This protocol demonstrates methods to enable extended in vitro culture of patient-derived xenografts (PDX). One step enhances overall viability of multicellular cluster cultures in 3D hydrogels, through straightforward removal of non-viable single cells. A secondary step demonstrates best practices for PDX culture in a perfused microfluidic platform.
Patient-derived xenografts (PDX), generated when resected patient tumor tissue is engrafted directly into immunocompromised mice, remain biologically stable, thereby preserving molecular, genetic, and histological features, as well as heterogeneity of the original tumor. However, using these models to perform a multitude of experiments, including drug screening, is prohibitive both in terms of cost and time. Three-dimensional (3D) culture systems are widely viewed as platforms in which cancer cells retain their biological integrity through biochemical interactions, morphology, and architecture. Our team has extensive experience culturing PDX cells in vitro using 3D matrices composed of hyaluronic acid (HA). In order to separate mouse fibroblast stromal cells associated with PDXs, we use rotation culture, where stromal cells adhere to the surface of tissue culture-treated plates while dissociated PDX tumor cells float and self-associate into multicellular clusters. Also floating in the supernatant are single, often dead cells, which present a challenge in collecting viable PDX clusters for downstream encapsulation into hydrogels for 3D cell culture. In order to separate these single cells from live cell clusters, we have employed density step gradient centrifugation. The protocol described here allows for the depletion of non-viable single cells from the healthy population of cell clusters that will be used for further in vitro experimentation. In our studies, we incorporate the 3D cultures in microfluidic plates which allow for media perfusion during culture. After assessing the resultant cultures using a fluorescent image-based viability assay of purified versus non-purified cells, our results show that this additional separation step substantially reduced the number of non-viable cells from our cultures.
Over the past decade, the field of cancer research has demonstrated renewed enthusiasm for patient-derived xenografts (PDXs) as a tool for assessing cancer cell pathway reliance and drug susceptibility1. The most common PDX models are established by subcutaneous or orthotopic implantation of human tumor cells—a tumor fragment, a cluster of dissociated tumor-derived cells, or a sample of isolated circulating tumor cells (CTCs)—into a rodent host. If the tumor “take” is successful, the xenograft cells will proliferate, vascularize, and otherwise interact with the host tissue to create a tumor, which can be harvested at an optimal size, subdivided, and re-implanted into other hosts. Among their many advantages as a model system, PDXs typically retain a substantial portion of the native tumor cell population’s heterogeneity and enable the assessment of human-specific pathways and cell responses2,3. The in vivo context enables tumor interaction with vasculature and other adjacent stroma and recapitulates tissue characteristics such as drug diffusion dynamics, oxygen tension, and extracellular matrix influence that biologically and mechanically impact tumor progression. A negative aspect of PDXs is their reliance on a rodent host, both for tumor expansion and ultimately for hypothesis testing. Because many PDXs cannot adapt to traditional two-dimensional (2D) culture on tissue culture polystyrene without losing many of their desirable characteristics, there has been minimal middle ground for researchers between this relatively controlled in vitro method, and the significant increase in expense, facilities, and time requirements for in vivo PDX use.
We have described multiple in vitro models that implement 3D cell culture within a supportive matrix, and recently expanded that work to demonstrate the ex vivo culture of multiple prostate cancer (PCa)-derived PDXs, both alone and in co-culture with bone marrow-derived fibroblasts4,5. Hyaluronic acid (HA)-based hydrogel matrices provide customizable and biologically-relevant support for both cell types, with facile control over hydrogel characteristics and optical clarity for imaging through the hydrogel depth6.
Mature PDX tumor tissues comprise a variable mixture of heterogeneous human cancer cells and mouse stroma (fibroblasts, endothelial cells, etc.). To study cell-type specific contributions to tumor progression in vitro, it can be advantageous to dissociate tumors, separate the cell populations, and experimentally incorporate them in an organized manner to dissect pathways of intercellular communication. The mixed cell populations within tissue digestates have differential compatibility with specific culture conditions. For example, tumor-associated fibroblast viability necessitates either surface adherence or 3D matrices functionalized with integrin ligands, while epithelial-derived PDX cells do not typically have these requirements, instead favoring cell-cell interactions. These differences can be exploited to achieve effective separation of PDX cells from contaminating mouse stromal cells. Rotation culture of tissue digestates allows stromal cell adherence to the tissue culture surface while cell-cell adhesions drive PDX cells floating above the rotating culture surface to form multicellular clusters in the supernatant in 24−48 h. The specific characteristics of these clusters vary with the PDX (e.g., large, tight, highly spherical clusters or smaller, looser aggregates resembling bunches of grapes), but are typically of biologically relevant sizes (50−250 µm diameter), sufficient for assessing cellular interactions that rely on intercellular contacts.
Tumor retrieval and processing inevitably results in some degree of collateral cell death, either due to short-term damage from mechanical/enzymatic disruption, or long-term incompatibility of subpopulations with the chosen culture conditions. Despite the utility of rotation culture as an initial bulk separation, dead or dying cells are inevitably transferred with the PDX clusters and can influence the resultant culture. These dead cells are often individual PDX cells that were not integrated into a cluster, mouse stromal fibroblasts that cannot survive in selected culture conditions, or particularly fragile endothelial cells. Such dying cells can influence experimental results from “survivors” and can substantially impact quantification, e.g., via fluorescent image-based viability screening assays. To improve the selection of live PDX cells from this method, we adapted centrifugation methods with density steps to easily remove individual dead/dying cells from PDX mixtures and retain predominantly live multicellular clusters.
To enhance the study of resultant PDX-derived clusters in 3D culture, we utilized a microfluidics-based perfusion culture platform, the OrganoPlate (Figure 1), which is a high-throughput organ-on-a-chip platform that allows for simultaneous culture of up to 96 individual perfused, 3D cultures on a 384-well microtiter plate-base (Figure 1A)7,8. In the 2-lane microfluidic plate, a single tissue chip is connected by two microfluidic channels (Figure 1B, gel channel: red, perfusion channel: blue) which span four wells in a row. The two microfluidic channels are separated by a short plastic ridge called a Phaseguide which prevents overflow of one channel into its adjacent neighbor channel, and simultaneously allows for a membrane-free interface between the contents of the gel and perfusion channel9. Because the bottom of the microfluidic plate is composed of microscope-grade glass, the cultures can be viewed in the observation window through the bottom of the plate with a standard or automated microscope. Perfusion is established in the microfluidic plate with a programmable rocker, using gravity to drive media through the microfluidic channels, between reservoir wells (Figure 1C). The perfusion flow-mimic more closely recapitulates the tumor microenvironment than static culture, allowing for the incorporation of shear stress and enhanced distribution of gases and nutrients. The benefits of maintaining a perfused cancer cell culture in the microfluidic plate have previously been described as perfused breast cancer cultures exhibited optimal viability as compared to a static 3D culture of the same cells7.
The present report describes an adapted density gradient centrifugation method for isolating live multicellular PDX clusters and demonstrates its utility in establishing 3D PDX cultures within perfusable microfluidic plates. Because an increasing number of research laboratories are seeking methods to facilitate PDX use, we anticipate that the protocols presented here will be of immediate utility.
Tumor tissue was obtained with patient consent and according to an approved Institutional Review Board (IRB) protocol. Xenografts were implanted, grown, and harvested according to an accepted Institutional Animal Care and Use Committee (IACUC) protocol.
NOTE: All work is to be performed in a sterile biological safety cabinet to maintain sterility. All steps should be conducted at room temperature unless otherwise specified.
1. Preparation of materials for PDX processing
2. PDX dissociation and initial purification of stromal component
3. Density gradient centrifugation-based separation of PDX-derived clusters from single cells
4. Hydrogel preparation and microfluidic plate seeding
5. Cell staining, imaging, and image quantification
A programmable perfusion rocker was prepared in a standard water-jacked cell culture incubator, and two-lane microfluidic plates were prepared in a standard biosafety cabinet for loading (Figure 1). An MDA-PCA-118b PDX tumor was expanded in vivo, harvested when it had reached a maximum size, and dissociated as described in protocol section 2 to create a slurry suspension of cells, at approximately a single-cell state (Figure 2A). The slurry was dispensed into 6-well tissue culture plates, and placed on an XY rotator as described, for 48 h. Mouse stroma attached to the bottom of the tissue culture plate, as has been previously described, and the human PDX cells remained floating in the medium, assembled into clusters (Figure 2B). Unassembled single cells were visible throughout the solution as well.
A step density gradient was established in a centrifuge tube using the methods in protocol section 3. The supernatant suspension of PDX cells was aspirated gently from the multiwell plate, loaded onto the step gradient, and centrifuged as described. After centrifugation, a hazy band, containing the PDX clusters, was visible near the interface between fractions 2 and 3, and single cells were identified at the interface between fractions 1 and 2 (Figure 2D). Fractions were collected as described, diluted further in HBSS, and centrifuged again to remove the density gradient solution. The supernatant was aspirated, and the resultant cell pellets were resuspended in PDX processing medium. A small aliquot of each processed fraction was assessed by microscope, to confirm that the expected fraction contained PDX cluster, and that the separate fraction contained primarily single cells.
HA hydrogel solutions were reconstituted as described in protocol section 4, and PDX clusters were resuspended in the hydrogel solution. PDX solutions were loaded into chips on the microfluidic plate and assessed for successful loading (Figure 3). In most cases, >90% of chips were loaded successfully after some practice with the technique and adjustment of loading volumes. After loading the desired number of chips, and incubating for gelation as described, PDX culture medium was added to the microfluidic plate, and the plate was cultured with perfusion.
Using fluorescent dyes and the methods in protocol section 5, and confocal microscopy, 3D microfluidic PDX culture viability and morphology were evaluated in both unseparated and density gradient centrifugation separated conditions (Figure 4A). Images of these plates were recorded as Z-stacks on the confocal microscope and assessed for cell viability in image analysis software. On day 1, those cultures which underwent the separation method exhibited 10-fold fewer single, dead cells (red), compared to unseparated cultures (Figure 4B). Importantly, the separated clusters primarily consisted of live cells (green). No statistically significant difference was identified for the cluster size distribution (Figure 4C).
Cultures were further maintained in the microfluidic plate for seven days (Figure 5A). Samples were assessed periodically, using the same imaging and quantification methods as above. The total number of live cells remained consistent (Figure 5B) and clusters retained ~80% viability (Figure 5C) over the life of the culture. The cell density in the unseparated condition is roughly one-third of the separated condition because single cells are alive during cell counting, but quickly die within the hydrogel. There is also more variability in cluster size without the density gradient separation.
Figure 1: The 2-lane perfusable microfluidic plate. (A) The 2-lane microfluidic plate is a standard 384-well microtiter plate with a modified bottom consisting of microfluidic channels embedded between glass plates. Each plate consists of 96 tissue chips for 3D cell culture. (B) As viewed from the top of the plate, a single 2-lane microfluidic chip is made up of 4 wells in a row connected by the gel channel (red) and perfusion channel (blue); the gel inlet, perfusion inlet, observation window, and perfusion outlet. (C) Perfusion is achieved in the microfluidic plate with a programmable rocker which uses gravity to drive media between perfusion inlet and outlet wells which are media reservoirs. Please click here to view a larger version of this figure.
Figure 2: PDX cluster isolation. (A) Dissect primary PDX tissue and dissociate into a single cell suspension containing both human PCa cells and mouse tumor stroma. Either cryopreserve dissociated cell mixture or (B) add to rotation culture to recluster tumor cells and allow mouse stromal cells to adhere to the tissue culture plate (48 h). (C) Use density gradient centrifugation to remove residual dead single cells, and (D) collect viable PDX clusters at the interface between density gradient layers 40% and 55% (red dashed box). (E) Recover PDX clusters, resuspend in hydrogel precursor solution, and dispense onto the plate with a repeating pipette. (F) Example calculations, corresponding to Table 1, demonstrate the necessary initial cell numbers required to reach a final desired cell concentration for all chips across a plate. Please click here to view a larger version of this figure.
Figure 3: Monitoring microfluidic plate loading. Loading success and errors, can be assessed by eye (in top and bottom views), with confirmation by microscope. (A) Successful loading is identified by an open (bright) perfusion lane in the top and bottom views, and a slightly darker gel lane in bottom view. Visualization by microscope confirms that the gel lane was filled completely with cells and gel precursor, without spilling over into the adjacent lane. (B) Cartoon demonstrating correct pipet tip placement during gel dispensing, with correct placement being just above the inlet port (left) and incorrect placement being off-centered (middle) or applying force to the inlet port (right). (C) Bright lanes in all views indicate a missed loading, suggesting that the pipet tip was not correctly placed within the gel inlet. (D) A partial fill of the gel lane (not filled to end) is identified by the red arrow, which shows where the advancing gel solution front was interrupted, either by a trapped air bubble, or a premature pause in loading. The red arrow in the microscope view identifies the same location. (E) The contents of the gel lane overflowed the PhaseGuide, spilling into the perfusion lane and ultimately blocking it. This overflow is visible in bottom view by the dark appearance of both gel and perfusion lanes. Scale bar = 200 µm. Please click here to view a larger version of this figure.
Figure 4: Cell staining and viability assessment in the microfluidic plate. (A) Using fluorescent stains, viability was assessed in unseparated and separated PCa cell clusters seeded in the microfluidic plate (all nuclei: Hoechst 33342, blue; dead cells: ethidium homodimer-1, red; and live cells: calcein AM, green). Scale bar = 50 µm. (B) Total dead cells were quantitated on day 1 of the culture and separated MDA-PCa-118b PDX cultures demonstrated a significant decrease in the number of dead cells. (C) Quantification of cluster size for PCa PDXs, in separated and unseparated cultures, showed a slight increase but no statistically significant difference. Bars and error bars in panel B represent the mean and standard deviation of 4 images per condition (2 chips with 2 image/chip). Asterisk (*) represents statistically significant differences (p < 0.05) compared to the unseparated condition using a student’s t-test. For the box and whisker plot in panel C, the cross icon indicates the mean, and the horizontal line represents the median, of 4 images per condition (2 chips with 2 image/chip). The box contains 50% of the data while the whiskers extend to the minimum and maximum cluster diameter for each condition. Please click here to view a larger version of this figure.
Figure 5: Characterization of separated MDA-PCa-118b in the microfluidic plate. (A) Separated (left) and unseparated (right) PCa cancer cell clusters were seeded in the microfluidic plate and perfused for up to seven days. Cultures were stained with three dyes specific for all nuclei (Hoechst 33342, blue), dead cells (ethidium homodimer-1, red), and live cells (calcein AM, green). Scale bar = 50 µm. (B,C) Quantitation of number of cells and culture viability from images over one week of culture at a seeding density of 8,000 cells/µL. Bars and error bars represent the mean and standard deviation of four images per condition. Please click here to view a larger version of this figure.
Final desired cell density in single chip (cells/µL) | Volume loaded per chip (µL) | Number of chips on microfluidic plate | Mutiplier for extra volume | Figure 2C: Number of separated cells required for a full microwell plate (step 3.8) | Multiplier for cell losses (stroma removal, cell death) |
Figure 2B: Starting # of dissociated PDX cells for protocol (step 2.8) |
2,500 | 1.5 | 96 | 1.3 | 4.7E+05 | 3 | 1.4E+06 |
5,000 | 1.5 | 96 | 1.3 | 9.4E+05 | 3 | 2.8E+06 |
10,000 | 1.5 | 96 | 1.3 | 1.9E+06 | 3 | 5.7E+06 |
20,000 | 1.5 | 96 | 1.3 | 3.7E+06 | 3 | 1.1E+07 |
Table 1: Approximate initial and final PDX material required to load a single 2-lane microfluidic plate.
Supplemental Figure 1: The 2-lane microfluidic plate layout. After seeding the microfluidic plate, record individual chip loading success (successful, missed loading, not filled to end, or overflow) using the 2-lane plate layout. Please click here to view a larger version of this figure.
Here we describe a method for processing and culturing viable PDX-derived tumor cells in a high-throughput, perfused 3D culture system. While this protocol utilizes PCa PDX tissue, it is equally effective for other epithelial-derived tumors. Tumor characteristics vary between individual PDX lines even within the same tissue of origin (prostate, breast, etc.). Some PCa PDX lines are more fibrotic and difficult to isolate viable cells from while others are more cellular. The tumor size noted here can be varied within IACUC guidelines to provide more tissue for particularly low-yield tumors. Additionally, tumor fragments can be collected after step 2.7, re-digested, and pooled with the initial digestion to increase yield. This is especially helpful for more extracellular matrix-heavy tumors. Additional rounds of digestion past two typically result in low viability and low yield and are not recommended.
The purification of PDX-derived populations, to remove dead/dying cells or contaminating host cells, is beneficial for the extended 3D culture and quantification of the desired cancer cells. The first step of the purification described here—suspension culture in a relatively large volume of media along with XY rotation at moderate speeds—promotes 1) the selective extraction of highly adhesion-dependent cells, typically mouse stromal fibroblasts, and 2) the formation of cell aggregates to supply pro-survival cell-cell contact. The efficiency of extracting host-derived mouse fibroblasts during the 48 h XY rotation step is very high, as described in Fong et al.5. Intentional PCa-fibroblast co-cultures are certainly feasible, as we have demonstrated, but require a tuned matrix (to support fibroblast adhesion to matrix) and a larger proportion of fibroblasts. The occasional PDX, usually those with more mesenchymal characteristics, will adhere more readily to tissue culture surfaces during the 48-hour rotation culture step. PDX lines should be closely monitored the first time this protocol is performed, and immunostained for HNA, to identify the proportion of human cells in the adherent and non-adherent populations. Non-tissue culture treated plates should be used for cluster formation with these PDXs. Mouse stromal cells will still adhere eventually to the surface, but separation will not be as efficient.
Following rotation culture, the supernatant containing the non-adherent population is processed through the described density gradient centrifugation method. The protocol reported here can be simplified to exclude at least one density step (e.g., using only the 20%, 40%, and 55% solutions), but all four should be used when establishing this method with each new PDX. Multicellular clusters are easily separated from individual cells by this method, and largely retain their clustered phenotype and other characteristics. The discarded single-cell population represents either (a) cells that were dead/dying prior to the 48 h rotation step, and thus never integrated into a cluster, or (b) cells that remained alive after the 48 h rotation, but still did not integrate into clusters. It is important to note that group (b) may include cells that some investigators find desirable; e.g., some reports have described primary cell cultures containing early stem/progenitor cells, or drug-resistant cells, which float within supernatant as poorly-adherent single cells above an otherwise adherent population10. Relevant to cancer studies, circulating tumor cells (CTCs), tumor initiating cells (TICs), or other similar metastasis-promoting cell types could be part of this single cell population. Therefore, investigators using our protocol should analyze carefully the discarded single cell population to ensure that any cells of interest have not been lost. In our hands, the vast majority of these single cells are either a small fraction of cancer cells that are dead/dying due to the initial tissue dissociation protocol, or rodent fibroblasts that are already proceeding through anoikis.
It is strongly recommended to practice the technique of seeding the microfluidic plate with less precious cells to ensure success when working with limited and valuable PDX material. Be aware of tip placement during dispensing, as improper technique is likely to result in 1) overflow if pressure is applied during loading, or 2) empty or partially filled channels if the tip is not centered over the inlet port. Premature gelation will also cause channels to not fill to the end. If this occurs, decrease the wait time between addition of the PEGDA crosslinker and dispensing the gel solution, or work with smaller aliquots of gel solution at a time. The dispensing volumes used in this protocol are well-optimized for success with HA hydrogels. Use of more viscous solutions such as Matrigel in this system requires an increase in the dispensing volume to ~2 µL for proper filling of the microfluidic channels. Note also that the choice of 50 µL increments in step 4.3, the preparation of four cell pellets in this same step, and the repeat resuspensions in step 4.10 are deliberate; although they yield more material than necessary, they also provide overage to account for losses from the repeating pipette.
It should be noted that each PDX is likely to have a unique morphology and size distribution within this 3D culture model. In one sense, this is expected, due to the diversity of patient disease, tissue history, matrix parameters, and many other factors. An expansive assessment of multiple prostate cancer lines in Matrigel demonstrated this potential diversity11. Still, investigators may be surprised to find a wide variation in the above parameters. In a manuscript in preparation, we present a broad look at several PDXs on this culture platform; the cells maintain a consistent response within each PDX type, but can vary substantially across specimens, resulting in clusters that are, for example, large with high cell density per cluster, or smaller, with looser intercellular connections. The specific behavior of each PDX should be assessed by investigators over multiple trials, to confirm a predictable and characteristic morphology. Investigators can expect specimens to follow the general behaviors outlined in this paper.
In summary, the presented protocol offers researchers a new method for employing PDXs ex vivo in a platform that allows for the fine tuning of culture conditions and incorporation of relevant 3D environmental cues such as tissue specific extracellular matrix (ECM) and perfusion flow, enabling extended cell survival over several days or weeks. Detailed characterization of these cultures was achieved by the staining and morphological analyses demonstrated here. In addition, if needed, cultures can be submitted to plate reader-based assays, immunofluorescent labeling, or used for the preparation of lysates enabling further molecular characterization. Although we have published previously some elements of PDX 3D culture within hydrogels, this application of multistep purification is new, easy to employ in a standard laboratory, and successful in retaining high-viability representative cultures. The OrganoPlate platform, in its 2-lane and 3-lane varieties, offers further complexity through its model of tissue perfusion, and a key opportunity to use even fewer cells per experiment. Compared to our previous models, the microfluidic platform reduced cell requirements by a factor of ~30, per experiment, which is an enormous benefit, given the scarce availability of PDX tumor tissue in most laboratories. The microfluidic system enables extended imaging, immunostaining, and facile imaging, across a cell population that is large enough (n = ~2,000−4,000 cells per imaging window) to permit quantification of phenotype within the context of spatial organization.
The authors have nothing to disclose.
This work was supported by National Institutes of Health National Cancer Institute SBIR Phase I (HHSN26120700015C) and P01CA098912.
1N NaOH | any suitable tissue culture grade | ||
60 mm round tissue culture dishes | any suitable | ||
6-well tissue culture plates | any suitable | ||
70 µm cell strainers | Corning | 431751 | or equivalent |
Centrifuge | Eppendorf | 5810R with suitable rotor and buckets for 15/50 mL conical centrifuge tubes | or equivalent |
Density gradient centrifugation solution | Millipore Sigma | P1644 | Percoll |
Dimethylsulfoxide | any suitable tissue culture grade | ||
Dissociation enzyme solution | StemCell Technologies | 07921 | ACCUMAX |
DMEM-F12 | ThermoFisher Scientific | 11039021 | or equivalent |
Forceps | any suitable | ||
HA hydrogel kit | ESI BIO | GS311 | HyStem (Hyaluronic acid-SH and PEGDA) |
Hanks Balanced Salt Solution | Lonza | 10-527F | or equivalent |
Heat-inactivated fetal bovine serum | Atlanta Biologicals | S11150 | |
Hemocytometer | Fisher Scientific | 02-671-51B Hausser BrightLine | or equivalent |
Hoechst 33342 | ThermoFisher Scientific | H1398 | or equivalent |
Image processing software | Oxford Instruments | Imaris 9.3 | or equivalent |
LIVE/DEAD Cell Viability/Cytotoxicity Kit (Calcein-AM/Ethidium Homodimer-1) | ThermoFisher Scientific | L3224 | or equivalent |
Microfluidic culture plate | Mimetas | 9603-400-B | 2-lane OrganoPlate |
Microscope | Nikon | A1R | or equivalent |
Multichannel pipette | Eppendorf | 3125000036 | or equivalent |
PDX-derived tumor tissue | obtained under IRB approval for human tissue and IACUC approval for animal host | ||
Penicillin-streptomycin | ThermoFisher Scientific | 15140-122 | or equivalent |
Perfusion rocker | Mimetas | OrganoPlate Perfusion Rocker Mini | |
pH strips (pH 5-9) | any suitable | ||
Phosphate-buffered saline solution | Lonza | 17-516F | or equivalent |
Razor blades | any suitable | ||
Rotating xy-shaker | VWR | Advanced 3500 Orbital Shaker | or equivalent |
Scalpel handle | any suitable | ||
Single channel repeating pipette | Eppendorf | 22260201 | |
Sterile, 15mL conical centrifuge tubes | any suitable | ||
Sterile, 50mL conical centrifuge tubes | any suitable |