A Simple Three-dimensional Hydrogel Platform Enables Ex Vivo Cell Culture of Patient and PDX Tumors for Assaying Their Response to Clinically Relevant Therapies Free

Ex vivo cell culture of resected tumors from patients and patient-derived xenografts (PDX) presents a unique opportunity to advance translational research and personalized medicine (1). Unfortunately, current in vitro techniques struggle to reliably grow ex vivo tissue and subsequently correlate drug responses with the clinical outcome. It has been shown that traditional two-dimensional (2D) cell culture systems may not be suitable models for investigating solid tumors and have shown inconsistencies with in vivo (2, 3). Furthermore, three-dimensional (3D) technologies such as spheroids or basement membrane extract (BME) do not provide the complete set of biomimetic microenvironments, such as appropriate scaffold architecture, or a defined, batch consistent extracellular matrix (ECM) for cell interactions, respectively (4, 5). BME, a cell culture matrix derived from a mouse sarcoma, has known issues with growth factor contamination and batch variability that may influence drug response and reproducibility (6). Spheroids in liquid suspension, that is, multi-cellular aggregates grown in hanging-drop or U-bottom plates, are cultured in a closely packed cell-only 3D environment, but lack external ECM-binding modalities which strongly impacts various aspects of tumor progression (e.g., ECM-dependent growth, invasion, and cell migration), especially important for recapitulating solid tumors (7).

Here, we analyzed the ex vivo culture of a renal cell carcinoma (RCC) PDX tumor as a proof of concept and, later, 5 patients with glioblastoma multiforme (GBM). GBM is the most prevalent and aggressive primary intracranial tumor. With an occurrence rate of 2 out of 1 × 10⁵ people, GBMs account for over 82% of malignant gliomas and 5%–20% of all primary intracranial tumors (8). Despite multimodal treatment, including radiation and chemotherapy following aggressive surgical resection, the prognosis remains poor with a median survival of 13–15 months (9, 10). Younger age, good functional status (Karnofsky Performance Score, KPS), extent of resection, IDH-1 status, and methylated O⁶-methylguanine DNA methyltransferase (MGMT) are the main predictors of a more positive prognosis (11, 12).

In this study, we sought to determine whether a batch consistent, growth factor free, 3D hydrogel culture system, VersaGel, enabled 3D ex vivo tumor growth and in vitro drug validation, compared with the liquid suspension MTT assay, MGMT promoter methylation and actual clinical responses to chemotherapy. Culturing tumor cells in 3D enables stronger physiologic relevance to in vivo due, in part, to the increased cell-to-cell and cell-to-ECM interactions (13, 14). More specifically, 3D scaffolds and ECM-mimicking hydrogels, may better simulate the native tumor microenvironment ECM and provide more accurate drug efficacy analyses (15). We propose a new scaffold-based 3D cell culture platform, VersaGel, that efficiently crosslinks to low-intensity UV (365 nm) and blue light (405 nm). This type of light-crosslinking approach been widely used in tissue engineering applications, and moreover the low intensity has been shown to have little effect of cytotoxicity (16–18). The objective of this study is to evaluate the efficacy of VersaGel in growing ex vivo patient and PDX samples, retrospectively predicting treatment efficacy and outcomes for 5 patients with GBM to temozolomide, first-line treatment in GBM, and prospectively testing known secondary therapeutics without clinical validation, as a proof of concept for personalizing therapies.

Temozolomide was purchased from Tocris Bioscience. Chloroquine, crizotinib, and sorafenib were all purchased from Selleckchem.

Tumor specimen diagnosed as GBM as defined by World Health Organization criteria, were collected for patients undergoing tumor resection. Informed consents were obtained from those patients, and the specimens were handled in accordance with Cedars-Sinai Medical Center Institutional Review Board (Los Angeles, CA). Within 3 hours of surgical resection, tumors were washed with PBS, mechanically minced with a scalpel, and enzymatically dissociated into single cells1. Tumor cells were then isolated using gradient centrifugation.

Tumor cells were cultured in neurosphere media, consisting of DMEM-based media with added recombinant bFGF (20 ng/mL), EGF (50 ng/mL), B27 supplement, heparin (160 ng/mL), and penicillin–streptomycin. Ultra-low attachment (ULA) dishes and flasks (Nunc, Thermo Fisher Scientific) were used as a culturing surface to induce cellular aggregation and neurosphere formation. Alternatively, Aggrewell-400 plates (Stemcell Technologies) could be used to grow more homogenously sized spheroids. Cells were cultured in neurosphere media for at least 3 days prior to use in VersaGel (Fig. 1).

Clinically patients were categorized as nonresponders to temozolomide if they have a progression-free survival (PFS) of less than 6 months, partial responders if their PFS was between 6 months and 1 year, and clinical responders if they had a PFS of more than 1 year.

The VersaGel solution, 15%, was warmed to 37°C for 10 minutes out of light. VersaGel was then mixed with spheroids and cell media in a 1:2 dilution for a final concentration of 5% (i.e., 200 μL cell media: 100 μL VersaGel; Fig. 1). To prevent mechanical disaggregation of the spheroid during the pipetting step, wide-bore tips and slow pipetting were employed during the process. The mixed VersaGel/cell media solution with spheroids was carefully pipetted into glass-bottom 24-well plates (Sensoplate, Greiner Bio-One) and placed on a custom low-intensity light apparatus. The light was turned on for 30 seconds to expose the light inside the wells and chemically crosslink the VersaGel solution. After the light was turned off, the plate was removed from the apparatus and wells were washed twice with prewarmed DPBS, followed by the addition of neurosphere media, and finally cultured at 37°C and 5% CO₂.

Tumor tissues resected from the PDX mouse were supplied by Stanford University (Stanford, CA) in cell media and mechanically dissociated using a razor blade in a petri dish with excess media. Small fragments were collected with a wide-bore tip and mixed with the VersaGel solution in the same manner as described, prior, with spheroids.

Tissue fragments from the above PDX tissue were further processed via mechanical and enzymatic dissociation using the gentleMACS Dissociator (Miltenyi Biotec) and Multi-Tissue Dissociation Kit (Miltenyi Biotec). Single cells were cultured on ULA plates (Nunc, Thermo Fisher Scientific) as previously noted to form spheroids. Alternatively, Aggrewell-400 plates (Stemcell Technologies) were used to produce more homogenously sized spheres.

RCC cells, spheroids, or fragments were grown in either DMEM (Corning, supplemented with 10% FBS, 1% penicillin–streptomycin) or EGM-2 media (Lonza).

VersaGel is a biocompatible, growth factor–free hydrogel for cell culture that forms interconnected 3D mesh networks when exposed to low-intensity UV (365 nm) or blue light (405 nm) for short durations (e.g., 30–60 seconds) and allows for cellular integrin binding and MMP degradation. Its light-activated chemical crosslinking mechanism provides enhanced matrix cross-linking and robustness, enabling long culture times (e.g., weeks to months). Its growth factor–free makeup is ideal for supporting serum-free conditions, for example, neurospheres, although growth factors may be added to the media before mixing with VersaGel, or in the liquid culture for molecule diffusion through VersaGel's micro-porous structure.

In addition, it was empirically determined that crosslinked VersaGel attaches to glass-bottom plates, providing an easier to use platform for media exchanges without damaging or accidentally lifting off the gel during pipetting.

Images of spheroids in each well were captured on a standard brightfield microscope and analyzed on ImageJ software. In ImageJ, the area around each sphere was traced and quantified for each timepoint and later compared with its day 0 area to determine percent (%) growth (Fig. 2A and B). This practice normalized the data irrespective of neurosphere size variability (Fig. 2C).

Brightfield images of neurospheres were captured using an EVOS Microscope (Thermo Fisher Scientific) and individual spheres were tracked over time. Spheroid size was quantified on ImageJ software by taking the sphere area. Percent growth of each sphere was calculated by quantifying its current day size (i.e., day n) and relating to its size at day 0:

In this way, each individual sphere was normalized to its initial day 0 size and average % growth could be determined for each timepoint and experimental condition.

Neurospheres in VersaGel were exposed to varying concentrations of temozolomide (0, 5, 25, 125, and 625 μmol/L) in 0.5% DMSO and neurosphere media. Drug was administered on day 0 and day 3, followed by standard neurosphere culture up to 14 days.

Cytotoxicity was qualitatively assessed using the LIVE/DEAD Assay (Thermo Fisher Scientific), calcein AM (GFP-channel of the EVOS microscope) for living cells, and ethidium homodimer (RFP-channel) for dead cells.

3D-IC₅₀ curves were created using the Prism 7 software package (GraphPad). Neurosphere size averages taken at day 10 or 14 for each experimental drug condition were used. Growth values for each temozolomide concentration were plotted in log₁₀, and as such, the control group was estimated to be 0.5 μmol/L for additional point of reference on the curve.

Upon drug treatment, GSC viability with and without concentrations of predicted drugs was quantified according to Vybrant Cell Proliferation Assay Kit instructions (Thermo Fisher Scientific). Briefly, cells were seeded at 2,000–5,000 per well in a 96-well plate format, and the conversion of MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) to insoluble formazan quantified per manufacturer's instructions after solubilization in SDS. Formazan concentration was determined by optical density at 570 nm. Cell number was determined by comparison with formazan levels in counted cell standards. Response or nonresponse determination was made by comparing test with published IC₅₀. Bottom of upper quartile of published IC₅₀s for individual drugs was used as a response/nonresponse threshold.

Similar to VersaGel, neurospheres were mixed with BME (Matrigel, Corning) in a 2:1 ratio, plated in 24-well plates and allowed to gel for 30 minutes at 37°C before adding neurosphere media. Spheroids in BME were imaged to assess 3D spheroid growth, invasion, and drug response. Within an hour after plating, spheres began disaggregating and invading the matrix with projections, complicating growth curves of the original tumor, as evidenced in Supplementary Fig. S2.

Imaging was performed on an EVOS Microscope (Thermo Fisher Scientific) and neurosphere size quantification in ImageJ software (NIH, Bethesda, MD).

Average neurosphere growth was quantified from at least three separate spheres for each experimental condition.

VersaGel enabled growth and cultivation of a RCC PDX tissue. The authors explored several ways to showcase PDX tumor growth. The tumor tissue was first resected from the mouse and mechanically dissociated into smaller pieces capable of VersaGel embedding. Alternatively, the tissue was further enzymatically dissociated into single cells and grown into spheroids as previously described with GBM samples (Fig. 3A). Spheroids were mixed with the VersaGel solution using conditioned media from the spheroid culture or with fresh media. Interestingly, VersaGel-embedded tumor fragments, and spheroids with conditioned RCC media, demonstrated invasive phenotypes, while VersaGel-embedded spheroids with fresh media showcased no such invasion, and instead grew as closely packed spheroids (Fig. 3B). Secreted proteins in the media or original tumor tissue may have contributed to the invasion, although more experimental evidence is required to determine the exact cause. It can be appreciated, however, that this invasive phenotype was only possible in a scaffold-embedded system, and that conditioned media could modulate the invasiveness in growth factor–free VersaGel.

In practice, VersaGel can plausibly work for any spheroid culture. Spheroids from other human cancer cell lines, HepG2 and LN18, were embedded in VersaGel and imaged over time (Supplementary Fig. S1). Interestingly, HepG2 spheroids grew as a tightly packed spheroid, while LN18 spheres grew and invaded VersaGel by day 22.

Five patients with GBM were included in this part of the study. Overall, median age of the cohort was 68 years at time of surgery, and 3 were female. All patients underwent standard of care GBM resection followed by temozolomide and radiation. Four out of five patient's tissue samples were obtained during their initial surgery. One patient sample was from a second repeat surgery following GBM recurrence after failure of temozolomide and radiation. Cell cultures via PCR indicated 2 of 5 patients had MGMT promoter methylation. Overall, median KPS of the patient cohort was 80.

Patient neurospheres embedded in VersaGel were exposed to temozolomide of varying concentrations (0, 5, 25, 125, and 625 μmol/L) on day 0 and day 3 and allowed to grow through day 14. Previous research has indicated this range as acceptable for GBM cell lines (e.g., U87), although interestingly the temozolomide C_max value, the maximum plasma concentration in the human, has been reported to be approximately 25 μmol/L (19, 20). Image analysis revealed varying growth patterns and 3D-IC₅₀ values in patients 1–5, correlating with the degree of clinical response and time to tumor progression (Table 1). Patient 1 and 2 were considered clinical nonresponders with a time to progression below 6 months. In VersaGel, these patients exhibited no statistical difference between the control groups and all temozolomide concentrations except 625 μmol/L, with 3D-IC₅₀ values > 125 μmol/L (Fig. 4A). Patient 3 had already failed temozolomide and radiation with a PFS of 5.8 months. Patient 3 tissue sample was obtained from the second surgery and hence, clinically Patient 3 was a nonresponder. In VersaGel, this tumor exhibited no statistical difference between the control groups and all temozolomide concentrations except 625 μmol/L, with a 3D-IC₅₀ value of 128 μmol/L. Patient 4 was clinically considered a partial responder with time to progression above 6 months and below 1 year (8.7 months). In VersaGel, this patients' tumor exhibited growth patterns that statistically varied in the higher 125 μmol/L and 625 μmol/L temozolomide groups but showed no statistical difference with lower temozolomide groups (Fig. 4B) and a 3D-IC₅₀ value of 49.8 μmol/L. Hence, patient 4 was classified as a partial responder per VersaGel 3D culture testing. Patient 5, the sole responder from a clinical standpoint with PFS of 16 plus months, demonstrated statically slower growth in all temozolomide concentration groups compared with the control (Fig. 4C), with a 3D-IC₅₀ value of 0.16 μmol/L. Zero growth was reported in the 625 μmol/L temozolomide group in all patients, however cells remained viable, suggesting the cells underwent senescence at this upper limit of the experimental design (Fig. 4D). Patient 4 showed zero growth in the higher concentrations of temozolomide (125 and 625 μmol/L) during treatment through day 7. After switching to neurosphere media for the latter 7 days of culture, the neurospheres in the 125 μmol/L-treated group began to grow again, suggesting an incomplete response at that concentration (Fig. 4E). Interestingly, this patient was a partial responder with time to progression of 8.7 months (compared with nonresponders of <6 month recurrence), and patient 5 was the only patient tumor to display an invasive phenotype in the control compared with temozolomide-treated group (Fig. 4F), while in temozolomide, the invasiveness was limited completely, suggesting that treatment was effective in vitro, and interestingly, the patient has been progression free for more than 16 months.

MTT assay results on cells cultured in liquid neurosphere media correctly correlated to the clinical outcome for only 3 out of the 5 patients (patients 1, 2, and 3, the nonresponders; Table 1; Fig. 4). Indeed, all patients (1–5) were universally resistant to temozolomide under these conditions, with none reaching IC₅₀. This low correlation may be due to phenotypic changes associated with growth of suspended cells or neurospheres in liquid culture as opposed to an ECM-mimicking scaffold-like VersaGel, to intrinsic limitations of the MTT assay itself, or to both.

BME likewise failed to report the correct response for patient 4 (the partial responder) and patient 5 (the responder; Supplementary Fig. S2). BME's rapid digestion by the tumor cells made measurements of individual neurosphere growth difficult to quantify. More importantly, as an example, neurospheres from patients 4 and 5 showed invasive phenotypes in both 125 μmol/L temozolomide and control groups in Matrigel, which opposed the clinical outcome, which may be due to the batch variability and growth factor–enriched nature of BME.

Nuerospheres derived from patient 3 (Table 1) were grown in VersaGel and subjected to varying concentrations of sorafenib, crizotinib, and chloroquine (Fig. 5). Sorafenib is a kinase inhibitor that affects tumor proliferation and is currently subject to phase I clinical trials as a combination therapy with temozolomide and radiation (21). Crizotinib is an ALK and cROS-1 inhibitor that is FDA approved to treat non–small lung cancer and is currently being explored as GBM treatment (22). Finally, chloroquine is a medication used to treat for malaria and is currently undergoing phase III trials for GBM treatment (23). The use of these therapies demonstrates the desire and willingness by the clinical community to repurpose existing drugs (24). Neurospheres in VersaGel were dosed appropriately with each drug and a 3D-IC₅₀ value based on sphere growth was determined similar to previous experiments. Interestingly, sorafenib's 3D-IC₅₀ was 27.11 μmol/L, exceedingly higher than the reported C_max value of 6.7 μmol/L. Crizotinb's 3D-IC₅₀ was 0.49 μmol/L with a reported C_max of 0.93 μmol/L (25). And, chloroquine's 3D-IC₅₀ was 4.71 μmol/L compared with a C_max of 4.8 μmol/L.

In translational and clinical cancer research, there is need for a simple to use, rapid in vitro system with the ability to grow ex vivo to tissue and correlate the clinical outcome. Current cell culture technologies have limited capacity to recapitulate the microenvironment, such as plastic plates, liquid suspension cultures of neurospheres, or growth factor–enriched BME. Spheroids in liquid culture also fail to provide a suitable biomimetic environment in which cells may attach, grow, and invade (all hallmarks of tumor progression), and additional assays like CellTiterGlo are required to assess drug response.

VersaGel provides a scaffold-based platform that recapitulates ECM cues for the tumor microenvironment in a growth factor–free manner. To demonstrate its feasibility with ex vivo tissue, a RCC PDX tumor directly following resection was cultured and grown in VersaGel. Interestingly, there was phenotypic variance between different preculture conditions. When tumor fragments (directly from resection) or spheroids that were grown following tissue dissociation and precultivation on ULA plates, were mixed with VersaGel and conditioned media, that is, media which came from the ULA plates, the tumor aggregates or spheroids in VersaGel displayed an invasive phenotype. Spheroids that were first washed with fresh media and then embedded in VersaGel grew as spherical shape with no invasive protrusions. This phenotypic alteration suggests that the secreted proteins in the conditioned media or original tumor tissue are critical for maintaining the invasive kidney carcinoma phenotype and could be an important consideration in future research with PDX models.

Clinically, temozolomide response is predicted by MGMT methylation status. Unfortunately MGMT does not correctly predict response to temozolomide in about 30% of cases (12). Therefore, cell assays offer a promising tool to verify temozolomide response prior to treatment; however technologies to date have failed to correlate response. In our own testing, liquid suspension cell culture using the MTT assay predicted the correct temozolomide response in only 3 of 5 patients, and neurospheres in BME were hindered by rapid invasion into the matrix surroundings, irrespective of temozolomide culture. It has become increasingly accepted that unsupported liquid cell culture has limited capacity in recapitulating the tumor microenvironment, and thus confers phenotypic and genetic alterations that affect drug response (2–5). BME contains varying amounts of growth factors that influence cell behavior, suggesting it may be a less reliable platform in this instance. As evidenced in patient 5, a clinical responder to temozolomide, the neurospheres in BME failed to demonstrate any response and instead reported an invasive phenotype similar to the control group.

VersaGel, on the other hand, correctly demonstrated response to temozolomide in all 5 patients, offering insight into the patient response in a rapid, in vitro setting. Interestingly, patient response in VersaGel, as noted by the degree of the reported 3D-IC₅₀, correlated to the degree of response and time to progression in the clinical setting. As a proof of concept, VersaGel was utilized with 1 patient with GBM (patient 3) in the prospective testing of several compounds. It is interesting to note that IC₅₀ values could be determined for all. While the data are early, the utility of repurposed drugs suggests a shifting trend with clinicians looking to find off-indication treatments for individual patients, especially after first-line treatment failure (24).

Although our tests were performed in standard 24-well plates, the platform is amenable to higher throughput and to high content analysis, making VersaGel a potential candidate for oncology drug screening and personalized diagnostics. This study provides a stepwise protocol for groups to explore various PDX and patient tumor types in VersaGel. Further patient testing should be considered before the assay is considered clinically validated and used in the clinical decision making process. The appropriate laboratory or university with access to patient samples may find this type of testing compelling when building a database of retrospective and prospective therapeutic responders. It is the hope of the authors that a synergy between 3D ex vivo tumor testing in VersaGel and other robust technologies like next-generation sequencing may assist translational oncology researchers and clinical decision makers to advance personalized medicine for oncology patients.

VersaGel enabled ex vivo growth of five patient tumors' and one PDX tumor. In all 5 patients, the assay correctly correlated the clinical outcome to temozolomide, showing that the 3D-IC₅₀ values related to time of progression in these patients. Interestingly, for prospective testing of patient 3, 3D-IC₅₀'s were lower than C_max for crizotinib and chloroquine, but higher for sorafenib. Expanded studies, both retrospective and prospective, will be necessary to broaden clinical validation for advancing personalized medicine and clinician decision making, however this tool could be readily explored for many cancers in oncology drug screening and translational research.

K.C. Hribar has ownership interest (including stock, patents, etc.) in Cypre Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.C. Hribar, C.J. Wheeler, A. Bazarov, K. Varshneya, R. Yamada, C.G. Patil

Development of methodology: K.C. Hribar, C.J. Wheeler, P. Buckley, C.G. Patil

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.C. Hribar, C.J. Wheeler, A. Bazarov, K. Varshneya, R. Yamada, P. Buckley

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.C. Hribar, C.J. Wheeler, A. Bazarov, K. Varshneya, R. Yamada, P. Buckley

Writing, review, and/or revision of the manuscript: K.C. Hribar, C.J. Wheeler, K. Varshneya

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.C. Hribar, C.J. Wheeler, R. Yamada, P. Buckley

Study supervision: K.C. Hribar, C.J. Wheeler, C.G. Patil

Other (all work on nonmatrix-cultured cells was performed under his direct supervision in his former laboratory at Cedars-Sinai Medical Center, in collaboration with C.G. Patil and K.C. Hribar): C.J. Wheeler

The authors graciously thank funding from the Precision Medicine Initiative for Brain Cancer Donor Grant (to C.G. Patil). We also thank Professor Donna Peehl, PhD, and her lab staff of Stanford University for access to renal cell carcinoma PDX tissue and Robert Bell, PhD, at Telo Therapeutics for access to their HepG2 and LN18 human cancer cell lines.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.