Human Colon Organoids and Other Laboratory Strategies to Enhance Patient Treatment Selection

INTRODUCTION

Colorectal cancer (CRC) is one of the leading causes of cancer and cancer-related deaths in the United States [1]. Over the last 10 years there has been a substantial increase in the clinical understanding of how molecular aberrations in CRC lead to differential treatment sensitivity, and thus, a more precision-guided approach to treatment selection. This includes using anti-epidermal growth factor receptor inhibitors for patients with KRAS/NRAS/BRAF wild-type cancers, anti-PD1 inhibitors for mismatch repair deficient cancers, BRAF inhibitor combinations for BRAF mutant cancers, and anti-HER2 agents for HER2-amplified cancers [2]. While these therapies are significant clinical advances for patients with CRC, there still remains a significant need to further personalize treatment strategies for patients. To date there are no clinically available tools to predict response for individual patients across possible treatment strategies. Organoid cultures are growing in popularity as a potential tool for this application.

Organoid cultures are three-dimensional (3D) cell cultures within a matrix substance, commonly Matrigel, that do not require the cells to adhere to plastic, which is a common necessity for 2D cultures. In addition, the cells in these cultures are not purely cell pellets, but actually form organized growing structures where the cells interact with each other (Figure 1). In general, these cultures can be readily generated with high efficiency and low cost making them the new model of choice for many translational cancer researchers. Unlike more classic patient-derived disease models, 3D organoids are proving to have a unique potential to be both accurate to tumor biology and high-throughput. In addition to modeling cancer cells, exciting new developments have been made in adding other cell types to these cultures, to more accurately reflect the tumor and its microenvironment. These models reflect a wide range of resistance mechanisms and provide ex vivo models to the growing field of immunotherapy. Here we review the ability of organoid cultures to model cancers and the feasibility and accuracy of these cultures to be used for personalizing treatment strategies for patients with CRC. Further developments into 3D organoid-based technology will improve patient outcomes by increasing the speed and accuracy with which we can predict an individual patient’s response to a wide range of therapies, allow an accurate preclinical model to identify novel therapies, and create a platform to identify novel biomarkers (Figure 2).

Figure 1.

Brightfield (A) and H&E (B) of the organized structure of patient-derived colorectal cancer organoids. Scale bar 200μM.

Figure 2.

Artistic schema of organoid development and response prediction pipeline.

PRECISION MEDICINE STRATEGIES

Tumor genetic profiling

The differences between patients in regards to tumor genetics is increasingly considered in treatment decisions. This phenomenon is reflected in the increase in genetic markers as criterion in clinical trials, and the development of specific inhibitors to block driver pathways. Tumor profiling is of special importance in highly heterogeneous cancers such as CRC, where patients display a wide range of initiating and driving mutations.

In 2017, it was estimated that less than 10% of patients actually benefitted from personalized medicine approaches [3]. While that number has certainly risen in the last couple of years, patient treatment selection could be greatly improved. Currently, tumor profiling in CRC does relatively little to determine course of therapy other than in select cases [3, 4]. Immunotherapies are approved for patients with mismatch repair deficiency (dMMR) or microsatellite instability (MSI-high) [5], which makes up less than 5% of metastatic CRCs [6]. KRAS/NRAS mutations, which make up nearly half of CRCs prevent treatment benefit from EGFR inhibition [7].

Most recently, BRAF mutant cancers, ~8% of metastatic CRCs were found to benefit from BRAF targeting combination therapies [8]. So, while tumor profiling can help guide patients to better therapy, the proportion of patients that can benefit from these therapies is still very small. The remainder of patients are treated with standard chemotherapy, radiation, or surgery based on disease stage [2]. Further developments in the relationships between tumor profile and treatment response will be necessary to expand the availability of targeted therapies. Additionally, clinical assays that can detect which patients with particular mutation profiles are the most likely to benefit from targeted therapies would be a major advance and allow for further treatment personalization, beyond using the molecular profile alone.

Circulating tumor DNA (ctDNA)

Tumors are rapidly dividing populations of cells with a high rates of cell turnover leading to the sheading of their DNA into the circulation. Technology advances have now lead to the ability to readily detect this DNA from patients as a less invasive means of tumor molecular profiling. ctDNA has many uses including tumor profiling at diagnosis, tracking of genetic changes in the tumor, and identification of tumor heterogeneity and subclonal populations within the tumor [9, 10]. Many questions still remain regarding the use of this technology to identify treatment options for patients including the mutant allele frequency needed to predict benefit and the accuracy/sensitivity of these assays to predict when possible resistance mutations are not present in the tumor.

HISTORIC MODELS FOR TREATMENT SENSITIVITY PREDICTION

Historically, 2D cell cultures have been derived from uncommon subtypes of cancer that are able to be grown adherent to culture dishes. These cell lines have been useful in developing a better understanding of cancer biology, but are not able to be utilized to identify treatment options for particular patients, as less than 10% of patient samples can be grown in 2D. Additionally, when cancer cells adhere to plastic this can change the phenotypic state of the cells including changes in RNA expression [10-12].

To overcome phenotypic changes stemming from association with plastic, researchers began creating 3D spheroid models. Several methods exist to establish 3D spheroid cultures including the hanging droplet method, aggregation, and magnetic levitation [13-18]. The hanging droplet method suspends isolated immortalized 2D cell lines or tumor derived cell suspensions in droplets allowing for the formation of 3D spheroid cultures [13-15]. The aggregation method involves centrifugation of single cell suspensions to establish compact aggregates that have the potential to form their own extracellular matrices, given the correct culture conditions [16]. Magnetic levitation involves the uptake of magnetic nanoparticles into the cells to use magnetic fields to generate aggregated cell cultures [17, 18]. These 3D culture methods result in less organized, pelleted-like structures that have limited proliferative ability [19].

In order to overcome the limitations of these classical cell culture techniques, some researchers have turned to patient-derived xenografts (PDXs), in which patient tumors are grown in immunocompromised mice. This system has shown to be more genetically and transcriptionally stable than 2D cultures, and to respond to treatment more similarly to the patient from which they are derived [10, 11, 20]. However, it is unlikely that these models will be useful for predicting an individual patient’s response in any high-throughput manner, as these models have a low take rate (about 50% [21]), require a significant amount of starting material, and take months to develop which increases the costs of these methods [20, 21].

PATIENT-DERIVED CANCER ORGANOIDS

Generation of organoid cultures

Patient-derived cancer organoids (PDCOs) have been developed as a translational cancer model to overcome many of the limitations identified with the historic models above. PDCOs can be derived from many types of samples, including surgical resections and even core needle biopsies with only a modest reduction in take rate [22, 23]. Procedures for isolation of organotypic cultures usually involves enzymatic digestion of patient tissue, either partially into spheres or into single cells that later form 3D organoids. Following digestion, cells or spheres are plated into 3D matrices that resemble the extracellular matrix, such as collagen, a major component of the extracellular matrix, or Matrigel [24]. While it is not entirely uncommon to digest tissues into single cells such that only stem cells are able to grow and form organoids, it’s most useful to keep cells together and not break them down [25].

Organoid cultures represent human disease

One concern with in vitro models is whether the cells as they exist in the plate truly reflect the biology in vivo. However, many have shown that organoids maintain both physical and molecular characteristics from the cancers that they come from (Table 1). Phenotypically, organoids maintain histological morphology and markers from the cancers they were derived from. Many authors report similarity in morphology, including retention of crypt-like or mucinous phenotypes [26]. Organoids maintain not only morphological features of the cancer from which they are derived, but clinical markers, such as cytokeratin 20 and CEA [27]. PDCO cultures have been shown to match their parent tumor transcriptionally [28]. Patient and matched organoids even have a distinct proteomic signature unique to the tumor-organoid pair [29].

Table 1.

Evaluation of organoids’ similarity to patient tumors

Reference	Cohort size (N)	Histologic similarity: Morphology matches tumor; markers	Genetic similarity
Van de Wetering 2015 [20]	20	Yes; N/A	88% across all samples, including 4 MMR deficient
Jeppesen 2017 [24]	22 primary resections 3 liver metastases	Yes; CK20, CEA	N/A
Weeber 2015 [19]	10	N/A; N/A	90% point mutations 81% copy number (8 sequenced)
Árnadóttir 2018 [27]	5 patients 13 organoid lines	N/A; N/A	40-70%
Pauli 2017 [4]	56 organoid lines; 8 CRC lines	Yes; N/A	“High” (exact numbers not reported)
Vlachogiannis 2017 [28]	23	Yes; CDX2, CK7	96%
Ashley 2014 [34]	5	N/A; CDX2, EpCAM, PR5D5	N/A
Dijkstra 2018 [48]	13 patients 15 organoid lines All dMMR	Yes; MHC I	N/A
Ganesh 2019 [29]	41 patients 65 lines	Yes; CDX2, nuclear β-catenin, MUC2, CK20, E-cadherin	92%
Kondo 2011 [22]	24	Yes; Villin, E-cadherin	Similar p53 status by staining 80% preservation of KRAS mutant status 100% preservation of BRAF mutant status
Neal 2018 [30]	20 CRC lines; 100 total	Yes; N/A	CRC sample with available sequencing maintained KRAS WT status
Pasch 2019 [23]	7	Yes; N/A	97% in MMR proficient cancers
Schumacher 2019 [60]	29 patients 49 organoids	N/A; N/A	“High”
Schütte 2017 [25]	46	Yes; Ki67, CDX2, CK7, CK20	60-100%

N/A, not available

Further, studies into the genetic changes that occur report that in most cases 80-90% identity is maintained between mutations found in organoids and the original tumor straight from the patient, with a few mutations found both in tumor-only and organoid-only [23, 25, 26, 30-32]. The exception to this rule is of course in the case of mismatch repair deficient cancers, where many small subclones exist and a great degree of genetic variation is expected to occur with each cell doubling [26]. Metastatic biopsies have been reported to maintain 90% shared mutations between biopsies and their organoids, with similar copy-number variants, and non-shared mutations being entirely passenger mutations [22]. Similar trends have been shown across many studies, with organoids maintaining key driver mutations and showing variation almost entirely within passenger mutations [23, 25, 26, 28, 30-32]. An important exception is in Pasch et al. who found a subclonal PIK3CA mutation in one organoid that was MMR proficient, and a subclonal KRAS mutation in a dMMR sample [26].

Organoids model tumor heterogeneity

Organoids also maintain cellular and molecular heterogeneity across a patient’s tumor. While most organoid culture is of pure epithelial cells [23-26], culture conditions can be altered to allow growth of other cell types, such as T cells, to be maintained [33].

Further, organoids maintain the subclonal architecture, at least of the biopsy from which they are derived [23, 26, 30]. Pasch et al. identified mutations in cancer-relevant genes at levels around 10% in organoids [26]. In their study of intratumoral heterogeneity, Árnadóttir et al. biopsied different parts of the same tumor, grew half of each biopsy into organoids, and half was used for DNA and RNA analysis. They found that each organoid culture most closely reflected the biopsy it was taken from, more so than the tumor as a whole. Each organoid/tumor pair contained founding mutations common to all samples from that patient, as well as their own genetic subclones [30]. This shows both a limitation and an advantage to the organoid model: while it appears necessary to get multiple biopsies to capture all subclones present in a tumor, it is possible to model subclones in 3D organoids, something that has not been seen in 2D culture that becomes clonal after a few passages [28]. Also, it is feasible to grow organoids from multiple biopsies, something likely too expensive, time-, and resource-consuming for PDX models.

Assessment of treatment response

Although organoid models are established to accurately represent patients’ cancers, there remain no standard assessment techniques to determine treatment response using these cultures. In 2D models, cells grow as nearly independent units, and as such confluency or metabolic activity of individual cells is a direct measure of cell growth, death, and therefore response. In 3D, evaluation of response is more difficult, and different measurements can have different biological implications. Below are discussed a few methods currently used to assess treatment response in organoids.

ATP-dependent assays.

CellTiter-Glo and CellTiter-Blue are commonly utilized response assessment tools that take advantage of the metabolic activity of cells in order to measure cell survival. The assay compares the levels of ATP present in the culture, and comparing amounts of ATP between conditions indicates different numbers of cells present, and therefore response. CellTiter-Glo and similar techniques are the gold standard in growth response in 2-dimensional cultures and have been widely used to assess response in 3D as well [4, 23, 28, 32, 34-36]. The advantage to CellTiter-Glo is that it is so widely accepted already as a measurement of treatment response. It gives a good overall reading of the metabolic activity of a well of organoids as a whole, and can therefore be a good tool for assessing an overall response. Another advantage is that the readout uses plate readers, which often have already been established to measure readout, which makes this readout easy to do in a high-throughput manner, consistent with one of the biggest potentials in organoids versus previous models such as PDXs.

However, there are several limitations to using CellTiter-Glo to measure response. Since CellTiter-Glo measures response on the whole-well level as the number of cells present and metabolically active after a treatment, it requires even plating of cells at baseline. In 2D cultures, this means counting single cells before plating, a simple task. However, many spheres do not survive being split into single cells for counting, and no other method currently exists to normalize cell numbers. Further, using metabolic activity in the well as a whole ignores the heterogeneity that organoids capture. For example, a 50% reduction in metabolic activity could mean half of the total number of spheres died or that half of the cells within each sphere died, with both answers carrying very different biological implications. Other colorimetric assays that have been used to measure organoid response to therapies include WST [37] and a combination of Sytox and Presto Blue [38].

Diameter or size.

One method that has proven robust in predicting treatment response is comparing the change in size of individual organoids after treatment, often by longest diameter or area [25-27, 39]. Measuring changes in size creates a natural form of normalization not effective in 2D culture; each sphere is normalized to its size at the beginning of the experiment, overcoming the need to plate wells evenly.

There are limitations, however. Sphere size does not always directly correlate with the number of live cells, as many spheres can have cores that are either necrotic or mucinous. Further, spheres are constantly changing, and may shrink due to reasons other than treatment response. Another limitation is the amount of time it takes to analyze spheres manually, making high-throughput methods with high numbers more likely to be significant difficult. This, however, is changing with the development of automated analysis algorithms.

Optical Metabolic Imaging (OMI).

OMI is a powerful technique to study cellular metabolism on an organoid or cellular level without the use of reagents or destruction of organoids allowing longitudinal analyses. It has the potential to measure the efficacy of a drug in patient derived organoids, which can be used to predict patient response [40]. By measuring nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD) ratio, this method serves to measure the metabolic signaling and therapeutic response. It is also an effective tool to identify heterogeneity at organoid and cellular levels [41]. One of the biggest challenges in cancer treatment is resistance [42]. In many cases there persists a subpopulation of tumor cells within a spheroid which is heterogeneous from rest of the cells and is the main culprit for tumor recurrence [43-45]. OMI can capture heterogeneous populations [44] in the setting of treatment to identify resistant cell populations. It is noninvasive [41], and time and cost efficient. Walsh et al. predicted drug response in breast cancer patients by quantifying OMI in primary tumor organoids [41, 42]. There is compelling evidence showing the robust nature of this technique used to predict the efficacy of drugs in colorectal [26] and other cancers [46-49], as well as detect drug resistance in subpopulations. The organoid system complemented with diameter study and OMI provides a robust platform for measuring drug efficacy, heterogeneity and weaving a path for personalized medicine [48]. As 3D cultures are further used for translational investigations, it is advisable to use multiple methods to determine therapeutic response [26, 37, 50].

Microenvironment

One highlight of using 3D organotypic cultures is the ability to model not only the cancer cells themselves but the surrounding environment, including cancer associated fibroblasts and immune cells. Not only can you model the signaling aspect as with 2D co-cultures, but you can add the dimensionality of movement through a matrix. At least for cell pellet-derived spheroids, it has been found that co-culture with both fibroblasts and endothelial cells can change treatment response, often decreasing treatment efficacy [35], and this trend is expected to be true in more representative cancer organoids as well.

Recently, Dijkstra et al. reported the ability to culture patient-derived tumor organoids with peripheral blood lymphocytes and enrich for tumor-reactive T cells [51]. These authors demonstrate the ability of using their platform to predict the ability of a patient’s population of T cells to generate an anti-tumor response in clinically relevant settings. Interestingly, they discovered that ~50% of MHC-I⁺ tumor organoids generated an activation in the PBMC-derived autologous CD8⁺ T cell population, whereas no MHC-I⁻ organoids generated a CD8⁺ T cell response.

In an effort to preserve the tumor-infiltrating immune microenvironment upon extracting tissue, several groups are investigating the use of a system which allows the researcher to preserve the 3D nature of the tumor organoid [52] as well as maintain viability of key immune cell populations such as CD4⁺ and CD8⁺ T cells and various myeloid populations [53]. The authors are able to demonstrate anti-PD1 responses to murine-derived organoids. Subsequent research by other groups using the same microfluidics system demonstrated similar responses to anti-PD1 therapy in murine-derived organoids [54]. This group has also demonstrated an enhanced response in anti-PD1-resistant murine organoids in combination with CDK4/6 inhibition, demonstrating the ability of this platform to investigate novel immuno-oncology drug combinations. The idea of creating culture conditions to maintain endogenous TILs has recently been achieved in human cultures as well [33]. Further research is urgent in this area as personalized medicine looks for an ex vivo platform to predict response to such agents.

High throughput screening

While patient-derived cancer organoids have been utilized for prospective and retrospective prediction of patient response to various clinically relevant agents as well as limited evaluation of novel agents and drug combinations, the use of high-throughput drug screening (HTS) with organoid models has been less-expansively described in the literature. Historically, HTS occurred though three-dimensional engineered aggregates of established lines or early passage primary cells (often referred to as “organoids”), two-dimensional cultures, short-term culture of ex vivo tumor sections, or xenograft tumors in mice [55, 56]. Such preclinical prediction models have been associated with high rates of failure upon clinical translation [56, 57].

More recently, self-assembling organoids from primary tissues including cerebral, hepatic, pancreatic, and intestinal – among others – have found favor as a potential alternative to the above models given improvement in maintaining the architecture and heterogeneity of the in vivo lesion [6, 9]. HTS of organoids has been demonstrated in multiple cancer types such as colorectal, ovarian, and endometrial utilizing different techniques including direct culture from patients as well as xenograft maintenance with subsequent tumor dissociation, nearly all ultimately utilizing a three-dimensional matrix like Matrigel for organoid suspension – whether as a droplet or peripheral ring in a plate well – with ATP-based luminescent viability assay read-outs [7, 10, 58]. In one study, three ovarian and one high-grade serous peritoneal cancer samples were screened with 240 kinase inhibitors using the ring method from biopsy/surgically obtained tissue such that results were available within one week from collection for potential clinical application [7].

In CRC, several recent studies have shown promise for the use of patient-derived cancer organoids in HTS ranging from 10s to 1000s of drugs per reported screen. For example, van de Wetering et al. screened 19 colorectal organoids – noted to be genetically similar to parental tissue – with 83 compounds and were able to connect genetic and drug sensitivity data [11]. Interestingly, a clustering of response was seen with drugs inhibiting the IGF1R and PI3K-AKT signaling pathways. Kondo et al. outlined that patient-derived organoids maintained in a xenograft model prior to ex vivo use resulted in fidelity of morphological and genomic features and could be automated in a 2427-drug HTS of two colorectal lines; this is the largest reported organoid HTS to date, though with noted variable sensitivity to the agents utilized given the limited number of lines employed [58]. Schütte et al. evaluated a panel of clinically relevant compounds including eight against 19 pairs of patient-derived organoids and xenografts and showed concordance for all by two if the drugs with some attribution of this to pharmacokinetic parameters, also noting that xenografts appeared slightly closer to the tumor molecular profiling than the organoids though likely from sampling heterogeneity with mutational pattern stability over multiple passages [9].

Established patient response data

Recent work has highlighted the clinical utility of patient-derived organoids to reflect tumor biology, including comparison to patient response. Studies and responses are summarized in Table 2. The feasibility for clinical prediction has been reported for radiation [26, 32] and all standard of care chemotherapy including 5-fluorouracil [26, 32, 36], oxaliplatin [26, 32], irinotecan [36], cetuximab [32], TAS-102 [31], and regorafenib [31]. These techniques require robust prosective validation including designation as an integrative biomarker in future trial designs.

Table 2.

Published Correlations with Patient Response

Reference	Cohort size	Method of response measurement in organoids	Correlation with patient response
Pasch 2019 [23]	1	Diameter and OMI	Intermediate response to FOLFOX predicted; 10% reduction clinically in primary site
Ganesh 2019 [29]	7 for 5-FU and FOLFOX 5 for XRT	Cell-Titer Glo	Spearman correlation r = −0.86 p = 0.024 between PFS in clinic and organoid response to FOLFOX or 5FU Correlation between categorized low- and high-responding organoids and patients undergoing XRT (measured by luminal occlusion)
Ooft 2019 [33]	10	Cell-Titer Glo	83.3% of patients correctly classified in irinotecan treatment No correlation for FOLFOX
Vlachogiannis 2017 [28]	21	Cell-Titer Blue	100% sensitivity 93% specificity 88% positive predictive value 100% negative predictive value

Pasch et al. describe the use of organoids to predict a sensitivity to FOLFOX in a patient who on long previous treatment courses had shown resistance to 5FU. Upon retreatment with FOLFOX after this discovery, the patient responded very well, showing the usefulness of organoids in predicting an individual patient’s response [26]. Ganesh et al. also saw a strong correlation with patient response and FOLFOX [32]. In contrast, Ooft et al. found a robust correlation between patient response and FOLFIRI, but not FOLFOX [36], though also from a smaller cohort, highlighting the need for development of both robust and consistent measures of response. Authors have also described a potential for these therapies to predict radiation response [32]. Vlachogiannis et al. have described a high specificity and sensitivity of organoid culture in responding to many diverse therapies in a cohort of 21 patients [31]. Further studies of larger cohorts is needed to better understand how these models can be used clinically to accurately predict patient response.

FUTURE APPLICATIONS

Current research has already shown the potential of organoids to reflect how individual patients will respond in clinic [26, 32, 36]. Technology is nearing timelines that will make it actually feasible to determine patient treatment options [32, 36]. Organoids can successfully be frozen in liquid nitrogen and brought back for further research [23]. Because of this nature, it is possible to create biobanks of several patient-derived organoids for research. Biobanks of colon organoids have already started as early as 2015 [23]. More recently, the Office of Cancer Genomics at the NCI has started the Human Cancer Models Initiative, which includes a bank of organoids, as well as clinical and omics information, available through ATCC [59]. Biobanks will allow researchers to connect treatment response across many clinical presentations of CRC. This, in turn, will lead to new discoveries of specific biomarkers for treatment response and improve selection of therapy.

High-throughput screens will allow for the use of these cultures to identify novel treatment combinations to be explored further in animal models and future clinical trials. Finally, the use of high-throughput yet biologically accurate 3D cultures allows for a better and deeper understanding of how patients with different molecular profiles respond to different therapies. In conjunction with tumor profiling, ctDNA, clinical results, and 3D cultures, we will finally be able to compare responses of actual patients across a wide range of molecular profiles and a wide range of treatments for each, such that future patients may only need to be profiled to understand what treatment is best for them.

SUMMARY

PDCOs are becoming a more viable option to assess treatment response for patient treatment selection. Several studies have already shown their potential to accurately predict a patient’s response to several agents, as evidenced by a correlation between the response seen in patient-derived cancer organoids and those same patients in clinic treated with the same agents.

There are several directions in which this technology is improving to help patient response predictions. In development are methods to improve response prediction to specific drugs, including more sophisticated analysis methods and models that include more components of the tumor microenvironment. Beyond analysis of a specific treatment, methods are being developed in order to screen patient samples in a high-throughput manner, with the intent to quickly test a large number of drugs and combinations to find the best match for a patient. Finally, with the development of biobanks of many patient-derived organoids with known genetic and even clinical information from a wide range of patients, researchers will have access to more clinically relevant samples and information than ever before. With access to samples from all sorts of different diseases, researchers will have the tools to find biomarkers to better predict patient response from patients without their own PDCO line.

PDCOs have several advantages over previous models used to study disease. 2D cell culture was not always able to be established from patient samples, and usually grew out of single clones, deleting any information about the effects of heterogeneity within the disease. Further, the environment of living in a monolayer on plastic is so removed from the native state of growing 3-dimensionally on a scaffold that the culture of these cells usually involved vast changes to the genetics and transcriptomics, bringing into question whether these cells truly reflected disease biology. PDX models much better model the true disease biology. However, mice are expensive, and PDX models are difficult to establish, rendering PDX-based studies feasible on only the most aggressive tumors of only specific disease types. In contrast, PDCOs have a relatively high rate of success, especially in CRCs, maintain the genetics, transcriptomics, and heterogeneity of the tumor from which they were derived, and can easily be manipulated in a high-throughput manner. While further advances are still needed, PDCOs show great promise as research tools and as a clinical tool to predict patient response.

OPINION STATEMENT.

Though many advancements in personalized medicine have been made, better methods are still needed to predict treatment benefit for patients with colorectal cancer. Patient-derived cancer organoids (PDCOs) are a major advance towards true personalization of treatment strategies. A growing body of literature is demonstrating the feasibility of PDCOs as an accurate and high-throughput pre-clinical tool for patient treatment selection. Many studies demonstrate that these cultures are readily generated and represent the tumors they were derived from phenotypically and based on their mutation profile. This includes maintenance of the driver mutations giving the cancer cells a selective growth advantage, and also, tumor heterogeneity, including molecular and metabolic heterogeneity. Additionally, PDCOs are now being utilized to developed patient biospecimen repositories, perform high to moderate-throughput drug screening, and to potentially predict treatment response for individual patients are undergoing anti-cancer treatments. In order to develop PDCOs as a true clinical tool, further studies are required to determine the reproducibility and accuracy of these models to predict patient response.