The Effect of Fibroblast Growth Factor 8, Isoform b, on the Biology of Prostate Carcinoma Cells and Their Interaction with Stromal Cells¹

The FGF³ family consists of an increasing number of structurally related polypeptide mitogens, which elicit their effects by binding to high-affinity tyrosine kinase receptors on the cell surface, encoded by at least four genes (FGFR1–4) in mammals (1, 2). FGF molecules are important in cell-cell interactions during embryogenesis and tissue differentiation, as well as during tumorigenesis. FGF8, the eighth member of this family, was originally identified as the androgen-induced growth factor from the CM of Shionogi mouse mammary carcinoma cell line, SC-3 (3). It was demonstrated to mediate the androgen-dependent growth of SC-3 cells. Its expression was correlated with murine and chicken embryogenesis in regions of outgrowth and patterning such as the elongating body axis, midbrain/hindbrain junction, limb, and face(4, 5, 6, 7, 8, 9, 10). Whereas knockout of the fgf8 gene resulted in early embryonic lethality in mice (11), FGF8 was identified by the use of the Cre/loxP system as an epithelial signal necessary for the outgrowth and patterning of the first branchial arch primordium (12). The fgf8 has been localized to mouse chromosome 19 and human FGF8 to chromosome 10q24–26 (6, 7, 13, 14). This gene is unusual in its first exon, which, in fact, consists of at least four exons compared with one exon in other FGF genes. Alternative splicing of these four exons in the mouse results in eight potential protein isoforms that vary in their amino termini (3, 6, 15, 16). The human FGF8 gene is similar to its murine counterpart in structure. However, only four protein isoforms (FGF8a,8b, 8e, and 8f) are predicted because of a blocked reading frame in the human exon 1B (17, 18). The FGF8 isoforms share the same signal peptide and identical COOH-terminal region. NIH3T3 cells transfected with fgf8b cDNA or treated with rFGF8b became highly transformed compared with those transfected with fgf8a or fgf8c (15, 19). Additionally, fgf8 was demonstrated to cooperate with the Wnt-1 gene as a murine mammary proto-oncogene in Wnt-1 transgenic mice (16).

Of the four possible isoforms, three (FGF8a, FGF8b, and FGF8e) were cloned in our laboratory from a human prostate tumor cell line, DU145 (20). The protein products of these cDNAs share extensive amino acid homology with mouse FGF8 isoforms in that FGF8a and FGF8b exhibit identical amino acid sequences to those of their murine counterparts. The human FGF8 isoforms, although weakly expressed in human adult tissues or cell lines, are nevertheless differentially expressed. FGF8b appears to be the primary species in prostatic epithelial cell lines (20). Consistent with previous reports, FGF8b, but not FGF8a or FGF8e, confers robust transforming and tumorigenic activities in NIH3T3 cells. This oncogenic activity becomes more relevant because evidence points to a significant up-regulation of FGF8b expression in high-grade prostate carcinomas (21, 22). A high frequency of FGF8 overexpression, which is associated with decreased patient survival and persists in androgen-independent disease, was detected by immunohistochemical analyses of prostate cancer specimens (22). Furthermore,targeted overexpression of this isoform in the mammary glands of FGF8b transgenic mice results in mammary tumorigenesis(23). In cultured human prostate cancer cells, expression of antisense FGF8b reduces their growth rate, inhibits their soft agar clonogenic activity, and decreases in vivo tumorigenicity(24). The above findings strongly suggest that FGF8b is involved in hormone-related carcinogenesis of the prostate and mammary glands.

These observations have prompted us to investigate the biological effects of overexpression of FGF8b in prostatic cancer cells. A lentiviral transfer vector was developed for transduction of the FGF8b gene along with a GFP marker gene into the weakly tumorigenic LNCaP prostate carcinoma cell line. The proliferation, anchorage-independent growth ability, and in vitro and in vivo invasion ability of transduced LNCaP cells were determined. To examine the epithelial-stromal interactions under the condition of FGF8b overexpression in LNCaP cells, we used a coculture system in which GFP expression served as a marker for the separate quantitation of the cellular components.

Transducing vector, packaging construct (pCMVΔ8.71), and VSV env-coding plasmids were originally obtained from Dr. Luigi Naldini (University of Torino, Torino, Italy; Ref.25). The transducing vector was modified to include an IRES-GFP marker gene cassette. Primary explant cultures of prostate and SV stromal cells were established from residual portions of radical prostatectomy specimens or suprapubic prostatectomy specimens after fresh pathological inspection. The tissues were freed of connective tissue elements by sharp dissection and minced into small pieces (approximately 1 mm³) using crossed No. 11 scalpel blades. They were first plated in a minimal volume of DMEM supplemented with 20% FBS and 1% penicillin/streptomycin and were incubated in a humidified atmosphere of 5% CO₂at 37°C. After primary outgrowth, the medium was changed to RPMI 1640 supplemented with 10% FBS. When the stromal cells had reached 70–80%confluency, they were detached using 0.25% trypsin/1 mm EDTA and replated at 20% densities.

Human FGF8b cDNA (20), which was flanked with the CMV immediate early promoter at the 5′ end and IRES-GFPat the 3′ end, was incorporated into the polycloning sites of the transducing vector. Plasmids were amplified in Escherichia coli and purified by Qiagen Maxi Prep kit. Using Superfect reagent(Stratagene, Inc.), human 293T cells, at about 80% confluency, were cotransfected with the transducing vector, packaging construct and VSV env-encoding plasmids at a ratio of 5:5:1. The transducing vector construct containing only the CMV-driven IRES-GFPcassette, without FGF8b, was used in parallel cotransfection to produce the control vector. The medium containing virions was harvested daily starting from 3rd day to the 5th day after transfection.

The lentivirus-containing medium was concentrated 10-fold by using M_r 300,000 molecular weight cutoff spin columns (Gelman). One ml of concentrated CMV-FGF8b-GFP or CMV-GFP vector lentiviral medium was applied to 80%-confluency LNCaP cells in T25 culture flasks. After 4 h of incubation, lentiviral medium was removed. Cells were washed with PBS twice and grown in complete medium for 2 days. The 8b-and vector-LNCaP cells were released from flasks by trypsinization and sorted by FACS on the basis of expression of GFP.

Total RNAs were extracted from 8b-, vector-, and noninfected LNCaP cells by using RNeasy Mini Kit (Qiagen). RNAs were separated by electrophoresis on a 1% denaturing formaldehyde agarose gel, and transferred to Hybond N membrane (Amersham Corp.). The blots were hybridized to a ³²P-labeled full-length FGF8b cDNA probe and exposed to X-ray film.

To produce CM for use in the transformation assay, fresh medium without FBS plus 10 μg/ml heparan sulfate was applied to transduced LNCaP cells when they were at 80% confluency. After 48–72 h of incubation,CM was collected and cell debris was removed by centrifugation. The CM,diluted 1:10 in DMEM medium, was added to NIH3T3 cells when they had grown to 70% confluency. After 2 days of incubation, cell morphology was examined microscopically as described previously (20, 24).

The 8b-, vector- and noninfected LNCaP cells were released by trysinization and suspended in 0.4% Seakem agarose at a cell density of 1 × 10⁴/2 ml. The suspensions were overlayed on 4 ml of 0.8% Seakem agarose in a 6-cm-diameter dish and incubated at 37°C in 5% CO₂ for 21 days. The visible colonies were counted. The colony-forming efficiency was calculated by dividing the number of soft agar colonies by the number of cells plated and multiplying by 100 to convert to a percentage.

Inserts of 8-μm pore-sized membranes for 24-well plates were prepared by coating with Matrigel basement membrane matrix (Becton Dickinson Labware, Bedford, MA) following the manufacturer’s instructions. Each chamber was separated into an upper and a lower portion by the insertion of a thin layer of Matrigel basement membrane matrix. The 8b-, vector-, or noninfected LNCaP cells were placed on the upper chamber at a cell density of 1 × 10⁵ cells/insert. The CM obtained by incubating NIH3T3 cells for 24 h in serum-free DMEM in the presence of 50μg/ml ascorbic acid was added to the lower chamber to serve as a chemoattractant. After 24 h of incubation, the upper surface of the inserts was wiped with cotton swabs, and the inserts were stained with H&E. Cells that migrated through the Matrigel and the filter pores to the lower surface were counted under a light microscope with five random high-power fields per insert (26).

The 8b-, vector-, and noninfected LNCaP cells were grown with complete medium in log phase and released from flasks by trypsinization. One million cells of each type were injected i.p. into athymic nude mice as described previously (26, 27). Each cell type was injected into three animals. After 9 weeks of incubation, mice were sacrificed. Diaphragms were fixed in 10% formalin overnight and embedded with paraffin. Sections were incubated with a purified goat polyclonal anti-GFP antibody (Santa Cruz) at 4 μg/ml concentration. The bound antibody was detected with biotinylated antigoat immunoglobulin. Sections incubated without primary antibody served as negative controls.

Approximately 1 × 10⁴ 8b- or vector-LNCaP cells and 1 × 10⁴stromal cells were mixed and seeded into six-well plates (Corning),whereas 1 × 10⁴ 8b-, vector-,noninfected LNCaP cells and stromal cells alone were also plated as controls. Each sample was seeded in triplicate. Total cell numbers of cocultures were counted at different time points. After counting, 8b-or vector-LNCaP cell and stromal cell cocultures were analyzed by flow cytometry on the basis of GFP expression. The fractions of transduced LNCaP or stromal cells were measured and multiplied by total cell numbers to calculate the exact number for each cell type in cocultures.

Chimeric lentivirions capable of a single cycle of infection were produced in a human embryonic kidney cell line, 293T, through three-plasmid cotransfection (Fig. 1; Ref. 25). LNCaP cells, responsive to rFGF8b treatment and only weakly tumorigenic in vivo, were infected with these virions, which carried either the FGF8b-GFP or only the GFP control gene. After infection and propagation,the cells were sorted twice by FACS on the basis of their GFP expression. Thus two types of populations, cells transduced with the human FGF8b gene (designated as 8b-LNCaP) and cells transduced with vector control construct (vector-LNCaP), were established. Under a fluorescent microscope, GFP was an excellent visible marker to determine whether the cells were indeed transduced(Fig. 2,A). The expression of FGF8b in 8b-LNCaP cells was readily detected with a Northern blot assay using a ³²P-labeled full-length FGF8b probe, whereas the expression in vector- or noninfected cells was too low to detect by the Northern technique (Fig. 2 B). To demonstrate FGF8b expression at the functional protein level, we conducted an NIH3T3 cell biological transformation assay using CM from transduced LNCaP cells. In agreement with our previous work(20), CM from 8b-cells displayed a strong ability to transform the NIH3T3 cells morphologically, whereas CM from vector-LNCaP cells did not (data not shown).

To avoid potential clonal variation, pooled populations of sorted cells rather than single clones of transduced cells were used in all experiments. First, the effect of FGF8b on growth of LNCaP cells was examined. The experiments were repeated four times with cells at different passages. Representative results are illustrated in Fig. 3,A. Clearly the overexpression of FGF8b in 8b-LNCaP cells increased the growth rate in comparison with that of vector-LNCaP cells. After 14 days of culture, the proliferation rate of 8b-LNCaP cells was 2-fold higher than that of vector-LNCaP cells. There was no difference in the growth rate between vector- and noninfected LNCaP cells (data not shown). Next, we determined whether FGF8b could function as a mitogen to stromal cells. Two different human primary prostatic stromal cell cultures and one primary stromal cell culture from human seminal vesicle were used to examine the effect of rFGF8b on stromal cell growth. Cells were treated with rFGF8b at concentrations of 10 ng/ml or 100 ng/ml. Each experiment was repeated twice. It was found that rFGF8b had no mitogenic effect on either prostatic (Fig. 3,B) or non-prostatic stromal cells (data not shown). In contrast, rFGF8b, consistent with a previous study (17),was able to stimulate proliferation of LNCaP cells. The stimulation noted with 10 ng/ml of the protein factor, however, could not be further enhanced by increasing the concentration by 10-fold (Fig. 3 C).

The soft agar assay was performed to investigate the effect of FGF8b on the anchorage-independent clonogenicity of transduced LNCaP cells. Each cell sample was seeded in triplicate and each assay was repeated twice. After 21 days of culture, the 8b-LNCaP cell group displayed clonogenic efficiency of 332 ± 45.74 colonies (3.32%), whereas a reduced efficiency of 250 ± 36.62 colonies (2.5%) was observed with the vector-LNCaP cell group. By Student’s ttest, the difference in clonogenic efficiency between these two cell groups was determined to be significant (P < 0.05). Additionally, the size of at least one-third of the colonies formed by 8b-LNCaP cells was generally larger than those formed by vector- or noninfected LNCaP cells (results not shown). The latter two cell types showed no significant difference in either clonogenic efficiency or colony size.

Next, the in vitro invasion ability of transduced cells was measured with the Matrigel invasion chamber assay. After 24 h of incubation, vector- (Fig. 4,A) or 8b-LNCaP cells (Fig. 4,B) that migrated through the Matrigel basement membrane matrix and the filter pores to the lower surface were counted by light microscopy. From 1 × 10⁵ seeded cells, 583 ± 37.47 8b-LNCaP cells were counted compared with 173 ± 60.70 vector-LNCaP cells (Fig. 4 C). Although the difference between these two cell types was significant (P = 0.001), there was no significant difference between vector- or noninfected LNCaP cells. This assay was repeated three times with cells at different passages, and each cell sample was done in triplicate.

To assess the survival ability of transduced LNCaP cells in vivo, one million 8b-, vector-, or noninfected LNCaP cells were injected i.p. into the athymic nude mice. Five mice were inoculated with each cell type. After 9 weeks of incubation, the mice were sacrificed and diaphragms were fixed with 10% formalin. After paraffin embedding, sections were stained with H&E (Fig. 5,A). Twelve different sections for each individual diaphragm were examined for tumor growth or tumor invasion under a microscope(26, 27). The fact that the attached cells were indeed tumor cells was confirmed by immunohistochemistry assay using an anti-GFP antibody (Fig. 5,B). The immunohistochemistry assay was repeated twice on the sections from each diaphragm. Although the analysis involved a limited number of tissue sections, it was found that at least four of five animals in 8b-LNCaP cell-injected mice were positive for tumor growth, whereas all five mice in vector- or noninfected LNCaP cell-injected mice were negative (Table 1). In two of these four diaphragms from 8b-LNCaP cell-injected mice,tumor cells also exhibited multifocal attached growth on the peritoneal surface of diaphragm. The sizes of areas that the tumor cells occupied in these diaphragms varied considerably. In 20% of areas examined,moreover, tumor cells did grow into the diaphragm and spread horizontally with evidence of early vertical penetration. The latter was characterized by tumor cells invading between muscle cells.

To exploit the potential FGF8b-mediated interactions between prostatic cancer cells and stromal cells, we used a coculture system in which GFP served as a marker to separate transduced LNCaP cells from stromal cells by the use of flow cytometry (Fig. 6,A). The same two human primary prostatic stromal cell cultures described above were also used in these analyses. The stromal cells in coculture spread out on the bottom of the culture flask providing a mesenchymal carpet, and LNCaP cells grew on the top of stromal cells forming gland-like structures. A microscopic view of this coculture is presented in Fig. 6,B. The LNCaP cells were rounder and much smaller when they were grown alone than their counterparts in cultures. As illustrated in Fig. 7,A, the growth of prostatic stromal cells in coculture with 8b-LNCaP cells increased significantly compared with that of stromal cells when grown with vector-LNCaP cells. The increase was approximately 2-fold after 14 days of culture. The proliferation rate of stromal cells in coculture with vector-LNCaP cells was, however,similar to that of stromal cells grown alone. The growth rate of 8b-LNCaP cells in coculture was also higher than that of vector cells cocultured with stromal cells. However, compared with the results with LNCaP cells cultured alone (Fig. 3,A), both 8b- and vector-LNCaP cells showed inhibited growth in coculture with stromal cells (Fig. 7,B). To investigate whether the stimulatory effect of stroma on 8b-LNCaP cells was prostate-specific, primary stromal cells from human SV were also grown with the transduced LNCaP cells. The cell morphology in coculture was similar to that of prostatic stromal cell and LNCaP cocultures. After 10 days of culture,the growth rate of these non-prostatic stromal cells in coculture with 8b-LNCaP cells increased 3-fold compared with that in coculture with vector-cells (Fig. 7,C). Again, there was no significant difference in growth rate between SV stromal cells when cultured with vector-LNCaP or cultured alone. The growth rate of 8b-LNCaP cells in coculture with SV stroma was slightly higher than that of vector-cells in coculture (Fig. 7 D), but both cell types showed remarkably restrained growth in the presence of SV stromal cells. It was also noted that 8b-LNCaP cells and non-prostatic stromal cells developed local confluency faster than those in coculture with prostatic stromal cells and 8b-LNCaP cells. It was quite unique that SV stromal cells became detached from the bottom of the culture flask once local confluence was reached. This observation was confirmed by repeat experiments with the same primary SV stromal cells.

Because rFGF8b had no significant direct effect on proliferation of stromal cells, it is likely that some soluble mediator(s) is being secreted by 8b-LNCaP cells to stimulate stromal cells in coculture. To test this possibility, 24 h CM from transduced LNCaP cells was applied to the primary prostatic stromal cells. As illustrated in Fig. 8, the growth rate of stromal cells treated with CM from 8b-LNCaP cells was clearly elevated compared with the use of vector-LNCaP cell CM(P = 0.009).

The interest in this work is 3-fold. First, our results demonstrate that FGF8b produced in the LNCaP cells can function as an autocrine stimulator of the proliferation of these cells. This effect was previously detected using bacterially expressed rFGF8b(17) which, unlike FGF8b produced in mammalian cells(3), lacks glycosylation. It is now documented that whether glycosylated or not, FGF8b can induce proliferation of LNCaP cells, thus implying that the receptor activation by the growth factor is likely to be independent of glycosylation. In this regard, it was reported previously that rFGF8b could efficiently activate the “c”splice isoform of FGFR2 or FGFR3, as well as FGFR4 (28, 29). Although these receptor isoforms are considered to be largely expressed in mesenchymal cells (30), there is evidence of aberrant expression of FGFR isoforms in prostate cancer cells (1, 31). It is also interesting to note that among the tissues of the male reproductive tract, only prostate appears to exhibit wide expression of the general classes of FGFRs. For example, a moderate level of expression of FGFR1 and FGFR2 is found in prostate epithelium and the microvasculature, whereas stromal smooth muscle cells exhibit a weak level of expression of FGFR3 (32). In another study (30) with primary cultures of human prostatic epithelial and stromal cells, FGFR3 was found to be the primary product in epithelial cells with a smaller amount of FGFR2,whereas stromal cells express primarily FGFR3 and smaller amounts of FGFR1 and FGFR2. Considering these observations, and recognizing that FGF8b is naturally overexpressed in aggressive prostatic carcinoma cells (21, 22), a scenario is presented for FGF8b-FGFR signaling in the regulation of prostatic epithelial growth.

Second, besides its ability to stimulate proliferation, it is also shown here that FGF8b can influence the various biological properties of the affected LNCaP cells. For example, in vitro parameters such as soft agar clonogenicity and matrigel invasion activity are significantly increased in 8b-LNCaP cells relative to the vector-LNCaP cells. An argument can be made that the observed changes of in vitro motility and invasion may be related partly to increased proliferation. However, this is unlikely to be a primary cause because the analyses were carried out only after 24 h, when proliferation is deemed to be quite limited. MMPs,which degrade extracellular matrix proteins, are known to be over-expressed in many types of cancers (33, 34, 35, 36, 37). There are several reports that describe induction of MMP expression by FGF proteins in cancer cells, including prostate cancer cells, but not in normal epithelial cells (31, 38, 39, 40, 41). Thus it is possible that the switch between expression patterns of FGFR isoforms that occurs in prostate cancer cells, and that may be critical for abnormal proliferation by FGF8b, may also be responsible for activation of MMPs, thereby facilitating tumor cell invasion. The pleiotropic effect of FGF8b signaling is clearly documented here in the study of the in vivo tumorigenesis/diaphragm invasion assay. Although LNCaP cells were not tumorigenic in this assay, FGF8b expression converted them to be not only tumorigenic but also invasive in some animals during the 9-week period of observation. Taken together,evidence is presented that FGF8b overexpression confers tumorigenic and invasive properties to LNCaP cells. However, it remains to be demonstrated whether results of the study using a single cell line might have broader validity when other prostate cancer cells or prostate cancer tissues are evaluated.

Finally, when we examined the effect of FGF8b on the epithelial-stromal interactions in the setting of coculture, two important aspects were uncovered. It is clearly evident that the proliferation of the parental or transduced LNCaP cells is remarkably inhibited when cocultured with stromal cells from prostatic or non-prostatic tissues. In contrast, the growth of the stromal cells is strongly up-regulated in the presence of FGF8b-producing LNCaP cells but not in the presence of the control LNCaP cells. Although the autocrine regulation of LNCaP cells by FGF8b could compensate for the stromal effect to a degree, the negative effect of stroma still remains quite pronounced. The stimulatory effect of 8b-LNCaP cells on stromal cells seems not to be mediated by the released FGF8b because rFGF8b is not capable of stimulating stromal cells. Considering that other soluble molecules or factors induced and secreted from the FGF8b-stimulated LNCaP cells might be responsible, we used CM on stromal cells to examine the possibility. Because there is some positive effect of the CM on stroma,although far less than that observed in the context of coculture, this potential remains. Additional work will be necessary to characterize the released stimulatory factors to obtain a better definition of the role of FGF8b in the environment of the prostate. However, a stronger case could be made for the importance of cell-cell contact between epithelial and stromal cells for the observed stimulatory effect on the stroma. This contact effect is explicitly dependent on the presence of FGF8b-producing LNCaP cells. Thus, FGF8b signaling, beyond the release of soluble factors, appears to be critical in stromal proliferation. It is tempting to consider alterations of cell surface molecular expression in LNCaP cells from FGF8b-FGFR interactions as the primary inducer of stromal growth.

In summary, the effect of overexpression of FGF8b in LNCaP cells and their interaction with stromal cells may have broad implications in the progression of human prostate cancer. The autocrine stimulatory loop of FGF8b is likely to provide advantages to the neoplastic prostatic epithelium with respect to their proliferation and invasive properties. Additionally, it is speculated that, during the course of invasion and metastasis, the FGF8b producing malignant cells when in physical contact with stromal cells may stimulate growth of stromal cells. The accelerated proliferation of the stromal cells has, in fact, been suggested to provide an amenable milieu for the development and progression of cancer (42).

We are grateful to Robert T. Tin and Heidi Miller for technical assistance. We thank Jiapeng Huang, Xiantuo Wu, and other members of the Roy-Burman laboratory for their help with some experiments. We are indebted to Lihua Zhang for manuscript preparation.