Impaired prostate tumorigenesis in Egr1-deficient mice

SARKI A. ABDULKADIRI¹, ZHICAN QU¹, EMILY GARABEDIAN², SHENG-KWEI SONC³, THOMAS J. PETERS¹, JOHN SVAREN⁴, JOSEPH M. CARBONE⁵, CATHY K. NAUGHTON⁵, WILLIAM J, CATALONA⁵, JOSEPH J.H. ACKERMAN^3,6, JEFFREY I. GORDON², PETER A. HUMPHREY¹ & JEFFREY MILBRANDT¹

¹Department of Pathology, ²Departments of Molecular Biology and Pharmacology,
³Department of Chemistry, Washington University, Box 8118, 660 S. Euclid Avenue, St. Louis, Missouri 63119, USA,
⁴Department of Comparative Biosciences, University of Wisconsin, Madison, Wisconsin 53706, USA,
⁵Division of Urologic Surgery. ⁶Departments of Medicine and Radiology, Washington University School of Medicine
Correspondence should be addressed to J.M.: email:jeff@milbrandt.wustl.edu

The transcription factor, early growth response protein 1 (EGR1), is overexpressed in a majority of human prostate cancers and is implicated in the regulation of several genes important for prostate tumor progression. Here we have assessed the effect of Egr1 deficiency on tumor development in two transgenic mouse models of prostate cancer (CR2-T-Ag and TRAMP). Using a combination of high-resolution magnetic resonance imaging and histopathological and survival analyses, we show that tumor progression was significantly impaired in Egr1^-/-mice. Tumor initiation and tumor growth rate were not affected by the lack of Egr1; however, Egr1 deficiency significantly delayed the progression from prostatic intra-epithelial neoplasia to invasive carcinoma. These results indicate a unique role for Egr1 in regulating the transition from localized. carcinoma in situ to invasive carcinoma.

Cancer mortality is principally associated with the consequences of tumor invasion and dissemination. Therefore, understanding the processes by which localized, carcinoma in situ evolves into invasive carcinoma will likely have major therapeutic implications. In prostate cancer, disease progression is thought to proceed through multiple steps, from prostatic intra-epithelial neoplasia (PIN) to locally invasive carcinoma to metastatic disease¹. Progression through the various stages of tumorigenesis presumably requires altered expression of oncogenes and tumor suppressor genes in the initiated cell. Ultimately, these genetic lesions change the cancer cell phenotype by altering global gene expression patterns.

One molecule repeatedly isolated in screens aimed at identifying genes overexpressed in prostate carcinoma is the transcription factor, early growth response factor 1 (EGRi)^2,3. EGRl is an immediate early gene product that has been implicated in multiple cellular processes including growth, differentiation, apoptosis, neurite outgrowth and wound healing^4-7. Egr1-deficient mice develop normally; however, the females are infertile due to a defect in luteinizing hormone-ß (LH-ß) production^R.

Two novel corepressor molecules, NAB1 and NAB2, can modulate EGRl transcriptional activity^9,l0. NAB1 is widely expressed at low levels while NAB2 displays a more tissue-restricted and regulated expression pattern¹⁰. NAB2 is also induced in a delayed fashion by the same stimuli that induce EGR1 and may be required to dampen EGR1 activity in a negative feedback loop^10,11.

Several observations point to the significance of EGR1 over-expression by prostate tumors. Levels of EGR1 correlate with the tumor Gleason grade³, and several EGR1 target genes are implicated in prostate tumor progression (for example, IGF-II, TGF-ß1 and PDGF-A chain)¹², In addition, a majority of primary human prostate carcinomas have lost expression of the corepressor NAB2 (Abdulkadir et al., submitted), indicating that EGR1 transcriptional activity is elevated and unrestrained in many prostate carcinomas.

To determine the functional significance of Egr1 overexpression in prostate tumors in vivo, we exploited the cryptdin-2-T-antigen (CR2-T-Ag) model of prostate cancer¹³. CR2-T-Ag mice express simian virus 40 (SV40) T antigen under the control of transcriptional regulatory elements from the mouse gene encoding cryptdin-2 (Defcr2). This targets transgene expression to the neuro-endocrine cells of the prostate, resulting in the development of carcinoma with 100% penetrance. Prostate cancer progresses in this model in a typical fashion, from a distinct in situ phase {PIN), to invasive carcinoma with metastatic spread to lymph nodes, liver and bone¹³. We generated and analyzed CR2-T-Ag mice that are deficient in Egr1 (ref. 8). Our results indicate that a lack of Egr1 does not affect cellular transformation or the development of PIN. Rather, the progression to invasive carcinoma is significantly impaired. We confirmed these results using the transgenic adenocarcinoma mouse prostate (TRAMP) model of prostate carcinoma, In which T-antigen expression is targeted to the luminal epithelial cells of the prostate¹⁴. Our findings show that Egr1 has an important role in regulating the progression from PIN to invasive carcinoma.

Aberrant expression of Egr1 and Nab2 in CR2-T-Ag mice

Overexpression of EGR1 and downregulation of the EGR1 corepressor NAB2 are common in human prostate carcinomas^2,3,12 (Abdulkadir et al., submitted). We investigated whether these gene expression changes also occur in the prostate tumors from CR2-T-Ag mice. Using immunohistochemistry, we found overexpression of Egr1 and loss of Nab2 expression in tumors and PIN lesions from CR2-T-Ag mice (Fig. 1a and b), To determine whether neuro-endocrine cells in the normal prostatic epithelium express Nab2, we used cryptdin-2-human-growth-hormone (CR2-hGH) mice that express a more neutral reporter (hGH), under the control of the same regulatory elements used to generate CR2-T-Ag mice^l3. CR2-hGH mice have no discernible phenotype^l3. In normal neuro-endocrine cells from CR2-hGH mice, hGH expression co-localized with Nab2 expression (Fig. 1c-f). Moreover, CR2-T-Ag PIN lesions contain occasional cells that co-expressed both Nab2 and T antigen (arrow in Fig. 1b). indicating that loss of Nab2 expression follows T-antigen expression and cellular transformation in these cells.

Increased tumor latency in CR2-T-Agi/Egr1^-/- mice

To examine the functional significance of Egr1 overexpresslon in prostate tumors in vivo, we crossed FVB/N CR2-T-Ag mice with C57B1/6 Egrl^-/- mice to generate CR2-T-Ag mice deficient in Egr1 (CR2-T-Ag/Egrl^-/- mice). Initial analysis of mice from these crosses indicated an effect of the C57BL6 genetic background on T-Ag expression. In the FVB/N background, PIN lesions are observed in the dorsolateral prostate by 8 weeks of age¹³, but in the F2 background, PIN lesions arise in the anterior prostate by 15 weeks in most mice (data not shown; see also Fig. 4e). Therefore, all subsequent analyses were conducted on littermates with equivalent FVB/N and B6 contributions to their genetic backgrounds.

Comparative analysis of prostate cancer-related mortality indicated a significant effect of Egr1 on the survival of CR2-T-Ag mice. CRZ-T-AgiEgr1^-/- mice survived for longer periods than either CRZ-T-Ag/Egr1^+/+ (P -0.0006) or CR2-T-Ag/Egr1^+/- (P = 0.0001) mice (Fig. 2). For CR2-T-Ag/Egr1^+/+ and CR2-T-Ag/Egrl^+/- mice, the mean survival times were 36.5 weeks and 35 weeks respectively, whereas the mean survival time for CR2-T-Ag/Egrl^-/- mice was 46.5 weeks. We saw no significant difference in survival between CR2-T-Ag/Egrl^+/+ and CR2-T-Ag/Egrl^+/- mice (P = 0.08). These data demonstrate that lack of Egr1 significantly delays prostate cancer progression in this model.

To facilitate further analysis of tumor development and progression in vivo, we developed a magnetic resonance imaging (MRI) assay for visualizing tumors in the mouse prostate. MRl permitted longitudinal analysis of tumor progression in the same mouse, including an estimation of the time of tumor appearance and measurement of the tumor growth rate. Using this assay, prostate tumors as small as 1 mm in diameter can be detected. Analysis of tumor incidence by MRI showed that both CR2-T-Ag/Egr1^+/+ and CR2-T-Ag/Egl^+/- (control) mice begin to develop detectable tumors by 20 weeks (Fig 3a). By 35 weeks, 90% of the control mice show evidence of tumor detectable by MRI (n = 23). By contrast, the appearance of tumors in CR2-T-Ag/Egrl^-/- mice was significantly delayed, with 70% of CR2-T-Ag/Egrl^-/- mice (n = 16) being free of MRI-detectable tumor at this age (P < 0.05). No discernible differences in tumor incidence were observed between wild-type and heterozygous mice (data not shown).

The increased tumor latency observed in CR2-T-Ag/Egrl^-/- mice could be due to: 1) a reduced rate of initiation, for example resulting from a reduced expression of T antigen or a decrease in T-antigen transforming efficiency in the Egr1-deficient background; 2) a reduction in the tumor growth rate; or 3) a reduction in the rate at which PIN lesions progress to carcinoma. We found no evidence of impaired expression of T antigen in CR2-T-Ag/Egrl^-/- mice. Quantitative reverse transcriptase (RT)-PCR analysis of T-antigen expression showed no significant differences between tumors in control and CR2-T-Ag/Egrl^-/- mice (data not shown), and comparable levels of T-antigen immunoreactivity were observed in lesions arising in both groups of mice (Fig. 4a and b). Moreover, CR2-T-Ag/Egrl^-/- mice developed PIN lesions at the same rate as control mice (see Fig. 4e), indicating adequate expression of T antigen in the absence of Egr1.

To investigate the effects of Egr1 deficiency on the rates of cellular proliferation and apoptosis in the prostate tumors, we used Ki-67 staining (to assess proliferation) and TUNEL assay (to assess apoptosis). PIN lesions and tumors from both groups of control mice showed evidence of high rates of proliferation and apoptosis. and quantitative analysis showed no significant differences between the two groups in these rates (data not shown). This indicates that Egr1 might not be exerting its effect by modulating the cellular growth rate. To address this issue directly, we measured the tumor-doubling time in vivo in cohorts of both control groups and CR2-T-Ag/Egrl^-l- mice using MRI. This analysis showed similar mean tumor-doubling times for mice of all three genotypes (12.83 + 0.18d. Fig. 3b).

The PIN-to-carcinoma transition is impaired in Egr1^-/- mice

Histo-morphometric analyses revealed that the development or PIN is not impaired in CR2-T-Ag/Egr1^-/- mice, as similar numbers or PIN lesions were observed in prostates or control and CR2-T-Ag/Egr1^-/- mice (Fig. 4e). The transition from PIN to invasive carcinoma is a key step in prostate cancer progression. Morphologically, PIN is distinguishable from carcinoma by the presence of an intact basal-cell layer. To examine whether the transition from PIN to carcinoma is affected in CR2-T-Ag/Egr1^-/- mice, cohorts of control and CR2-T-Ag/Egr1^-/- mice were killed at 15, 25 or 35 weeks of age. Serial sections of their entire prostates were examined for the presence of foci of carcinoma. At 15 weeks, none of the mice of either genotype had developed invasive carcinoma (n = 8). By 25 weeks, however, 5/6 of the control mice had foci or invasive carcinoma, compared with only 1/5 among the CR2-T-Ag/Egr1^-/- mice. By 35 weeks, 6/6 of the control mice had foci of invasive carcinoma compared with 2/5 of the CR2-T-Ag/Egr1^-/- mice. The differences between control and CR2-T-Ag/Egrl^-/- mice at 25 and 35 weeks are highly significant (P < 0.001). Additionally, the average tumor diameter in the control group is higher than in CR2-T-Ag/Egr1^-/- mice (1.84 mm versus 0.05 mm at 25 wk and 9.07 mm versus 1.83 mm at 35 wk; n = 5-6). This likely reflects the fact that at a given age, fewer CR2-T-Ag/Egr1^-/- mice have developed invasive carcinoma. These results therefore indicate a delay in the ability of CR2-T-Ag/Egr1^-/-, mice to progress from PIN to invasive carcinoma.

Angiogenesis is closely related to tumor invasion, as adequate neovascularization is required for tumor invasiveness¹⁵. Using antibody against von Willebrand factor to highlight blood vessels, we found similar microvessel densities (MVD) among prostates from 15-week-old control and CR2-T-Ag/Egr1^-/- mice with PIN. By 25 weeks, an increase in the microvessel density was apparent in the control mice but not in the CR2-T-Ag/Egr1^-/- mice. In the control mice, MVD increased from 5.75+ 1.9 at 15 weeks to 15.67+ 3.77 at 25 weeks (P < 0.05; n = 4 for each time point). In CR2-T-Ag/Egr1^-/- mice, however, MVD was 6.33+0.94 at 15 weeks and 8.67+1.8 at 25 weeks (not sig., n = 4 for each time point). These results show a correlation between the degree of angiogenesis and the presence of invasive carcinoma in this model. We have not, however, excluded the possibility that the lower rate of progression to invasive carcinoma In CR2-T- AgiEgrl^-/- mice is the cause, rather than a result, of the lower microvessel density.

Altered patterns of gene expression in CR2-T-AgIEgr1^-/- Tumors

The PIN-to-carcinoma transition is a complex process involving loss or the basal-cell layer, disruption or the basement membrane, alterations in cell-cell and cell-matrix interactions, cellular motility and angiogenesis. As several Egr1 target genes may function in these processes, we examined target gene expression in CR2-T-Ag tumors. In situ hybridization analysts showed increased expression of TGF-ß1 in tumors and high-grade PIN lesions from CR2-T-Ag/Egr1^+/+ but not CR2-T-Ag/Egr1^-/- mice (Fig. 5a-c). Similar results were obtained for PDGF-A (Fig. 5d-f). These results were confirmed by quantitative RT-PCR analysis (Fig. 5g). TGF-ß1 can induce the invasive phenotype in many carcinoma cells in a cell-autonomous manner¹⁶, and PDGF promotes cell motility¹⁷. Taken together, these results indicate that the PIN-to-carcinoma transition in this model is mediated at least in part by aberrant expression of these known Egr1 target genes.

Nab2 represses Egr1-mediated target gene activation

To establish a mechanistic link between Nab2, Egr1 and Egr1-dependent gene expression in prostate carcinoma cells, we used adenovirus-mediated gene transfer to overexpress these molecules in the LAPC4 human prostate carcinoma cell line¹⁸. LAPC4 cells were infected with recombinant adenovirus expressing either Egr1 or an Egr1 mutant (Egr1-I293F) singly or in combination with a Nab2-expressing adenovirus. Egr1-I293F does not bind Nab2 and is thus resistant to repression by Nab molecules^9,10,19. After 24 hours of infection, visualization or Green Fluorescent Protein (GFP) expression (which is expressed from an independent transcription unit in the same adenovirus) demonstrated that greater than 95% or the cells were infected with the adenovirus.

Expression or Egr1 or Egr1-1293F significantly induced the endogenous TGF-ß1 gene (Fig. 6}. Co-expression of Nab2 inhibited the activation or TGF-ß1 expression in response to Egr1, but did not affect the activation mediated by Egr1-J293F, demonstrating the specificity or the interaction. These results indicate that the ability of Egr1 to activate target genes is modulated by Nab2 in human prostate carcinoma cells.

Impaired prostate tumorigenesis in TRAMP/EGR^-/- mice

To extend our studies of the role of Egr1 in prostate carcinogenesis to a non-neuro-endocrine cell-derived tumor model, we generated TRAMP mice deficient in Erg1 (TRAMP/Egr1^-/- mice). TRAMP mice express the SV40 T antigen under control of the promoter of the probasin gene and develop prostate tumors that originate from the luminal epithelial cells of the prostate¹⁴. Analysis of mice at 20, 25 and 35 weeks of age revealed a significant delay in tumor formation in TRAMP/Egrl^-/- mice (Fig. 7). To determine whether this delay in tumor progression is due to an effect on the rate or PIN formation, we compared the number of PIN lesions in TRAMP/Egrl^-/- mice to that in control TRAMP/Egrl^+/+ and TRAMP/Egr1^+/- mice. We chose to analyze mice at 20 weeks, because at this age all mice have PIN but most of the lesions have not yet progressed to form large tumors. The number of PIN lesions per section per mouse was 29.2 + 13.6 foci in TRAMP/Egrl^-/- mice (n = 4), compared to 39.7 + 15.8 foci in the control mice (n = 5). The difference between the two groups was not statistically significant (P = 0.33). These results indicate that the role of Egr1 in tumor progression extends to tumors derived from prostate luminal epithelial cells.

Discussion

Here we provide evidence for the involvement of the transcription factor Egr1 in regulating the PIN-to-carcinoma transition. EGR1 expression is frequently elevated in human prostate cancer, and levels of the EGR1 corepressor NAB2 are down regulated in a significant proportion (> 80%) of prostate cancers (Abdulkadir et al., submitted). This imbalance in the levels of an activator (EGR1) and its corepressor (NAB2) is likely to result in high EGR1 transcriptional activity in prostate carcinoma cells. Our adenovirus-mediated gene transfer experiments in prostate carcinoma cells provide direct support for this hypothesis. Consistent with this notion, several EGR1 target genes are overexpressed in primary prostate carcinomas^12,20,21.

Our experiments with two different prostate cancer models provide genetic evidence that Egr1 promotes the PIN-to-carcinoma transition, Though Egr1^-/- mice developed PIN in response to the expression of T antigen at the same rate as their wild-type or heterozygous counterparts, the development of invasive lesions in Egr1^-/- mice was significantly delayed. This difference is not due to detectable differences in the expression levels of T antigen or in the tumor growth rates. Furthermore, although Egr1^-/- mice have slightly lower levels of LH, their serum testosterone levels are within normal limits²², making it unlikely that the dramatic effects on tumor development we observed are secondary to differences in the levels of testosterone. In addition, the tumors that arise in CR2-T-Ag mice are resistant to castration, and the neuro-endocrine cells from which they arise do not express the androgen receptor¹³. Our results therefore indicate a role for Egr1 in prostate carcinogenesis, in which the PIN-to-carcinoma transition is delayed in the absence of Egr1, without a significant effect on tumor growth rate.

Expression of the Egr1-target genes TGF-ß1 and PDGF-A was altered in tumors from CR2-T-Ag transgenic mice. TGF-ß1 is commonly overexpressed in prostate carcinomas, and has been implicated in various aspects of tumor progression, including the regulation of tumor invasiveness^16,23-25. TGF-ß1 also mediates the ability of cyclosporine to confer an invasive phenotype on non-transformed cells¹⁶. PDGF is mitogenic for prostate cancer cells and is overexpressed in prostate carcinomas and PIN lesions²¹. PDGF can also stimulate cellular motility¹⁷, which is an important component of the tumor invasion process.

Cancer invasion is thought to represent a deregulated form of a physiological invasion process required for neurite outgrowth, tissue remodeling and wound healing²⁶. Due to the many similarities between the wound healing process and tumorigenesis, tumors have been called (wounds that do not heal)²⁷ and, intriguingly, Egr1/Nab2 have been implicated in both wound healing and neurite outgrowth. Nab2 overexpression blocks neurite outgrowth in nerve gr:owth factor-treated PCl2 cells in an Egr-dependent manner⁸, and the inhibition of Egr1 expression by DNA enzyme targeting slows vascular wound healing²⁸. In physiological settings, Egr1 activity is transient through regulation by Nab2, as Egr1 induces expression of the Nab2 gene in a negative feedback loop. Therefore it is unexpected that Nab2 expression in prostate tumors is reduced or absent in the face of high levels of Egr1. Egr1 and Nab2 may represent examples of transcription factors that normally regulate controlled invasive processes like healing and neurite outgrowth. but which get co-opted by tumors to promote uncontrolled invasion during the tumorigenic process.

Both of the transgenic models used in our study utilize the SV40 T antigen to induce prostate carcinoma. As T antigen does not appear to have a role in human prostate cancer, the role of Egr1 in tumors induced by oncogenes and/or tumor suppressors implicated in human prostate cancer remains to be determined as do the roles of Egr1/Nab2 in other neoplasias. Egr1 is overexpressed in multi-step cardnogenesis of the skin²⁹. Egr1 also interacts with the Wilms tumor gene product WT-l, as the two proteins bind similar DNA elements, and ectopic expression of Egr1 antagonizes the tumor suppressor activity of WT-I (ref. 30).

For the analysis of tumor progression in vivo, we used MRI technology which offers distinct advantages, including the ability to monitor tumor behavior (growth or regression) in the same mouse over time, thereby increasing the statistical power of the analysis reducing the number of mice required for study.

In conclusion, we have explored the roles or Egr1 in tumorigenesis using transgenic mouse models of prostate cancer. Our data indicate that Egr1 regulates a gene program that promotes the progression from carcinoma in situ to invasive carcinoma, and provide a basis for the investigation of the involvement of Egr1 in other epithelial cancers.

Methods

MRI analysis. Anesthetized mice were scanned with an MRI system composed of a Varian INOVA console interfaced with an Oxford Instruments 4.7 Tesla magnet equipped with a 16-cm inner-diameter. actively-shielded gradient coil. A fast spin-echo imaging sequence was used to acquire multi-slice images (29 slices without gap) centered on the prostate gland and covering the whole abdominal cavity. MRI acquisition parameters were: TR = 5s; TE = 20ms; field of view, 3 cm x 3 cm; data matrix, 256 x 256 (in-plane resolution 117 µm x 117 µm); slide thickness, 300 µm. Post-mortem examination of over 30 mice was used to verify MRI findings. Details of the MRI assay will be presented elsewhere (Song et al., submitted). For determinIng tumor volume (TV) , Varians Image Browser software was used. The calculation of TV is based on the summation of the manually segmented tumor areas from all images across the prostate. Tumors grew exponentially as evidenced by linearity in the semi-log plot of TV_Norm (tumor volume at any time D2, in days, divided by the first detected tumor volume at time D1) versus elapsed time (D2 - Dl). When the data are represented in thls manner, the tumor-doubling time, DT equals Log 21/slope, as determined by least-squares analysis. Tumor initiation volume (TV₀₁) was set as 5 µl. and the tumor initiation age (D1) was estimated with knowledge or DT from the following equation:

LogTVD2-LogTVD1

D1=D2-__________________ DT

Log2

where TVD_D2 is tumor volume at time D2. In some cases, more than one tumor nodule was detected from the MRI scan, in which case tumor volume was taken as the sum of all nodules present.

Immunohistochemistry. Immunohistochemistry was performed as described¹³. The following primary antibodies were used: SV40 T antigen (from D. Hanahan); Ki67 (Vector, Burlingame, California); synaptophysin and von Willebrand (DAKO. Carpinteria, California); hGH (Cortex Biochem, San Leandro, California); Egr1 (Santa Cruz Biotech, Santa Cruz, California); Egr1 6H10³²; Nab2³³ (from J. Johnson). For quantitation of angiogenesis, vessels were highlighted using rabbit antibody against von Willebrand factor and a blinded observer counted at least 5 fields (magnification: x200), n = 4 mice for each genotype per time point.

Proliferation and apoptosis. Ki67 immunoreactivity and the TUNEL assay were used to assess proliferation and apoptosis respectively. The numbers or positive cells per 1000 prostate epithelial cells were counted (magnification: x400). The numbers of cells counted in some early lesions were occasionally < 1000. For the TUNEL assay, areas of necrosis within tumors were excluded from quantification. For each time point, at least 3 mice of each genotype were examined.

Morphometric analysis. Prostates were serially sectioned in their entirety. and every tenth section was stained by H&E and evaluated by the blinded pathologist (PAH) for the presence of PIN and invasive carcinoma. The extent of PIN was quantified by recording the number of separate foci of PIN (refs. 13, 34) per section. The size of carcinomas was determined with an ocular micrometer.

Adenovirus Infection and quantitative RT-PCR analysis. Details of the adenovirus constructs used and infection of the LAPC4 cells have been reported¹². For co-infections, viruses were used at a 1:1 ratio. Total RNA was Isolated using Trizol (Gibco) and cDNA prepared as described ³⁵. Quantitative. real-tlme RT-PCR (TaqMan, Applied Blosystems, Norwalk, Conecticut} using SYBR-GREEN was performed as reported^12,36. Primers and details of PCR conditions are available upon request.

In Situ Hybridization. Fixed frozen sections (16 µm} were treated with active 0.1% DEPC for 15 min twice, then immersed in 5x SSC for 5 min. Hybridization was performed at 58^oC for 40 h In 50% formamide/5x SSC (pH 7.5}. A 445 bp fragment of rat TGF-ß1 cDNA and a 260 bp fragment of rat PDGF-A chain cDNA were used to generate digoxigenin-labeled cRNA probes. After washing, signal was detected using alkaline phosphatase-coupled anti-digoxigenin antibodies (Boehringer} and development with NBT/BCIP. E18 mouse liver and spinal cord served as positive controls for TGF-ß1 and PDGF-A expression respectively. H&E stained adjacent sections were used for orientation, and sense probes served as negative controls.

Acknowledgments
We thank N. Greenberg for TRAMP mice and G. Gavrilina, T. Gorodinsky, S. Audrain and C. Bollinger for technical assistance. This work was supported by NIH grants 5 P01 CA49712-OB (J.M) R24 CAB3060 {J.J.A.H.)and by grants from The Association for the Cure of Cancer of the Prostate (J.M.). S.A.A. was supported by NIH grants 5 T32CA09547-13 and 1KOBCAB790101.

RECEIVED 29 JUNE; ACCEPTED 8 NOVEMBER 2000

1. Nagle, R.B., Petein, M., Brawer, M., Bowden, G.T. & Cress, A.E. New relationships between prostatic Intra-epithelial neoplasia and prostatic carcinoma. J. Cell. Blochem. Suppl. 26-29 (1992).

2. Thigpen, A.E. et al.lncreased expression of early growth response-1 messenger ribonucleic acid in prostatic adenocarcinoma. J. Urol. 155, 975-981 (1996)

3. Eid, M.A., Kumar, M.V., Iczkowski, K.A., Bostwick, D.G. & Tindall, D.J. Expression of early growth response genes in human prostate cancer. Cancer Res. 58. 2461-246B (1998).

4. Gashler, A. & Sukhatme, V.P. Early growth response protein 1 (Egr.1): prototype of a zlnc-finger family of transcription factors. Prog. Nucl. Acid Res. Mol. Bioi. 50. 191-224 (1995).

5. Khachigian, L.M., Lindner, V., Williams. AJ. & CollIns, T. Egr.1-induced endothelial gene expression: a common theme in vascular injury. Science 271. 1427-1431 (1996).

6. Qu. Z. et al. The transcriptional corepressor NAB2 inhibits NGF-induced differentiation of PC12 cells. J. Cell. Biol. 142. 1075-1082 (1998).

7. Silverman, E.S. & Collins, T. Pathways of Egr-1-mediated gene transcription in vascular biology [comment]. Am. J. Pathol. 154, 665-670 (1999).

8. Lee, S.L et al. Luteinizing hormone deficiency and female infertility in mice lacking the transcription factor NGFI-A (Egr-1). Science 273, 1219-1222 (1996).

9. Russo. M.W., Sevetson, B.R. & Milbrandt, J. Identification of NAB1. a repressor of NGFI-A and Krox20 mediated transcription. Proc. Natl. Acad. Scl. USA 92,6873-6877 (1995).

10. Svaren, J. et al. NAB2, a Corepressor or NGFI-A (Egr-1) and Krox20, is induced by proliferative and differentiative stimuli. Mol. Cell. Biol. 16,3545-3553 (1996).

11. Miano, J.M. & Berk, B.C. NAB2: a transcriptional brake for activated gene expression in the vessel wall? [comment]. Am. J. Pathol. 155, 10O9-1012 (1999).

12. Svaren, J. et al. EGR1 Target genes in prostate carcinoma cells identified by microarray analysis. J. Biol. Chem. In Press.

13. Garabedian, E.M., Humphrey, P.A. & Gordon, J.I. A transgenic mouse model of metastatic prostate cancer originating from neuro-endocrine cells. Proc. Natl. Acad. Sci. USA 95,15382-15387 (1998).

14. Greenberg, N.M. et al. Prostate cancer in a transgenic mouse. Proc. Natl. Acad. Sci. USA 92, 3439-3443 (1995).

15. Skobe, M., Rockwell, P., Goldstein, N., Vosseler, S. & Fusenig, N.E. Halting angio-genesis suppresses carcinoma cell invasion. Nature Med. 3, 1222-1227 (1997).

16. Hojo, M. et al. Cyclosporine induces cancer progression by a cell-autonomous mechanism [comments]. Nature 397, 530-534 (1999).

17. Pedersen. P.H. et al. Heterogeneous response to the growth factors [EGF. PDGF {bb). TGF-a, bFGF. IL-2] on glioma spheroid growth, migration and invasion. Int. J. Cancer 56, 255-261 (1994).

18. Klein, K.A. et al. Progression of metastatic human prostate cancer to androgen independence in immunodeficient SCID mice. Nature Med. 3, 402-408 (1997).

19. Russo, M.W., Matheny, C. & Milbrandt, J. Transcriptional activity of the zinc finger protein NGFI-A is influenced by its interaction with a cellular factor. Mol. Cell. Biol. 13, 6856-6865 (1993).

20. Barrack. E.R. TGF-ß in prostate cancer: a growth inhibitor that can enhance tumorigenicity. Prostate 31, 61-70 (1997).

21. Fudge, K., Wang, C.Y. & Stearns, M.E. Immunohistochemistry analysis of platelet-derived growth factor A and B chains and platelet-derived growth factor a- and ß- receptor expression in benign prostatic hyperplasias and Gleason-graded human prostate adenocarcinomas. Mod. Pathol. 7, 549-554 (1994).

22. Tourtellotte. W.G., Nagarajan, R., Bartke. A. & Milbrandt. J. Functional compensation by Egr4 in Egr1-dependent luteinizing hormone regulation and Leydig cell steroidogenesis. Mol. Cell. Biol. 20, 5261-5268 (2000).

23. Ivenovic. V., Melman, A., Davis-Joseph. B., Valcic, M. & Geliebter. J. Elevated plasma levels or TGF-beta 1 in patients with invasive prostate cancer [letter]. Nature Med. 1, 282-284 (1995).

24. Steiner. M.S. & Barrack, E.R. Transforming growth factor-beta 1 overproduction in prostate cancer: effects on growth in vivo and in vitro. Mol. Endocrinol. 6. 15-25 (1992).

25. Cui. W. et al. TGF.ß1 inhibits the formation of benign skin tumors, but enhances progression to invasive spindle carcinomas in transgenic mice. Cell 86, 531-542 (1996).

26. Liotta, L.A., Steeg, P.S. & Stetler-Stevenson, W.G. Cancer metastasis and angiogenesis: an imbalance of positive and negative regulation. Cell 64, 327-336 (1991).

27. Dvorak, H.F. Tumors: wounds that do not heal. Similarities between tumor stroma generation end wound healing. N. Engl. J. Med. 315, 1650-1659 (1986).

26. Santiago, F.S. et el. New DNA enzyme targeting Egr-1 mRNA inhibits vascular smooth muscle proliferation and regrowth after injury. Nature Med. 5, 1436 (1999).

29. Riggs, P.K., Rho, O. & DiGiovanni. J. Alteration of Egr-1 mRNA during multistage carcinogenesis in mouse skin. Mol. Carcinog. 27, 247-251 (2000).

30. Schamhorst, V. et al. EGR-1 enhances tumor growth and modulates the effect of the Wilms tumor 1 gene products on tumorigenicity. Oncogene 19, 791-800 (2000).

31. Garabedian, E.M., Roberts, L.J., McNevin, M.S. & Gordon, J.I. Examining the role of Paneth cells in the small intestine by lineage ablation in transgenic mice. J. Biol. Chem. 272, 23729-23740 (1997).

32. Day, M.L., Fahrner, T.J., Ayken. S. & Milbrandt, J. The zinc ringer protein NGFI-A exists in both nuclear and cytoplasmic forms in nerve growth factor-stimulated PC12 Cells. J. Biol. Chem. 265, 15253-15260 (1990).

33. Kirsch, K.H.. Korradi, V. & Johnson. J.P. Mader: a novel nuclear protein over expressed in human melanomas. Oncogene 12, 963-971 (1996).

34. McNeal, J.E. & Bostwick, D.G. Intraductal dysplasia: a premalignant lesion of the prostate. Hum. Pathol.17. 64-71 (1986).

35. Lee, S.L. Wang. V. & Milbrandt,J. Unimpaired macrophage differentiation and activation in mice lacking the zinc finger transplantation factor NGFI-A (EGR1). Mol. Cell. Biol. 16, 4566-4572 (1996).

36. Morrison, T.B., Weis, J.J. & Wittwer, C.T. Quantification of low-copy transcripts by continuous SYBR Green l monitoring during amplification. Biotechniques 24, 954-962 (1998).