Characterization of a New in Vivo Hollow Fiber Model for the Study of Progression of Prostate Cancer to Androgen Independence ¹

Prostate cancer is the second most prevalent cause of death from cancer in American males (1). Localized prostate cancer can be treated surgically or by radiotherapy. The only effective systemic therapy available for metastatic prostate cancer is androgen deprivation. The inability of androgen deprivation to completely and permanently eliminate all prostate cancer cell populations is manifested by the predictable pattern of initial response and relapse with the ultimate progression to androgen independence. Androgen deprivation is associated with a gradual transition of prostate cancer cells through a spectrum of androgen dependence, androgen sensitivity, and ultimately androgen independence.

The tissue-specific marker that is used to monitor treatment responses, prognosis, and progression to androgen independence in patients with prostate cancer is PSA.

Numerous animal models have been developed to investigate the mechanisms underlying the development and pathogenesis of the prostate that include hormone-induced carcinogenesis models, transgenic and reconstitution models, and xenograft models. However, to date no ideal model for prostate cancer exists. This is because the ideal tumor model would have to mimic the clinical presentation, i.e., be of human origin; have a slow doubling time; initially be androgen dependent or respond to androgen; secrete PSA; metastasize to lymph nodes and bone; and hormonally progress to androgen independence after castration (5). Several human xenografts have been developed for the study of prostate cancer including the PC-82 (6), Honda, (7), LuCaP (8), LAP4 and LAPC-9 (9, 10), CWR22 (11), and LNCaP tumor model (12). The LNCaP xenograft can be passaged as a cell line, secretes PSA, and progresses to androgen independence after castration, thereby resembling the pathogenesis of human prostate cancer (12–14). A beneficial aspect of this tumor model is that PSA serum levels can be measured in mice before and after castration to show the transition from androgen sensitivity to androgen independence. However, to determine the molecular events involved in the progression of prostate cancer to androgen independence, homogeneous populations of cells are required for meaningful results. Harvested LNCaP xenografts contain considerable and variable contamination with host tissue. This prevents the use of these tumors for DNA, RNA, and protein extraction. To purify these tumors before they can be used for molecular biology experiments (such as cDNA arrays and proteomics), a labor- and time-intensive protocol is required to isolate the epithelial cells from contaminating host tissue (blood and stroma). Thus, xenograft tumors do not provide an easy model for identifying the molecular events involved in the progression of prostate cancer to androgen independence.

The hollow fiber model was developed by investigators at the National Cancer Institute for in vivo cultivation of tumor cells in retrievable packages and prevention of contamination by host cells (15). Although this model has been used for pharmaceutical applications, to date there are no reports of using this model to obtain pure populations of endocrine cells during various stages of hormonal progression. Facilitation of applications using advances in genomic and proteomic technologies requires rapid and pure isolation of tumor cells during a specific stage of tumor development. The aim of this work was to develop an in vivo model that encompasses the use of hollow fibers to retrieve uncontaminated packages of prostate cancer LNCaP cells (tumors) that can be used for subsequent molecular biology analyses of the progression of prostate cancer to androgen independence. The manipulation of the tumor cells and the ease with which they can be isolated without host contamination make this model ideal for detailed genomic and proteomic analyses.

Male athymic Nude mice (BALB/c strain), 6–8 weeks of age, were obtained from Charles River Laboratory (Montreal, Quebec, Canada). Male SCID mice, 6–8 weeks of age, were obtained from the breeding program at the Joint Animal Facility of the British Columbia Cancer Agency. All animals were free of known pathogens at the time of use and maintained in isolator cages. All procedures were performed in compliance with regulations on the humane use and care of laboratory animals under an appropriate animals license issued by the University of British Columbia (Vancouver, British Columbia, Canada). LNCaP cells were kindly provided by Dr. L. W. K. Chung (University of Virginia School of Medicine, Charlottesville, VA) and maintained as a monolayer culture in RPMI 1640 supplemented with 10⁵ μg/l penicillin, 10⁵ μg/l streptomycin, and 5% FBS (Life Technologies, Inc., Burlington, Ontario, Canada).

Polyvinylidene difluoride hollow fibers (M_r 500,000 molecular weight cutoff; 1-mm internal diameter) were purchased from Spectrum Laboratories (Laguna Hills, CA). Preparation of the fibers was performed as described previously (15). Before being filled with LNCaP cell suspensions, the fibers were individually flushed with 70% ethanol and soaked in 70% ethanol for 4 days. After rinsing and autoclaving in deionized water, the fibers were filled with RPMI 1640 supplemented with 20% FBS and incubated at 37°C for 24 h. Subconfluent cultures of LNCaP cells were trypsinized using 0.25% trypsin-EDTA, pelleted by centrifugation, resuspended in RPMI plus 20% FBS, and loaded into the fibers with the aid of a 20-gauge needle at a seeding density of 1 × 10⁷ cells/ml unless stated otherwise. Extremities of the fibers were fused together by heat sealing. Three days after culture in vitro, the fibers were sectioned into ∼3-cm pieces unless stated otherwise, heat sealed again, and implanted s.c. in Nude or SCID mice.

Mice were anesthetized using methoxyfluorane as an inhalant for all surgical procedures. A small skin incision was made in the hind flank region of the animal to allow the insertion of a 10-gauge trocar. The trocar containing the LNCaP-filled fibers was inserted caudally through the s.c. tissues, thereby allowing the fibers to be deposited during its withdrawal. Unless stated otherwise, a total of 16 fibers were implanted in each animal, in bundles of four fibers at four different regions in the animal. Castration of mice was performed by making a small incision in the scrotum to excise each testicle after ligation of the cord. Surgical suture was used to close the incision. Time of castration varied and is indicated for each experiment.

Intermittent androgen suppression experiments used testosterone pellets (2.5 mg; Innovative Research of America, Sarasota, FL). For each cycle, these pellets were inserted s.c. for 1 week and removed subsequently for 2 weeks, before the cycle was repeated. Serum PSA was monitored over a period of 4 months.

Levels of PSA in the blood of implanted animals were determined weekly using an enzymatic immunoassay kit (Abbott IMX, Montreal, Quebec, Canada). Blood samples were obtained by a small incision in the dorsal tail vein to collect ∼50 μl of blood using a hematocrit capillary tube. No anesthesia was required. Fifteen μl of mouse serum were diluted with 135 μl of diluent to perform the assay.

Fibers were retrieved at different times after implantation, and cell viability was assessed by a modified MTT dye conversion assay as described previously (15, 16).

A parallel set of samples was fixed in 10% buffered formalin and embedded in paraffin. Longitudinal sections of fixed tissue were stained with H&E to evaluate the morphology of LNCaP cells cultivated in vivo within the fibers.

Total RNA was isolated from LNCaP cells grown inside the fibers using Trizol (Life Technologies, Inc.) and fractionated by electrophoresis on a 1% denaturing agarose gel containing formaldehyde before blotting onto Hybond-N⁺ filters (Amersham, Oakville, Ontario, Canada). The 1.4-kb EcoRI fragment of PSA cDNA (17) and cDNA probe for human glyceraldehyde-3-phosphate dehydrogenase were labeled with [α-³²P]dCTP using the Random Primers DNA Labeling kit (Life Technologies, Inc.). Hybridization was performed as described previously (17). The mRNA bands were visualized and quantified using the Storm 860 PhosphorImager (Molecular Dynamics Inc., Sunnyvale, CA).

Nuclear and whole cell proteins (25 μg/lane) were separated by 8% SDS-PAGE and transferred to Immobilon-P membranes (Millipore). The membrane was blocked with 5% (w/v) milk in 20 mm Tris containing 500 mm NaCl and 0.3% Tween 20 (TBST) for 1 h and then incubated 1 h with 0.7 μg/ml antibody to the androgen receptor (SC-7305; Santa Cruz Biotechnology, CA). The membrane was washed and incubated for 1 h with a 1:10,000 dilution of secondary antibody (SC-2005; Santa Cruz Biotechnology). The antibodies were diluted in 5% milk (w/v) in TBST. All incubations were performed at room temperature. Androgen receptor protein was detected using enhanced chemiluminescence kit (Amersham).

The National Cancer Institutes developed the hollow fiber assay to facilitate its screening program for chemotherapeutic agents (15, 18). However, for the chemotherapeutic agents to show efficacy, blood supply to the implanted hollow fibers must be established. In this regard, induction of angiogenesis and development of blood vessels to s.c. implanted hollow fibers has been shown and is dependent on the presence of tumor cells (19). The typical appearance of the fibers containing LNCaP cells that have been implanted s.c. in male Nude mice is shown in Fig. 1.

Elevation of levels of PSA in the serum of men with prostate cancer has been suggested to be directly correlated to the tumor volume (20, 21). To test whether LNCaP cells grown in fibers could proliferate, we used the reported optimal seeding density for other cell lines that was 1 × 10⁷ cells/ml (15) and measured serum levels of PSA in the implanted mice. Male Nude mice prior to implantation with fibers containing LNCaP cells had undetectable levels of serum PSA (<0.02 ng/ml of serum) as expected because PSA is a human-specific protein that is secreted by prostate epithelial cells (22). PSA could be detected as early as 3 days after implantation (data not shown), although earlier times were not tested. At a cell density of 1 × 10⁷ cells/ml, a continuous rise in serum PSA levels was measured over the duration of the experiment (Fig. 2). Fourteen days after implantation of the fibers, serum PSA levels were 4 ng/ml. These PSA levels continued to rise to ∼16 ng/ml by day 56 after implantation of the fibers. Thus, a 4-fold increase in serum PSA levels were measured in the intact animal over the period of 42 days, giving a velocity of 2 ng/ml/week. Serum PSA values obtained from a parallel set of mice that were seeded with a higher density of LNCaP cells (1 × 10⁸ cells/ml) showed an initial sharp rise, followed by a plateau. Fourteen days after implantation of fibers containing LNCaP cells, a mean serum PSA level of 10 ng/ml was detected (Fig. 2). By day 28, serum PSA had continued to rise to a level of 25 ng/ml, representing a velocity of 7.5 ng/ml/week. By day 42, no further rises in serum PSA were observed.

Xenograft experiments in Nude and SCID mice require the inclusion of Matrigel for LNCaP cells to form tumors (14). To test whether Matrigel had an effect on proliferation of LNCaP cells grown in the hollow fiber model, we examined the levels of serum PSA in the Nude mice containing: (a) LNCaP cells without Matrigel and preseeded 3 days in fibers maintained in vitro before implantation into mice; (b) LNCaP cells not preseeded and without Matrigel; or (c) LNCaP cells not preseeded in the presence of Matrigel. Serum PSA levels were comparable between the animals containing fibers that were not preseeded, regardless of the inclusion or absence of Matrigel (Fig. 3). Serum PSA levels were ∼7.5 ng/ml at 7 days after implantation of fibers containing LNCaP cells either with or without Matrigel. Serum PSA levels were lower in the mice containing fibers that were preseeded 3 days earlier and maintained in vitro before implantation. However, by 2 weeks after implantation there was no difference in serum PSA levels among the three groups of animals. All three sets of animals responded similarly to castration with a >85% drop in serum PSA by 1 week after castration. All three sets of animals also showed androgen-independent increases in serum PSA levels over the nadir at 4 weeks after castration. These results suggest that Matrigel is not required for LNCaP cells to grow, respond to androgen, and become androgen independent inside hollow fibers maintained in vivo as determined by serum PSA responses. Similar trends were observed in SCID mice (data not shown).

Androgen deprivation therapy in most patients with prostate cancer results in an 80% drop in serum PSA levels (23). Here we examined the effects of androgen deprivation on serum PSA levels in mice implanted with fibers containing LNCaP cells that were surgically castrated after serum PSA levels were shown to be stable or rising. After castration, serum PSA levels were monitored for an additional 14 weeks. Results from three separate experiments (four Nude mice/experiment) showed that castration resulted in a 80% drop in serum PSA levels by 2 weeks after castration (Fig. 4). Four weeks after castration, serum PSA levels rose again, signifying the emergence of androgen independence. By ∼14 weeks after castration, serum PSA levels were increased by 3.5-fold over the PSA nadir.

To determine whether there were differences between Nude and SCID mice in the time to androgen independence and serum PSA response to castration, we performed parallel studies using Nude and SCID mice. All fibers implanted were prepared at the same time; hence, no variables between the experiments could be attributed to differences from the LNCaP cells (e.g., density, passage number, handling, and others). Results in Fig. 5A show that castration in Nude mice (n = 6) caused a >90% decrease in serum PSA from precastrate levels. The PSA nadir was reached ∼1 week after castration. Five weeks after castration, serum PSA levels were elevated by ∼8-fold the nadir value. When comparing serum PSA levels in parallel-treated SCID mice (n = 4), the only difference was the 80% decrease in serum PSA levels from precastrate values by 1 week after castration (Fig. 5B). Five weeks after castration, serum PSA levels were 3.6-fold the nadir value, suggesting that the LNCaP cells had become androgen independent. Time to androgen independence was similar between the two sets of animals.

Using the LNCaP xenograft model of prostate cancer, increases in PSA gene expression in androgen-independent cells is established at the level of transcription (17). Both serum PSA and tumor mRNA levels are down-regulated when testosterone is withdrawn and up-regulated when it is replaced. However, when the tumor becomes androgen-independent, PSA mRNA is constitutively up-regulated despite the continuing absence of testosterone. To test whether this occurs in the LNCaP hollow fiber model, we performed Northern blot analyses on LNCaP cells harvested from fibers in SCID mice at the following time points: (a) day 5 when the serum PSA was elevated prior to castration; (b) 4 days after castration when the serum PSA was dropping; and (c) 24 days after castration when the serum PSA was elevated (Fig. 6A). Results from Northern blot analyses of RNA isolated from LNCaP cells harvested at these time points in a single animal are shown in Fig. 6B. Northern blot analysis of PSA mRNA levels from three to four different animals, normalized with 18S RNA and bands quantified by phosphorimaging, are shown in Fig. 6C. These data indicate that PSA mRNA levels are: (a) elevated in the presence of androgen at time point 1 (Intact), which corresponds to elevated serum PSA at this time point; (b) decreased at time point 2 (CX), in the absence of androgen, when serum PSA levels are dropping; and (c) elevated in the absence of androgen (AI) when serum PSA levels are elevated, signifying androgen independence.

A high percentage of clinical specimens of human prostate cancer show expression of androgen receptor both in primary tumors and in hormone-refractory recurrent tumors (24–26). To determine whether androgen-independent LNCaP cells still expressed androgen receptor in this hollow fiber model, whole cell lysates and nuclear extracts were prepared from androgen-independent cells harvested at time point 3 shown in Fig. 6A, and Western blots were performed. Androgen receptor was detectable in LNCaP cells in both whole cell lysates (Fig. 6D, Lane 1) and nuclear extracts (Lane 2) harvested after castration when the PSA become elevated again, signifying androgen independence. Thus, androgen receptor was still expressed in androgen-independent LNCaP cells.

One of the most beneficial aspects of this model is that cell number within fibers can be determined at any point in the experiment. To determine whether LNCaP cells proliferate in this model, we assessed the seeding and harvested cell densities using the MTT assay to calculate the net percentage of growth (15). In the group of intact animals, cells seeded at a density of 1 × 10⁷ cells/ml continued to proliferate over the duration of the experiment at 48 days after implantation of the fibers (Fig. 7). The rate of proliferation was relatively constant during this time, and 196.8 ± 18.8% net growth was calculated for the duration of the experiment. In the group of animals that was castrated at 21 days after implantation, the initial rate of proliferation in the presence of testosterone was similar to that observed in the intact animals, as would be expected. However, after castration, the rate of proliferation was markedly decreased as indicated by the change in slope after 21 days. A reduced net percentage of growth of 121.5 ± 12.1% was calculated for LNCaP cells maintained in animals castrated at day 21 at the duration of the experiment, as compared with the net percentage of growth for the intact animals.

In ∼30% of animals, the cells grown in fibers failed to become androgen independent after extended periods in the castrated host. To test whether this was attributable to the loss of cell viability from possibly necrosis or apoptosis, we inserted a testosterone pellet s.c. to determine whether the cells would still respond to androgens after these lengthy periods of cultivation in the animals. Results presented in Fig. 8A show that in Nude animals that did not become androgen independent 10 weeks after castration, the implanted LNCaP cells still responded to testosterone, as indicated by the rapid and robust increase in serum PSA to values greater than precastrate levels. Upon removal of the testosterone pellet, serum PSA levels dropped by >90%. Similarly, LNCaP cells maintained in fibers in SCID mice that did not become androgen independent 7 weeks after castration still responded robustly to testosterone as indicated by the 2-fold increase in serum PSA levels over precastrate values (Fig. 8B). Animals that did not become androgen independent were next examined to test their response to several cycles of androgen withdrawal and replacement. When the testosterone pellets were inserted, each of the animals responded with a robust increase in serum PSA levels above precastrate values (Fig. 8C). When the testosterone pellet was removed, serum PSA returned to nadir levels. Serum PSA response to testosterone could be observed over numerous intervals reflecting the clinical scenario observed in patients treated with intermittent androgen suppression (27).

Harvested fibers containing cells were fixed in 10% buffered formalin, embedded in paraffin, sectioned (longitudinal and cross-section), and stained with H&E to evaluate the cells cultivated within the fiber. These studies demonstrated that initially the cells grew along the wall of the fiber. In intact animals, the cells tended to stack up, forming layers and scaffold structures resembling a solid tumor with areas that contained necrotic populations (Fig. 9A). Cells grown in the presence of Matrigel and maintained in vivo in fibers is also shown (Fig. 9B). Surprisingly, some fibers removed from castrated animals with elevated levels of serum PSA contained spheroids of cells that had detached from the fiber wall (Fig. 9C).

One potential application of this model is to develop new sublines of cells with different requirements for androgen. Presently, there are limited prostate cancer cell lines that are androgen independent and still express androgen receptor (28). To develop such a cell line using the hollow fiber model, androgen-independent cells were harvested from castrated animals with increasing serum PSA and recultured and maintained in vitro. While reculturing in vitro in RPMI 1640 supplemented with 10% DCC-FBS, many cells did not attach to the plate surface, and clumps of cells floated on the surface of the medium. However, when the cells were able to attach and grow, they resembled neurons, with a small cell body and long dendrite-like outgrowths (Fig. 10).

Development of LNCaP hollow fiber model for the study of progression of prostate cancer to androgen independence is shown here for the first time. In the course of developing this model, several observations were made. These include: (a) Matrigel and direct cell-cell interactions between LNCaP and stroma cells are not required for proliferation or progression of LNCaP cells to androgen independence; and (b) androgen-independent increases in serum PSA in castrated animals are uncoupled from proliferation and androgen receptor continues to be expressed in androgen-independent cells.

The LNCaP cell line was established from a prostate metastatic lesion obtained from a patient’s lymph node (29). These cells express androgen receptor and markers for prostatic epithelium such as PSA and prostatic alkaline phosphatase. LNCaP cells can be passaged as a monolayer and can form s.c. and intraprostatic tumors in immunocompromised mice (14, 29, 30) that will progress to androgen independence in castrated animals, thereby making this cell line extremely attractive for prostate cancer research. Here we have expanded the applications of this cell line by development of the LNCaP hollow fiber model, which prevents host cell contamination as measured by reverse transcription-PCR for a murine-specific gene using RNA isolated from the fibers (data not shown) and as reported previously (15). At a seeding density of 1 × 10⁷ cells/ml in 16 fibers of 3-cm lengths, for a total of approximately 4.8 × 10⁶ cells/animal, PSA can easily be measured in the blood of SCID and Nude mice. This cell density falls into the range reported previously for other cell lines to achieve optimal cell growth in hollow fibers maintained in vivo (15). LNCaP cells seeded at this cell density provided measurable serum PSA levels that increased in intact (noncastrated) animals, decreased by 80% to a nadir after castration, and subsequently increased by 4 weeks after castration. In vivo proliferation of LNCaP cells in the fibers continued in the presence of androgens and continued to increase, albeit at a slower rate, in the castrated animals. Data from this fiber model for both serum PSA and proliferation of LNCaP cells are consistent with those data obtained with the LNCaP tumor model (13). In the LNCaP tumor model, the time to androgen-independent progression was shown to be variable, ranging from 14 to >100 days, with a mean of 28 days (31).

Stromal and epithelial interactions are important in the development of the prostate (32, 33). Interactions between these two cells types have also been suggested to play an important role in carcinogenesis and progression (13, 34–37). However, little evidence has been shown as to whether direct cell-cell interactions are required or whether merely soluble factors are sufficient. One suggestion as to the underlying requirement for Matrigel for the LNCaP xenograft to be successful is that Matrigel provides essential growth factors required for the development of LNCaP tumors (38). However, in the original report by Horoszewicz et al. (29), neither the inclusion of Matrigel nor coinoculation of stromal cells with LNCaP cells were required for tumor take. The LNCaP hollow fiber model that we present here provides evidence that direct stromal-epithelial interactions are not required to achieve androgen-independent increases in PSA. Nor are these direct interactions required for proliferation of epithelial cells in the presence or absence of androgens because MTT results indicate that LNCaP cells still proliferate inside of the fibers regardless of castration. Whether soluble factors from surrounding host cells are required for this proliferation and androgen-independent expression of PSA remains to be determined. There are several lines of evidence that support the role of soluble growth factors from other cells influencing the proliferation of prostate cancer epithelial cells. The first line of evidence can be drawn from the fact that prostate epithelial cells are stimulated to grow in vitro by soluble growth factors released from fibroblasts (39). The second line of evidence can be drawn from observations made in the present study that not all of the LNCaP cells maintained in fibers in a set of mice became androgen independent after castration of the host. These differences are most likely attributable to the individual animals because each set of the animals was implanted with fibers prepared at the same time; hence, no variables from the cells were introduced.

In the intact noncastrated animal, serum PSA continued to rise when seeded at the optimal density in fibers (1 × 10⁷ cells/ml) for the duration of the experiment (Fig. 2). Comparison of the rises in serum PSA after implantation (Fig. 2) to the MTT data measuring proliferation of cells (Fig. 7) suggests that the rises in PSA are attributed to proliferating LNCaP cells in the presence of androgens. However, this correlation does not hold true for the castrated animal. In these animals, serum PSA drops by at least 80% within 2 weeks after castration (Figs. 3,4,5,6 and 8). This drop in PSA does not correspond to the MTT data generated that suggest that proliferation still occurs, albeit at a slower rate, after castration (Fig. 7B). Thus, when serum PSA and mRNA levels are greatly attenuated in castrated hosts, the cell numbers are increasing, thereby uncoupling proliferation from PSA gene expression. This suggestion is consistent with previous observations that LNCaP tumor volume stabilizes but does not decrease after castration, despite the 80% decrease in serum PSA values (14).

Histology studies of LNCaP cells grown in vivo inside of fibers showed that these cells form a solid tumor. This is consistent with the reported growth of other cell lines cultivated in vivo in hollow fibers (15, 18). In the presence of androgens, LNCaP cells tended to stack up in layers along the fiber wall. At the cell density used for these experiments (300,000 cells/3-cm fiber), at no time was the fiber completely filled with cells to indicate volume constraints. Cells cultivated in fibers for long periods of time in castrated animals tended to form spheroids that no longer grew along the fiber wall. This is the first report to show this phenomenon. Consistent with this was the observation that upon reculturing of androgen-independent cells, many of the cells did not attach to the surface of the plate and formed clumps of cells that floated on the surface of the medium. When androgen-independent cells were recultured, their morphology was altered such that the cells resembled neurons. This change in morphology was similar to that described for LNCaP cells maintained in monolayer and exposed to interleukin 6 (40) and compounds that increase cellular levels of cAMP (41). This change in morphology has also been associated with the differentiation of LNCaP cells into neuroendocrine cells such that markers for neuroendocrine cells, chromogranin A and neuron-specific enolase, become expressed (41). Neuroendocrine cells have been reported to play a role in the progression of prostate cancer to androgen independence (reviewed in Ref. 42). Thus, the observed changes in the morphology of LNCaP cell to resemble neuroendocrine cells in response to castration support the suggestion that androgen ablation may lead to a more aggressive phenotype of prostate cancer (43).

In summary, we have developed the LNCaP hollow fiber model, which provides a reproducible means of obtaining “pure” populations of LNCaP cells during different stages of progression to androgen independence for molecular analysis requiring uncontaminated RNA and protein extracts. However, other possible applications for this model include studies of the transcriptional regulation of genes that are altered during progression to androgen independence and developing new sublines of cells with differing requirements for hormone. Application of this model is not intended to replace xenografts because it cannot reflect the contributions of cell-cell interactions or angiogenesis on tumor biology.