Patient-Derived Cancer Organoids for Precision Oncology Treatment

The emergence of three-dimensional human organoids has opened the door for the development of patient-derived cancer organoid (PDO) models, which closely recapitulate parental tumor tissue. The mainstays of preclinical cancer modeling include in vitro cell lines and patient-derived xenografts, but these models lack the cellular heterogeneity seen in human tumors. Moreover, xenograft establishment is resource and time intensive, rendering these models difficult to use to inform clinical trials and decisions. PDOs, however, can be created efficiently and retain tumor-specific properties such as cellular heterogeneity, cell–cell and cell–stroma interactions, the tumor microenvironment, and therapeutic responsiveness. PDO models and drug-screening protocols have been described for several solid tumors and, more recently, for gliomas. Since PDOs can be developed in clinically relevant time frames and share many characteristics of parent tumors, they may enhance the ability to provide precision oncologic care for patients. This review explores the current literature on cancer organoids, highlighting the history of PDO development, organoid models of glioma, and potential clinical applications of PDOs.


Introduction
Precision oncology seeks to identify genetic, epigenetic, and transcriptomic interand intratumoral changes and match cancer therapy based on an individual's distinctive cancer biology [1,2]. This translational paradigm is dependent on the coordination between clinicians and researchers, wherein both parties use their expertise to evaluate potential drug therapies using patient-specific cancer modeling systems. Traditionally, preclinical models in oncology have relied on two-dimensional cell culture and murine models, which have been a mainstay in attempts to advance basic science discoveries to the clinic through clinical trials [3,4]. However, many of the drugs investigated in traditional preclinical models have yielded suboptimal results in clinical trials, culminating in an FDA approval rate as low as 3% for new oncologic therapies [5].
The failure of new cancer therapies may be due, in part, to the intrinsic limitations of two-dimensional cancer cultures and current animal models. Two-dimensional can-cer culture techniques are limited by oligoclonal cell populations, genetic homogeneity, and lack of cell-cell or cell-stroma interactions [3,4,6,7]. Two-dimensional cultures also lose tumor-specific heterogeneity through genetic drift that occurs during long-term culture [8]. Furthermore, certain cancer-associated mutations are under-represented in tumor cell lines, necessitating suboptimal efforts to introduce such mutations ectopically into foreign cellular contexts to enable studies of their function [9,10]. In certain cancers that are highly complex, infiltrating, and heterogeneous (such as glioblastoma and pancreatic cancer), these two-dimensional cell lines lack critical inter-and intratumoral complexity and are unable to reflect the epigenetic complexity that is a signature challenge in developing novel therapeutics [3,4,11,12]. Furthermore, mounting evidence suggests close interdependence between tumors and their microenvironments, including surrounding vasculature, immune cells, and nerves, that are not captured in traditional two-dimensional models [13,14]. Gliomas, which range from grade I to grade IV by means of increasing aggressiveness, are infiltrating tumors that invade the brain parenchyma [15]. Currently, few if any cell lines exist that reflect grade II or grade III gliomas, which invariably progress to higher-grade tumors and are linked to a low median survival [15,16]. This effectively precludes in vitro study of cancer progression in this oncologic subtype. Even in vitro models of higher grades of glioma, such as glioblastoma (WHO grade IV), have significant limitations [4]. The concept of directly deriving cell lines from patient tumors has increased the ability to demonstrate genetic diversity in early preclinical studies. However, these remain challenging cell lines to create, with overall yields of 5-10% when attempting to establish patient-derived high-grade glioma cell lines. Isocitrate dehydrogenase (IDH) mutant oncogenes, which are found in the progression of lower-grade gliomas to highergrade gliomas, are almost completely absent from patient-derived lines, thus prompting efforts to exogenously express these oncogenes in IDH wild-type tumor lines. While this approach represents an important advance in our ability to study these oncogenes, it also has shortcomings, given that many tumor cell phenotypes are shaped by complex interactions between genomic alterations and epigenetic states that may not be fully recapitulated in engineered cell culture models [17][18][19][20]. This particular concern notwithstanding, 2D cell lines offer clear benefits for cancer biology research. These include their highthroughput capacity, thus enabling systematic investigations of promising therapeutics and large-scale drug target identification efforts. Furthermore, animal models of cancer, particularly murine models, offer a closer replication of clinical microenvironments but are limited by the need for immunodeficient mice, time and resource intensity, oligoclonality, and low-throughput capabilities [3,7,21,22]. Together, the limitations of these models suggest that the development of alternative, complementary models may improve precision oncology care.
The recent development of organoids, defined as in vitro three-dimensional models of human tissues, has allowed for more accurate recapitulation of parent organs [6][7][8][9][10][11][12][13][14][15][16][17]. The available literature describing patient-derived cancer organoids suggests that organoid cancer models consistently reflect in vivo tumor genotypes, phenotypes, intertumor heterogeneity, and cell-cell and cell-stroma interactions and are less resource and time intensive compared to existing mouse models [23][24][25][26][27][28][29][30][31][32]. Organoid cancer models may mitigate challenges presented by existing cancer models and augment the implementation of precision oncology by providing personalized platforms to study each patient's tumor. Herein, we review the history of organoid model development and the use organoids for cancer modeling, with an emphasis on glioma models. We also discuss the clinical implications of cancer organoid models for precision oncology applications.

History of Organoid Development
Historically, the term "organoid" referred to any in vitro culture whose characteristics resembled a primary organ. However, within the past 15 years, an organoid has been redefined as a self-organizing three-dimensional structure made from murine or patientderived stem cells or primary tissue [32][33][34][35]. Advances in stem cell science have facilitated the development of organoid cultures, particularly through the isolation of embryonic and pluripotent stem cells, which provided the tinder for the creation of early original organoid models [36][37][38][39]. Sato et al. (2009) were the first to describe organoids using intestinal stem cells for the induction and perpetuation of a human intestinal organoid [36]. By using Lgr5 + intestinal stem cells embedded in an extracellular matrix substitute (Matrigel) on media enriched with r-spondin 1, Noggin, and epidermal growth factor, the group was able to create crypt-villus organoid structures that retained the cellular organization and distribution of newly isolated small-intestinal crypts for over 8 months [36]. Since 2009, organoid models have been used to study the development of several tissues, including the brain, retina, lung, thyroid, blood vessels, small intestine, liver, mesodermal kidney, taste buds, pancreas, and ovaries [32,37,[39][40][41][42][43][44]. Notably, Lancaster et al. (2014) were the first to describe cerebral organoid models derived from human pluripotent stem cells that could produce a functional cortex, retina, and choroid plexus with discrete brain regions in a 1-2-month time frame [37,45]. Following this, fused dorsal and ventral cerebral organoids demonstrated directional migration and connectivity of GABAergic neurons [44]. Together, the cerebral organoids created by these groups not only offered distinct advantages over standard neuronal culture models but also showed that creation of non-epithelial organoid models was possible and could recapitulate in vivo tissue architecture.
Several organoid models have now been described with protocols that use induced pluripotent, embryonic, or adult stem cells [26,30,36,37,[45][46][47][48][49][50][51][52]. The organoid creation protocols for induced pluripotent stem cells and embryonic stem cells are similar, wherein isolated stem cells are exposed to tissue-specific growth factors to create desired embryoid body-like aggregates [34]. The aggregates are then embedded in a scaffold that acts as the extracellular matrix, most commonly Matrigel or collagen, as this allows the stem cell aggregates to develop the three-dimensional architecture that is critical for the development of tissue polarity from cell-cell and cell-stroma interactions [34,53]. The embedded embryoid bodies are then cultured in media containing various tissue-specific growth factors to produce mature organoids [34]. Alternatively, adult stem cells can also be used to create organoid models, but unlike induced pluripotent and embryonic stem cells, adult stem cells do not require genetic transduction [32,38,54,55]. Therefore, organoids created from adult stem cells can be immediately embedded in an extracellular matrix substitute and cultured in media enhanced with growth factors [32]. The use of stem cells to create organoids, however, is limiting, as many require genetic engineering to produce models of pathologic tissue, which may not recapitulate in vivo disease, prompting groups to create tissue-derived organoid models of disease [56,57].
Organoids have now been developed to model pathologies, including genetic diseases, infectious diseases, parasitic infections, and cancer [8,58]. A primary focus of organoid disease models has been in modeling of cancer, which has resulted in several cancer organoid models derived from epithelial tumors-prostate, liver, pancreas, ovary, bladder, lung, gastrointestinal, and breast, among others (Table 1) [23][24][25][26][27][28][29][30][31]59]. Some of the initial cancer organoid models used genetic modification of stem cells to produce cancer-like organoids, but recent work has focused on developing patient-derived cancer organoid (PDO) models from fresh cancer tissue [26,47,49,56,57,60]. The production of PDOs involves direct coordination between physicians performing tumor biopsy and laboratory personnel in order to create and propagate patient-specific cancer organoids, which may better represent the parent tumor compared to stem-cell-derived cancer organoids [23][24][25][26][27][28][29][30][31]59,61,62]. Biopsied tumor samples used for organoid production are optimally collected from specimens that are highly cellular, not cauterized, necrotic, or marginal samples with surrounding normal tissue as these factors are critical to organoid production success rates and clinical relevance [61,62]. One method to produce PDOs from patient tissue samples entails enzymatic digestion to isolate single cells, especially cancer stem cells, which are then propagated like adult stem cells [35]. While the use of cancer stem cells is one way to create PDOs, this method has been criticized for limiting the clonal heterogeneity of resultant PDOs; hence, most PDO protocols now describe the use of fresh tissue samples that are parcellated but not enzymatically digested to single cells [24,26,[30][31][32]. After being parcellated, the tissue samples are cultured in specific media with growth supplements, often embedded in Matrigel, and incubated for weeks to months to produce mature PDOs ( Figure 1) [24,26,[30][31][32]. For example, pancreatic ductal adenocarcinoma, which remains treatment refractory despite advancements in multimodality therapy, has been modeled with PDOs to test potential treatments [24]. Breast cancer has also been modeled with PDOs, with a landmark study by Sachs et al. (2018) establishing culture conditions necessary for mammary epithelial organoid creation and propagation [28].  [26] Abbreviations: EGF, epidermal growth factor; FGF, fibroblast growth factor; TGF, transforming growth factor; BMP, bone morphogenic protein; BME, basement membrane extract.

Current Models of Patient-Derived Cancer Glioma Organoids
Several glioma organoid models have been described in the literature ( Table 2). The methods used to create glioma organoids include isolation and culture of patient-derived stem cells, CRISPR/Cas9-mediated mutagenesis of non-cancer cerebral organoids, co-culture of GBM spheroids with normal cerebral organoids, and direct culture of patient-derived glioblastoma samples [56,57,62,82,83]. Hubert and colleagues (2016) first described a GBM organoid model produced from fresh tumor stem cells that recapitulated hallmark features such as hypoxia gradients, selective radiosensitivity, and invasiveness when orthotopically xenografted [57]. The glioblastoma-like organoids created from stem cells with or without directional mutagenesis were also capable of recapitulating some aspects of gene expression and phenotypic behavior of invasive brain tumors [56,68,83,84]. Further, co-culture techniques of glioma stem cells (GSCs) and spheroids with normal cerebral organoids offer significant advantages over xenograft models [82,83,85]. Single-cell RNA sequencing of co-culture of patient-derived GSCs with cerebral organoids demonstrated more accurate representation of the parent tumor compared to glioma spheres, Organoids derived from patient tissue appear to have diverse cellular backgrounds that recapitulate parent tumors. PDOs may also maintain the cellular heterogeneity found in human parental tumors [28,29,52,[62][63][64][65][66][67][68]. Weeber et al. (2015) demonstrated that 90% of the somatic mutations found in metastatic colorectal cancer biopsy tissue correlated to the corresponding colorectal PDOs [52]. Concordance of patient tumor tissue and PDO heterogeneity has also been demonstrated by several other groups, who found that interand intratumoral heterogeneity was maintained in their PDOs based on transcriptomic analysis and exome sequencing [28,29,52,62,65,[67][68][69]. The use of immunohistochemical, genomic, and transcriptomic analyses has allowed for characterization of the constituent cells in different PDO models [25][26][27]. Rosenbluth et al. (2020) used mass cytometry to demonstrate that breast cancer organoids maintain important epithelial subpopulations, including basal and luminal progenitor cells, capable of differentiating into all mammary cell lineages in long-term culture [67]. Together, these results indicate that PDOs may capture parent tumor cellular heterogeneity and allow for investigation of unique cell-cell and cell-stroma interactions.
The ability to implement advances in cellular and molecular techniques has facilitated progress in organoid creation. For example, Takebe et al. (2013) created functional, vas-cularized human liver buds through co-culturing human induced pluripotent stem cells differentiated into hepatic endodermal cells with human umbilical vein endothelial cells and human mesenchymal stem cells [70]. The resultant liver buds produced physiologic levels of albumin 10 days after ectopic transplantation in mice, demonstrating the ability of in vitro functional organ bud development [70]. CRISPR/Cas9 has also been employed by several groups to enhance PDO creation [49,50,68]. Drost et al. (2015) used this method to introduce APC, KRAS, P53, and SMAD4 mutations into mature human small-intestinal stem-cell-derived organoids [49]. The introduction of these mutations allowed human gut stem-cell-derived organoids to grow into colorectal carcinoma organoids with invasive features [49]. Assembloids, combinations of organoids representing different cell populations or regions in an organ, are also used to study the interaction between cellular populations [71,72]. Assembloids can be leveraged, as done by Kim et al. (2020), who created assembloids with patient-derived bladder tumor tissue, cancer-associated fibroblasts, immune and endothelial cells, and muscle to evaluate the epigenetic changes caused by the association of cancer cells with the surrounding stroma and their influence on tumor plasticity [72].  [26] Abbreviations: EGF, epidermal growth factor; FGF, fibroblast growth factor; TGF, transforming growth factor; BMP, bone morphogenic protein; BME, basement membrane extract.

Current Models of Patient-Derived Cancer Glioma Organoids
Several glioma organoid models have been described in the literature ( Table 2). The methods used to create glioma organoids include isolation and culture of patient-derived stem cells, CRISPR/Cas9-mediated mutagenesis of non-cancer cerebral organoids, coculture of GBM spheroids with normal cerebral organoids, and direct culture of patientderived glioblastoma samples [56,57,62,82,83]. Hubert and colleagues (2016) first described a GBM organoid model produced from fresh tumor stem cells that recapitulated hallmark features such as hypoxia gradients, selective radiosensitivity, and invasiveness when orthotopically xenografted [57]. The glioblastoma-like organoids created from stem cells with or without directional mutagenesis were also capable of recapitulating some aspects of gene expression and phenotypic behavior of invasive brain tumors [56,68,83,84]. Further, co-culture techniques of glioma stem cells (GSCs) and spheroids with normal cerebral organoids offer significant advantages over xenograft models [82,83,85]. Single-cell RNA sequencing of co-culture of patient-derived GSCs with cerebral organoids demonstrated more accurate representation of the parent tumor compared to glioma spheres, GSC-derived organoids, and xenograft models, indicating maintenance of the tumor microenvironment [85]. However, these stem-cell-based models are limited by the need for exogenous growth factors (which may lead to population drift), reliance on mutagenesis to produce tumor-like phenotypes, use of Matrigel rather than the native extracellular matrix, and relatively slow growth rates [56,57,68,[83][84][85].  Jacob et al. (2020) recently described successful GBM organoid production and biobanking from minimally processed fresh GBM samples [61,62]. Their model represents a significant advance from prior models due to several methodologic advantages and biological similarities to human tumors [62]. These GBM organoids were created with fresh tissue samples taken from the operating room in a medium formulated for neural tissue, clipped to appropriate size, and washed with minimal processing [61]. The pieces were subsequently cultured in basal media with added supportive supplements, insulin, and antibiotics and, notably, without an extracellular matrix substitute or exogenous growth factors [61,62]. These organoids formed small, round spheres by 1 week and cell-dense organoids within 2-4 weeks [61,62]. The authors demonstrated that these GBM organoids recapitulated many characteristics intrinsic to human tumors [62]. Their GBM organoids displayed characteristic histologic and immunohistochemical hallmarks of GBM, such as hypoxia gradients, nuclear atypia, and molecular heterogeneity [62,86]. The presence of hypoxia gradients is a hallmark of GBM, as hypoxia-induced signaling through notch and calcineurin pathways is known to contribute to proliferation of GSCs and correlates with patient survival [87][88][89]. RNA and exome sequencing of these organoids demonstrated similarities in gene expression, allele frequency of somatic variants, and similar copy number ratios between the parent tumor and resultant organoids [62]. Intratumor heterogeneity was also preserved in subregion organoid samples, where there was differential expression of gain-of-function epidermal growth factor receptor vIII (EGFRvIII) mutations within single organoids [62,84,90]. Furthermore, these GBM organoids retained the spatial heterogeneity and clustering of neoplastic and immune cells (T cells and macrophages) of their parental tumors, a unique feature that previous models lacked [44,62,68].
Importantly, these GBM organoids exhibited comparable tumor biology to their parental tumors, with similar ex vivo responses to both tumor-specific (tyrosine kinase inhibition) and non-specific (radiation and temozolomide) treatments as compared to their parent tumors [62]. However, the MGMT methylation status did not consistently predict the response to radiation and temozolomide in GBM organoids described by Jacob et al., as is seen in the clinic [62,91]. Lastly, the GBM organoids engrafted into immunodeficient mice within 1-3 months and recapitulated the invasive morphology of human GBM tumors [62]. Moreover, organoid-derived xenografts displayed ipsilateral and contralateral hemispheric invasion, parent-tumor-dependent satellite phenotypes, and neoangiogenesis by host endothelial cells [62]. Recent work by Golebiewska and colleagues (2020) also demonstrated that xenografted glioma organoids maintain numerous genetic glioma subtypes [63]. Together, these results highlight the similarities of histologic and phenotypic properties of GBM organoids to respective parent GBMs, a significant advance from prior glioma organoid models [56,[61][62][63]68].
Despite the rapid progress and interval advances in GBM organoid models, numerous limitations and questions remain. First, low success rates in IDH1-mutant (66.7%) and recurrent GBM (75%) indicate additional methodological optimization is needed [61,62]. Further, no data in published research currently exist on the ability to create PDOs from low-grade gliomas, which uniformly progress to higher-grade gliomas (WHO grade III and IV), or non-glioma brain tumors [15,16]. While the immune microenvironment seems to persist in current GBM organoid models, the extent to which it is represented is still in question as we continue to understand GBM-immune interactions [62,92]. The immune microenvironment has historically been difficult to demonstrate in vitro, even in organoid models where most immune cells have not been captured except when artificially replaced [93]. It is also unclear whether nerve-tumor interactions are retained in organoid models, which are a known driver of high-grade glioma proliferation [94]. Data are also underpowered in some experiments, limiting the ability to make definitive conclusions on the success of xenografts for different tumor subtypes and clinical correlation of therapeutic responses [62,63]. Further, glioblastoma is also marked by migration along blood vessels, which has not been demonstrated to date in current xenograft data [95]. Despite the need to address these uncertainties, the existing evidence suggests that patient-derived glioma organoids may offer certain advantages over standard patient-derived cancer models.

Clinical Applications
The goal of patient-derived cancer cultures is to provide a high-throughput and efficient method to accurately study cancer biology and treatment. Clinically, patientderived cultures have the potential to help physicians create personalized cancer treatment plans and inform upcoming clinical trials for novel therapies. However, existing in vitro culture models have had limited clinical utility [3,4]. There is mounting evidence indicating that PDOs offer distinct advantages over traditional in vitro methods that can be applied to clinical practice. PDOs can be generated rapidly and faithfully recapitulate characteristics of their parental tumor. These include the ability of PDOs to display similar biomarkers as parent tumors, similarity in treatment responses to human tumors, and ability to be split and propagated for high-throughput drug screening. As such, they may offer the ability to pre-screen multiple therapeutic options in a model system specific to each patient, allowing for individualized treatment selection based on the results of personalized preclinical testing. This is especially valuable for heterogeneous and poor-prognosis tumors such as GBM and pancreatic cancer, where there may be numerous clinical trial options to choose from and few, if any, predictive biomarkers that inform selection among these options.
In the past few decades, there has been a surge in the identification of key biomarkers and mutations that influence cancer progression and are targets for novel treatments. Organoids may offer the ability to recapitulate the parent tumor expression of relevant biomarkers and, in turn, serve as effective surrogates to predict responses to treatment in patients [96]. Numerous studies have reported PDO drug-screening protocols, with preliminary results indicating that organoid drug responsiveness may mimic in vivo responses [28,31,59,62,63,66,79,97]. Sachs and colleagues recently showed that xenografted breast cancer organoids mimic HER2 receptor status and tamoxifen responsiveness to parent tumors [28]. In GBM organoids, Jacobs and colleagues found that drug responsiveness to gefitinib for tumors with EGFR alterations, trametinib for NF-1-mutated tumors, everolimus for PI3K-mutated tumors, and EGFRvIII-targeted CAR-T cells could be generally predicted based on the presence of these mutations in the parent tumor [62]. Importantly, PDO versus clinical outcome studies have demonstrated that even with therapy-dependent biomarkers, drug resistance is recapitulated in subsets of clinically non-responding tumors [28,31,62,63,66,98,99]. This indicates that mutational analysis may not accurately predict drug response [28,31,62,63,66,98,99]. In addition to these examples, drug-screening PDO protocols have been described in pancreatic, gastrointestinal, lung, prostate, ovarian, and bladder cancers [28,29,31,59,62,66,79,[97][98][99][100][101]. Reasonably, using PDOs to test treatment efficacy could predict human tumor responsiveness or inform additional research to delineate underpinnings of drug sensitivity. Drug screening in patient-derived organoid models offers advantages in both their similarity to in vivo tumor biology and the increased efficiency of organoid production methodology. Many groups have also described successful biobanking of cancer organoids that can be reanimated from frozen stocks, thus allowing for the development of PDO-matched clinical databases [27][28][29][30]59,61,62].
The increase in efficiency to create PDO models compared to traditional patientderived cancer cultures makes them well suited for clinical implementation. While there is some variance in production efficiency between tumor types and the cellularity of samples, the establishment of initial organoids tends to take less than 4 weeks, with success ranging from 60% to >90% [30,31,61,62,79,97,100]. Centenera et al. (2018) notably described an efficient patient-derived explant model in which fresh prostate and breast cancer samples could be cultured on gelatin scaffolds within 6 days with high creation success rates [101]. Additionally important for the establishment of numerous organoids from a single patientderived tumor sample is the splitting and propagation of organoids to facilitate drugscreening protocols, which typically takes between 0 and 7 days for re-establishment, depending on tissue type [97]. The longer time frames needed for the production of previous in vitro models and their lack of representation to parent tumors have rendered these models less clinically useful to guide treatment [3,4] The shortened time frame for PDO production may allow clinicians to initiate drug-screening protocols directly after initial tumor biopsy or resection in order to optimize adjuvant therapy. While the time to complete drug screening will vary based on cancer type, the number of therapies tested, and success rates of organoid establishment, most drug-screening protocols can be completed within 4-6 weeks [97]. Together, organoids and ex vivo models may be especially helpful to guide treatment choices in the setting of pancreatic cancer and GBM, where the median overall survival is short, even with maximal resection and adjuvant therapy [102][103][104]. The low median survival time in aggressive or late-stage cancers creates a narrow window in which physicians can make meaningful changes to treatment modalities. The 1-4-week time frame needed for the creation and propagation of GBM organoids provides an ideal time frame for drug screening to be performed while patients are recovering from primary tumor resection and receiving the universal protocol of temozolomide and concurrent radiation [61,62,102]. Tiriac and colleagues describe a similar drug-screening protocol for pancreatic cancer, which can be completed in as little as 6 weeks, albeit with lower PDO creation success rates (<80%) [66]. High PDO creation success rates will be critical to implementation of screening to ensure that numerous organoids can be propagated from the same tumor efficiently.
Prior to the production of organoids, cancer xenograft models suffered from time intensiveness, low success rates, and lack of representation of polyclonality and phenotypic characteristics [3,4,105]. Xenografts of GBM organoids into mice have shown quick engraftment in 1-3 months with recapitulation of parent tumor characteristics [62,63]. Recently, Golebiewska and colleagues demonstrated that organoid-based glioma xenografts respond predictably to temozolomide, EGFR and CDK4/6 inhibitors, and dianhydrogalactitol, an alkylating agent [62,63]. However, xenograft success rates were lower at 86% for GBM and 25% for grade III gliomas [63]. Conceivably, the production of murine models for gliomas wherein tumor genotypic and phenotypic properties are maintained will allow for quicker testing of preclinical in vivo models. This could speed up clinical trial development and aid in enrollment in ongoing clinical trials. Since PDO xenografts retain important tumor characteristics, drug responses in these models may better represent human tumor responses compared to prior culture techniques [62,63]. However, due to the time to propagate an organoid from a fresh tumor sample (2-4 weeks) and engraft it into a murine model (1-3 months), organoid xenograft models may be challenging to incorporate in bench-to-bedside individualized precision care [22,59,[61][62][63].
In the future, organoids may also help inform the development of and enrollment in clinical trials. A major limitation in the progress of human oncologic trials has been the use of preclinical models that may not represent in vivo tumor characteristics [3,4,66]. The biological discordance between heterogeneous human tumors and classical preclinical xenograft models used to test experimental therapeutics may conceivably contribute to therapy failure in clinical trials [3,11,106,107]. The high-throughput capabilities allowed by PDOs could be leveraged to select robustly effective therapies that will succeed in clinical trials via testing numerous biobanked PDOs as opposed to fewer xenografts. Further, PDOs could act as a "pre-screen" to clinical trial enrollment, wherein a patient's PDOs are screened for a response to the novel therapy prior to enrollment. Together, these strategies may lead to improved preclinical drug screening and clinical trial patient selection to increase the odds of finding beneficial therapies for specific patient cohorts.

Conclusions
Organoids offer distinct advantages compared to previous in vitro methods of culture for human tissue, including recapitulation of three-dimensional tissue-specific structures, cell-cell and cell-stroma interactions, and genetic and cellular heterogeneity. In the decade following the first description of normal gastrointestinal tissue organoids, these models have been produced in multiple tissue types and cancers. PDOs can be created from fresh tumor tissue samples and produced rapidly compared to other cancer modeling systems. Early data suggest that PDOs can be produced in clinically relevant time frames, recapitulate characteristic biologic features of the parent tumor, and respond to oncologic therapy similarly to the parent tumors. These advantages may provide an avenue for the implementation of drug-screening approaches that allow clinicians to test tumor responsiveness to multiple chemotherapy, radiation, and targeted cancer treatment modalities. The recent development of PDOs for GBM is also important, given the historical limitations of brain tumor culture and xenograft models. While initial studies on PDOs are promising, limitations of cancer organoid model implementation in clinical oncology care include variable success rates in PDO establishment, sparse knowledge on the persistence of tumor-associated immune cells in PDOs, and limited prospective, comparative clinical data. Given the current trajectory of research in this field, PDOs appear poised to contribute to the implementation of precision oncology care for patients.