C/EBPα is expressed in many tissues including white and brown adipose tissue, myeloid cells, lung, prostate, ovary, and colon [11, 30]. The role of C/EBPα in adipose, myeloid, and hepatocyte terminal differentiation has been well established. In myeloid cells, for example, the expression of C/EBPα is regulated and C/EBPα expression leads, in turn, to expression of genes necessary for cell differentiation and to limited myeloid proliferation [4, 7, 31, 32]. However, all hematopoiesis is not affected similarly. While C/EBPα stimulates the differentiation of myeloid cells, such as granulocytes, increased expression of C/EBPα significantly blocks the differentiation of erythroid precursors . The regulation by C/EBPα in proliferation and differentiation might contribute to the development of the malignant phenotype. The aberrant expression of C/EBPα has been described in myeloid leukemias [10, 13]. In a subset of myeloid leukemias mutations of C/EBPα result in the synthesis of abnormal proteins that act as dominant negative proteins to block the action of C/EBPα with a resulting block of differentiation and loss of the limits on proliferation . In other leukemias the fusion protein formed from the AML1-ETO translocation down-regulates transcription of the C/EBPα gene .
However, only to varying extents has the role of C/EBPα in the differentiation of some of the other tissues been defined. To even a lesser extent has the presence of aberrant expression of C/EBPα in malignancies of these various organs been described. For example, C/EBPα has a gradient of expression in the small intestine of mice with the greatest expression in the most proximal portions of the intestine and without any expression detected in mouse colonic epithelium . In the mouse intestine C/EBPα was found in the epithelium of the villi and not in the crypt cells suggesting that here too C/EBPα expression is associated with terminal differentiation. In human colon and colonic cancers, however, C/EBPα can be detected suggesting that there is some species specificity and a slight decrease of expression has been detected in more advanced colon cancer . In the rat ovary expression of C/EBPα increases with differentiation of the follicular cells ; to date there are no reports of altered C/EBPα expression in ovarian cancer. In addition to the detection of mutations in C/EBPα in acute myelogenous leukemia of subtype FAB classification M2, mutations have been detected in myelodysplastic disorders but only silent mutations have been detected in non-hematologic tumors including one lung cancer and one prostate cancer . In the lung C/EBPα is expressed in the basal layer of normal lung tissue and expression of C/EBPα is recognized to contribute to type II cell differentiation . However the role of expression of C/EBPα on the differentiation of lung cells is still unknown.
The studies presented here provide additional information to the study that examined C/EBPα expression by gene array  and in which expression of C/EBPα RNA was increased by three fold in cancerous prostate epithelium versus normal epithelium. C/EBPα expression at the protein level in malignant and non-malignant prostate glands has not been previously investigated. In the normal prostate, C/EBPα is predominately expressed in the basal layers of the epithelium with little C/EBPα detected by immunohistochemistry in the more differentiated secretory epithelium expressing AR and PSA. This is a distinctly different distribution than seen in other tissues where C/EBPα is expressed in differentiated cells. It is interesting that in our immunohistochemistry studies, p63 was found to co-localize with the expression of C/EBPα in normal prostate gland. p63 is a marker of the prostate gland basal layer cell and a potential marker for prostate stem cells, which are characterized by active cellular proliferation [37, 38]. Further, in prostate cancer the detection of C/EBPα in the basal layer was lost, but C/EBPα staining could be detected in cancerous cells throughout pseudoglandular structures. In advanced prostate cancer with the loss of glandular like structures, C/EBPα was still expressed in some cancer cells with a similar intensity of staining as in the non-malignant epithelium. Interestingly, in the immunohistochemical staining of prostate tissue arrays for C/EBPα and AR, the ratio of C/EBPα to androgen receptor expression increased with increasing Gleason scores (data not shown) supporting our observation that forced overexpression of C/EBPα in LNCaP cells induced loss of AR expression.
To address the questions derived from our immunohistochemical staining, we first selected the PSA promoter/enhancer as a molecular marker to investigate the role of C/EBPα in the prostate. PSA seemed a reasonable choice: PSA is as a marker of prostate differentiation and is expressed in the terminally differentiated epithelium lining the lumen of tubules in the normal prostate gland as is AR; expression of PSA is androgen-dependent; and increased levels of PSA in the serum is an important biomarker for prostate cancer. The observation that C/EBPα was expressed in the basal epithelium suggested that C/EBPα may suppress PSA expression. In view of very low level of constitutional expression of C/EBPα protein in the LNCaP cells, a widely used prostate cancer cell line that expresses AR and PSA, exogenous C/EBPα was expressed by use of a retrovirus system and two pooled clones of LNCaP cells with stable expression of C/EBPα were established. Clone1 was found to express neither AR nor PSA immediately upon establishment of the clone. Clone 2 developed epigenetic loss of AR and PSA expression with serial passage. Given the rapidity in which AR expression is lost it is unlikely that the retrovirus itself caused the loss of AR expression. In addition, we have examined AR and PSA expression in LNCaP cells stably transfected with pcDNA3-C/EBPα and also noted decreased expression of PSA and AR (data not shown). However, the stable expression of C/EBPα introduced with a non-virus vector could not be maintained in LNCaP cells despite continued selection pressure with geneticin.
In the LNCaP clones with stable expression of C/EBPα, concomitant with the expression of C/EBPα we observed the loss of AR expression. Both C/EBPα and AR are important for the control of growth of prostate cells. The continued growth of the clones overexpressing C/EBPα itself was a surprise. In lung cancer cells and other cell types restoration or increased expression of C/EBPα caused growth arrest . In contrast, in the LNCaP clones expressing C/EBPα we did not observe any signs of growth arrest. The resistance to C/EBPα-induced cell growth arrest may have taken place during clonal selection. However, the establishment of stable expression of C/EBPα itself in several prostate cell lines meant that cells could escape from the negative regulation of cell proliferation by C/EBPα. This may explain the difference of cell fates between transient and stably expression of C/EBPα. A recent similar example has been observed in hepatocytes where the suppression of cell proliferation by C/EBPα is lost during oncogenesis by dephosphorylation of C/EBPα which leads to failure of C/EBPα to interact with cdk2 and E2F as well as sequestering retinoblastoma protein [39, 40]. Similar loss of the interaction between C/EBPα and cell cycle related proteins was found in established prostate cell lines (date not shown). In addition, the expression of C/EBPα in the actively proliferating basal cell layer of the prostate suggests that C/EBPα does not cause growth arrest in normal prostate epithelium.
The loss AR expression occurs in several prostate cancer cell lines, such DU145, PPC1 and PC3, and occurs frequently during the clinical evolution of prostate cancer [41, 42]. The data from gene expression profiles shows that expression of AR RNA is absent in 30 to 40 % of prostate cancer samples (GEO accession: GSE1431) and the proportion of prostate cancers not expressing AR is increased in metastatic prostate cancer (GEO accession:GSE3325). LNCaP cell lines with loss of AR expression exhibit stronger tumorgenicity and a greater metastatic potential in xenografts than AR positive LNCaP cells . In another prostate cancer cell line that lacked AR expression restoration of that expression restored AR-dependent cell growth . We would predict that the loss of AR in our C/EBPα overexpressing LNCaP clones would cause a transformation of cells from androgen-dependent to androgen-independent growth and that the cells would exhibit more aggressive growth, invasion, and metatstatic potential. Indeed, our initial analysis of cell growth showed a more rapid growth of the C/EBPα expressing Clone 2 LNCaP cells compared with the control cells at passage 8 when AR was barely detectable. In addition, our immunohistochemical findings of reduced expression of AR in advanced prostate cancer specimens especially those that express C/EBPα is in agreement with recent demonstrations of altered AR expression especially in androgen-independent prostate cancer .
The mechanism(s) for the loss of AR expression both in prostate cell lines and prostate cancer tissue are not completely understood and expression may vary in the same patient at different sites of recurrence . Methylation of the AR gene promoter has been reported in advanced hormone-independent prostate cancer tissue  and the addition of 5-Aza-2'-deoxycytidine(5-AZA) into DU145 cells with a heavily methylated AR promoter restored AR mRNA expression . In our C/EBPα overexpressing LNCaP cells, treatment with 5-AZA at 10 μM increased expression of AR and PSA in early passages of clone 2 but not clone 1 cells (data not shown). However, increased expression of AR by 5-AZA was seen in the corresponding control cells suggesting that methylation affecting AR expression was occurring in both C/EBPα expressing and non-expressing cells.
The immunohistochemical result demonstrating C/EBPα expression in the basal layer of the prostate epithelium with no expression of PSA in these cells led us to examine the effect of C/EBPα on the expression of the PSA promoter/enhancer. The inhibition of expression was not cell type specific as similar studies in CV1 and COS7 cells of co-transfection of C/EBPα and PSA promoter reporter gene demonstrated similar results (data not shown). A recent report shows that inhibition of PSA expression in LNCaP cells was dependent on an interaction between AR and C/EBPα  while our results indicated that the inhibition was independent of any direct interaction with AR. With transient expression of C/EBPα in prostate cells, ALVA 101 and the non-prostate cell line CV-1 we could not show inhibition of the MMTV promoter which contains a cryptic AR (data not shown). Nor did we observe that an increased expression of AR affected the suppression of the PSA promoter by C/EBPα. In addition, a standard search for transcription factor binding sites by TESS or TFSEARCHING showed that more than 10 potential C/EBPα binding sites in the PSA promoter. Preliminary studies with a gel shift assay indicated that C/EBPα could bind to at least 3 of these sites. Therefore, direct binding of C/EBPα to PAS promoter may explain the suppression of the PSA promoter by C/EBPα. We can not exclude that C/EBPα might also stimulate expression of coactivators. On the other hand, the results also suggest that the C-terminal leucine zipper region was more important than the transactivating domains.
We are beginning to explore, amongst the various genes whose expression in prostate cancer cells is uniformly altered by forced expression of C/EBPα, those genes that may contribute to the loss of AR expression and which may provide therapeutic targets for the difficult clinical problem of hormonal independence.