Progress in building clinically relevant patient-derived tumor xenograft models for cancer research

Abstract

Patient-derived tumor xenograft (PDX) models, a method involving the surgical extraction of tumor tissues from cancer patients and subsequent transplantation into immunodeficient mice, have emerged as a pivotal approach in translational research, particularly in advancing precision medicine. As the first stage of PDX development, the patient-derived orthotopic xenograft (PDOX) models implant tumor tissue in mice in the corresponding anatomical locations of the patient. The PDOX models have several advantages, including high fidelity to the original tumor, heightened drug sensitivity, and an elevated rate of successful transplantation. However, the PDOX models present significant challenges, requiring advanced surgical techniques and resource-intensive imaging technologies, which limit its application. And then, the humanized mouse models, as well as the zebrafish models, were developed. Humanized mouse models contain a human immune environment resembling the tumor and immune system interplay. The humanized mouse models are a hot topic in PDX model research. Regarding zebrafish patient-derived tumor xenografts (zPDX) and patient-derived organoids (PDO) as promising models for studying cancer and drug discovery, zPDX models are used to transplant tumors into zebrafish as novel personalized medical animal models with the advantage of reducing patient waiting time. PDO models provide a cost-effective approach for drug testing that replicates the in vivo environment and preserves important tumor-related information for patients. The present review highlights the functional characteristics of each new phase of PDX and provides insights into the challenges and prospective developments in this rapidly evolving field.

BACKGROUND

Mouse models in preclinical and translational cancer research are important for screening drugs, evaluating the effectiveness of treatments, identifying biomarkers, and performing molecular subtyping.¹ To develop precision medicine, patient-derived tumor xenografts (PDX), involving the surgical extraction of tumor fragments from cancer patients and their subsequent transplantation directly into immunodeficient mice, have proven to be a useful model.² The PDX model is highly effective in simulating tumor progression and evolution in human patients and is considered one of the most promising preclinical models for addressing complex clinical challenges such as identifying biomarkers, exploring intratumor heterogeneity, and evaluating new drugs.³ Toolan et al. transplanted one biopsy specimen of a human epidermoid carcinoma into originally a single cortisone-treated rat in 1953, which was the first study involving the transplantation of human-derived tumor cells in mice and passing them on, opening the door to the world of PDX models.⁴ In 1969, Rygaard and Povlsen initially mentioned the PDX model. They managed to transplant a slice of sigmoid carcinoma from a 71-year-old patient's tumor into nude mice (exposure to X-ray irradiation, cortisone treatment, induction of immunological tolerance, or neonatal thymectomy) for 76 passes.⁵ To circumvent tumor engraftment rejection in mouse models, conventional PDX models are commonly established by utilizing immunocompromised mice. Different animal strains can have an impact on tumor transplantation survival rates, such as nude mice (lack of thymus and T lymphocytes), severe combined immunodeficient (SCID) mice (lack of both functional T and B lymphocytes), SCID/Beige mice (severely reduced natural killer [NK] cell functioning), nonobese diabetic (NOD) mice (lack of complement and impaired NK, macrophage, and dendritic cell [DC] functions), NOD/SCID mice (lack of functional T lymphocytes and residual NK activities), NOD-scid IL2Rγ^null mice (complete loss of NK cells), and Balb/c Rag2−/-Il2rg−/-SirpaNOD (BRGS) mice (lack of mature T- and B cells as well as lack of NK cells).⁶

The utilization of human-derived tumor cells in PDX models ensures remarkable fidelity and preservation of primary tumor cell information. Previous peer-reviewed articles have extensively demonstrated that PDX models effectively retain essential characteristics originating from the patient's primary tumor, encompassing molecular attributes, genomic rearrangements, and gene expression profiles, thus confirming the sustained genetic stability of these models throughout their life span.^7–9 The latest reports also validated that view.^10–14 For instance, Shi et al. discovered the established neuroendocrine prostate cancer PDX retaining the essential histopathological and molecular characteristics of donor tumors.¹³ A comprehensive molecular characterization study of PDX models sourced from 536 patients found that the driver TP53 missense mutation was conserved in both the human primary tumor and the PDX. More than 90% of PDX mutations in the human primary tumor suggest that the PDX models are genuinely derived from human primary tumors.¹⁴ Furthermore, the PDX model could keep the tumor characteristics unchanged during the passages.^13,15,16 The researchers observed that the metabolic characteristics of both tumor cells and stromal cells exhibited stability over a minimum of four passages, with replacement occurring specifically during the second passage.¹⁷ This investigation provided evidence, suggesting that the substitution of human stroma could be considered an acceptable limitation during the initial stages of PDX research.¹⁸ Meanwhile, the PDX model has also shown excellent drug response sensitivity characteristics. Currently, the PDX model stands as the most efficacious preclinical model for faithfully replicating the intratumor heterogeneity of cancer, maintaining the intrinsic architectural characteristics of tumors, and facilitating investigations into drug response and resistance mechanisms.^19–23 Globally, numerous PDX biobanks housing a wide range of cancer samples with diverse genomic and transcriptomic profiles have been established.¹⁸ These biobanks provide invaluable resources for preclinical studies aimed at developing novel therapeutic strategies. By utilizing PDX models, researchers can identify predictive markers and biomarkers that can indicate the efficacy of specific treatment options for individual patients. Therefore, PDX models show tremendous potential for drug research in a clinical setting.

Subcutaneous or renal capsule implantation models have long been the most popular ones. They also provide advantages such as high fidelity, tumor characteristic retention during the transmission process, and high drug sensitivity. Still, their engraftment rate varies tremendously, so it does not highlight the benefits of the classic PDX model. For example, non-small cell lung cancer (NSCLC) showed a remarkable 90% engraftment rate in regulating tumor implantation in the renal capsule, compared to 25% after subcutaneous implantation.

To improve the effectiveness of tumor transplantation, patient-derived orthotopic xenograft (PDOX) models with higher engraftment rates would be a preferable choice for studying transplant characteristics. Besides, apart from the intrinsic elements of the tumor microenvironment (TME), the interaction between tumor and immune cells plays an important role in cancer patient survival and tumor characteristics. However, such interactions cannot be fully recapitulated in conventional PDX models. The lack of a human-like immune system in immunodeficient mice is the issue that has to be resolved. Humanized mouse models are being developed to simulate the human immune system against tumor tissue by adding human immune hematopoietic stem cells (HSC) to stimulate mice to produce immune cells such as T cells, B cells, and NK cells. Another considerable obstacle for PDX is time consumption. Patients wait too long, from months to years, and most patients do not have such a time frame. It is a significant, astonishing benefit that zebrafish patient-derived tumor xenograft (zPDX) has been developed to identify an effective drug testing approach in about 5 days for model generation and less than a month for drug efficacy testing. The aim of this comprehensive review is to illuminate the functional characteristics exhibited in each progressive phase of PDX research, explore the challenges encountered throughout, and provide valuable insights into the potential advancements and prospective developments within this dynamically evolving field.

PATIENT-DERIVED ORTHOTOPIC XENOGRAFTS

Similar to the PDX model process of tumor transplantation, PDOX is designed to advance translational research in precision medicine by surgically removing tumor fragments from cancer patients and transplanting them into the corresponding locations in immunodeficient mice, such as breast cancer tissue transplanted into the mammary glands of mice (Figure 1). As early as in 1982, Moore developed the first PDOX model in nude mice and subsequently discovered the aggressiveness of the tumors.²⁴ Compared to the subcutaneously implanted model, orthotopically implanted PDX tumors may have higher retransplantation rates, larger tumor sizes, and more substantial growth.²⁵ Although currently there is limited adoption of the PDOX model among researchers, its distinct advantages are undeniable.³ Its lack of popularity is primarily attributed to the higher demand for surgical expertise and the expensive nature of imaging technologies.²⁶ The functional characteristics of the PDX orthotopic model and its applications are presented in Table 1.

FIGURE 1. Overview of the PDOX (patient-derived orthotopic xenograft) modeling process and its applications in cancer research. Orthotopic transplantation (F1) involves inserting tumors from cancer patients into the corresponding locations, such as the lung, breast, colon, ovary, or pancreas. Once developed, secondary recipients (F2) receive the tumors for growth. Then, the expanded tumors can be cryopreserved or transplanted into F3 mice to serve as avatar mice. The avatar model can evaluate and rank those potential treatments that could be given to the patient as part of individual drug therapy.

TABLE 1 Comparison between PDX heterotopic and orthotopic models.

	Orthotopic models	Heterotopic models
Fidelity	High	Low
Transplant rate	High	Low
Established PDOX models	Small	Large
Effectiveness of drug treatment	High	Low
Metastasis	Yes	No
Difficulty of operation	High	Low
Technology expense	Large	Small

Abbreviations: PDOX, patient-derived orthotopic xenograft; PDX, patient-derived tumor xenograft.

High fidelity

Elaborating on this characteristic is important because of the substantial body of evidence that has consistently demonstrated the exceptional fidelity of PDOX.^9,27,28 Using immunostaining of 49 evaluable tumors, Stewart et al. found that 98% (48/49) were consistent between PDOX and the patient's tumor; 36 PDOX tumors of those observed using transmission electron microscopy showed that PDOX retained the cellular and subcellular characteristics of the original tumor.²⁹ This study also analyzed the characteristics of the genomic landscape, sublines, and regional clonal heterogeneity, yielding the same results. Meanwhile, the high engraftment rate was demonstrated by Wang's study in the early 20th century.³⁰ These outcomes may be related to mice in vivo providing a similar anatomical environment and retaining the ability to transfer spontaneously.³¹ Recent glioma research shows that glioma PDOXs provide for the long-term multiplication of patient tumors and serve as therapeutically pertinent patient avatars that preserve the histological, genetic, epigenetic, and transcriptome characteristics of parental tumors. Even the information remains preserved between passages.²⁷ In a study on sarcoma, Kiyuna et al. demonstrated that between three generations, the average tumor red fluorescent protein was 20 ± 7%, 28 ± 11%, and 27 ± 13%, respectively, when compared to noncolored mice. The consistency of the retained information indicated a high information transfer rate in the orthotopic model.³² These shreds of evidence confirm that PDOX models can increase the incidence of maintaining the in vivo structure, further indicating a possible association with extracellular matrix (ECM) and carcinoma-associated fibroblasts (CAF). Yoshida summarized that CAF increases the expression of α-SMA and RhoA by secreting TGF-β1, which enhances fiber organization and stromal stiffness, ultimately leading to ECM remodeling.²

Effectiveness of drug treatment

Recent medication investigations implementing PDX models revealed that the effects of the drugs are promising. The PDOX model has revealed the efficacy of recombinant methioninase (rMETase) under different modes of drug administration,³³ including intraperitoneal injection,^34,35 oral administration,^36,37 rMETase combined with chemotherapy,^38–42 and integrated treatment with rMETase and bacterial therapy.^40,43 In research on chimeric antigen receptor (CAR) T-cell therapy, which is remarkably effective in treating hematological malignancies, mice with PDOX of mesothelin-positive pancreatic tumors exhibited a substantial suppression of tumor development and more prolonged survival after treatment with IL-7/CCL19-producing CAR-T cells, compared to treatment with standard CAR-T cells.⁴⁴ All evaluated PDOX tumor models responded favorably to treatment with Salmonella typhimurium A1-R against malignancies, demonstrating more tumor-specific efficacies than chemotherapy or molecular-targeting therapy.⁴⁵ It can be noted that the PDOX model plays a critical role in drug sensitivity for efficacy. The exquisite sensitivity to cisplatin and gemcitabine in four BRCA PDX models was skillfully illustrated by Lohse et al., emphasizing the importance of additional chemotherapy drugs like gemcitabine.²² These pieces of evidence suggest that drug treatment in PDOX models is more effective than traditional subcutaneous transplantation methods.

Metastasis

It has previously been shown that orthotopic transplantation can cause metastases and invasive local growth that resemble those seen in patients.^31,46 Studies in the 1990s found that orthotopic transplantation of different types of human tumors leads to metastatic results.^47,48 PDOX models from clinical patients with colon,⁴⁹ lung,³⁰ pancreatic,⁵⁰ stomach,⁵¹ ovarian,⁵² and breast cancer⁵³ and mesothelioma⁵⁴ were developed during that time, producing primary and metastatic tumor development that was strikingly identical to those patients. A recent sarcoma study validated this idea.⁵⁵ Conversely, if human solid tumor cells are implanted subcutaneously, metastasis does not generally occur.³¹ A comparative study between orthotopic and subcutaneous xenografts of pancreatic ductal adenocarcinoma found metabonomic variations in tumor cells, and the difference in ABC-transporters between the two mouse models partly accounts for liver metastases.⁵⁶ These differences can be attributed to variations in the TME by distinct implant locations. Hiroshima et al. demonstrated that primary and metastatic tumor tissues in nude mice have histology similar to patient-derived tumors by establishing a PDOX model with HER-2-positive cervical cancer.⁵⁷ Metastatic lesions in nude mice, including peritoneal dissemination, liver metastases, lung metastases, and lymph node metastases, enabled us to infer the metastatic pattern of the tumor donor.³¹

Established PDOX models

Using the aforementioned advantages, PDOX models are being developed for an increasing number of disease types. Early in the 1990s, PDOX models were developed from individuals with breast,⁵³ lung,³⁰ ovarian,⁵² colon,⁴⁹ pancreatic,⁵⁰ and stomach cancer⁵¹ and mesothelioma⁵⁴ by Kentaro team's laboratory. Recently, they presented their laboratory experience with PDOX models of sarcoma, showing promising results consistent with those in Table 1.⁵⁵ In 2017, Lai et al. summarized the representative potential therapeutic agents and their objectives in various types of cancer PDOX models, such as colorectal cancer, acute myeloid leukemia (AML), prostate cancer, colon cancer, lymphoma, cholangiocarcinoma, gastric cancer, pancreatic adenocarcinoma, chordomas, and melanoma.⁵⁸ This implies that the PDOX model serves as an effective carrier for tumor preclinical drug development and is an important tool for researching molecular interactions in vivo. The PDOX model is now being used in an increasingly wide range of applications. Veeranki et al. successfully developed the gastroesophageal junction (GEJ) PDOX model, which captures the specific tissue microenvironment inside the local GEJ architecture. They demonstrated exceptional tumor fidelity to actual illness, providing a unique approach for technology transfer from research on human GEJ cancer.¹¹ This model may improve the development of metastatic spread and develop brand-new therapeutic strategies for treating GEJ cancer. Furthermore, the successful development of PDOX models using endoscopic biopsy without undergoing surgical resection has been reported.⁵⁹ For malignant peripheral nerve sheath tumors (MPNST), the MPNST-PDOX model has been used for real-time therapy guidance, and its results indicate the feasibility of developing models and genomic characteristics to guide clinical treatment.⁶⁰ One study, using the orthotopic xenograft model to establish a medulloblastoma (MB) mouse model, validated the role of the blood–brain barrier in tumor development. It highlighted the functional variations in the integrity of the blood–brain tumor barrier and tumor vascular phenotype across frequently used preclinical models of MB, with significant ramifications for the preclinical assessment of potential treatments for MB.⁶¹ This review summarizes the current PDOX models applied to different cancer types (Table 2).

TABLE 2 Current application of different cancer types on PDOX models.

Cancer types	Mouse strain	Year	References
Adenoid cystic carcinoma	–	2019	[62]
Myxofibrosarcoma	Nude mice	2018	[63]
Osteosarcoma	Nude mice	2021	[64]
Gastroesophageal junction cancer	SCID or athymic mice	2019	[11]
Pediatric sporadic malignant peripheral nerve sheath tumor	Nude Harlan mice	2020	[60]
Medulloblastoma	NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice	2021	[61]
Meningioma	SCID mice	2020	[65]
Undifferentiated spindle-cell sarcoma	Nude mice	2018	[66]
Pediatric brain tumors	NSG mice for medulloblastoma, atypical teratoid rhabdoid tumor, ependymoma, and embryonal tumor with multilayered rosette. CD1 nude/nude for gliomas	2021	[67]
Malignant glioma	SCID mice	2021	[68]
Endometrial cancer	Athymic nude-Foxn1nu mouse	2022	[69]
Dedifferentiated liposarcoma	Nude mice	2022	[70]

Abbreviations: NOD, nonobese diabetic; NSG, NOD/SCID gamma null; PDOX, patient-derived orthotopic xenograft; SCID, severe combined immunodeficient.

HUMANIZED MOUSE MODELS

Understanding the interactions between immune cells and tumors, as well as the changes in immune cell phenotype and function after anticancer treatment, is a complex task due to the intricate nature of the human immune system homeostasis and the processes within the TME.^71,72 These complexities pose substantial challenges for researchers in this field, raising an increasing focus on humanized mouse research. Humanized mice are traditionally constructed by transplanting CD34+ HSC, peripheral blood lymphocytes (PBL), or bone marrow–liver–thymus (BLT) tissues into highly immunocompromised mice,⁷³ creating an immune microenvironment that will be closer to human-specific tumor research and contribute to precision medicine⁷⁴ (Figure 2).

FIGURE 2. This figure shows three models to develop humanized mice. CD34+ hematopoietic stem cell (HSC) model: Immunodeficient mice are administered intravenous (i.v.) or intraperitoneal (i.p.) injections of HSCs from bone marrow, fetal liver, granulocyte, or umbilical cord blood. Peripheral blood lymphocyte model: peripheral blood mononuclear cells (PBMC) are injected intravenously or intraperitoneally into an adult immunodeficient mouse. Bone marrow–liver–thymus (BLT) model: HSCs taken from the fetal liver are injected intravenously, and fetal liver and thymus pieces are inserted under the renal capsule in nude mice.

CD34+ HSC transplantation

The most widely utilized humanized mice globally are human CD34+ HSC-transferred animals, and HSCs could be acquired from bone marrow (BM), umbilical cord blood, fetal liver, and granulocyte.⁷⁵ However, humanized mice do not precisely replicate the human immune system.⁶ For instance, the histocompatibility system 2–restricted human T cells that grow in humanized mice are trained to recognize mouse major histocompatibility complex (MHC), leading to interactions between human leukocyte antigen (HLA) and MHC that mismatch with human antigen-presenting cells that express HLA.⁷⁶ Cultured CD34+ HSCs differentiate into myeloid, B-lymphatic, and erythroid lineages. However, they do not produce any or few T lymphocytes,⁷⁷ which lack advantages in terms of yield and purity, proliferation potential, transplantation efficiency, T-cell function, and multilineage hematopoietic development.^77–79 The capacity for engraftment of CD34+ HSCs and the establishment of highly human immune cell chimerism are limited in terms of functional development of innate and adaptive immune cells before the introduction of immunodeficient mice harboring mutations in the Il2rg locus.⁸⁰ For example, NOD-SCID mice are readily engrafted with CD34+ HSCs. However, the CD34+ cell levels produced from their BM are significantly low, leading to limited and variable development of the peripheral human immune system.^81–84 With the development of the mice microenvironment, this problem has been given attention, and the progress of functional HLA-restricted T cells has been reported.⁸⁵ Compared to cultured CD34+ HSCs, fresh cord blood CD34+ stem cells (CD34+ HSCs) injected into mice develop into mature human immune cells without the onset of graft-versus-host disease (GVHD) within 4 weeks.⁸⁶ Over time, peripheral blood, spleen, and BM levels of repopulated human B cells, T cells, NK cells, DCs, and myeloid-derived suppressor cells increased,⁸⁶ which is discussed in the following text.

Based on recent advances in humanized mouse models of CD34+ HSCs, a wide variety of studies on the first humanized mouse models of related cancers have emerged. The first humanized mouse model of CD34+ HSCs was established for the study of adrenocortical carcinoma,⁸⁷ nasopharyngeal carcinoma,⁸⁸ myelofibrosis,⁸⁹ estrogen receptor-positive metastatic⁹⁰ and triple-negative breast cancer,⁹¹ and ovarian cancer.⁹² Notably, in the context of the COVID-19 pandemic, Sefik et al. produced a new coronal humanized mouse using adeno-associated virus to transport human ACE2 to the lungs of humanized mice of M-CSF^h/h IL-3/GM-CSF^h/h SIRPA^h/m TPO^h/h Rag2^null/Il2Rγc^null (MISTRG6) strain.⁹³ The results first describe the pathological course of the lung after viral infection, with an increase in macrophages, monocytes, resident and infiltrating Tγδ cells, and bypass-activated memory Tαβ cells, especially CD14+. Early detection of IgM+ B cells is followed by a decline and replacement by IgG+ B cells. These evidences showed that interferon regulates the innate and adaptive immune response to COVID-19. Regarding treatment, TNF-α plays a more prominent role in the acute phase of the disease, and dexamethasone treatment is also effective. Furthermore, regarding the humanized mouse models that have been developed, scientists delve further into the mechanisms of the disease^94,95 as well as the mechanisms of drug action.^96,97

Novel humanized mouse models have emerged one after another. The NK cell-PDOX model, which supports human NK cell growth in the PDOX model, was developed in a neuroblastoma study as a new tool for PDX research by injecting CD34+ hematopoietic progenitor cells into MISTRG mice.⁹⁸ To create an intraepithelial tumor-infiltrating lymphocyte (TIL) microenvironment associated with overall survival in high-grade serous ovarian cancer, Gitto et al. established a TIL-PDOX model using T cells from the same patient's tumor.²⁸

Humanized mice are evolving. Since the 1990s, there has been a significant amount of research into human peripheral blood stem cells in the immunodeficient mouse model. Human peripheral blood leukocytes, BM, or cord blood cells have been proven to result in long-term stable restoration of a functioning immune system for more than 6 months in mice with severe combined immunodeficiency.⁹⁹ However, it has been shown that the engraftment level of human cells in the NOD/SCID model was 5- to 10-fold higher than in the conventional SCID mice after the transplantation of peripheral blood mononuclear cells (PBMC).¹⁰⁰ Compared to similarly treated NOD-scid mice, engraftment of human HSCs with NOD-scid IL2Rγ^null mice results in a sixfold more significant percentage of human CD45+ cells in the host BM involved with B cells, myeloid cells, NK cells, plasmacytoid DCs, and HSCs. De La et al. summarized different strains, resulting in humanized mouse effects, and the proteins involved up to July 2018. Based on this, we have updated parts of the research progress of humanized mouse models for 2017–2022 in Table 3.⁷³ The table summarizes that there is a shift in the humanized mouse models, mainly from NOG and NOD/SCID gamma null (NSG), to newer strains of mice, particularly the MISTRG strain. However, it is important to note that engraftment of CD34+ HSPCs in standard NSG, NRG (Rag1 and Il2rg knockout), NOG, and BRG models is characterized by certain limitations. These include the incomplete development of monocytes, macrophages, DCs, and NK cells, along with the absence of HLA expression.¹⁰¹ The change in strain application may be due to the fact that the internal environment of immunodeficient mice is more diverse and closer to the human immune environment, thus increasing the demands on the experimental animals. The new advances in experimental mice are a major accelerator for humanized-PDX. Previous works have concluded that the main types of mice widely used are Il2rg^null mice, and their accessibility is thoroughly examined elsewhere.^75,102 Nonetheless, mice carrying the Il2rg mutation exhibit impaired development of lymphoid tissue–inducer cells, resulting in restricted lymph node structure formation.^103,104 This characteristic is considered as an important factor in the inability of Il2rgnull mice to generate potent antibody responses upon CD34+ HSPC engraftment. In contrast, BRGST mice, when compared to BRGS mice, demonstrate improved development of lymph node and thymus structures after CD34+ HSPC engraftment.¹⁰⁵ This enhancement subsequently leads to an increased proportion of follicular helper T cells and facilitates the production of antigen-specific antibodies and T-cell responses. As mentioned earlier, the mouse model is constantly improved. For the use of humanized-PDX in immunodeficient mice, De La et al. presented a detailed table summarizing the various sources of humanized mice and their transplantation results and preclinical research using HIS models for immuno-oncology studies.⁷³ Moreover, a related study has standardized the modeling technique methods for humanized mice.¹⁰⁶

TABLE 3 Genetic modification of humanized mice to increase HSC and PBMC engraftment and to decrease Xeno-GVHD.

Mouse strain	Proteins involved	Effects	Year	References
NOG-dKO mice	IL-2, IL-4, IL-10, TNF-α, IFN-γ, TGF-β1, IL-17	Upregulation of NK cell	2017	[107]
NOG	GM-CSF, IL-3	Increased differentiation and maturation of mast cells, basophils, and eosinophils	2018	[108]
TKO-BLT	–	Increased T cells (CD3+) and loss of B cells	2018	[109]
BRGS	IL-21	TSLP increases NK cells Development of mature B cells	2018	[105]
NSG	IL2	Increased differentiation of T cells, B cells, NK cells, DCs, MDSCs, and macrophages	2019	[86]
NSG	IL-7 and IL-15	Increased differentiation of TILs	2019	[28]
NCG	IFN-γ and IL-2	Increased differentiation of macrophages; DCs decrease with time	2020	[110]
MISTRG	IL-2 and IL-15	Development of NK cell development; ADCC(Antibody-dependent Cellular cytotoxicity) evaluation	2021	[98]
NSGS	SCF, GM-CSF, and IL3	Increased myeloid population, T cells, and HSCs (CD34 + KIT+CD38-)	2021	[89]
Hu-BLT-hIL34	IL34	Myeloid cell reconstitutions in the brain tissues	2021	[111]
Rag2−/-IL2rg−/-129xBalb/c (MISTRG6)	M-CSF, IL-3, SIRPα, thrombopoietin, GM-CSF and IL6	Increased differentiation of macrophages, T cells, and B cells	2022	[93]
NBSGW	M-CSF	Development of tumor-associated macrophages and tumor-infiltrating lymphocytes	2022	[112]
MISTRG6	IL6	Increased NK cells, CD33+ myeloid cells, hCD33+ myeloid cells and hCD34 + CD38- and fewer mouse B cells, lin(−)cKit+ (LK) and lin(−)Sca1+cKit+ (LSK) cells	2022	[113]
NOG	–	Increased functional T-cells	2022	[114]

Abbreviations: BLT, bone marrow–liver–thymus; BM, bone marrow; BRGS, Balb/c Rag2−/-Il2rg−/-SirpaNOD; DC, dendritic cell; HSC, hematopoietic stem cell; MDSC, myeloid-derived suppressor cell; MISTRG, M-CSF^h/h IL-3/GM-CSF^h/h SIRPA^h/m TPO^h/h Rag2^null/Il2Rγc^null; NK, natural killer; NOD, nonobese diabetic; NSG, NOD/SCID gamma null; PBMC, peripheral blood mononuclear cell; TIL, tumor-infiltrating lymphocyte; TSLP, thymic-stromal-cell-derived lymphopoietin.

Peripheral blood lymphocyte transplantation

PBMCs are delivered intravenously or intraperitoneally into an adult immunodeficient animal. Human CD3+ T cells, which consist of CD4+ and CD8+ subsets, are the most abundant population of cells that survive during the engraftment process. On the contrary, the human innate cell population, including myeloid cells and NK cells, exhibits transient survival in the murine host, lasting only for a few days before becoming undetectable in both blood and tissues.¹¹⁵ The Hu-PBL (human peripheral blood lymphocyte) model serves as a valuable testing platform for drugs targeting the suppression of human T-cell responses. These drugs encompass antibody-based therapies,^116–118 Treg cell-based therapies,^119–121 and cytokine-based therapies.^117,122 The Hu-PBL model provides an effective means to evaluate the efficacy of these therapeutic approaches. However, the engraftment of human T-cell populations into SCID mice in the PBL model can lead to the development of acute xenogeneic GVHD mediated by the recognition of mouse MHC molecules.¹¹⁵ T-cell-mediated GVHD is often fatal and significantly restricts the experimental time window. In NSG and NOG mice, the targeted disruption of genes encoding mouse MHC class I and class II molecules has been demonstrated to effectively limit the development of GVHD.^{107,123–125} This genetic modification gives rise to NSG-MHC-dko and NOG-dko models, which provide an extended experimental time window by mitigating T-cell-mediated GVHD. In numerous research studies, the transgenic expression of HLA class I and/or class II molecules in humanized mouse models has been demonstrated to significantly enhance the development and viability of human CD8+ and CD4+ T cells.^126–130 Consequently, this augmentation effectively strengthens the antigen-specific immune responses. When HLA-A2+ and HLA-DR1+ PBMCs are engrafted in humanized mice, a stable antigen-specific antibody response is elicited from human CD4+ and CD8+ T cells, as well as CD19+ B cells, without the occurrence of GVHD.⁸⁵ Humanized mice derived from PBMCs have proven to be valuable models for evaluating the effectiveness of drug treatments, consistently yielding desired outcomes in various studies. For instance, in a mouse model of ulcerative colitis established in NOD/SCID IL2R-null mice, treatment with oxelumab or adalimumab led to considerably lower clinical, colon, and histological scores; lower blood concentrations of IL-6; and lower frequency of the signal of splenic human effector memory T cells and switched B cells.¹³¹ The efficacy of these two medications was compared using orthogonal partial least squares discrimination analysis, revealing that patients with elevated levels of OX40L (CD134, also known as TNFRSF4) might benefit from therapy with oxelumab. The findings have implications for the usage of clinical oxelumab. Another factor that proves the point is that the effectiveness of PD-L1/PD-1-targeted immunotherapies can be accurately predicted regardless based on the injection of PBMCs¹³² or the immune microenvironment of IFN-γ.¹¹⁰ Comparing two PBMC humanized models, cell-line-derived xenograft (CDX) and PDX in immune reconstructed lung cancer showed that the PBMC-PDX model was superior to the PBMC-CDX model in preserving tumor heterogeneity and saving construction time.¹³³ PBL models have emerged as indispensable tools in unraveling the mechanisms of various diseases. These models have facilitated significant advancements in our understanding of disease pathogenesis, providing valuable insights into the underlying mechanisms involved. Kebir et al. established a valuable humanized mouse model of Rasmussen's encephalitis (RE) using PBMC from patients. This model, monitored by video electroencephalography, developed severe cortical seizures and showed intense astrocytic lesions and accumulation of human IFN-γ- and granzyme B-expressing T lymphocytes in the brain, providing an essential vehicle for the development of RE mechanisms.¹³⁴ And the CD200 mechanism of overexpression in PBMC-humanized mouse models was studied in AML stem cells.¹³⁵ Another instance is that the PBMC-PDX model was used to investigate the role of IL-34 in immunotherapy resistance and to provide an optimized solution from a single unit.¹³⁶ These recent studies point to the active role played by the PBMC-CDX model in deeper mechanisms.

BLT transplantation

An inherent limitation of the HSC model is the absence of human thymic epithelium, which plays a critical role in the HLA-based education and selection of human T-cell populations. This deficiency impedes the full recapitulation of the complex processes involved in human T-cell development and maturation in this mouse model.¹³⁷ Indeed, this approach potentially represents the most comprehensive yet simultaneously complex engraftment method available. BLT Hu-mice are produced via autologous transplantation of fetal liver and fetal thymic explants into immunodeficient mice and peripheral injection of CD34+ HSCs. This method facilitates the establishment of a human thymus-like structure capable of supporting the development of HLA-restricted T cells. BLT models are used to research human immunodeficiency virus (HIV) in human lymphoid tissues and blood.^138–140 The benefit of the BLT model is that it promotes the maturation of MHC-restricted T lymphocytes by providing an autologous human thymic milieu¹⁴¹ used for research on tumor growth and development,¹⁴² drug treatment,^143,144 and prediction of disease development.¹⁴⁵

The current focus of widespread public attention has remained on HIV research. Infected with HIV-1 via the vaginal, rectal, and oral routes,^146–149 the BLT model was able to show a more substantial in vivo immune response and repopulate multiple lines of immune cells (T cells, B cells, NK cells, DCs, neutrophils, and monocytes). They are distributed across various organs, including the BM, lymph nodes, spleen, thymus, liver, lungs, gastrointestinal tract, and reproductive tract.¹⁵⁰ Using the model's excellent immune response, current research is focused on the vaccine and immune development for HIV. The ability of the BLT model to recapitulate human pathogen–specific T-cell responses and immune responses was confirmed in a study by Claiborne's team. The data on increased plasma concentrations of immunomodulatory cytokines and chemokines further confirm the potential utility of BLT humanized mice for HIV-1 vaccine development but suggest that further refinements to the model are needed.¹⁵¹ The Hu-BLT-hIL34 NOG mouse¹¹¹ targeting human microglia in the central nervous system for infection and the TKO-BLT humanized mouse model,¹⁵² which is resistant to wasting syndrome and GVHD, have been established. Related studies have shown that antiretroviral treatment of HIV using the TKO-BLT mouse model prolongs its latency and results in a rapid rebound of the virus after interruption of treatment.¹⁰⁹

Recently, immune microenvironment modulating drugs were applied for the study of against HIV infection.   On the one hand, HIV infection results in characteristic changes in the immune environment in vivo and that may become the future drug targets. For example, the γ-δ (γδ) T cells with independent antigen recognition and cytotoxic capacity had a reduced proportion of subpopulation Vδ2 T cells, but an increased proportion of Vδ1 T cells in HIV-infected BLT humanized mice, which suggests the Vδ2 T cells may serve as an early target for HIV replication.¹³⁸ On the other hand, attention was also attracted to drug research to improve immune cell performance. The TANK-binding kinase 1 (TBK1) could be used to modulate the intrinsic immune activation of DCs against HIV-1. TBK1-primed DC-vaccinated mice had a higher proportion of less depleted CD4+ T cells after HIV-1 infection, which may be beneficial  for subsequent vaccine studies.¹⁵³ In another study, the PolyI: C and STING were used as agonists on DC induction, and the addition of the adjuvant combination improved the expression of IL-12, IFN-β, and especially TBK-1 phosphorylation on DCs. This characteristic increased cytokine capacity to activate HIV-1-specific CD8+ T cells, contributing to less CD4+ T-cell depletion.¹⁵⁴ The macrophage cells are also inseparable deeply involved in the process of HIV infection. When compare to using antiretroviral therapy only, the apolipoprotein A-I mimetic 6F treatment of HIV-infected mice exhibited reduced macrophage activation (h-sCD14, h-sCD163) and intestinal barrier dysfunction (m-IFABP, LPS, LBP, and m-sCD14). In turn, the risk of comorbidities was decreased.¹⁵⁵

Overall, the BLT model is advantageous for HIV-related studies. However, the emergence of GVHD-like wasting syndrome, which can markedly abbreviate the life span of BLT mice, is an important concern associated with this model.^156–158 In addition to C57BL/6 Rag2null IL2rgnullCD47null mice, the presence of the human thymic organ environment seems to contribute to the progression of chronic GVHD in the majority of BLT models, leading to tissue inflammation and fibrosis observed 15 weeks after transplantation.^159,160 Besides, with the exception of the United States and China, many nations have banned the breeding of BLT mice because of the moral and legal concerns surrounding the use of aborted fetal material. The utilization of fetal sources is becoming increasingly limited, even in the United States.¹⁶¹

zPDX MODELS

As a novel model, the first drug testing experiment using zPDX was reported 18 years ago.¹⁶² The findings of the zPDX avatars (cell death, angiogenesis, and micrometastasis detection) can be obtained in just 4 days, which is compatible with the time frame required for clinical decisions.¹⁶³ The time from obtaining the tumor tissue from the patient through biopsy or surgery to the final result takes only 12 days.¹⁶³ The larval xenograft test might be the only functional assay for highly aggressive cancers where the patient survival rate may be only a few months, surpassing other patient-derived xenograft (PDX) models in this regard. We have summarized the comparative characteristics of traditional murine PDX models and zPDX models in Table 4.

TABLE 4 Comparison between murine-PDX and zPDX models.

	Murine-PDX models	zPDX models
Modeling time	Months, even years	Days
Number of transplanted cells	Large	Small
Graft-versus-host disease	Yes	No
Effectiveness of drug treatment	Yes	Yes
Species differences	No	Yes
Precision medicine	Yes	Yes
Body temperature	37°C	33–35°C
Transplantation difficulty	Low	High
Network databases	Yes	Yes

Abbreviations: GVHD, graft-versus-host disease; PDX, patient-derived tumor xenograft.

Advantages of zebrafish

Recent studies have shown that zPDX has advantages such as a simple morphological structure, short breeding time and high reproduction numbers, short modeling time, low need for patient-derived tissue, and high drug sensitivity, which are discussed here (Figure 3). Morphologically, zebrafish are tiny and vigorous, usually capable of producing hundreds of eggs with a transparent appearance.¹⁶⁴ This self-contained optically transparent morphological advantage simplifies the transplantation process.¹⁶⁵ Due to the small, translucent fry's ideal form for imaging, it is possible to see an animal entirely and determine the extent of metastatic spread at the single-cell level.^166–168 Additionally, new cutting-edge imaging techniques reveal cancer cellular activities at unprecedented resolution. For example, Fior et al. integrated high-resolution microscopy methods and a powerful battery of complementing analytic tools.^166,169 With the advancements in microscopy speed, spatial resolution, and automated image analysis, simple CRISPR-mediated knockout techniques in primary cells, cell lines, or living model organisms are now possible. These new live imaging methods and biophysical strategies will be capable of capturing details of the dynamic cancer evolution at the single-cell or molecular level, even including the development of clonal genomes and epigenetics. With unique characteristics, the number of patient tumor cells required for successful transplantation is exceptionally minimal,¹⁷⁰ with fewer than 500 cells,^166,171,172 or even as few as 100–200 cells¹⁶⁴ required for patient-specific chemosensitivity analysis. Most impressive is the speed with which zebrafish get modeling and treatment results, as a recent zPDX study on NSCLC yielded treatment results in just 3 days.¹⁷³ In contrast to mouse xenografts, which take several months to produce sufficient numbers of animals for study, the experimental time advantage of zebrafish is significant.¹⁷⁴ It took such a short time because of the benefit of the specific environment and rapidly developing growth in zebrafish larvae. Their immune system is immature in the first weeks, which makes them ideal for cell transplantation and live imaging, and the transplantation process circumvents GVHD and does not require immunosuppression.¹⁷⁵ It is also possible to efficiently transplant human patient-derived tumor cells into adult immunodeficient zebrafish.¹⁷⁶

FIGURE 3. Setting up an experiment to produce zebrafish xenografts. After adult zebrafish spawn embryos at a rate of 300 eggs/week, they grow into larvae on the fifth day. The patient's tumor cells (~500 cells) are subcutaneously injected into the perivitelline space or the yolk sac of zebrafish larvae. Zebrafish can be used for live imaging in ~5 days for research on cancer metastasis, molecular mechanisms, and preclinical drug testing.

The results of most drug trial studies are encouraging, with a large amount of data showing positive results from the zPDX oncology study.¹⁷⁷ The supporting evidences make it suitable for the elucidation of tumor pathogenesis and preclinical testing of novel antitumor compounds¹⁷⁸ and lead to a deeper comprehension of the initiation of cancer, its relationship to the microenvironment, its spread, and resistance to treatment.^179,180 Explant cultures of head-and-neck and colon squamous cell carcinomas were used in the study, and they exhibited a strong association with the patient's response when cultured on a scaffold tailored to each tumor's matrix. Additionally, it has been shown that 82% of human proteins that cause illness have a zebrafish ortholog.^181,182 The short duration and high drug sensitivity make zPDX a better choice for the preclinical testing of personalized drugs as a predictive drug testing tool.^{164,183–185} In vivo, zPDX drug screening is a potential addition to functional pharmaceutical testing in personalized medicine platforms. Gatzweiler et al. first paired the drug sensitivity profile with 76 specific clinical anticancer drugs of three culture models aimed at standard-of-care therapy.¹⁸⁵ Although mouse PDX usually successfully establishes a long-term organoid culture in this study, mouse PDX models are severely limited by the time frame to grow over weeks or months, which may need to be faster in a clinical environment. In another comparison among patient-derived avatar models, mouse patient-derived xenografts, organoids, zebrafish xenografts, and Drosophila avatars, the results of zPDX are superior in terms of time, minimum cell number used, and cost.

Established zPDX models

The development of zPDX models for various diseases is ongoing. For gastric cancer, AGS and SGC-7901, two human gastric cancer cell lines successfully xenografted into zebrafish embryos by Wu et al., were examined for 5-FU sensitivity both in vitro and in vivo. SGC-7901 demonstrated greater 5-FU sensitivity than AGS in vitro and in vivo (6.3 ± 0.9 vs. 10.5 ± 1.8 μM).¹⁸⁶ All showed potential for metastatic, angiogenic, and proliferative growth in the live embryos, and drug sensitivity was reflected in the chosen chemotherapeutic medicines. For glioblastoma (GBM), Ai et al. found invading GBM microtumors in model animals' and people's proportionately bigger brains¹⁸⁷ after creating zebrafish GBM xenografts that can sustain cerebral engraftment while maintaining their distinctive histological characteristics. Subsequent single-cell RNA-sequencing research demonstrated considerable transcriptional similarity to those of invading GBM microtumors. This model predicts a 20-day long-term prognosis for GBM patients. In another study, the 11 GFP-tagged patient-derived GBM IDH wild-type cell cultures (PDCs) exhibited proliferation, invasion, and survival heterogeneity, and tumor initiative related highly with matched mouse PDX counterparts (Spearman's R = 0.89, p < 0.001) after Almstedt implanted 11 PDCs in 1-day-old zebrafish embryos.¹⁷⁰ The findings demonstrated a strong correlation between transplanted tumor cells and donor blood vessels in three PDCs. In vivo testing of the medication marizomib, which is now undergoing clinical trials for the treatment of GBM, revealed an effect on fish survival that was consistent with PDC in vitro and in vivo marizomib sensitivity. To create the zPDX model for pancreatic cancer, Di Franco skillfully used tiny pieces of CM-Dil-stained tissue and implanted these pieces into the yolk of zebrafish embryos. This improved established model may one day predict how each patient will respond to chemotherapy.¹⁸⁸ Based on research conducted by Alors-Perez on the effects of pladienolide-B treatment on zebrafish and mice, pladienolide-B equally compromised cancer stem cells by decreasing their stemness capability and increasing their susceptibility to chemotherapy. Pladienolide-B also inhibited pancreatic xenograft tumor development in mice and zebrafish in vivo.¹⁸⁹ A recent in vitro study in leukemia xenograft drug testing demonstrated the enhanced effects of MEK- and flt3 inhibitors in combination in vitro.¹⁶⁵ It is worth highlighting that Costa's team summarized the various types of disease zPDX models until 2020, with tables including details of cell lines, readouts, zebrafish stages, and experimental temperatures.¹⁶³

Overall, zPDX has a promising future as a new therapy for personalized clinical medicine and remarkably reduces the ethical issues that come with mice. However, not all patients have tumor changes and tumor microenvironmental responses, and many patients are not qualified to receive such “new therapies”.¹⁶³ Therefore, most patients are still treated with conventional chemotherapy/radiotherapy and surgery based on clinical guidelines.

Metabolomics, pharmacogenomics, and proteomics techniques provide the still state of the cell. However, these approaches do not consider all possible genetic interactions that may occur between different subclones or with the TME and lack quantification of their function and response to direct interference.¹⁶³

Limitations

Although the direct detection of drug sensitivity in patient biopsies using zebrafish holds significant promise as a powerful application in clinical cancer therapy, these techniques present challenges, and the number of patients tested is relatively low. Thus, conducting larger-scale clinical studies is necessary to thoroughly validate the zPDX model. Combined with zebrafish's extremely short incubation time, this may be considered too short for detecting drugs that act through cell-cycle interference or epigenetic mechanisms, which typically take much longer to take effect. The temperature for the growth of zebrafish larvae has also been one of the concerning limits.¹⁸⁵ Zebrafish larvae do not grow well at 37°C, which is the optimal temperature for human tumor cell growth, leading to cardiac issues and edema that threaten zPDX survival.¹⁹⁰ Zebrafish xenografts are commonly raised at 33–35°C, which may have an effect on the physiology of the fish and the biology of the tumor cells because human cells thrive at 37°C, whereas zebrafish are typically raised at 28–29°C.¹⁸⁰ Human cancer cells continue to multiply at 34°C but with a somewhat longer population doubling time. This might make it more challenging to identify medications that interfere with the cell cycle, underestimating the impact of those treatments.¹⁹¹

Websites

In the Internet age, an overview of zebrafish used in biomedical research is provided in an award-winning short video for a knowledgeable lay audience in the Internet era (https://www.zebrafishfilm.org). The zebrafish genome has been fully sequenced, and gene analysis tools are well developed. Resources are accessible through the Welcome Trust Sanger Institute (https://www.sanger.ac.uk/science/data/zebrafsh-genome-project) and the Zebrafish Model Organism Database (ZFIN; https://zfin.org/).

CHALLENGES AND DEVELOPMENTS

The PDX model can largely retain the original tumor's genetic landscape and clonal heterogeneity. Therefore, conducting therapeutic efficacy tests and investigating drug resistance in PDX can more accurately reflect the actual scenario of cancer. With the great benefit of preserving the original tumor characteristics, PDX models are treated with the same treatment regimen as the primary tumor (both clinical and preclinical), namely co-clinical trials, also called avatar models or “mirror models,” as preclinical personalized medicine, which is the direction researchers are working on.^14,192,193 The method of the co-clinical trial utilizing the PDX model makes it easier to prioritize the best treatments, quickly categorize responders, find biomarkers, and identify resistance mechanisms.¹⁹⁴ In particular, the development of humanized PDX models has enabled the in vivo simulation of the intricate interplay between tumors and the immune system. This bidirectional influence between the tumor and the immune system results in mutual adaptation and changes over the course of their interaction, illustrating the complex dynamics inherent in the cancer-immune microenvironment. Humanized mice may serve as future personalized therapeutic vehicles.¹⁹⁵

However, a comparison among several models reveals that creating a nude mouse model is quite expensive, that setting up PDX takes time (between 6 months and 2 years), that success rates vary (10%–90%), and that it is challenging to recover all of the patient data.⁶ Even if patient-derived tumor tissue is successfully transplanted into mice, there may still be a deficiency of human-derived tumor-associated cells and a constant replacement of human stroma with mouse stroma.^196,197

Sometimes, there are cases that make no sense. For example, the time taken to generate the number of PDOXs necessary for preclinical drug testing (generations 2–3) is typically longer than the survival time of the majority of high-grade glioma patients, making the co-clinical use of glioma PDOXs as patient avatars for therapy prediction challenging.²⁷ Overall, the development of the mouse PDX model over more than 50 years has been limited by three significant constraints: time, successful transplant rates, and the replacement of mouse stroma. With an awareness of data sharing, there is a consensus to store the data obtained from the PDX mouse model on a dedicated website and to build a PDX-specific bioinformatics bank. The various databases are presented in Table 5.

TABLE 5 PDX database website.

Website	Website content	Year	References
https://pdx.wustl.edu/pdx	“PDX Variant Viewer” provides a web portal with a comprehensive pan-cancer PDX genome landscape and source code, including cancer type, derived PDX samples, corresponding clinical information (gender, patient age range, self-reported race, and treatment status of PDX specimens), models, and the type of variant calling pipeline used (tumor-only or tumor-normal).	2021	[14]
http://www.pdxfinder.org	The authors' effort to address the substantial difficulty in finding PDX models pertinent to particular cancer research questions used PDXfinder to federate dispersed data and combine it.	2019	[203]
https://www.europdx.eu/	EurOPDX, consisting of more than 1500 subcutaneous and orthotopic models from more than 30 different pathologies, provides a platform for population-scale studies to discover low-prevalence genetic alterations and standardize biobanking procedures.	2013	[204]
https://pdmr.cancer.gov/	These PDMs are early passage and therapeutically labeled with molecular data accessible to the extramural community on a public website.	2012	–
https://preclinicalpivot.org/	The website improves information on in vivo testing of pediatric anticancer treatment candidates.	2017	–
http://www.cogcell.org/xenografts.php	A database run by the Children's Oncology Group that provides verified cell lines and PDXs from pediatric cancers.	–	–
https://www.cbioportal.org/	An open-source collection for PDXs165 from leukemia and lymphoma.	2016	–
www.stjude.org/CSTN/	Stewart et al. placed the data results on the Childhood Solid Tumor Network; 168 individuals provided tumor samples, and 67 PDOX models for 12 different kinds of cancers were developed, including gene-expression patterns and epigenomes of the tumors.	2017	[29]
https://www.jax.org/jax-mice-and-services/in-vivo-pharmacology/oncology-services/onco-hu	The Onco-Hu mouse strain was developed by Jackson Laboratory and is dual engrafted with human CD34+ hematopoietic stem cells and the therapeutically useful PDX.	2018	[205]

Abbreviations: PDOX, patient-derived orthotopic xenograft; PDX, patient-derived tumor xenograft.

It is worth noting that patient-derived organoids (PDO) exhibit remarkable potential in the realm of biomedical applications.^198–200 They emulate the intricate pathophysiological environment of the human body in vitro, using supporting techniques such as three-dimensional (3D) co-culture assays to construct 3D tumor tissue structures. Similar to humanized mice, PDOs provide a complex cellular environment for genetic and drug research. Researchers endeavor to replicate cancer cell–stromal interactions, thereby creating a “representation” of specific diseases. Moreover, PDO models have advantages such as simple operation, cost-effectiveness, and preservation of essential cancer-related information.²⁰¹ This presents a robust and promising approach for drug development and precision therapy in the field.

Nonetheless, in vitro models are not suitable for novel therapy testing in preclinical drug evaluation due to their inability to accurately replicate various in vivo properties, such as pharmacokinetic performance. PDOs may not accurately represent the diverse array of cell types found in native tissues, including nonparenchymal cells such as immune cells and stromal cells.²⁰² Furthermore, they are limited in their ability to capture the impact of environmental exposures and organismal aging on human organs. To ensure the reliability of the model, it would be necessary to employ high-throughput analytical assays that provide information and define quality standards regarding cellular composition, differentiation, cellular states, and responsiveness to stimuli. In the future, a comprehensive catalog of human-like organs with clearly defined characteristics should encompass a wide range of replication techniques and biological sources to accurately quantify variability. To overcome the limitations of PDOs, scientists have embarked on the Organoid Cell Atlas pilot project.²⁰² By utilizing cutting-edge techniques such as single-cell sequencing and spatial profiling, they aim to construct a comprehensive reference map of all human cells in the future. This ambitious endeavor serves as a foundational resource for understanding human health and forms the basis for diagnosing, monitoring, and treating diseases.

AUTHOR CONTRIBUTIONS

Weijing Wang and Yongshu Li wrote the review. Xiaokang Wang, Kaida Lin and Yanyang Tu contributed to literature research and graphics preparation. Zhenjian Zhuo and Yongshu Li orientated and proofread the review.

ACKNOWLEDGMENTS

Thanks to the animals that were killed in science discovery.

FUNDING INFORMATION

The Science, Technology and Innovation Commission of Shenzhen (JCYJ20220531093213030) .

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest except Zhenjian zhuo. Zhenjian Zhuo is an editorial board member of AMEM and a coauthor of this article. To minimize bias, he was excluded from all editorial decision making related to the acceptance of this article for publication.

ETHICS STATEMENT

Ethics statement should be not applicable.