Patient-derived organoids (PDOs) and PDO-derived xenografts (PDOXs): New opportunities in establishing faithful pre-clinical cancer models

One of the major bottlenecks in advancing basic cancer research and developing novel cancer therapies is the lack of in vitro pre-clinical models that faithfully recapitulate tumor properties in the patients. Monolayer cultures of cancer cell lines usually lose the heterogeneity of the parental tumors, while patient-derived xenograft (PDX) suffers from its time- and resource-intensive nature. The emergence of organoid culture system and its application in cancer research provides a unique opportunity to develop novel in vitro cancer pre-clinical models. Here we review the recent advances in utilizing organoids culture system and other related three-dimensional culture systems in studying cancer biology, performing drug screening, and developing cancer therapies. In particular, we discuss the advantages of applying xenograft initiated from patient-derived organoids (PDOs) as a faithful cancer pre-clinical model in basic cancer research and precision medicine.


Introduction
Cancer remains one of the major health problems worldwide despite the ever-increasing knowledge and investments in cancer related research and therapeutic development.One of the major bottlenecks is the lack of 3D pre-clinical models of cancer that can faithfully recapitulate tumor properties to study the disease progression and develop novel treatments.For decades, the most widely applied cancer models are transgenic mouse models and monolayer culture of cancer cell lines.Although important insights of cancer biology have been gained through these models, both experimental systems have obvious drawbacks.For example, genetically modified mouse models of cancer are in general time-consuming and may not fully recapitulate the disease progression of cancer patients, particularly for certain features such as genetic heterogeneity and histological complexity. 1 On the other hand, the in vitro 2D culture models exert selective pressure on cancer cells, and thus only a handful of rapidly proliferating cancer cell clones may expand from tumor samples derived from human patients.Furthermore, the 2D cultured cancer cell lines gradually lose the heterogeneity of the parental tumors as they are passaged.Due to all these limitations of conventional cancer models, therapeutic treatments developed from these models often fail to be translated into clinical settings, suggesting that these are not reliable pre-clinical models.
Patient-derived models of cancer (PDMC), such as patient-derived xenografts (PDXs) and patient-derived organoids (PDOs), have been developed aiming to address the limitations of the conventional models.Both PDO and PDX capture the clinical heterogeneity of cancer in terms of stage, molecular features, and genetic background.4][5][6] Therefore, PDO and PDX are becoming the gold standards in pre-clinical therapeutic development.
PDX is an in vivo model by implanting human patient tumor materials into immunodeficient mice, 7 , 8 where they develop into xenografts in the recipient animals.The patient tumor materials can be either minced tissues or single-cell suspensions.The most common form of implantation is via subcutaneous injections, while other approaches including orthotopic injection are also widely performed. 8The PDX model is not a new concept, with some initial attempts that could be traced back to the 1980s. 9 , 10These early studies demonstrated that PDX models and their corresponding patients showed similar responses to certain therapeutic treatments.The more recent clinical needs for developing new therapeutic options for cancer have brought the PDX model back to the spotlight.PDX can be established from many types of cancer, in-Table 1 Summary of PDX success rate in different cancer types.cer research and clinical trials, substantially increasing the options for modeling different types of cancer.
One of the advantages that PDX holds over conventional PDO is the presence of in vivo stromal component, which allows for studying the tumor-stroma interaction albeit the apparent differences between mouse and human patient stroma.However, the longstanding drawbacks of PDX limit its applications in precision medicine and tumor biobanking.In contrast, since PDO is an in vitro culture system, it can be established from patient tumor samples with a much higher success rate.It is also less expensive, requires less labor and propagates faster than PDX.Therefore, PDO permits high-throughput drug screening for patient tumors with significantly reduced cost and time.Moreover, there are significant limitations in applying PDX models in immuno-oncology studies since the xenografts are commonly inoculated into immunecompromised mice.Although recent progress in generating humanized mice have enabled the reconstitution of human immune cells in mouse models using human CD34 + cells or peripheral blood mononuclear cells (PBMC), certain limitations persist when using PDXs in such humanized mouse models to study immuno-oncology, including the HLA matching issue and the low fraction of CD4/8 + T cell populations. 47 , 48In this regard, while the traditional 3D submerged PDO selectively enriches for tumor cells, numerous newly developed PDO models have started to incorporate TME into the experimental setting, such as putting the immune cells and cancer-associated fibroblasts (CAF) into culture.0][51] Notably, organoids can be used to test unwanted side effects in order to prevent adverse drug reactions and determine the appropriate dosage of chemotherapy.Organoids have been deployed to assess drug toxicity in many organs. 52][55] Furthermore, xenografts derived from PDO (PDOX) have been successfully established.Similar to PDX, PDOX have been demonstrated to retain key pathological features of the parental tumor, such as the mutational profiles and the level of tumor heterogeneity. 56 , 57The establishment of PDOX grants more applicability to PDO and may overcome the limitations of both PDX and PDO ( Fig. 1 ).
In this review, we will summarize recent advancements in PDO and PDO-derived culture systems, including immune cell and CAF coculture systems, ALI systems and microfluidic murine-and patientderived organotypic tumor spheroid (MDOT/PDOT) systems, as well as their applications in cancer research and precision medicine.In particular, this review will discuss the potential of PDOX as a new frontier of establishing faithful pre-clinical cancer models.

Developing PDO for different cancer types
PDO have been established for many different types of cancer ( Table 2 ).One of the most widely established and studied cancer PDOs is the CRC organoid.In 2015, a biobank of 26 (22 + 4) CRC organoids was established from resected colon segments with a success rate of 90%. 34hese organoids could be readily expanded and cryopreserved, with the typical survival rate after thawing to be > 80%, representing a signif- icant advantage compared with PDX models.Shortly afterwards, CRC PDOs were successfully established from 18-gauge biopsies of metastatic lesions with a success rate > 70%.Importantly, these PDOs preserved genetic diversity of their parental metastases. 37In addition, niche factors that were often required for normal human colonic organoid culture, including epithelial growth factor (EGF), Noggin, TGF- inhibitor A83-01, and p38 inhibitor SB202190 (ENAS), as well as Wnt3A/R-spondin1 and hypoxia, were carefully examined in the culture of CRC organoids. 35The niche dependency of CRC organoids indicated that CRC tissues largely retain the requirements of stem cell homeostasis in the normal intestinal epithelium.Decreased niche factor requirements in CRC organoids were observed in the transition from adenoma to carcinoma, reflecting the accumulation of genetic mutations in this process.It is worth noting that by establishing eight combinatorial culture conditions with different niche factors, a 100% success rate could be reached when deriving and propagating CRC organoids from different subtypes of CRC tissues.Other cohorts of cancer organoids established from esophageal cancer, metastatic CRC and gastroesophageal cancer patients also demonstrated a high degree of similarity in phenotypic and genotypic features to the original tumors. 36 , 58These results, together with other studies in which additional CRC PDOs were successfully established, 59 , 60 demonstrate that this pre-clinical model can be widely adapted in the field of CRC research.
PDOs for other cancer types have also been developed.Boj et al. demonstrated that organoids can be derived from surgically resected pancreatic cancer at a success rate of 80%. 61They further demonstrated the feasibility of deriving PDOs from fine needle aspirations, which significantly enhanced the applicability of PDO in modeling PDAC since 85% of the patients are not eligible for surgical removal of their tumors. 62Moreover, PDAC organoids can also be generated from PSCs after experimental transformation via lentivirally transducing the KRAS G12V and TP53 R175H mutation, which serves as an alternative way to develop PDOs for PDAC.In addition, a large PDO biobank of breast cancer was established in 2018. 44This study included 155 primary and metastatic breast cancer tumors and established 95 breast cancer organoids.They demonstrated that the PDOs could recapitulate different subtypes of breast cancer that differ significantly in their histopathology, hormone receptor status and epidermal growth factor receptor 2 (HER2) status.For liver cancer organoids, their development was initially hindered because of the non-tumoral, stromal cell contamination.Certain stromal cell types from liver cancer patient specimens, such as cholangiocytes, tend to overgrow in Matrigel and become the dominant cell type.This issue was later addressed by removing R-spondin-1, Noggin and Wnt3a in the culture medium while optimizing the digestion and culturing protocol. 40Subsequently, organoids from common primary liver cancers (PLC), including hepatocellular carcinoma (HCC), cholangiocarcinoma (CC) and combined HCC/CC could be propagated following the new protocol.In addition, lung cancer organoids can be established from five subtypes, including lung adenocarcinoma, squamous cell carcinoma, adenosquamous carcinoma, small cell carcinoma, and large cell carcinoma. 63Prostate cancer organoids were first established in 2014 from biopsy specimens and circulating tumor cells, with a relatively low success rate at 15% − 20%. 3 In particular, neuroendocrine prostate cancer, a rare and aggressive prostate cancer subtype with few available pre-clinical models, has also been successfully modeled by organoids. 64On the other hand, PDOs can also be derived from endometrial diseases ranging from hyperplasia to low-and high-grade cancer and have been proven to maintain long-term expandability, genomic and transcriptomic stability and patient heterogeneity. 65 , 66On top of that, ovarian cancer (OC) can also be modeled by PDO.Kopper et al. established 56 organoid lines to represent all main subtypes of OC, which faithfully recapitulate the histological and genomic features of the parental tumor. 45Similarly, PDO can also be applied to bladder cancer studies.In 2018, 22 bladder cancer organoids were generated and found to faithfully represent tumor evolution and treatment responses of their parental tumors. 67Kidney cancers, the Wilms tumor, or nephroblastoma, can also be modeled by PDO. 68Head and neck squamous cell carcinoma (HNSCC) is another type of cancer that can be modeled by PDO and be tested with novel treatment options. 55Glioblastoma (GBM) or brain metastases initiated by esophageal adenocarcinoma can also be modeled by organoids. 69Hubert et al. revealed that GBM organoids could be directly derived from surgically removed patient tumors, and organoids derived from distinct tumor regions retained their divergent phenotypes.Importantly, brain metastases from esophageal adenocarcinoma patients, which are known to be notoriously difficult to culture in vitro, can be cultured as organoids. 69Other methods of establishing From pathology amterial Can be used for in vitro photodynamic therapy testing Driehuis et al. 55 Abbreviations: CRC, colorectal cancer; GEP-NENs, Gastroenteropancreatic neuroendocrine neoplasm.GSCs, glioma stem cells; hESC, human embryonic stem cell; KC, LSL-Kras G12D/ + ; Pdx1-Cre; KPC, LSL-Kras G12D/ + ; Trp53 fl/ + ; Pdx1-Cre; PDO, patient-derived organoid.
brain organoids have also been reported.For example, co-culturing the patient-derived glioma stem cells (GSCs) with human embryonic stem cell (hESC)-derived cerebral organoids demonstrated that GSCs could deeply invade the human cerebral organoid, forming tumors that phenotypically recapitulate patient GBMs. 70PDO can be further applied to model rare cancer types such as the gastroenteropancreatic (GEP) neuroendocrine neoplasm (NEN).Independent cohorts of GEP-NEN have been established to lay the foundation for understanding and developing therapeutics for this orphan disease. 71 , 72ost protocols for tumor organoid establishment rely on tumor tissues to be freshly retrieved from patients, but other surrogate or indirect methods are available should the direct method fail.In 2014, a single ALI culture method of organoids was developed to model the oncogenic transformation process of multiple cancer types.Recapitulating the clinical observations, pancreatic or gastric organoids bearing Kras G12D mutant gene and p53 loss were seen to form adenocarcinoma as xenografts, while primary colon organoids need combinatorial Apc, p53, Kras G12D and Smad4 mutations to gain tumorigenicity. 73Later, Hans Clevers 74 and Toshiro Sato 75 groups used CRISPR-Cas9 to introduce oncogenic mutations into human intestinal organoids and intestinal stem cells.All of these engineered organoids showed tumor-like features after inoculation into immunodeficient mice.Brain tumor organoids can also be established by genetically engineering normal cerebral organoids, which were transformed to glioblastoma-like or central nervous system primitive neuroectodermal tumor (CNS-PNET)-like neoplasms. 76Moreover, similar methods could also be applied to iPSC-initiated organoids. 77The iPSC-derived CRC organoid model has been used for testing therapeutic options such as chemotherapy.In addition, it could also be applied to basic cancer research, such as studying the origin of certain mutations. 78owever, it remains unknown whether these organoids recapitulate the original tumor complexity as those PDOs derived from patients via the direct method.
Most organoids are cultured in the synthetic medium with defined molecular and growth factors.Except for the common growth factors such as EGF, which may act to inhibit cellular senescence, 79 , 80 some other factors like Noggin (BMP antagonist), R-Spondin (Wnt inducer), A83-01 (TGF- inhibitor), SB202190 (p38 inhibitor) are also required for most organoid developments.On top of this, additional specific factors may be added to the growth medium based on different cancer types.For example, Neuregulin 1, a ligand of human EGF receptor tyrosine kinases involved in mammary development and tumorigenesis, was shown to allow efficient generation of breast cancer organoids and sustain long-term expansion for > 20 passages. 44Individual fibroblast growth factor (FGF) family ligands contribute differently to lung organoid formation.For instance FGF7 and FGF10 induced more organoid branching as compared to FGF2 and FGF9. 81Moreover, the higher concentration (1 nM) of dihydrotestosterone (DHT) is important in maintaining the viability for specific types of prostate cancer organoids. 82We have summarized the composition of growth media for culturing PDOs of most common tumor types in Table 3 .Nevertheless, further optimization for medium composition based on cancer types, stages, genetic mutations, metastatic loci and other potential factors is required to enhance the success rate of PDO establishment.

Advanced models of PDO
PDOs have been shown to maintain the histopathological, transcriptomic and genetic characteristics of the parental tumors.However, the traditional PDOs tend to contain only cancer cells and lack components of the TME, such as fibroblasts, endothelial cells, and immune cells among others.Therefore, there are ongoing efforts focusing on adding the TME components to the organoids to recapitulate the microenvironment of the parental tumors.
For example, CAF are a critical component of TME, which play a crucial role in tumorigenesis and progression by remodeling the extracellular matrix (ECM) and producing extensive cytokines and chemokines to interact with cancer cells and infiltrating immune cells. 83Moreover, CAFs may induce drug resistance in tumors by either producing certain secretory molecules that protect cancer cells from drug-induced cell death or depositing matrix components that provide survival signals to cancer cells via integrins. 84Meanwhile, the immune system is another crucial component of TME.][89] Nonetheless, poor responses to immunotherapy are not uncommon in the clinic, which can be attributed to a variety of mechanisms.Therefore, several organoid culturing systems have been developed to coculture organoids with CAFs and/or immune cells.These more advanced PDO models could 1) be deployed as a model to study the interactions between cancer cells and key components of the TME and 2) accurately predict a patients's sensitivity to chemotherapy and immunotherapy.

PDO co-culture with CAF
Reconstituting the TME by adding the CAFs to PDO models is one approach to address the need for studying interactions between cancer cells and the TME, 49 , 90 especially for cancers like PDAC, which normally includes up to 90% of stroma component in the tumor mass. 91When studying the inter-tumor heterogeneity of PDAC based on their stem cell niche factor dependencies, Seino et al. revealed a Wnt-non-producing subtype of PDAC that requires Wnt from CAFs, 90 which demonstrated the necessity to introduce the matched CAF into organoids culture for this subtype.Meanwhile, pancreatic stellate cells (PSCs) were shown to differentiate into SMA-expressing CAFs when co-cultured with PDAC-PDOs. 92Interestingly, in addition to the SMA-high and myofibroblastslike CAFs, PSCs were found to differentiate into another CAF subtype named inflammatory CAFs using the co-culture platform of PSCs and PDAC-PDOs.The inflammatory CAFs in turn support tumor cell survival and growth through paracrine signaling via IL-6, IL-11 and LIF. 92oreover, CAFs could also contribute to the malignant progression of CRC. 93Luo et al. successfully established a CRC PDO -CAFs co-culture system by embedding both cell types in 3D hyaluronan (HA)-gelatin hydrogel, which is believed to recapitulate the in vivo CRC ECM. 94Further characterization of this system demonstrated that the CRC PDO -CAFs co-culture promoted PDO growth, recovered biological pathways that are absent in the conventional PDO culture but present in patient tissues, and could be used as a platform for predicting drug responses.These studies demonstrate the great potential of PDO-based models in studying the symbiotic interactions between cancer cells and the TME components.

PDO co-culture with PBMC derived T cells
Another important feature that is missing in the conventional PDO model is the immune components.A recently established PDO co-culture system allows the expansion and selection of tumor-reactive T cells from peripheral blood for predicting their anti-cancer properties. 51In this study, mismatch repair-deficient (dMMR)/MSI CRC organoids that are major histocompatibility complex (MHC) class I proficient were first developed.Then autologous PBMCs were isolated from the matched patients and cultured together with the corresponding tumor organoids.The co-culture involves pre-treatment of PDOs with Interferon (IFN)  to enhance antigen presentation, and plate-bound anti-CD28 and IL-2 to provide co-stimulation and to support T cell proliferation.Increased tumor-reactive CD8 + T cell populations were detected in 50% (4/8) of the co-culture models and notably, T cell-organoid co-culture could expand previously undetectable tumor-reactive T cell populations.Coculture of PBMCs and autologous tumor organoids could also be used in NSCLC to expand tumor-reactive CD8 + T cells. 51These autologous tumor-reactive CD8 + T cells were seen to substantially reduce the survival of the matched PDOs, indicating the potential of this PDO coculture model in evaluating the responses of immuno-oncology therapies in a personalized manner and in generating patient-specific tumorkilling T cells.It should be noted that although the enriched CD8 + T cells were largely tumor specific, a group of expanded CD4 + T cells was found to recognize the murine original culture matrix (Geltrex), which was used for organoid culture in this model.To address this issue, synthetic matrices might be more appropriate in future T cell-organoid co-culture systems to avoid the possible T cell cross-reactivity to nontumor antigens. 95Admittedly, deriving tumor-reactive T cells directly from tumor-infiltrating lymphocyte (TIL) can be more straightforward and less time-consuming, 96 but the absolute number of TILs may not be sufficient in sparsely infiltrated "cold-tumors ".Therefore, deriving tumor-reactive T cells from peripheral blood permits broader applications, especially when PDOs could be successfully established from FNB.

a) ALI as a holistic model for studying immune component in organoids
Besides the submerged Matrigel embedded organoids, another major form of organoid culture adopts the ALI, which was first deployed in culturing murine organoids. 73In this approach, PDOs grow in the collagen gel in the inner Transwell dish, which is exposed to air via the ALI to have sufficient oxygen supply, while the culture medium in the outer dish can diffuse into the collagen gel via the permeable Transwell.Using this system, diverse gastrointestinal cancer organoids were established by transforming primary organoids from the mouse colon, stomach and pancreas.Extending this method, Neal and Li et al. successfully cultured PDOs from surgically resected primary and metastatic tumors. 50This method was reported to show an overall 73% success rate and the established PDOs could be robustly cryorecovered.One notable advantage of this system is that it preserves the stromal components from the original normal or cancerous tissue specimens without the need of reconstitution, though the stromal cells are gradually lost over passaging.It is worth noting that in this system, CD4 + and CD8 +  34 ; Weeber et al. 37 Vlachogiannis et al. 36 Boj et al. 61 Sachs et al. 44 Kim et al. 63 Broutier et al. 40 Gao et al. 3 Lee et al. 67 Hubert et al. 69 Abbreviation: PDO, patient-derived organoid.
TILs were found to be at a substantial amount at day seven and could be sustained for more than one month when IL-2 was added.These T cells preserve the intratumoral T cell receptor (TCR) repertoire of the parental tumor.CD11b + tumor-associated macrophages (TAMs) were also observed in the co-culture system at day seven along with other immune cells, including B cells and NK cells.It should be noted that recent studies revealed that a substantial fraction of TILs does not exhibit tumor-reactive properties. 97 , 98Therefore, it remains to be seen whether tumor-reactive T cells from PBMCs or the TILs from tumor tissues provide a better model in predicting the efficacy of immunotherapy in personalized medicine.Nevertheless, the diverse types of PDOs and PDO-derived co-culture models hold great promise in addressing these issues and in providing faithful in vitro models for immuno-oncology studies.

b) Microfluidic organotypic spheroids as a holistic model for studying immune component in organoids
Despite the various advantages of PDOs in modeling tumor progression and the tumor microenvironment as mentioned above, there remains a few technical challenges associated with the use of Matrigel in organoid culture, including the heterogeneous nature of organoids' size and shape, and the limited access to the lumen space of organoids for other cell types or reagents.On the other hand, another in vitro cell/tissue culture system named microfluidic organ-on-chips, in which organotypic spheroid grows in 3D microfluidic culture units, could provide precise control of certain cellular, chemical and biophysical parameters and address some of these difficulties. 99In a typical setting of microfluidic organ-on-chips, normal tissue or tumor spheroids grow in the center of the microfluidic device, which are embedded in cell-supporting 3D gel and supplemented by media flow in the microfluidic channels.1][102] MDOTs/PDOTs from syngeneic mouse models and patient tumors preserved the autologous lymphoid and myeloid cells. 103MDOT and PDOT were reported to preserve heterogeneity in culture and could respond to immune checkpoint blockade (ICB) treatments, allowing this method to be deployed to study combinatorial immunotherapies.Although MDOTs and PDOTs can be developed from diverse cancer types, including melanoma, Merkel cell carcinoma and CRC, this application is also confined by several limitations.For example, MDOTs/PDOTs can only be maintained in short-term culture for around 1-2 weeks, and PDOTs tend to have low reproducibility due to the issue of tumor heterogeneity.Moreover, due to the resource-intensive nature of MDOTs and PDOTs, applying high-throughput therapeutic screening is challenging on these devices. 104n addition, droplet-based patient-derived micro-organospheres (MOS) have been demonstrated to have the potential for highthroughput screening. 33The principle of this method involves a benchtop machine for generating the MOS, where dissociated cells from primary tissues were added to the Matrigel in a microfluidic chip.Then, the cell-containing Matrigel is mixed with a biphasic liquid oil to generate the droplet MOS, which is directly dispensed into a MOS recovery vessel.A prospective clinical study was conducted on this system involving eight CRC patients.On average, this system only requires about ten days to complete a high-throughput drug screening, and the drug-screening results reliably predict the clinical outcome.In addition, immune cells could be well preserved in this system and remain functional, including CD4 + , CD8 + T cells, CD56 + Natural Killer cells and CD11b + cells.Therefore, this system shows great potential for testing immune-oncology therapies, including programmed cell death-1 (PD-1)/PD-ligand 1 (PD-L1) blockade and T-cell therapies for cancer patients within a time frame that is suitable for guiding cancer treatment in the clinic.Both the microfluidic organ-on-chips and the MOS technologies are compatible with PDO models to some extent.It is interesting to see in the future how PDO models could be further improved by incorporating these technologies in order to better recapitulate the tumor microenvironment in patients.

Applying PDO for precision medicine
Biomarker-directed selection of therapies in many cancer types has been demonstrated to significantly improve patient survival.For example, dMMR and high microsatellite instability (MSI-H) define a subset of CRC patients who are suitable for anti-PD-1/PD-L1 immune checkpoint blockade therapy. 105In addition, hormone receptors and HER2 expression status guide the application of hormone therapy and HER2targeted therapy respectively in breast cancer. 106However, conventional biomarkers fail to differentiate rare subpopulations of patients, and both single and "omics "-based biomarkers often suffer for their low sensitivity and/or specificity. 107Therefore, there is an urgent clinical need for personalized treatment plans that can be developed based on an accurate and timely prediction by pre-clinical models.PDO shows obvious advantages in precision medicine due to its high success rate of establishment, relatively short turnaround time of the procedure and accurate prediction of sensitivity to therapies in corresponding patients.As discussed earlier, PDOs can be generated from limited starting tumor material, such as that from fine needle aspiration.Compared to PDX models that commonly take months to develop, PDOs can be established in weeks or even days utilizing some new methods.As a result, it is possible for patients to get timely treatment advice based on the drug screening results using PDO models derived from the same patient.As summarized below and in Table 4 , many studies have validated the accuracy of using PDO as the drug screening model.

Applying PDO for chemotherapy screening
The first attempt to use PDO for predicting treatment responses in patients was performed on a cohort of metastatic CRC (mCRC) and gastroesophageal (mGOC) patients recruited in phase 1/2 clinical trials. 36he results revealed a high confidence of using PDOs in predicting responses to targeted therapies and chemotherapies in patients with 100% sensitivity and 93% specificity.This study highlighted the advantage of using PDOs to predict clinical outcomes, which is better than using molecular biomarker panels alone. 36The follow-up TUMOROID study that focused specifically on mCRC revealed that it is clinically feasible to deploy a PDO-based screen to predict responses to irinotecanbased chemotherapy. 59However, this study showed conventional PDO models failed to predict responses to oxaliplatin-based chemotherapy.These findings suggest that the conventional PDO models may not fully recapitulate the cancer cell state in the parental tumor and thus have different accuracies in predicting the responses to different treatments, which in turn highlight the importance of including immune and stromal components in PDO models to rebuild the faithful tumor microenvironment.While most studies in this category were observational and summarized the correlation between clinical outcomes and PDO-based drug screening in a non-interventional manner, the APOLLO trial in 2020 was the first to offer patients the PDO-informed treatment options. 108his study demonstrated that medium to high-throughput drug screening on PDOs could reveal novel therapeutic options with or without validated biomarkers in addition to the standard of care.For CRC with peritoneal metastases (CRPMs), FDA-approved targeted drugs on other cancer types could be potentially repurposed to treat this poor-prognostic disease should standard care be exhausted.
While CRC remains one of the most extensively studied cancer types using PDO models, similar drug screening studies on PDOs have been applied to several other cancer types, including breast cancer, 44 , 109 pancreatic cancer, 110 and ovarian cancer. 111In all of these cancer types, drug responses recorded using the in vitro PDO models highly resemble the clinical outcomes, portending that PDO-informed treatment can lead to promising personalized oncology.

Applying PDO for predicting responses to radiation-based therapy
3][114] The radiation treatment on PDOs was commonly performed by placing the culture dish into the irradiator cabinet to be treated with a total dose of 1-10 Gy based on the tumor type, and then recovering by several days before measuring the cell viability.Similarly, the in vitro response of PDOs to radiotherapy/chemoradiation could be assessed within weeks, and the results were highly correlated with the patient responses across diverse cancer types. 38 , 115-117Although the results need to be further confirmed in larger cohorts, the promising results suggest the PDOs could be used in evaluating other treatment modalities in addition to chemotherapies.

Applying PDO for predicting responses to immunotherapy
The aforementioned PDO-immune cell co-culture models can be used to test the clinical response to ICI.In the NICHE study, Chalabi et al. reported that in the presence of PD-1 blocking antibodies, CRC PDOs derived from clinical responders could be recognized by autologous T cells. 118However, in vitro CD8 + T cell reactivity in the PDO-autologous T cell co-culture system was not seen in all clinical responders.Thus, the current platform remains to be improved to accurately predict the response to ICI therapy, though it is suitable to gain biological insights of mechanisms of resistance to ICI therapies and to develop new methods to overcome the resistance.In addition, the submerged 3D Matrigel PDO system can also be used as a platform for testing the efficacy of chimeric antigen receptor (CAR) T therapies.A recent study demonstrated that engineered CAR T cells could be co-cultured with bladder cancer organoids and induce antigen specific CAR T cell-mediated cytotoxicity. 119However, the accuracy of predicting CAR T cell therapy in solid tumors by PDO -CAR T co-culture platform remains to be further validated using larger patient cohorts and in more cancer types.
As mentioned above, ALI PDO culture permits the tumor cells to grow largely in their native state, retaining a significant proportion of the tumor stroma including diverse types of infiltrated immune cells.Using this method, Neal et al. demonstrated that the TILs in the NSCLC PDOs exhibited nivolumab-dependent tumor cytotoxicity, suggesting that the PD-1-dependent ICI therapy can be recapitulated by this coculture system.A follow-up study demonstrated that for clear cell renal carcinoma (ccRCC), ALI PDO could be used to evaluate the response of combination therapy with cabozantinib and nivolumab, suggesting the potential of using ALI as a holistic model for precision medicine. 120

Applying PDO for basic cancer research using high-throughput sequencing
Since tumor PDO can be easily biobanked and passaged at a relatively low cost compared with PDX, it is beneficial to expand precious patient tumor biopsies in vitro by establishing PDOs, thus enabling many downstream applications that could not be done on patient biopsies per se.For example, high-throughput sequencing for profiling genetic and epigenetic features can be easily applied to PDO, offering the possibility for a comprehensive understanding of the parental tumor.In Table 5 , we summarized a list of recent publications that involved performing high-throughput sequencing on PDO.Although this list is not exhaustive, it already shows that for many cancer types, different methods of sequencing could be applied to PDO to advance our knowledge of disease progression, to improve the model development, and to provide mechanistic insights for future therapy development.
Genetic profiling using whole-exome or whole-genome sequencing provides information on copy number variation (CNV), single nucleotide polymorphism (SNP) and insertion or deletion (INDELs), which is widely applied in PDO based research to ensure the faithful representation of the parental tumor.Moreover, assessing the genomic landscapes of clonal organoids allows for inferring the phylogenic tree. 51 , 121Genetic mutations in the subclones could vary as a result of clonal evolution during PDO passage, resembling the genetic clonal evolution process in human cancers in vivo . 51urthermore, applying transcriptomic profiling by RNA sequencing (RNA-seq) and single-cell RNA-seq (scRNA-seq) on PDO not only allows for assessing the similarity between PDOs and the parental tumor, 122 but also allows for investigating the inherent tumor heterogeneity.For example, Togasaki et al. applied organoid models to demonstrate that the two often-coexisting subtypes of diffuse-type gastric cancer (GC), including the signet-ring cell carcinoma (SRCC) and non-SRCC subtypes, were actually clonally identical. 123This was achieved by performing RNA-seq on GC organoids representing these two subtypes, which identified Wnt signaling as the key regulator of their morphological differences.In addition, single-cell level analysis on PDOs and their matched tissue samples enabled the identification of specific cancer cell states and their drivers.For example, by performing scRNA-seq on matched PDAC tumor tissue and organoids, Raghavan et al. demonstrated a selectively enriched scBasal state with transcriptional evolution in the in vitro PDOs compared with the parental tumors. 122Moreover, the culture media formulation was found to alter the transcriptional states of PDO cancer cells and their drug sensitivity, suggesting that it is essential to use parental tumors as a benchmark and correct the culture-related biases on cancer cell state in order to establish a faithful pre-clinical cancer model.
Epigenetic profiling methods can also be applied to PDO for mechanistic studies.By performing assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) and chromatin immunoprecipitation followed by sequencing (ChIP-seq) on PDAC organoids, Roe et al. identified an enhancer reprogramming process that could promote PDAC metastasis. 124By comparing the epigenome of matched primary and metastatic organoids, a more active state in metastatic organoids was revealed, and the pioneer factor FOXA1 was identified as the driver for this metastasis-promoting enhancer reprogramming.Furthermore, examining the epigenetic and transcriptomic profiling on organoids before and after certain drug treatments may reveal resistance mechanisms.For example, Tung et al. performed the integrated profiling of ATAC-seq and RNA-seq on CRC PDOs treated with the standard-of-care therapy oxaliplatin. 125They demonstrated that the oxaliplatin-resistant tumor cells underwent patient-specific transcriptomic and chromatin alterations to increase the FGFR1 and OXTR.Silencing both genes via CRISPR/Cas9 reversed the phenotype of oxaliplatin resistance.

PDOX -a potential substitute for PDX?
PDOXs can be derived by directly injecting PDOs into the immunecompromised mice.Similar to the conventional PDX, PDOX preserves many parental tumor features, including the histopathological traits of disease progression, drug sensitivity and tumor invasiveness, which were demonstrated in multiple cancer types, including prostate cancer, CRC, glioblastoma, HNSCC, and PDAC. 3 , 35 , 44 , 61 , 116 , 117 As summarized in Table 6 , in previous studies, PDOXs were mainly used to validate the in vitro observations 1 for cancer cell invasiveness, 69 , 115 , 117 metastatic potential, 40 , 126 niche factor requirements, 35 drug sensitivities, 44 , 51 and mutation profiles. 117t is worth noting that PDOX enables the orthotopic transplantation of PDO, which has considerable merits since subcutaneous PDXs often fail to recapitulate tumor invasion and metastasis. 127Orthotopic transplantation also avoids the problems associated with conventional genetically engineered mouse models (GEMMs) of CRC, of which the high tumor burden and tumor forming in the small intestine (rather than the colorectum) have long been questioned.The orthotopically transplanted tumor can be easily modified by CRISPR-Cas9-based editing to carry oncogenic genes, which can closely recapitulate the CRC disease stage progression of adenoma-carcinoma-metastasis. 126 , 128Similarly, ortho- topically transplanted breast cancer organoids were demonstrated to not only faithfully recapitulate the tumor heterogeneity but also to preserve the drug response of the original tumor. 129These studies highlighted the importance of selecting the tumor injection site when deriving PDOXs from PDOs.
Cancer cells can be interconverted between organoid cultures and xenografts with high efficiency.Two separate studies have shown that tumor cells underwent sequential derivation as PDOs and PDOXs maintain their major mutational spectrum and are suitable for drug screening in both bladder cancer 67 and breast cancer. 130In one study, organoids were established from multiple subtypes of breast cancer PDXs.These organoids, as well as new xenografts derived from these organoids, showed high fidelity to the originating patient tumor and the initial PDXs in terms of the growth rate, proliferation pattern and histopathological characteristics. 130More interestingly, when studying bladder cancer-derived organoids, Lee et al. found that despite a group of organoid lines that stably maintains their phenotypes, there was a second group of organoid lines showing distinct phenotypes when cultured as PDOs but reverting to the phenotype of their parental tumors when cultured as PDOXs. 67Our recent work also revealed that the transcriptome and epigenome (as reflected by ATAC-seq) of PDOXs better resemble the matched PDXs and patient tumors compared with PDOs. 131aken together, these results suggest that cancer cells from PDOs retain the ability to reconstitute the TME when implanted in vivo and through interacting with components of this cancer-cell-driven TME, to re-establish their original phenotypic cell state.In other words, culturerelated deviation of cancer cell state as discussed previously may be addressed by generating PDOXs.
Based on these studies and our own results, we think that PDO and PDOX represent better pre-clinical models for cancer research and precision oncology.There are obvious advantages of this method compared with traditional 2D culture systems and the PDX models: 1) Establishing PDOs and PDOXs from patient samples show a high success rate with a reasonable temporal and monetary cost.As discussed above, the success rate of establishing PDOs from patient biopsies is much higher than that of PDXs.In many cancer types, PDOs can be established with a high success rate using FNB that contain limited patient materials, making it suitable to establish pre-clinical cancer models in broad clinical settings.In addition, numerous studies have demonstrated that xenografts deriving from organoid culture have a significantly higher success rate of establishment as compared to the xenografts directly deriving from patient samples. 35 , 67 , 117 , 130In our own hands, we observed a close to 100% success rate for PDOXs from established PDOs, 131 a finding consistent with previous studies.Moreover, PDOs can be established at a lower cost and with a shorter time compared with PDXs; they are also easy to maintain because of their ability to be easily cryopreserved, which is highly challenging for PDX models.2) PDOXs faithfully resemble the cancer cell state in the parental tumor and are thus suitable for examining the responses to a wide range of therapeutic agents.Compared with 2D cultured cancer cell lines, PDOs show significant improvements in maintaining the genetic and phenotypic heterogeneity of the parental tumor.However, culture-related deviation of cancer cell state in PDOs has been reported, 67 , 122 which may mislead the drug screening results.On the other hand, cancer cells in PDOXs are known to revert to the original cancer cell state by reconstituting the TME.Moreover, for certain targeted therapies such as VEGF inhibitors, it is not appropriate to test in the PDO model, 132 because the agents function through regulating the TME.In these occasions, only PDOXs can be used to evaluate the efficacy of these therapies.In summary, PDOX models hold great promise in accurately predicting the responses to a variety of therapeutic agents in their corresponding patients.3) PDOs and PDOXs can be easily modified genetically, allowing their broad applications in basic cancer research.Similar to 2D cultured models, PDO can be easily modified genetically, 2 and subsequently form PDOX to test the effects of these manipulations in vivo . 133These features make PDOs and their derived PDOXs suitable for studying the biological processes of tumor initiation and progression.In fact, PDOs have been used to gain novel biological insights on both the oncogenic transformation process as well as the invasion-metastasis cascade.For example, these models were applied to study the relationships between oncogenic mutations and niche factors 90 and to identify the epigenetic regulators that drive metastasis-promoting enhancer reprogramming. 124

Future perspectives
Based on a retrospective study, the pooled sensitivity and specificity of PDO in predicting drug response across different tumor types were 0.81 and 0.74, suggesting that personalized tumor response to anti-cancer therapies using ex vivo screening on PDO could serve as a strong biomarker in predicting the clinical outcomes. 134However, for some rare cancer types, the success rate of establishing PDOs can go as low as 16%, 64 suggesting that PDO-informed decision is currently not available for all patients.PDO developed from pre-treated metastases have an even lower success rate of establishment than from materials of treatment-naïve tumors.Future studies should focus on increasing the success rate of PDO establishment by optimizing the culture conditions and incorporating other advanced cell/tissue culture technologies such as the microfluidic organ-on-chips and the droplet-based patient-derived micro-organospheres.
Furthermore, it remains challenging to apply current PDO and PDOX models to examine the contributions of immune components of the TME on cancer development, which are pathologically important for disease progression.Further improvements are needed for organoids-immune cells co-culture systems and humanized mouse models so immunooncology studies can be faithfully performed using in vitro cultured models or mouse models.
PDOXs have been established and examined in several recent studies by injecting PDOs back into immunodeficient mice. 3 , 35 , 61As summarized above, this type of pre-clinical model shows several obvious advantages compared with PDX.However, more studies are needed to convincingly show PDOX still maintain essential in vivo features of patient tumor in different cancer types.In particular, more work should be done to determine whether the PDOX maintains the genetic and phenotypic heterogeneity of the parental tumor.If the level of heterogeneity varies in PDOX, it should be carefully examined whether such change will lead to significant differences in drug responses.Nevertheless, based on all the available experimental evidence, we believe that PDO and PDOX have a great potential to be faithful pre-clinical cancer models for both basic cancer research and precision medicine ( Fig. 2 ).

Table 2
PDO established in different cancer types.

Table 3
A summary for PDO culture media in different cancer types.

Table 4
PDO applied in therapeutic screening in different cancer types.

Table 5
PDO applied in high-throughput screening.

Table 6
PDOX established in different cancer types.