Vascular endothelial growth factor (VEGF) expression in prostatic tumours and its relationship to neuroendocrine cells.

Vascular endothelial growth factor (VEGF) expression was examined by immunohistochemistry in 45 prostatic carcinoma specimens and ten benign prostatic tumours (BPH). The majority of carcinoma specimens exhibited cytoplasmic staining for VEGF and showed a trend of increasing expression with dedifferentiation (2p = 0.003). Immunoreactive VEGF was also seen in the prostatic carcinoma cell lines, the order of staining intensity was PC3 > DU145 > LNCaP. Intense granular cytoplasmic staining for VEGF was observed in neuroendocrine-like cells which were seen focally in many of the prostatic specimens. Consecutive sections were incubated with a chromogranin A antibody to confirm the neuroendocrine phenotype of these cells. A significant correlation (P < 0.0001) between the total number of intensely stained VEGF-positive cells and chromogranin A-positive cells was found. A subpopulation of neuroendocrine-like cells also showed intense immunoreactivity for transforming growth factor alpha (TGF-alpha). A correlation was observed (2p = 0.0092) between the intensity of VEGF and TGF-alpha immunostaining in carcinoma cells which were not of neuroendocrine differentiation. The presence of these two angiogenic factors may aid the neovascularisation of carcinomas and their increased expression in tumour-associated neuroendocrine cells may contribute to a more aggressive phenotype. ImagesFigure 2

Angiogenesis, the sprouting of new capillaries, in the vicinity of the tumour cells is a prerequisite of solid tumour growth (Folkman, 1987). A multitude of factors have been shown to influence the proliferation, migration and differentiation of endothelial cells in this process (Folkman, 1990;Zagzag 1995). Measurement of microvessel density (MVD) by immunolocalisation of factor VIII antigen has been used to determine the angiogenic activity of a variety of tumours including breast (Bosari et al., 1992), lung (Yamazaki et al., 1994) and prostatic cancer (Weidner et al., 1993) and been found to have prognostic value. Compounds that inhibit the angiogenic process have shown promise in animal model systems, reducing solid tumour growth in athymic nude mice and in rat prostatic cancer (Kim et al., 1993;Vukanonic et al., 1995).
Vascular endothelial growth factor (VEGF) is a highly specific mitogen for endothelial cells and a potent microvascular permeability factor (Plate et al., 1994;Neufeld et al., 1994) and as such may play an integral role in angiogenesis and thus in the potentiation of solid tumour growth. Indeed, evidence from studies on brain tumours (Weindel et al., 1994), renal cell carcinoma (Takahashi et al., 1994), breast (Brown et al., 1995) and bladder cancer (O'Brien et al., 1995) indicates that VEGF is often up-regulated when compared with the normal tissue counterpart.
In prostatic cancer, measurement of MVD has been found to be prognostic (Hall et al., 1994), whether a relationship exists between MVD and the expression of VEGF and/or basic fibroblast growth factor (FGF) in malignant disease of this gland may have implications for therapy. In this respect assays on breast tumours (Toi et al., 1994) and epidermoid lung carcinomas (Mattern et al., 1996) have demonstrated a correlation between VEGF expression and MVD measurements. Recent studies have indicated that measurement of intratumoral levels of VEGF may be useful in assessing the activity of tumour angiogenesis (Toi et al., 1996). In a study on renal cell carcinomas, VEGF was up-regulated, but not basic FGF (Takahashi et al., 1994). To date, VEGF has not been identified in prostate tissue so it was considered important to determine if this growth factor is involved in neoplastic changes in this organ. A series of prostatic carcinomas, benign tumours and prostatic cell lines were therefore examined for the presence of VEGF.
In the course of this study it became apparent that a subpopulation of cells with neuroendocrine characteristics were expressing immunoreactive VEGF in large amounts. Neuroendocrine cells, which are known to secrete a variety of peptides and biogenic amines (Di Sant'Agnese, 1992), have been found in 50% of prostate cancers (Di Sant'Agnese and de Mesey Jensen, 1987) and their presence indicates a poor prognosis in this malignancy (Cohen et al., 1991). In view of these findings, immunohistochemical assays were undertaken on consecutive sections of the prostatic tissue to determine the neuroendocrine cell distribution and to establish the coexpression of VEGF and chromogranin A in this particular cell type.
Another growth factor which appears to have potential as a prognostic marker in malignant disease (Bebok et al., 1994;Sauter et al., 1995;Reinartz et al., 1994) and indeed has been demonstrated in increasing amounts with dedifferentiation in prostatic adenocarcinoma (Harper et al., 1993), is transforming growth factor a (TGF-cx). Its presence has been noted in prostate cancer cells with neuroendocrine morphology (unpublished data) and in the light of these observations, coupled with the reported angiogenic properties of TGF-a (Yamamoto et al., 1994;Ono et al., 1992), expression of this factor was examined in relation to chromogranin A and VEGF distribution. The role that neuroendocrine cells may play in tumour progression is probably multifold (Aprikian et al., 1993), one important aspect of which might involve secretion of angiogenic substances, such as VEGF and TGFax, aiding capillary growth and vascular permeability. underwent transurethral resection (TURP) of their tumours, while the remaining two patients had open prostatectomy operations. A representative sample of the curettings from the operations was fixed in formal saline followed by paraffin wax embedding for subsequent histopathological examination and the VEGF, chromogranin A and TGF-c immunohistochemical assays. Grading of the tissues was carried out using the WHO classification (Mostofi et al., 1980). Prostatic carcinoma cell lines The DU145, LNCaP and PC3 cell lines were cultured as described previously (Glynne-Jones et al., 1994). For use in immunocytochemical analysis the cells were seeded onto sterile TESPA (3-aminopropyl-triethoxy-silane; Sigma Chemical Co., Dorset, UK) coated coverslips in tissue culture dishes. After 72 h in culture (i.e. still in logarithmic growth) the monolayers were washed in unsupplemented medium and fixed in formal saline (10 min at room temperature), transferred to 70% ethanol (2 x 5 min) and then to phosphate-buffered saline (PBS).
Immunocytochemical analysis Immunocytochemical analysis was performed on 5 ,m consecutive sections of the formal saline-fixed wax-embedded tissues and the cell line monolayers. Sections were dewaxed and taken through graded alcohols to PBS (0.01 M, pH 7.4) followed by incubation for 5 min with 0.3% hydrogen peroxide to block endogenous peroxidase for both the TGF-a and chromogranin A assays. After dewaxing and rehydration the sections assayed for VEGF immunostaining were microwaved 2 x 10 min at maximum power in sodium citrate buffer, pH 6. These sections were washed with PBS, 2 x 5 min, before blocking endogenous peroxidase with hydrogen peroxide.
Blocking of endogenous peroxidase in the coverslip cultures was carried out by incubation with 50 jiM phenylhydrazine in PBS for 10 min at 37°C followed by 20 min incubation at 37°C with 0.005% hydrogen peroxide in 50 gM phenylhydrazine in PBS. The coverslips were then washed, 2 x PBS, before incubation with the VEGF antibody. VEGF assay Immunostaining for VEGF used a rabbit polyclonal antibody (Santa Cruz Biotechnology, supplied by Autogen Bioclear UK, Devizes, Wiltshire, UK). The primary antibody (1 ,ug ml-') was incubated for 1 h at room temperature. After washing the antigen was detected using the streptavidinbiotin Universal kit system (Diagnostic Products, Caernarfon, UK), polyclonal reagents as per protocol and immunolocation of VEGF was visualised with the chromogen 3-amino-9 ethyl carbazole. Sections were lightly counterstained with 10% haematoxylin and mounted in Aquamount (BDH Chemical Co., Poole, UK).
Semiquantitative analysis of cytoplasmic VEGF staining was carried out by two independent observers. Each section was scanned in its entirety and a score was assigned according to the intensity of the immunoreaction in the prostatic epithelial or carcinoma cells (excluding those of neuroendocrine differentiation) (0, no staining; +/-, very low; 1 +, low; 2+, moderate; 3 +, strong). Neuroendocrinelike cells were intensely stained, equivalent to 4+ to 5 + on the above scale of intensity. The total number of these cells were counted.
Chromogranin A assay Sections were washed in PBS before incubation for 1 h with a monoclonal antibody to chromogranin A (Novacastra Laboratories, Newcastle-upon-Tyne, UK) at a dilution of 1/50. Antigen binding was detected using the streptavidin -biotin Universal kit (Diagnostic Products) as per protocol with the chromogen 3-amino-9ethyl carbazole. Sections of human pituitary and adrenal were included in these assays as positive control tissues for the presence of chromogranin in neuroendocrine cells.
A semiquantitative score of 0, + / -, 1 +, 2 + or 3 + was assigned according to the relative abundance of positive cells. In addition the total number of positive cells was counted per section.
TGF-a assay Assay procedures using a monoclonal human TGF-a antibody (Oncogene Science, distributed by Cambridge Bioscience, Cambridge, UK) and the indirect peroxidase-conjugated streptavidinbiotin detection method have been described previously (Harper et al., 1993). The intensity of staining was scored as 0, + / -, 1 +, 2 + or 3 +. In addition the total number of neuroendocrine cells per section identified by their intense TGF-a immunoreaction, was counted.
Absorption studies To test the specificity of the VEGF immunostaining observed, the antibody was incubated overnight with various compounds: VEGF peptide, a 20 amino acid synthetic peptide corresponding to residues 1 to 20 of the amino terminus of VEGF (Santa Cruz) at 10 ig ml-1 final concentration; PDGF-AB human recombinant (Sigma Chemical Co.) at a final concentration 5 jg ml-1; TGF-c, human recombinant (R&D Systems Europe, Abingdon, UK) at 10 jg ml-' final concentration; bovine serum albumin (BSA) (Sigma Chemical Co) at 10% final concentration. Consecutive sections of two prostatic carcinomas with neuroendocrine TGF-a Grade of cancer Figure 1 Histographical analysis of cytoplasmic staining for VEGF and TGF-ac in benign glandular epithelium or carcinoma cells which are not of neuroendocrine differentiation. The percentage of specimens at the various intensity scores is plotted from 0 to 3 +. A trend towards increasing expression with dedifferentiation can be seen in carcinoma specimens. differentiation, a BPH specimen and the prostatic cell lines cultured on coverslips were incubated with the various absorbed and unabsorbed antibody solutions at 1 ig ml-' final concentration and assay by the standard procedure. The VEGF peptide was able to completely absorb the staining seen with the VEGF antibody on the sections. None of the other peptides or proteins diminished the staining on any of the specimens. Furthermore, the staining seen with the TGF-a and chromogranin A antibodies was not altered by prior absorption with the VEGF peptide.

Statistics
Spearman's rank correlation was used to analyse the data obtained. Linear regression plots of the number of neuroendocrine cells detected by chromogranin A, VEGF and TGF-a immunodetection were performed.

Results
Clinical samples VEGF immunoassays Diffuse cytoplasmic staining of 0 to 1 + intensity was observed in BPH secretory epithelium with intense staining of occasional neuroendocrine-like cells within the glandular structures. Neuroendocrine cells present in the TSH-and FSH-secreting cells of the anterior human pituitary and the chromaffin cells of the adrenal medulla also exhibited intense VEGF staining. The majority of carcinoma specimens exhibited cytoplasmic staining for this growth factor which showed a trend of increasing VEGF expression with dedifferentiation ( Figure 1) (2p = 0.003). This result was similar to that seen for TGF-oa cytoplasmic staining in the consecutive sections ( Figure 1) (2p = 0.0349). In the carcinoma specimens a correlation was found between the expression of these two growth factors (2p = 0.0092). Intense (3+) cytoplasmic granular immunolocalisation of VEGF in a neuroendocrine-like cell population was also demonstrated in carcinomas (Figure 2 a and c). These cells were often seen focally in the moderately differentiated carcinomas particularly in cribriform tumours. These strongly stained VEGF cells were observed in 25/45 carcinomas (55%) and 18/22 (81%) BPH specimens.
Another cell type that was intensely stained with the VEGF antibody was that of polymorphonuclear neutrophils as seen in the blood vessels and in the stroma, particularly noticeable in those attached to vessel walls in areas of inflammation or infection in the BPH specimens. VEGF in prostatic tumours ME Harper et al 913 A heterogeneous distribution of neuroendocrine cells was detected in the adenocarcinomas with a range of positivity ( Figure 3). Up to 5% of the tumour cells were chromogranin A-positive in areas of focal neuroendocrine differentiation. Occasionally isolated positive cells were seen in the stroma of infiltrative carcinomas. The incidence of chromogranin Apositive cells was greatest in grade 2 tumours (Figure 3) and it was noticeable that the majority were associated with a cribriform morphology (Figure 2d). There was also evidence, within cribriform areas of the tumour, of disintegration of individual neuroendocrine cells, their contents immunologically detectable in the intraluminal spaces. In benign tissue chromogranin A-positive cells were localised within the glandular epithelium. Generally only a few scattered positive cells were seen with focal immunoreactivity in dysplastic acini, often associated with infection.
Regression analysis of the number of neuroendocrine cells identified by the chromogranin A and VEGF assays in both adenocarcinomas and BPH specimens showed a close correlation (P<0.0001) (Figure 4). Similar distributions of VEGF and chromogranin A-positive cells could be seen in the individual tumours (Figure 2a Figure 3 Relative abundance of chromogranin A-positive cells in BPH and carcinoma specimens showing the degree of neuroendocrine differentiation. A semiquantitative score from 0 to 3 + was assigned and the percentage of specimens at each score is plotted for BPH and the three grades of cancer (Mostofi et al., 1980).  (Figure 4). Some neuroendocrine-like cells, as judged morphologically, stained intensely for TGF-a but were not chromogranin A-positive and vice versa. The remainder of the neuroendocrine cells showed a similar level of TGF-a immunoreactivity as the secretory epithelium of BPH glands or the cytoplasm of the carcinoma cells within the tissue section, as given by the TGF-ax score. There was no correlation between this TGF-a score and the level of neuroendocrine differentiation, as defined by the chromogranin A score. Cytoplasmic staining for TGF-a was seen in the majority of carcinomas which related to histological grade in agreement with previous studies (Harper et al., 1993).
Cell lines VEGF expression was observed in the cell lines PC3, DU145 and LNCaP examined. The staining was diffuse, located in the cytoplasm but not associated with any subcellular fraction. Absorption experiments indicated that this staining detected with the antibody was specific for VEGF. The staining seen was not diminished by incubation of the antibody with PDGF, TGF-a or BSA. The order of expression of VEGF as judged by intensity of staining was PC3 > DU145 > LNCaP.

Discussion
A variety of factors have been found to influence the proliferation, migration and differentiation of endothelial cells but one of the more specific mitogens of capillary endothelium is a member of the platelet-derived growth factor (PDGF) family, vascular endothelial growth factor/vascular permeability factor. Although increased expression of this factor has been detected in several carcinomas when compared with their normal tissue counterparts (Takahashi et al., 1994;O'Brien et al., 1995), its presence has not previously been reported in prostate tumour specimens or prostatic cell lines. This immunohistochemical study indicates that VEGF is present in both the benign and malignant prostate specimens, with the carcinomas having a slightly increased expression as judged by their cytoplasmic staining intensity. It would appear that a source of this particular cytokine is present within the majority of these tumours and may play a role in the neovascularisation essential for tumour growth and metastatic spread.
One of the ways in which the angiogenic activity of a tumour is assessed is by measurement of the microvessel density (MVD) in the tissue specimen by using antibodies to blood vessel markers such as factor VIII. Increased MVD has been correlated with a higher incidence of metastasis and a worse prognosis in both breast (Weidner et al., 1992;Horak et al., 1992) and prostatic carcinoma (Vesalainen et al., 1994;Wakui et al., 1992). Two studies in prostate cancer, one in invasive carcinoma (Weidner et al., 1993) and the other in clinically localised disease (Hall et al., 1994), concluded that MVD had prognostic potential. In non-small-cell lung carcinoma the MVD measurements also related to progression (Macchiarini et al., 1992).
There is some evidence that angiogenic activity is more important at earlier stages of growth when the tumour changes from an avascular to vascular phase (Vesalainen et al., 1994). Experiments on tumour transplantation into nude mice indicate that tumour take and growth are very dependent on this feature, more so than mitotic rates, and furthermore, administration of antibodies to VEGF has been shown to slow xenograft growth (Kondo et al., 1993;Kim et al., 1993).
It is possible that alternative angiogenic pathways may be operating in the various tumours. In bladder cancer, VEGF expression was higher in superficial tumours than invasive tumours whereas platelet-derived endothelial cell growth factor expression was higher in the invasive compared with the superficial bladder tumours (O'Brien et al., 1995). In this same study VEGF expression in the superficial bladder cancer patients predicted progression.
Basic fibroblast growth factor (FGF) is also a potent mitogen of endothelial cells as well as many other cell types and is expressed in a variety of tumours (Benharroch and Birnbaum, 1990). In the transplantable human prostatic carcinoma line, Tenl2, which is extremely angiogenic in nude mice, VEGF expression was not detected, but its secretion of basic FGF however was found to be 5-fold higher than that of the other prostate cell lines, PC3, DU145 and LNCaP (unpublished work). All three in vitro prostatic cell lines expressed VEGF as detected immunohistochemically. It is possible that the prostatic carcinomas which were negative for VEGF (15%) in this study may express alternative angiogenic compounds such as basic FGF which aid their neovascularisation. In a study of benign and malignant breast tumours higher levels of acidic and basic FGF mRNAs were found in the benign specimens although approximately 25% of the carcinomas had equally as high concentrations of basic FGF mRNA (Anandappa et al., 1994). This indicates that carcinomas, within an organ type, have the ability to express and possibly use different cytokines for growth and angiogenesis.
Of interest in this current study was the observation that the neuroendocrine cells present in the prostatic tumours exhibited intense staining for VEGF. A correlation was found between the total number of chromogranin A-positive cells and the total number of VEGF-containing neuroendocrinelike cells in individual tumours. Furthermore, consecutive sections indicated that these were almost certainly the same cells (see Figure 2 a-d). These VEGF-positive neuroendocrine cells were more frequently observed in cribriform tumours which is reminiscent of the distribution of panethlike cells described in prostate tumours (Adlakha and Bostwick, 1994).
Neuroendocrine cells were found in 55% of the carcinoma specimens in this current study which is a similar incidence to that reported by other groups (Cohen et al., 1990). They express prostate-specific antigen (PSA) (Aprikian et al., 1993;Cohen et al., 1992) and are believed to represent a third epithelial type originating from a common stem cell. As these cells do not appear to proliferate (Bonkhoff et al., 1991), it is assumed that the increased numbers seen in the tumours originate from prostatic carcinoma cells which are induced to undergo neuroendocrine differentiation. Prostatic neuroendocrine cells lack androgen receptor and it has been suggested that they could represent an androgen-insensitive cell population that can expand in some cases to populate the tumour (Krijnen et al., 1993). There does, however, appear to be three types of prostatic tumours with neuroendocrine differentiation (Di Sant'Agnese, 1992) and the most common pattern, adenocarcinoma with focal neuroendocrine differentiation, is the one seen in this study. Small-cell carcinoma of the prostate is, however, prostate-specific antigen (PSA) and androgen receptor negative and, as elsewhere, is a highly expansive and aggressive tumour. The majority of the prostatic neuroendocrine cells secrete serotonin and to lesser extents TSH, calcitonin and gastrinreleasing peptide (Abrahamsson et al., 1987). Many of these peptides are thought to influence progression of the tumours via their growth-regulatory activities. More importantly the number of neuroendocrine cells in prostatic cancer correlated with poor prognosis (Cohen et al., 1991;Turbat-Herrera et al., 1988). The presence of a whole variety of neuropeptides and cytokines in these cells, which could act in a paracrine manner on the neighbouring cancer cell population, probably accounts for their relationship to tumour aggressiveness and progression. Increased expression of VEGF in neuroendo-915 cnne cells could also influence progression via its local angiogenic activity and. because of its action on vascular permeability, might facilitate metastatic spread. Expression of the angiogenic factors PDGF and basic FGF has been demonstrated in neuroendocnrne tumours of the digestive system (Chaudhry et al.. 1993) and marked vascular proliferation observed in brain metastases of small-cell lung carcinomas which related to basic FGF expression (Ito et al.. 1993) but no evaluation of VEGF expression was made in either study.
The observation that TGF-2 is expressed in a subset of the neuroendocrine cells would suggest that this factor may contribute to the unfavourable prognosis of the particular patient. TGF-z, in addition to having angiogenic properties (Schreiber et al.. 1986). has prognostic relevance (Bebok et al.. 1994: Sauter et al.. 1995 and is also capable of stimulating prostatic cell growth in an autocrine and or paracrine manner via the EGFR receptors which are present in many of the prostate adenocarcinomas (Harper et al.. 1995). Recently TGF-x has been reported to be expressed in all 30 neuroendocrine tumours examined, which included carcinoid tumours of the midgut. medullary thyroid carcinomas and phaeochromocytomas (Nilsson et al., 1995). Large amounts of TGF-x were also found to be secreted into the medium by primary cultures of carcinoid tumours and phaeochromocytomas (Nilsson et al., 1995). Several other growth factors. IGF-1 IGF-I1 PDGF and TGF-fi, have also been detected in carcinoid tumours (Beauchamp et al.. 1991: Chaudhry et al.. 1992: Nilsson et al.. 1992. There are reports that the cell lines PC3 and LNCaP display certain neuroendocrine characteristics expressing markers of the neural crest in addition to epithelial markers. In response to cyclic AMP analogues they were shown to differentiate terminally, with epithelial cell markers being down-regulated whereas neurone-specific enolase was elevated (Bang et al.. 1994). Furthermore. all three prostate cell lines have been shown to be growth inhibited by selective serotonin receptor antagonists such as pindobind (Abdul et al.. 1994). PC3. DU145 and LNCaP cell lines are known to express TGF-.7 (Schurmans et al.. 1991: Ching et al.. 1993: Liu et al.. 1993 and in the latter line the response is modulated by androgens. It was noticed that some TGF-a-containing neuroendocrine-like cells were negative for chromogranin A and may represent a cell subpopulation which is chromogranin B positive. In some poorly differentiated prostatic tumours chromogranin B is the major component (Schmid et al.. 1994).
Polymorphonuclear leucocytes were also strongly stained with VEGF antibody which appeared specific according to the absorption studies. They were noticeable in vessels adjacent to infection and or inflammation. It is possible that the vascular permeability function of this factor may be more relevant in this situation.
In conclusion, this study indicates that the majority of prostatic tumours express VEGF which may aid their neovascularisation. The neuroendocrine cells present focally in many of the tumours appear to express VEGF in large amounts and a subpopulation of these secrete TGF-x. In addition to their production of a variety of compounds which can act in a paracrine manner to stimulate adjacent cancer cell growth, the storage of these two angiogenic factors in such cells provides another pathway by which neuroendocrine cells could influence the progression of this disease.