantation in 4–8-week-old immunodeficient mice, and the tumor tissue sample is placed in each of the surgical areas. The remaining tumor tissues can be stored (− 80 °C for short term and in a liquid nitrogen freezer for long term storage) for genomic and protein evaluations compared with the xenograft tumor. The first generation F1and the subsequent generations are named F2, F3. Although designation can be done in other ways such as (G1, and so on) [22‐24]. Reports suggest that the efficiency of nu/nu athymic mice is 75% and that NOD/SCID mice are more commonly used in F1 [25, 26]. The creation of NOD/SCID/IL2rγnull (NSG) mice resulted in higher efficacy of 95–100% in tumors that are difficult to transplant [27] (Fig. 3).

Xenograft tumors are assessed at least twice a week by a vernier caliper to measure tumor length and width. The time required to reach cancer depends on type of cancer, the location of the transplant, and the recipient strain. It takes an average of 2–4 months, and the transplant has not been successful if no tumor is detected over 6 months. When the size of the tumors reaches 1–2 cm³, we can begin the tumor passage. Finally, the third generation (F3 or G3) can be used for drug treatment and commonly employed in studies. However, genetics and histology (rather than merely the number of passages) should be the main determinants of PDX derivation from the patient’s tumor [28‐30].

In recent years, the development of PDX models in humanized mice has been considered, in particular form in humanized PDX models, human immune system and human tumor tissue both together will be engrafted to the model and then human immune system reconstitute in an immunodeficient mice along with patient tumor engraftment [31].

There are three main classes of humanized mice; (A) human gene transgenic mice that are modified by specific human gene expression; (B) humanized organ mice which in this model the mice carrying a specific organ of human such as human hepatocyte infusion; (C) humanized immune system mice which are established in immunodeficient mice and human cells will reconstitute the mice immune system which describe briefly in Table 2 [32, 33].

PDX Clinical Trials (PCTs) are substantial for clinical decision-making before human clinical trials and the development of anti-cancer agents. PCT is referred to as “phase II type clinical trial-like models”. In 2015, Gao et al. designed a high-throughput in vivo drug screening method, “1 × 1 × 1”, which intended one animal per model per treatment using a large number of PDX models. However, two or three animals per model per treatment (2 × 1 × 1 or 3 × 1 × 1) PDX clinical trials because of more representative of generalizable drug response have recently become more common [34] (Fig. 4A).

PDX models have become a powerful tool for evaluating drug efficacy and drug sensitivity, also known as PDX clinical trials. As shown in Table 3, several studies have tested the response rate of different drugs to different cancers in PDX models. These studies have shown that the response rate of PDX models to the drug is correlated with clinical outcomes [35]. In fact, before a human clinical trial, PDX clinical trials are critical to evaluate anti-cancer therapies. Accordingly, these cancer model features have led research centers and pharmaceutical companies to develop PDX repositories. The Novartis Institute of Biomedical Research has launched the PDX library. In this project, Novartis has created more than 1000 Avatar mice of various cancerous tissues, technically, there is no difference between Avatar and PDX, but an Avatar is mainly used in cases where the goal is personalized medicine; they use the predictive capacity of this model for particular patients, in contrast, PDXs are more commonly used in preclinical studies of drug evaluation and co-clinical trials. Accordingly, research centers and large pharmaceutical companies seek to establish a PDX repository, to be used in the development of preclinical studies on anti-cancer drugs [36]. Europeans have established a consortium called EurOPDX to store PDXs, which currently holds more than 1500 PDXs samples [37, 38]. Since 2017, the NCI has also been developing a national repository of Patient-Derived Models (PDMs), including PDX models. The overall goal of NCI is to establish a long-term storage site for at least 1000 PDX models so that researchers have sufficient biological and clinical diversity to conduct their studies [39, 40].

This model has helped increase our understanding of the response of tumor cells to drugs, which leads to effective treatment strategies. Other applications of such models include creating PDX resistance to treatment, in which the consecutive administration of a particular drug can lead to the production of drug-resistant PDX models. These resistant tumors are more consistent with cancerous cells than cell lines, which can be used to trace drug resistance biomarkers and investigate drug resistance mechanisms [41]. For example, ovarian cancer exposed to long-term cisplatin induces resistance to this substance, Similar to clinical conditions. These models are used to find new anti-cancer agents, to evaluate suitable drugs for patients with platinum resistance [42]. Intratumoral heterogeneity of primary tumors in patients is recapitulated by PDX tumors, increasing evidence indicating that tumors have a distinct subset called Cancer Stem Cells (CSCs). Another advantage of this model is the maintenance of CSCs. These cells have been purified and characterized by several types of PDX tumors through specific surface markers, which can be used for drug screening and discovery studies [43].

PDX models have accelerated the development of medications and the phases of clinical trials. Currently, the prediction value of PDX models is used in co-clinical trials. Phases 1/2 of the clinical trials take more than 5 years. Due to limitations in analysis and data integration, co-clinical trials have been suggested at this time (mouse hospitals are manifested), whereby new drug therapies are performed on experimental tumor models concurrently with clinical trials. Finally, the pre-clinical and clinical data are aggregated [44, 45]. The approach of the co-clinical trial using the PDX model facilitates the prioritization of optimal medications, facilitates rapid classification of respondents, determines biomarkers, and detects resistance mechanisms [44]. This format of trials has been or is being conducted for some cancers, such as Sarcoma [46], and Solid tumors [47] (Fig. 4B).

PDX enables the growth of a tumor from any patient in an in vivo system called “Avatar models” or “mirror model”. Avatar is mainly used in personalized medicine approaches. They use the predictive capacity of this model for a particular patient. This model allows for the evaluation of drug sensitivity and precision drug efficacy and eliminates the cost of non-targeted therapy by providing a personalized regimen [48]. For example, Kopetz et al. established an Avatar model of colorectal cancer with BRAF mutation, which by using vemurafenib as an oral BRAF inhibitor, drug resistance was observed due to KRAS (Kirsten Rat Sarcoma Virus) and NRAS (Neuroblastoma RAS Viral Oncogene Homolog) mutations at low allele frequency in this tumor subset [49]. Each patient can have an equivalent animal, assuming that the animal model (PDX) has a similar function in the patient. Additionally, the acquisition of tumor profiles at different times with different therapies by PDX models allows for the evaluation and understanding of various molecular factors, signaling pathways, the flow of tumor growth metabolism, and molecular changes leading to metastases and drug resistance [48].

In recent years, immunotherapy has achieved widespread success against various malignancies, so the limitations of the xenograft cell line-based model and Genetically Engineered Mouse Model (GEMM) have led researchers to become more inclined towards PDX models by reconstitution human immunity in PDX models and to create humanized models, scientists will be able to evaluate immunotherapies. In this approach, the correlation between the hematopoietic system and the tumor of the same patient is necessary. Humanized PDX models by reconstituting the human immune system and tumor growth enable the investigator to study tumor biology and immune system functionality. Zhao et al. have been developed humanized PDX model by type I human leukocyte antigen matched human immune system in NOD-SCID Il2rg−/− (NSG) mice as an immuno-oncology model to study immunotherapy approaches which both therapeutic and side effects of pembrolizumab and ipilimumab had been investigated in this model [50]. Immune check point investigation in gastric cancer humanized avatar model by engrafting the human tumor and PBMC injection showed delayed tumor growth by using anti-check point monoclonal antibodies [51]. These Humanized models have been used to evaluate many targeted therapies such as anti-PD1/PDL-1, CTLA-4, and other checkpoint inhibitors [52‐54].

In 2010, the autologous CAR-T effect on PDX models of the same patient was evaluated for the first time [55]. In another study, PDX models of three hepatocellular carcinoma patients were developed, all of which retained the primary tumor characteristics, CAR-T inhibited tumor growth in the first group, and the tumor was eliminated in the other two groups [56]. In a 2017 study, the pancreatic cancer PDX models were used to evaluate anti-mesothelin CAR-T cells, which effectively inhibited the tumor [57]. Another application of PDX models in cell therapy is assessing various aspects of CAR-T cell treatment and biology. One of these aspects is the interaction between CAR-T and other immune cells such as Tregs and Myeloid-Derived Suppressor Cells (MDSCs) in the tumor microenvironment. These immune-inhibiting cells could be one of the causes of the negative results of some CAR-T trials. In the presence of these immune-inhibiting cells, PDX models will show more accurate and acceptable results, eventually, models with Patient immune systems such as humanized PDX models have received more attention from immunotherapy researchers [58]. Humanized PDX models are future tools in personalized medicine and will support clinical decisions. In an avatar (hIL2-NOG mice) of human melanoma patients, anti-PD-1 (Programmed Cell death Protein 1) antibody response and tumor-infiltrating T cells supported the clinical decision for immunotherapy [59]. However, the humanized immune-PDX models still need to be more validated. Still, they can eventually be a suitable approach for personalized medicine and clinical decision, especially for novel cancer immunotherapies (Fig. 5).

In xenograft models used in immunotherapy, the type of immunodeficient mice should be considered. As shown in Table 4, NOD/SCID /IL2rγnull (NSG) mice suitable for immunotherapy, these mice lack the production of cytokines such as IL-2, IL-4, IL-7, IL-9, IL-15 and, IL-21, which cause the loss of immune cells B, T, and NK, so these rodents are essential for transplantation of primary human tumors, while other immunocompromised mice are suitable for the development of cancer cell line-derived models [60, 61].

PDX of Non-Small-Cell Lung Carcinoma (NSCLC) by imitating the molecular pattern of mutations and histopathological characteristics had been reported on 127 stable models by Wang et al. [62]. In the following, they showed that these emulator models are as same as their donors in gene expression profiling [62]. One hundred surgically resected specimens had been used to develop the PDX model of NSCLC by Ilie and colleagues. Among them, 35 models were established, and anything the team expects from these models was met, similar to tumor specimen [63]. These models have been used to evaluate common drug resistance such as docetaxel, cisplatin and can also be used to identify therapeutic predictive biomarkers [28, 64]. Lung tumor tissue samples were implanted under the sub-renal capsule in immunodeficient mice at a ˃ 95% rate [65].

Lung cancer patients who are at greater risk of recurrence after surgery are predictable, Due to the success of the PDX model’s engraftment rate. Similarly, in a study of 63 of 157 tissue samples of lung cancer transplanted to mice, they were compared with the clinical outcomes of patients, which were predicted to have a lower survival rate than others in a multivariate analysis [25].

PDX models provide a robust basis to study this fertile soil. First, cancer-related female death is not a single disease but a network of mutations’ which may lead to the development of new target therapies that, in turn, a preclinical in vivo model is needed [66]. The first transplantation of mammary tumor lines had been established in 1903, till 2003 that the efficient engraftment of human tumor tissue had been done in immunodeficient mice, CDX models of HER2+ (Epidermal growth factor Receptor 2) models, GEMM model of the BRCA1 (Breast Cancer Gene 1) mutation, and spontaneous mammary adenocarcinoma in transgenic mice were the most available in vivo models to study targeted therapies [67].

Initially, models of hormone-dependent cancers had many challenges. But over time and further research into the orthotopic transplantation of breast cancer tissue in the mammary fat pads of immunodeficient mice, estrogen supplementation has achieved some favorable engraftment rates [68, 69] (Table 2). The different parts of tumor tissue had been reported to use for Xenograft model establishment; primary breast, pleural effusion, ascites, and metastatic sites like the ovarian, brain, Nodal, skin, and soft tissue. Statistically Controversial metastatic site was found between the PDX model and human tumor; bone and brain metastasis are more frequent in patients but lymph nodes and lung in PDXs [18]. Triple-negative (ER-negative, PR-negative, and HER2-negative) tumor tissues have higher engraftment rates due to the aggressiveness of this type of cancer [69].

Various studies have shown that PDX models maintain heterogeneity of the Colorectal Cancer (CRC) primary tumor [70, 71]. The results of drug evaluations are most similar to the clinical settings. The rate of engraftment is dependent on the immunodeficient mice strain and the method of delivery. Tumor tissue is usually transplanted heterotopically into the subcutaneous or sub-renal site [72, 73]. However, several studies have isolated tumor tissue cells by enzymatic digestion and used them to develop this model [74, 75]. In other research, orthotopically, scientists developed CRC PDX models with endogenous metastases that can migrate to the lungs and liver similar to a patient’s primary tumor. Subcutaneous engraftment does not metastasize to other organs [74, 76, 77]. PDX models with NOD/SCID mice have the highest engraftment rate in this cancer [9]. PDX CRC models maintain important gene mutations such as KRAS, BRAF (v-Raf murine sarcoma viral oncogene homolog B), PIK3CA (Phosphatidylinositol-4, 5-bisphosphate 3-kinase), gene expression, copy number changes, and microsatellite instability of the primary tumor [78, 79]. Almost all studies on preserving the stromal structure, histological differentiation, and histopathological subtypes by these models are in agreement [74, 76]. PDX model of colorectal cancer in the evaluation of response rates of drugs such as cetuximab and other systemic chemotherapeutic agents [76, 79, 80] as well as the cell line production [81, 82], drug and biomarker discovery [83, 84], further understanding of tumor biology [83] and colosphere production [85] have had broad applications.

Prostate cancer PDX models have been introduced as models with molecular diversity and cellular heterogeneity that have similar histology compared to primary tumors [86, 87]. Prostate cancer PDX models can be used to evaluate the efficacy of anti-cancer drugs [88]. However, the transplantation rate of human prostate cancer was low, and only advanced and high-growth cancerous tissue was successfully transplanted [89]. The sub renal capsule (SRC) site has a high transplantation rate because it has highly vascularized potential than the subcutaneous site for primary human prostate cancer; however, orthotropic models have shown the best expression of androgen receptor and Prostate specific antigen (PSA) [90, 91]. Another approach recently considered to increase the rate of transplantation is the use of CTCs, which have been used to produce various cancers, including the PDX prostate model. In this type of PDX model, there is no need for surgery to remove the tumor tissue [92]. The prostate tumor take rate is low 10–20% and requires Supplementing mice with a specific dose of exogenous androgens [87]. Different PDX models of prostate cancer for performing pre-clinical studies and evaluating biological processes such as identification of pluripotent stem cells [93], angiogenesis [94], response to androgen ablation therapies [95], and Chromosomal abnormalities [96] have been created over the years.

Gastric cancer cell lines and animal models derived cell lines were used for drug screening; although these models had advantages, most importantly, the low predictive value of these models for clinical outcomes makes them inappropriate [97, 98]. The primary way to overcome these limitations is to use PDX that allows the transplantation of a patient’s tumor tissue into immunocompromised mice to form a population of mice with a tumor of the same molecular pattern. Suitable for evaluating targeted and personalized therapies [37, 80]. The most important achievement of an ideal preclinical model is preserving the genetic and histological features of the patient’s tumor, which are kept in the PDX model of gastric cancer for several generations [99‐101]. Various factors have contributed to increasing the success rate of PDX modeling, including engraftment site [28, 37] using Matrigel [102] and type of immunocompromised mice [37, 103]. However, the diffuse type of gastric cancer is reported to have a low engraftment rate due to insufficient tumor cells [104]. Epstein bar virus (EBV)-related B-cell lymphomas infect approximately 68% of PDX tumor models [102, 105]. These are due to the presence of EBV in the patient's primary transplanted B cells, which are activated in the body of immunocompromised mice and cause lymphoma instead of the tumor [106]. PDX models have helped to improve gastric cancer therapies in recent years. Because of their in vivo benefits, and have been used in numerous pre-clinical trials to evaluate treatment response and identify biomarkers [38].