A patient‐derived explant (PDE) model of hormone‐dependent cancer

Breast and prostate cancer research to date has largely been predicated on the use of cell lines in vitro or in vivo. These limitations have led to the development of more clinically relevant models, such as organoids or murine xenografts that utilize patient‐derived material; however, issues related to low take rate, long duration of establishment, and the associated costs constrain use of these models. This study demonstrates that ex vivo culture of freshly resected breast and prostate tumor specimens obtained from surgery, termed patient‐derived explants (PDEs), provides a high‐throughput and cost‐effective model that retains the native tissue architecture, microenvironment, cell viability, and key oncogenic drivers. The PDE model provides a unique approach for direct evaluation of drug responses on an individual patient's tumor, which is amenable to analysis using contemporary genomic technologies. The ability to rapidly evaluate drug efficacy in patient‐derived material has high potential to facilitate implementation of personalized medicine approaches.

Breast and prostate cancer research to date has largely been predicated on the use of cell lines in vitro or in vivo. These limitations have led to the development of more clinically relevant models, such as organoids or murine xenografts that utilize patient-derived material; however, issues related to low take rate, long duration of establishment, and the associated costs constrain use of these models. This study demonstrates that ex vivo culture of freshly resected breast and prostate tumor specimens obtained from surgery, termed patient-derived explants (PDEs), provides a high-throughput and cost-effective model that retains the native tissue architecture, microenvironment, cell viability, and key oncogenic drivers. The PDE model provides a unique approach for direct evaluation of drug responses on an individual patient's tumor, which is amenable to analysis using contemporary genomic technologies. The ability to rapidly evaluate drug efficacy in patient-derived material has high potential to facilitate implementation of personalized medicine approaches.

Introduction
Preclinical models that accurately replicate the native architecture and microenvironment of primary human tumors are urgently needed to improve our understanding of basic tumor biology and to facilitate the development of improved therapeutic approaches (Thompson et al., 2008). The use of immortalized human cell line models for investigating novel therapies in vitro or in vivo is convenient but also a major reason for the high failure of new drugs entering clinical trials. To move toward more clinically relevant model systems, researchers have adopted patientderived approaches such as organoids (Drost et al., 2016;Gao et al., 2014) and xenografts (PDX) (Lawrence et al., 2013;Wang et al., 2005;Whittle et al., 2015). We report an alternative approach of culturing freshly resected breast and prostate cancer tissue as patient-derived explants (PDE). Ex vivo culture of human prostatic tissue has been used since the 1970s with varying degrees of success, and protocols ranging from total immersion of tissue pieces in medium to culture of tissue pieces or slices on grid or sponge scaffolds, reviewed by Centenera et al. (2013). Despite this long-standing history and general acknowledgment of the potential of ex vivo cultured tissues to increase the clinical relevance of laboratory research (Centenera et al., 2013;Kim, 2005;Pretlow et al., 1995;Risbridger et al., 2018;Vescio et al., 1991), the PDE method has not been widely adopted to study solid tumors. The purpose of this study was to highlight advantages of the PDE model and demonstrate how it can be applied to interrogate hormone-dependent cancers such as those of the breast and prostate.

Ethical approval for research using human tissue
All research conducted in this study conformed to the standards set by the Declaration of Helsinki.

Australia
The use of freshly resected human tissue samples for this study was approved by the Human Research Ethics Committees of the University of Adelaide, the Royal Adelaide Hospital, and Burnside War Memorial Hospital. All material is collected with written informed consent from patients, and data are de-identified according to National Health and Medical Research Council of Australia guidelines for human research. Prostate tissues were collected as part of the Australian Prostate Cancer BioResource.

United States
Prostate tissues from UT Southwestern Medical Center at Dallas and Thomas Jefferson University hospitals were obtained with written informed consent from patients undergoing radical prostatectomy for highvolume cancer, under Institutional Review Boardapproved protocols for the respective institutions.
Clinical characteristics and histopathology of patients who donated tissue to this study are outlined in Table S1.

PDE culture of solid tumors
The PDE technique utilized in this study employs a gelatin sponge platform (Centenera et al., 2013) (Fig. 1A). This method was selected for two key reasons: firstly, the use of a substrate for explant culture prevents cellular outgrowth that is frequently observed when tissues are cultured without support and completely submerged in media (Pretlow et al., 1995;Varani et al., 1999); secondly, the gelatin sponge used in this study is a commercial medical device developed for hemostasis (Ferrosan, 2014) that is readily available, cost-effective, and simple to use, making the method feasible for widespread adoption in translational cancer research laboratories. Figure 1A illustrates the PDE method using prostate tumors as an example. Following surgical tissue removal, the surgeon or clinical pathologist resects a sample of the presumed malignant or nonmalignant region and the specimen is transported to the research laboratory in cold phosphate-buffered saline on ice, typically within 1-2 h of surgery. Under sterile conditions, tissue was placed onto the lid of a 10-cm plate along with the saline it was transported in. Using a surgical blade, a 1-mm-thick longitudinal section of the tissue sample is cut and placed into neutral-buffered formalin for paraffin embedding. This tissue is called the Day 0 sample and is used to determine the cancer content of the tissue following staining with hematoxylin and eosin (H&E). The remaining tissue is dissected into 1-mm 3 pieces, called explants, and placed in triplicate or quadruplicate (depending on amount of tissue received) on top of a media-soaked gelatin sponge (Ethicon; Johnson & Johnson, Somerville, NJ, USA) inside the wells of a 24-well plate containing 500 lL RPMI 1640 medium containing 10% fetal bovine serum, 19 antimycotic/antibiotic solution, 0.01 mgÁmL À1 hydrocortisone, and 0.01 mgÁmL À1 insulin. The appropriate vehicle, treatment, or shRNA was added directly to the media inside the appropriate tissue culture well at the indicated concentrations, allowing direct comparison of treatments and controls. Explants were cultured at 37°C and 5% CO 2 for various time points and then formalin-fixed and paraffinembedded, snap-frozen, or preserved in RNAlater (Invitrogen, San Diego, CA, USA) depending on the desired downstream analysis. For assessment of BrdU incorporation, BrdU (10 lM) was added to the culture medium 2 h (prostate) or 24 h (breast) prior to harvest.
In our collective experience, all tissues containing epithelial cells can successfully be cultured, and the limiting factor for analysis is instead the presence of sufficient numbers of epithelial (for benign tissue studies) or malignant cells (for cancer tissue studies). For this reason, routine hematoxylin and eosin (H&E) staining of all Day 0 and cultured tissues is an essential part of our protocol to confirm the presence and approximate percentage of benign/malignant cells within the specimens before proceeding with further analyses. Between the three prostate cancer laboratories, our collective experience indicates that approximately 10% of tissues from high-volume disease and 20-30% of tissues from low-volume disease do not contain malignant cells, and this is largely due to sampling.

Immunohistochemistry
Immunohistochemical staining was performed on 2-to 4-lm sections that were deparaffinized, rehydrated, and blocked for endogenous peroxidase before being subjected to heat-induced epitope retrieval. Breast tissues were additionally treated to block for endogenous biotin. Sections were blocked and incubated with the appropriate primary and secondary antibodies, then developed using 3-3 0 -diaminobenzidine chromogen, and counterstained with hematoxylin. Appropriate positive and negative controls were included in all runs. Table S2 lists pertinent information for all antigens. Immunostaining in the Tilley laboratory was performed manually, and in the Raj laboratory, immunostaining was performed on the DAKO autostainer (DakoCytomation, Carpinteria, CA, USA). Tissues cultured by the Knudsen laboratory were stained by the Raj laboratory. The percent positivity and intensity of nuclear staining for each antigen were quantified manually by an independent pathologist or observer, who was blinded to the treatments/conditions.

RNA extraction
Tissues preserved in RNAlater (Qiagen, Hilden, Germany) were homogenized in ice-cold PBS with a Precellys 24 Tissue Homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France), and RNA was extracted using Trizol according to the manufacturer's instructions. DNAse treatment was performed using the TurboDNase kit (Ambion, Austin, TX, USA) according to manufacturer's instructions. RNA was quantified using a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription was performed on 400 ng total RNA using the iScript kit (Bio-Rad Laboratories, Hercules, CA, USA) according to manufacturer's instructions.

Quantitative real-time PCR
qPCR was performed using 2 lL of cDNA from whole tissue extracts and SYBR green for prostate explants, and gene-specific TaqMan assays were performed for breast explants. Primers and assay IDs are listed in Table S3. Relative gene expression was calculated using the DCt method. Expression of PSA in prostate explants was normalized to PPIA, L19, TUBA1B, ALAS1, and GAPDH. Progesterone receptor (PGR) expression in breast explants was normalized to IPO8 and PUM1.

Ablation of AR by shRNA
The shAR and shControl lentiviral constructs were packaged using Lenti-X HT (Open Biosystems, Dharmacon Inc., Lafayette, CO, USA) according to the manufacturer's instructions and added to the tissue (C) HIF1a staining from three independent laboratories showed no significant difference in prostate cancer tissue oxygenation after 6 days of PDE culture. Staining intensity was manually assessed by a single pathologist (P. Kapur). Data are presented as mean AE SEM, n = 3. (D) Expression and signaling of steroid receptors critical for prostate and breast carcinogenesis were maintained in PDEs cultured in complete media for 6 days, as demonstrated by immunostaining for AR and the AR-regulated protein PSA in prostate PDEs, and ERa and the ERa-regulated protein PGR in breast cancer PDEs. Scale bars represent 50 lm. culture medium for 48 h. Lentiviral media was then removed and replaced with fresh media containing no lentivirus and cultured for an additional 48 h. Tissues were harvested and formalin-fixed prior to immunohistochemistry staining for androgen receptor (AR) and prostate-specific antigen (PSA) as described above.
2.7. Chromatin immunoprecipitation followed by massively parallel DNA sequencing (ChIP-seq) Breast cancer PDEs cultured in complete medium containing vehicle (EtOH; 0.1%), 17b-estradiol (E2; 10 nM), or synthetic PGR agonist R5020 (10 nM) for 72 h were harvested, crosslinked, and processed for ChIP-seq analysis as described previously (Ross-Innes et al., 2012). The estrogen receptor alpha (ERa) HC-20 antibody (Santa Cruz Biotechnologies, Dallas, TX, USA) was used for immunoprecipitation of ERa from PDE tissue lysates. Sequences generated by an Illumina HiSeq 2000 were processed by the Illumina analysis pipeline version 1.6.1 and aligned to the Human Reference Genome (hg19) using BWA version 0.5.5 (Li and Durbin, 2009). Reads were filtered by removing those with a BWA alignment quality score less than 15. Enriched ERa binding regions were identified by comparing ERa ChIP DNA samples to total ChIP input DNA. ER ChIPsequencing was then performed from each PDE tissue and treatment. Peak calling was performed using MACS2 version 1.4.0rc2 (Zhang et al., 2008).

Statistical analysis
Data are displayed as the mean AE standard error. Differences were determined using Student's t-test or oneway analysis of variance (ANOVA). The correlation between Ki67 and PSA expression was analyzed using the Spearman correlation test with an accompanying P value. All statistics and generation of heat map were performed using GRAPHPAD PRISM Software version 7.02 (GraphPad Software, La Jolla, CA, USA). A P-value ≤ 0.05 was considered statistically significant.

Morphology, viability, and molecular signaling are sustained in PDEs
Histopathological evaluation of tissue architecture and cellular appearance demonstrated that the morphology of PDEs cultured for up to 6 days on gelatin sponges is consistent with the original, uncultured (i.e., Day 0) tumor tissue, as illustrated by representative H&E staining of prostate cancer and breast cancer (Fig. 1B).
Tumor cells are present in the surrounding stroma, demonstrating maintenance of the tumor microenvironment (Fig. 1B). Immunohistochemical staining for the hypoxia-inducible factor (HIF)-1a showed no evidence of hypoxia in PDEs cultured using the sponge method (Fig. 1C). Importantly, equivalent HIF-1a staining was observed in prostate tumors cultured in three independent laboratories (Fig. 1C) and evaluation of serial sections of individual tissues indicated no discernible difference in histology or antigen staining between the air or sponge interface (data not shown).
The AR and ERa are critical drivers of prostate and breast cancer, respectively; therefore, AR and ERa were assessed as examples of major oncogenic signaling pathways in the PDE model. Epithelial cell positivity for immunoreactive AR and ERa was sustained within malignant prostate and breast PDEs, respectively, for up to 6 days in complete media (Fig. 1D). Expression of the androgen-regulated protein, PSA, and the estrogen-regulated protein PGR, indicative of functional signaling by AR and ERa, respectively, was also evident after 6 days of culture (Fig. 1D).

Proliferative capacity of PDE tumor tissue
Cellular uptake of the thymidine analog 5-bromo-2deoxyuridine (BrdU; Fig. 2A) and its nucleoside analog 5-ethynyl-2 0 -deoxyuridine (Fig. S1) from the culture media was observed in a subset of prostate tumors, indicating that cultured tissues have the capacity for de novo proliferation. BrdU uptake was evident throughout the PDE samples ( Fig. 2A) and was concordant with Ki67 positivity (Fig. 2B). Ki67 positivity in breast and prostate explant tissues was investigated further to assess suitability of the PDE model to evaluate proliferative responses to hormonal stimuli, growth factors, or therapeutic agents. The number of Ki67positive epithelial cell nuclei (i.e., the cellular proliferative index) in Day 0 tissues ranged from 0 to 16% and increased to a range of 0 to 43% (P < 0.05) in matched tissues after 48 h of culture and remained stable for at least 96 h (Fig. 2C). The time-dependent increase in proliferation observed here and in previous organotypic culture models (Nevalainen et al., 1993;Olbina et al., 1998) is likely due to growth-promoting factors in the media or to the release of an inhibitory serum influence that was present in vivo, but absent ex vivo.

Modulation of AR signaling in prostate cancer PDEs
AR is the driving transcription factor in prostate cancer and the primary therapeutic target for systemic treatment. To demonstrate modulation of this critical signaling pathway, a lentiviral-based shRNA approach was used for isogenic suppression of AR gene expression. Lentiviral transduction of cultured prostate PDEs with an AR-directed shRNA for 48 h decreased the steady-state protein levels of AR by approximately 50% (Fig. 3A). Importantly, AR inhibition resulted in a 40% decrease in expression of the AR-regulated protein PSA (Fig. 3A), which demonstrates that the AR signaling pathway is functional in PDE tissues.
Clinical inhibition of the AR signaling pathway is achieved through androgen deprivation therapy or AR antagonists (Attard et al., 2011). To demonstrate capacity of the PDE model for evaluating therapeutics, prostate cancer PDEs were treated with the clinically used AR antagonist bicalutamide. Proliferative response was assessed using the immunohistochemistry marker Ki67, with a response considered significant when treatment induced a change from vehicle of ≥ 25%. Proliferative responses were observed after only 48 h of culture; therefore, this time point was selected to allow short-term, high-throughput preclinical testing to be performed. Similar to Fig. 2C, a large range of baseline Ki67 (2-58%) was observed across the cohort (Fig. 3B). In prostate cancer PDEs cultured in the presence of 10 lM bicalutamide for 48 h, the proliferative index was reduced in 10/23 tissues (44%), increased in 6/23 tissues (26%), and had no significant effect on proliferation in 7/23 tissues (30%) compared to vehicle-treated PDEs (Fig. 3B). Similar heterogeneity in response was observed with the newer generation antagonist enzalutamide in a smaller, independent cohort of PDEs (Fig. S2). A change in apoptosis was not observed in PDEs cultured for 48 h with 10 lM bicalutamide (Fig. 3C). Collectively, these findings illustrate the heterogeneity of clinical samples and demonstrate the importance of using a model that reflects this natural variation when evaluating solid tumors. A critical aspect of this research will be relating the observed heterogeneity with clinical outcomes of these patients, but we were unable to assess this in the current study. To date, only one patient (Fig. 3A patient no. 18) from our cohort of predominantly lowto-intermediate risk patients (median follow-up time of 48.3 months; range 23-64.4 months) has experienced biochemical recurrence (BCR). This is consistent with the reported 5-year BCR rate of 10-20% in low-tointermediate risk patients undergoing robotic-radical prostatectomy (Diaz et al., 2015) and highlights the need for longer-term follow-up to correlate PDE results with individual patient outcomes. A more feasible clinical context to ascertain whether PDE culture is correlated with clinical outcome is neoadjuvant clinical trials that allow direct comparison of pre-and posttreated tumors.
In a subset of 12 PDEs from Fig. 3B, where sufficient matched tissue was available to extract RNA, changes in PSA gene expression in response to bicalutamide treatment were evaluated by qPCR. Similar to Ki67, heterogeneity in PSA response to bicalutamide was observed across the 12 explants (Fig. 3D). Importantly, in 10/12 tissues Ki67 positivity and PSA expression increased or decreased concordantly, demonstrating a significant positive association (r = 0.657; P = 0.0238; Fig. 3E). Transcript analysis of additional prototypical AR-regulated genes FKBP5, KLK2, NKX3.1, and TMPRSS2 shows a similar pattern of expression to PSA and Ki67 (Fig. 3E), indicating general inhibition of androgen signaling by bicalutamide in prostate cancer PDEs. No correlation was observed between proliferative response and tumor grade, stage, or presurgery serum PSA. However, we acknowledge that a larger range of samples may be required to determine this conclusively as the majority of our specimens were of Gleason grade 7, stage PT3A/B, and PSA < 10 (Table S1).

Analysis of ERa signaling in breast cancer explants
Estradiol (E2)-activated ERa binds to cis-regulatory elements of target genes such as the PGR (Ross-Innes et al., 2012). Candidate gene analyses performed on RNA extracted from ERa-positive PDEs (n = 14) cultured for 24 h with E2 revealed a range of responses to hormone treatment. E2 stimulation decreased PGR expression by ≥ 50% in 6/14 tissues (43%), increased PGR expression by ≥ 50% in 5/14 tissues (36%), and had no significant effect on PGR in 3/14 tissues (21%) compared to the matched vehicle controls (Fig. 4A). To further investigate ERa signaling, we used ChIPseq to evaluate E2-treated breast cancer PDEs and compared whole-genome ERa binding events with primary breast cancer tissues and traditional breast cancer models, including the most commonly used ERa-positive breast cancer cell line, MCF7, grown in vitro or in vivo as xenografts. Figure 4B depicts an example ERa binding site shared by all models [retinoic acid receptor-a (RARA)], a binding site identified in patient-derived tissues only (SLCO5A1), and binding sites identified in cell line models only (50 kb to TOB1/SPAG9). These findings clearly highlight how cell line models do not accurately represent clinical breast cancer and the importance of utilizing tissuebased models for interrogation of mechanisms of carcinogenesis.
We have previously shown that PGR is not only an ERa-target gene but is also an ERa-associated protein that can reprogram ERa DNA binding and transcriptional responses in breast cancer and, importantly, used the PDE model to study the transcriptome and growth effects of this ERa reprogramming by PGR (Mohammed et al., 2015;Singhal et al., 2016). To further investigate these findings, herein we report successful ERa ChIP-seq in ERa-positive breast cancer PDEs cultured with E2 in the presence or absence of synthetic PGR agonist R5020. The number of ERa binding events in each E2-treated and E2 + R5020treated PDE is shown in Table S4. ERa binding could be mapped in all PDEs, but total peak intensity and the number of identified binding events differed between PDEs and treatments. To systematically quantify these observations, consensus ERa chromatin binding sites were compiled for each treatment by including peaks found in at least two of the three cases. Under estrogenic conditions alone, there were a total of 22 658 consensus ERa binding sites, and under estrogenic conditions with R5020 supplementation, there were a total of 34 051 consensus sites, with 18 790 shared sites between the two conditions (Fig. 4C). Further 3868 and 15 261 ERa binding sites were identified to be specific to either estrogen alone or estrogen plus R5020 treatment, respectively (Fig. 4C). Representative images of ERa chromatin binding at the RARA-positive control locus demonstrate robust peaks common to both treatment conditions (Fig. 4D, upper panel), as well as binding peaks specific to individual hormone treatments, such as those observed 10 kb upstream of the nuclear receptor coactivator 3 (NCOA3) locus (Fig. 4D, lower panel).

Discussion
Patient-derived explant tissue culture on a gelatin scaffold retains many features of human solid tumors, including the native microenvironment and cellular interactions that are critical for carcinogenesis but are lacking in many preclinical models. This study demonstrates that PDEs are not only viable in culture but can be manipulated using hormones, siRNA, or cancer drugs and that the response to those interventions can be assessed using an array of techniques, including immunohistochemistry, real-time qRT-PCR, and genomewide molecular analyses of cistromes. The PDE method is robust, as shown through observations of sustained tissue morphology, viability, and maintenance of critical cancer-related signaling pathways in independent laboratories. It is also applicable to multiple solid tumor types; to date, the methodology has successfully been applied to all cancers tested, including some not represented in the current study (ovary, endometrium, renal, sarcoma; personal communication). The de novo cellular proliferation observed in cultured tissues indicates that the system is not static and makes this a particularly useful model to assess the growth inhibitory activity of new or emerging therapeutic agents, as demonstrated by our respective teams using PDEs from prostate cancer and breast cancer (Centenera et al., 2012;Comstock et al., 2013;Dean et al., 2012;de Leeuw et al., 2015;Schiewer et al., 2012).
One of the most exciting aspects of the PDE model is the potential to uncover information about the diversity of tumor biology that is not possible using cell lines or cell line xenografts due to their clonality. Observed changes in PSA and PGR expression in prostate and breast tumors as markers of AR and ERa signaling, respectively, reflected the natural tumor heterogeneity observed clinically for each tumor type examined (Arnedos et al., 2013). Variation in PSA response to bicalutamide, the most common antagonist used in locally advanced disease (Heidenreich et al., 2014), in prostate PDEs is consistent with outcomes of the TERRAIN and STRIVE clinical trials. As an AR antagonist, bicalutamide inhibits expression of ARregulated genes such as PSA, as well as genes involved in cellular proliferation, differentiation, and survival (Furr, 1996;Maucher and von Angerer, 1993). In TERRAIN and STRIVE, a serum PSA decline of ≥ 50% from baseline was achieved in 31% and 21% of bicalutamide-treated patients, respectively (Penson et al., 2016;Shore et al., 2016). In our cohort of PDEs treated with bicalutamide, a similar rate of PSA decline was observed, with 25% of PDEs showing ≥ 50% decrease in PSA gene expression compared with vehicle treatment. PSA increase from baseline was also observed in a subset of patients in both clinical trials, which was again similar to our results where PSA increased ≥ 50% compared with vehicle treatment in 16% of cases.
Ki67 expression is a measure of cells in the active phase of the cell cycle and is a widely used marker of proliferating cells (Scholzen and Gerdes, 2000). Ki67 proliferative index is an independent predictor of prostate cancer outcomes (Fisher et al., 2013;Pollack et al., 2004;Zellweger et al., 2009); however, the wide range of Ki67 expression naturally observed in clinical prostate tumors and a lack of consensus on appropriate cutoff points have prevented utilization of Ki67 as a marker in the clinic (Penault-Llorca and Radosevic-Robin, 2017). Similarly, we observed wide variation in baseline Ki67 expression in PDE tissues, reflecting the diverse genetic heterogeneity of prostate tumors (Barbieri et al., 2013). Further, we observed a positive correlation between Ki67 immunostaining and androgen signaling in bicalutamide-treated PDEs, consistent with the effect of bicalutamide on prostate tumor PDE proliferation being AR-mediated. Despite observing significant effects of bicalutamide on PDE proliferative index, apoptotic response to bicalutamide was also evaluated, using the immunohistochemistry marker cleaved caspase-3. We have previously reported apoptosis induction in prostate cancer PDEs cultured for 48 h with other therapeutic agents (Centenera et al., 2012). It is therefore likely that higher doses of bicalutamide than the 10 lM used in this study are required to induce apoptosis in prostate cancer PDEs.
Breast cancer is highly heterogeneous, clustering into 10 different molecular subgroups based on an integrated analysis of genomic aberrations and transcriptional profiling (Curtis et al., 2012). Tumors that express ERa represent the majority (≥ 70%) of all cases (Curtis et al., 2012). Assessment of ERa and PGR status by immunohistochemistry guides treatment decisions for breast cancer, as PGR is an ERa-regulated gene and used as a biomarker of ERa activation (Lee and Gorski, 1996). We found that ERa signaling is not only sustained in breast cancer PDEs, but that the estrogenic response of ERa-positive PDEs significantly varies in terms of PGR regulation. Although an established ERa-regulated gene, PGR expression at the mRNA level can be influenced by factors that modulate the transcriptional activity of ERa (e.g., E2 metabolism, expression of transcriptional cofactors, degree of receptor phosphorylation) or regulate PGR expression independent of ERa (e.g., environmental levels of progesterone, insulin-like growth factors), reviewed in Ref. (Grimm et al., 2016). Exposure to exogenous hormones through endocrine therapy or menopausal hormone therapy is another factor that affects ERa and PGR status (Ali and Coombes, 2002). Therefore, the heterogeneity we observed is likely representative of individual tissue microenvironments, and the PDE model provides an avenue to more accurately dissect how the tumor setting influences ERa signaling. In support of this concept, we used genomewide profiling of hormonetreated breast cancer PDEs to capture ERa binding events and demonstrated reprogramming of ERa binding by synthetic progestin R5020, a phenomenon we recently reported using cell line models and clinical samples (Mohammed et al., 2015;Singhal et al., 2016).
Collectively, our data demonstrate the major advantage of the PDE model, which is the capacity for quantitative evaluation and comparison of different pharmacological agents in matched patient material. Traditionally, within-patient comparisons have only been possible between pre-and post-treatment samples from neoadjuvant studies or through utilization of diagnostic needle biopsies (Beltran et al., 2017). Obtaining this type of material is notoriously difficult without direct access to clinical trials or due to the limited amount of quality tissue available from biopsy after diagnostics are complete. The utilization of clinical material for profiling the genome, transcriptome, and proteome has remarkably advanced our understanding of the molecular features of breast and prostate cancer. The potential now offered by the PDE model is -omic analysis of matched treated and untreated samples using the PDE model. Our teams have published transcriptomic and proteomic analysis of PDE tissues (Armstrong et al., 2018;Nguyen et al., 2018;Pishas et al., 2014), and here, we demonstrate the added capacity for cistromics. Expanding the repertoire to include the lipidome, metabolome, kinome, and secretome will provide further critical insight into tumor biology and better define the mechanism of action of pharmaceutical agents preclinically, which will lead to the identification of more clinically relevant therapeutic targets and improved translation of research findings.
Preclinical models that more accurately represent human disease are needed to improve translational cancer research outcomes. The appreciation for patient-derived models in this context is increasing rapidly in parallel with organoid and PDX techniques. The major difference between explant and organoid culture is that the latter are generated from tumors that are minced and enzymatically digested to dissociate cells for seeding in Matrigel (Gao et al., 2014), whereas explant tissue is cultured in its native format without disruption to the tissue architecture or tumor microenvironment. In this way, explants are more analogous to PDX models but without the significant associated costs, timeframes, and complication of infiltrating mouse stroma (Whittle et al., 2015). Another disadvantage of PDX models is that the take rate is biased toward more aggressive tumors such that engraftment may even be indicative of poor patient outcome (DeRose et al., 2013). The poor take rate of primary tumors means that PDX lines do not necessarily represent all disease states. In contrast, tissues from benign, primary, and advanced disease stages have been successfully cultured using the PDE technique with an extremely high take rate that is dependent on the presence of epithelial cells rather than the aggressiveness of the tumor cells. The PDE method therefore enables rapid, high-throughput, and costeffective analyses of diverse human tumors and disease stages. We propose that incorporation of the PDE model into preclinical drug development programs will facilitate better selection of agents for clinical trials and provide biological insight into key molecular pathways of oncogenesis. The major criterion for implementation of the PDE model is ready access to fresh tissue samples, necessitating a strong collaboration between surgeons, pathologists, and scientists. The reliance on fresh tissue also means that PDEs cannot be passaged or revived from frozen material. What remains to be proven over the longer term is whether this approach can indeed predict in vivo clinical responses and lead to new, more effective cancer treatments or biomarkers of treatment response. Toward this goal, a recent study in non-small-cell lung cancer reported a significant correlation between cisplatin sensitivity in ex vivo cultured tissue and patient survival (Karekla et al., 2017). Future neoadjuvant clinical studies comparing pre-and post-treatment tissues to include parallel ex vivo cultures of the pretreatment tumor tissues will validate the PDE model using clinically relevant end points. This will provide an exciting opportunity to investigate novel mechanisms of treatment resistance and identify biomarkers of treatment response that are essential for the realization of personalized cancer medicine.

Conclusion
PDE culture of breast and prostate tumors is a highthroughput and cost-effective technique that retains the tissue's native architecture, microenvironment, and key oncogenic drivers. This approach allows direct within-patient comparison to rapidly evaluate efficacy of therapeutic agents in a personalized manner and is amenable to analysis using contemporary molecular technologies.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Clinical and pathological characteristics of tumors used in this study. Table S2. Summary of antibodies and conditions for immunohistochemistry. Table S3. Primer sequences/Taqman Probe Assay IDs. Table S4. Number of ER binding peaks identified by Mac Peak caller. Fig. S1. Representative images of EdU staining in prostate cancer PDEs. Fig. S2. Quantitation of EdU staining in prostate cancer PDEs.