The expression and localisation of beta-nerve growth factor (beta-NGF) in benign and malignant human prostate tissue: relationship to neuroendocrine differentiation.

beta-NGF is a determinant of sympathetic innervation and a neural differentiation factor. In the present study, we have examined 15 benign prostatic hyperplastic and 15 prostate cancer patients for the expression and localisation of beta-NGF by reverse transcription-polymerase chain reaction (RT-PCR), Western blotting, immunohistochemistry and ELISA. We have correlated the beta-NGF concentrations to prostate morphometry and neuroendocrine differentiation. The presence of beta NGF mRNA transcripts was confirmed by RT-PCR where a 542 bp product was found with specific primers for the human beta-NGF cDNA sequence. The presence of the peptide was also confirmed by Western blot analysis which showed a protein co-migrating with recombinant human beta-NGF. Our results demonstrate that beta-NGF is localised to prostate epithelium, and the concentrations of the peptide were not significantly different in malignant (mean +/- s.d.; 3100 +/- 1502 pg g-1 wet weight of tissue) than in benign tissues (1992 +/- 684 pg g-1, P = 0.512). We were, however, unable to correlate the concentrations of beta-NGF to neuroendocrine differentiation in malignant tissues. Clearly, the present study demonstrates that beta-NGF is a product of the prostate and may be involved in the control of the sympathetic innervation of the human prostate.

Beta nerve growth factor (f,-NGF) stimulates the growth of sympathetic and some sensory neurites; in vitro, NGF was found to stimulate the growth of dendritic processes from the undifferentiated PC-12 rat adrenal phaeochromocytoma cellline (Tischler and Greene, 1975) while, in vivo, in adult mammals, tissue NGF concentrations correlated with the density of tissue sympathetic innervation (Korsching and Thoenen, 1983). NGF also acts as a positive chemotaxin for neurones and may facilitate their contact with their target tissues (Gundersen and Barrett, 1980). NGF production, therefore, is a means whereby tissues influence the density and distribution of their sympathetic and sensory innervation.
Mouse /3-NGF is a homodimer of two protein molecules of 14.5 kDa each. f,-NGF purified from mouse salivary glands (the most abundant source of the hormone) is shortened during purification to a 26.5 kDa dimer known as 2.5S mNGF (Bradshaw, 1978). Mouse and human f-NGF show 92% sequence homology and are immunologically similar, such that antibodies raised against the mouse protein will recognise the human (Nikolics, 1993;Naher-Noe et al., 1993). The mouse f3-NGF gene is expressed as a number of splice variants (Nikolics, 1993). However, only one transcript of the highly homologous (Ullrich et al., 1983) human gene has so far been described; this mRNA transcript is spliced from just two exons. The whole sequence of the mature hormone is coded in one exon (Borsani et al., 1990). ,B-NGF dimers are non-covalently bound, and the hormone appears as a monomeric form under denaturing conditions. Study of f,-NGF in neoplasia has emphasised its possible role in tumours derived from the neural crest and in neurogenic tumours. In the D54 glioma, U1 18 and U251 glioblastomas, TE671 medulloblastoma and Hs294 melanoma cell lines, /B-NGF treatment leads to variable differentiation of these cells with a reduction in growth rates (Yaeger et al., 1991). All of these cell lines bear /3-NGF receptors, with melanoma cell lines having a particularly high concentration of these receptors (Buxser et al., 1984). The capacity of ,B-NGF to inhibit growth of neurogenic tumours is supported by the in vivo finding that neuroblastomas which express high-affinity f,-NGF receptors have a better prognosis than those which do not (Kogner et al., 1993).
In spite of the role of adrenergic innervation in outflow tract obstruction (Lepor and Shapiro, 1990) and recent interest in prostate neuroendocrine cells as a prognosticator in prostate adenocarcinomas (Cohen et al., 1991), ,B-NGF has received little attention in the human prostate. In the prostates of rodents, it may be present in high concentration and is localised to prostate epithelium (Harper and Thoenen, 1980;Shikata et al., 1984;MacGrogan et al., 1990). Recently, MacGrogan et al. have found f,-NGF expression and variable ,B-NGF concentrations (0-1720 pg g-') in human benign hyperplastic (BPH) tissue (MacGrogan et al., 1992).
Furthermore, Djakiew's group have localised f,-NGF to the stroma of the human gland by immunohistochemistry (Graham et al., 1992) -a surprising result in view of the high degree of conservation of the hormone and its epithelial localisation in other mammals. In cell culture studies, this group have demonstrated that proteins produced by prostate stromal and epithelial cells that are immunologically related to, but heavier than, /3-NGF (42-65 kDa), stimulate the growth of both stromal and epithelial cells in culture (Djakiew et al., 1991).
The aims of this study were to examine the expression of fl-NGF in benign hyperplastic and malignant human prostate tissue at the protein and mRNA levels. f3-NGF concentration and neuroendocrine cell content of prostate adenocarcinomas were also compared.

Tissues
Transurethral resection of the prostate (TURP) 'chips' were obtained from 15 cases each of BPH and prostate cancer. For control studies, an adrenal gland was obtained from an organ donor (aged 15 years) after familial consent. Tissues were snap-frozen in liquid nitrogen and stored at -70°C.
Randomly selected prostate chips from prostate adenocarcinoma specimens were examined histopathologically and tumour grade assessed by Gleason scoring. Paraffinembedded tissue from the 15 prostate adenocarcinomas was obtained. Paraffin-embedded autopsy pancreas tissue was also obtained and islets of Langerhans used as a positive control for neuroendocrine stains.
Reverse transcription-polymerase chain reaction (RT-PCR) for f3-NGF Total cellular RNA was prepared from prostate tissues by the method of Chomczynski and Sacchi (1987). Briefly, 5 jig of sample RNA were reverse transcribed by the addition of 20 jil of 25 mm magnesium chloride, 10 Mil of transcriptase buffer, 10 jil of 10 mM dNTP mixture, 100 U RNAsin, 2.5 jil of oligo (dT) and 67.5 U AMV reverse transcriptase ('Reverse transcription system' kit, Promega Corporation, Southampton, UK). The mixture was incubated for 1 h at 42°C. f,-NGF primer sequences were 5'-GACCCAAGCT-CAGCTCAGC-3' and 3'-TATTCTGGTGGCGGTGTCTG-5', defining a 542 bp fragment designed from the human /3-NGF cDNA sequence (Borsani et al., 1990). PCR amplification was carried out in a volume containing 20 jil of the RT product mixture, 8 jil of Taq polymerase buffer, 10 jl of 5' and 3' primers, 16 jil of 1.25 mM dNTP mixture and 0.2 ,ul of Taq polymerase (Promega 'Taq polymerase' kit). Reaction mixtures underwent denaturation at 94°C for 90 s, annealing at 58°C for 90 s and extension at 72°C for 150 s for 35 cycles using a Hybaid 'Thermal Cycler' (Hybaid, Teddington, UK). As a positive control, RNA isolated from adrenal tissue was subjected to RT -PCR amplification in the same way as prostatic RNA. Dral, EcoRI and PstI (Promega) digests were carried out at 37°C for 1 h in 35 jil volumes containing 20 pl of PCR product, 2.5 pl of endonuclease buffer, 2.5 jil of acetylated bovine serum albumin (BSA), 9.5 jil of ultrapure water and 0.5 jil of the endonuclease. RT-PCR products, endonuclease digests and a 100 bp molecular weight ladder were size fractionated and visualised by electrophoresis on 1% (w/v) agarose, 1 x TBE gels with ethidium bromide.
Briefly, slides were incubated in 20% normal sheep serum (Scottish Antibody Production Unit, Carluke, UK) in TBS for 30 min before exposure to the primary antibody for 30 min. After a TBS wash, sections were subsequently incubated with a secondary sheep anti-rabbit IgG biotinylated F(ab')2 fragment (Boehringer Mannheim Biochemica, Lewes, UK), at a dilution of 1: 400 in TBS/BSA/azide for 30 min. After a further TBS wash, slides were incubated with streptavidin-linked alkaline phosphatase ('extravidin', Sigma, 1: 1000 in TBS/BSA/azide) for 30 min. The substrate colour reaction was developed with fuchsin chromogen solution ('New Fuchsin', Dako, High Wycombe, UK) for 15 min before counterstaining with Mayer's haematoxylin for 1 min. Slides were air-dried and mounted with DPX medium and cover-slips. The adrenal and pancreatic tissues provided positive controls for f3-NGF and PGP9.5 staining, respectively, while negative controls were carried out with the omission of the primary antisera from the TBS/BSA/azide preparation.
Sections were examined for positive staining for ,B-NGF and PGP9.5 by an independent pathologist. Tumours were simply classified as positive or negative for PGF9.5.
Enzyme-linked immunosorbent assay (ELISA) The concentration off,-NGF in prostate tissues from patients with BPH and prostate cancer was measured as follows: 230 -550 mg of frozen prostate tissue was pulverised in a 'mikrodismembrator', (Braun Medical, Aylesbury, UK) and the powdered tissue was suspended in 500 jil of 100 mM tris-HCI containing 400 mm sodium chloride, 2% (w/v) BSA, 0.1% (w/v) sodium azide, 4 mM EDTA, 1 mM phenylmethylsulphonyl fluoride (PMSF) and 7 ug ml-' aprotinin and homogenised with an 'Ystrall' homogeniser (Ystrall GmbH, Dottingen, Germany). For recovery, 500 c.p.m. of [1251]2.5S mNGF 1500 Ci mmol-1 (NEN, DuPont, Stevenage, UK) were added to each homogenate and allowed to equilibrate for 2 h at 40C, following which homogenates were ultracentrifuged at 100 000 g for 10 min. Gamma emissions from the supernatant and pellet were counted and the recovery fraction for NGF assessedc.p.m.supernatant! C.p.m.supematant+pellet. Supernatants were diluted 1: 1 with a solution containing 20 mm calcium chloride and 0.2% (w/v) Triton X-100. Aliquots of 100 jil were incubated along with rh#-NGF (Boehringer Mannheim) standards (0- Tissue morphometry Frozen sections were cut from a portion of each prostate chip subject to ELISA. After haematoxylin and eosin staining, these sections were used for tissue morphometry. Sections were examined with a microscope attached to an Olympus 'CCD' camera and Olympus 'Cue-2' (Olympus Optical, Tokyo, Japan), image analysis software. The entire area of the sections was analysed to avoid problems of random sample selection in tissue. The total area of each tissue section and the area of its contained acini and ducts were measuredreferred to here as the 'glandular' area.
To ensure that the frozen sections accurately reflected the relative portions of glandular and stromal tissues in the assayed tissue, they were taken immediately adjacent to the assayed tissue itself. The small size of prostate tissue fragments provided by TURP precluded taking a number of frozen sections through the assayed tissue. It was assumed that areas of tissue components measured on a tissue section reflect the underlying volumes of those tissue components in the tissue studied. Furthermore, it was assumed that the relative volumes of tissue components accurately reflects the relative weights of those tissue components. These assumptions have been verified by Rohr et al. (1976). Thereby, as the weight of each prostate chip used in ELISA was known, the percentage volume of 'glandular' tissue (i.e. prostate ducts and acini) and its weight could be calculated in each ELISA sample.
The correlation (Pearson) between % volume of 'glandular' tissue and f3-NGF concentration was calculated.
The ELISA results were recalculated as pg fl-NGF 100 mg-' 'glandular' tissue. Protein extracted from three BPH tissues and three prostate cancers were analysed by Western blot for the presence of ,B-NGF. The prostate specimens with the highest fl-NGF concentrations, as measured by ELISA, were used for Western blotting. A single protein band was found in all the specimens analysed which co-migrated with a sample of positive control rh,B-NGF (Figure 2). Like the RT-PCR, the Western blots were not quantitative. In gels not shown, the protein band found in prostate tissue was run sufficiently far into the resolving gel to exclude the presence of heavy, NGFlike proteins (45 kDa or heavier).
Immunohistochemical staining f3-NGF In BPH tissue specimens (n = 15, i.e. 15 patients), specific and intense staining for ,B-NGF was confined to the epithelium with only occasional stromal cells showing faint staining ( Figure 3). Staining was most intense at the apical border of prostate epithelium (Figure 4). In cancer tissue (n = 15), staining was also confined to epithelium of adenocarcinomas while surrounding stroma was not stained ( Figure 5). Positive staining for ,B-NGF was seen in the adrenal gland controls, and no staining was seen in the negative controls (not shown).
Neuroendocrine cells (PGP9.5) Fifteen prostate cancer specimens were examined for PGP9.5 staining. Positive staining in adenocarcinomas varied from sparse focal groups of cells to a widespread staining as shown in Figure 6. Tumours, however, were simply classified as positive or negative and, of the 15 specimens examined, seven tumours were PGP9.5 positive and eight were PGP9.5 negative. All pancreatic positive control tissues showed staining in cells of the islets of Langerhans, while negative controls showed no staining.

f3-NGF concentrations in prostate tissues
The concentrations of fl-NGF in 15 adenocarcinomatous and 15 BPH specimens were measured using an ELISA assay. The detection limit of the assay was 5 pg ml-1 and the recoveries varied between 27% and 51% (mean+ contain 300 gg each of protein prepared from BPH tissue. Lanes 5, 6 and 7 contain 300 ,ug each of protein prepared from adenocarcinoma tissue. Protein was detected with polyclonal rabbit anti-mouse 2.5S NGF primary antibody. All prostate tissues showed the indicated protein band which co-migrated with recombinant human fl-NGF.  s.d. = 33.9 + 4.879). The distribution of f3-NGF concentrations as expressed per gram wet weight of tissue in BPH and prostate cancer is detailed in Figure 7. The mean + s.d. concentration in BPH tissue (1992+684 pg g-') was lower than the concentration in prostate cancer (3100 + 1503 pg g-') but, because of the significant overlap in the two groups, the difference was not significant (P=0.512). Furthermore, there was no significant correlation between the ,B-NGF concentrations in adenocarcinoma and Gleason score.

Discussion
The data presented in this study demonstrated the immunohistochemical localisation of JJ-NGF to the epithelial elements in prostate tissue. This contrasts to earlier studies showing that fJ-NGF localised to the prostate stroma. While both studies employed the same primary fl-NGF antiserum, there were differences in the techniques employed with regard to blocking agents. The earlier report by Graham et al. (1993) used ovalbumin with a goat secondary antibody, whereas, in the present study, we employed 20% sheep serum along with a sheep secondary antibody. Whether this would account for the differences seen is not clear, but it is well established that the use of blocking and secondary sera from the same species   Figure 7 is reversed. That is, BPH glandular tissue contains more ,B-NGF than does prostate cancer tissue. produces less non-specific staining (Bancroft and Cook, 1994). This, however, does not explain the failure by * Graham et al. (1992) to stain the epithelium in their prostate specimens. Significantly, however, f-NGF is a highly conserved protein, and in support of our findings are previous studies demonstrating that fJ-NGF is localised to )o -*-> @prostate epithelium in other species (Harper and Thoenen, 1980;Shikata et al., 1984;MacGrogan et al., 1990). In addition, in the case of prostate adenocarcinoma, such epithelial staining has been previously described by DeSchryver-Kecskemeti et al. (1987), who localised ,B-NGF to epithelial elements of five human prostatic adenocarcino-  ; 1995 is suggested that the interpretation of many differences in biochemistry of benign and malignant prostate tissues would be affected by the use of morphometric data.
The original report of Cohen et al. (1991) that neuroendocrine differentiation was a powerful prognosticator in prostate adenocarcinoma has not been supported by other authors (Allen et al., 1995). Although the presence of these cells may not be a prognosticator, there is evidence that as prostate adenocarcinomas dedifferentiate (Allen et al., 1995), progress (Aprikian et al., 1994) and become androgen independent (Berner et al., 1993), they contain a higher portion of neuroendocrine elements. It is also clear that these cells do not bear androgen receptors in normal hyperplastic or adenocarcinomatous prostate tissues (Iwamura et al., 1994;Bonkhoff et al., 1993) and that some of their secreted products may be growth promoting for prostate epithelial cells (Power et al., 1991;Pinski et al., 1993). Therefore, it can be postulated that androgen depletion therapies confer a survival advantage to neuroendocrine cells over other prostate cancer cells and that neuroendocrine cells may then promote androgen-independent prostate cancer growth (di Sant-Agnese, 1995).
Because of local experience in staining bronchogenic carcinomas for neuroendocrine elements with antibodies against PGP9.5, that antibody was used to stain neuroendocrine cells in prostate tissue. As far as we are aware, this is the first report of such a use of this antibody. PGP9.5 is a 25 kDa hydrolase which is specifically associated with neural and neuroendocrine cells (Wilson et al., 1988) and which is responsible in those cells for the C-terminal hydrolysisactivation of ubiquitin (Jentsch, 1992).
The proportion of adenocarcinomas showing PGP9.5 staining in the present study was 46.7%, and this was similar to the figure found by di Sant-Agnese and Jensen (1987), using the argyrophil reaction and a battery of immunohistochemical stains (47%). Interestingly, we were unable to detect any correlation between the concentration of the neural differentiation factor f,-NGF and the presence of neural differentiation, as demonstrated by positive PGP9.5 staining.
Western blot analysis of prostate proteins with the primary antiserum used in immunohistochemistry identified a single protein band which co-migrated with a sample of rh,B-NGF. Earlier studies by Djakiew et al. (1991) have also described heavier NGF-like protein products (42-65 kDa) from prostate cell cultures, but there was no evidence of these heavy NGF-like proteins in the human prostate tissues analysed here. NGF-like proteins in the human prostate represent fl-NGF itself, while heavier NGF-like proteins were not detected in native human tissue. fl-NGF in peripheral tissues has predominantly been regarded as a determinant of sympathetic and sensory innervation. Recent studies have demonstrated that, in addition, NGF is involved in the paracrine control of prostate epithelial growth through the production of NGF by prostate stroma (Graham et al., 1992); these act mitogenically via specific receptors on the epithelium (Djakiew et al., 1991).
However, our own data do not support this view as we have immunohistochemically localised ,B-NGF to prostate epithelium and confirmed this by demonstrating the presence of a positive correlation between f,-NGF concentration and the epithelial content of prostate tissue. Malignant prostate epithelia contain less ,B-NGF than do benign epithelia, but the concentration of /3-NGF in malignant tissue, as a whole, is higher than that in benign tissue.
The data presented here suggest that fl-NGF is an endogenous product of the prostate gland and f,-NGF concentrations present in the gland (1-6 ng g -') are similar to concentrations causing neuronal differentiation and growth in vivo (Tischler and Greene, 1975). Graham et al. (1992) and MacGrogan et al. (1992) have localised the low-affinity nerve growth factor to the prostatic epithelium, in keeping with the view that f,-NGF acts upon those cells. As fl-NGF also stimulates the growth of sympathetic neurones, we postulate that, by producing ,B-NGF, prostate epithelium may recruit sympathetic nerves -and hence smooth muscle. That is, that f,-NGF may be one mediator of the stromal-epithelial relationship, produced by the epithelium and acting upon the stroma (Tenniswood, 1986).