Background
The progression of epithelial tumors to invasive carcinomas involves changes in cell polarity, adhesion and motility that permit the detachment of cancer cells from the epithelial layer, their invasion into adjacent tissue layers and eventually their spread throughout the body. These processes require reorganization of the cellular cytoskeleton and altered expression of proteins that connect it to the cell membrane as well as remodelling of the extracellular matrix including changes in the composition and processing of its constituents.
The 4.1 proteins, encoded by the
EPB41 (erythrocyte protein band 4.1) genes, are components of the cortical cytoskeleton underlying the cell membrane [
1,
2]. The family of 4.1 proteins consists of the eponymous 4.1R protein first identified in erythrocytes (gene:
EPB41), 4.1N (
EPB41L1), 4.1G (
EPB41L2), 4.1B (
EPB41L3) as well as the less closely related members NBL4 (
EPB41L4A), EHM2 (
EPB41L4B) and EPB41L5 (
EPB41L5). They form nodes in the cell cortex connecting further components of the cortical cytoskeleton like spectrins, actin and transmembrane adhesion proteins, receptors and transporters with each other. In this fashion 4.1 proteins contribute to the organization of cell polarity, adhesion and motility, and affect transport through the membrane and responses to growth factors.
The 4.1B protein is most strongly expressed in neurons and is enriched in the basal cells of certain epithelia [
2,
3]. In addition to spectrins and actin, known binding partners comprise the adhesion molecule CD44 that binds hyaluronic acid in the extracellular matrix [
1] and the candidate tumor suppressor disc large 1/DLG1 [
4]. The 4.1B protein is downregulated in several carcinomas, including prostate cancer [
5], likely by deletion or promoter hypermethylation of the
EPB41L3 gene promoter [
6]. Mouse models of prostate cancer progression suggest that it acts as a metastasis suppressor [
7]. In contrast, EHM2 is conspicuous in tumor cells with high migratory potential, such as metastatic melanoma and fibrosarcoma cells [
8,
9]. In prostate cancer, EHM2 has been reported to be overexpressed [
6,
10] and to diminish adhesion of prostate cancer cells to collagen [
10]. The most recently discovered 4.1 family member, the product of the
EPB41L5 gene (also called Limulus), regulates cell adhesion during development [
11], but has not yet been investigated in the context of human cancer.
The 4.1 proteins are part of a larger protein family characterized by FERM domains, of which many have related functions. For instance, the FERM domain protein ezrin, encoded by the
VIL2 gene, connects CD81 at the cell membrane to the actin cytoskeleton. Ezrin has been shown to mediate invasion of prostate cancer cells [
12,
13], but whether it is overexpressed in prostate cancer is not known. The more distantly related protein dematin, too, interacts with the actin cytoskeleton and growth factor receptors. It is encoded by the
EPB49 gene on 8p21.1, a region frequently deleted in prostate cancer. Overexpression of dematin in PC3 prostate cancer cells changed their morphology towards a more epithelial phenotype [
14], but no investigations of
EPB49 expression in prostate cancer tissues have been published.
The fundamental reorganization of the cytoskeleton, its attachment to the cell membrane and the extracellular matrix during cancer progression are orchestrated by transcription factors that activate cellular programs for cell migration and invasion, which are physiologically employed during embryogenesis, wound healing and tissue regeneration [
15,
16]. In cancer, such transcription factors, like Snail/SNAI1, Slug/SNAI2 and ZEB1, often become deregulated and promote tumor progression.
In prostate carcinoma, transcription factors of the ETS family are prominent candidates for oncogenes driving this facet of tumor progression. Specific members of this protein family are activated towards oncogenes by chromosomal translocations [
17] placing a structural ETS transcription factor gene under the control of an androgen-responsive promoter, resulting in its deregulation and overexpression. The most common translocation, found in 30-70% of all cases, creates a fusion gene placing the androgen-responsive promoter of
TMPRSS2 in control of the
ERG structural gene encoding an ETS family transcription factor. This genetic aberration results in the androgen-driven overexpression of intact or amino-terminally truncated ERG proteins in prostate epithelial cells. ERG oncoproteins influence tumor cell proliferation, but exert a more pronounced effect on migration and invasion through broad changes in gene expression [
18‐
20]. In accord with a function in promoting tumor progression,
TMPRSS2-ERG translocations are observed in a significantly lower fraction of high-grade prostate intraepithelial neoplasias, a non-invasive precursor stage, than in invasive carcinomas [
21].
We have previously reported downregulation of
EPB41L3 encoding protein band 4.1B and upregulation of
EPB41L4B encoding EHM2, respectively, in prostate cancer [
6], in accord with observations by other groups [
7,
10]. In the present study, we have investigated the expression of further members of the family as well as selected genes encoding related or interacting proteins such as disc large 1, ezrin and dematin, and the relation of the changes to activation of oncogenic ERG.
Discussion
Previous studies have concurrently reported changes in the expression of
EPB41L3/4.1B and
EPB41L4B/EHM2 in many prostate cancers. The other members of the 4.1 family had not been studied yet in prostate cancer, but previous investigations have indicated potential functions for the related proteins ezrin and dematin [
12‐
14,
26,
27]. Despite these hints, we did not observe significant changes in ezrin/
VIL2 or dematin/
EPB49 mRNA expression in prostate cancer compared to benign tissues. Instead, the present investigation indicated that the expression changes of
EPB41L3 and
EPB41L4B are quite specific among cortical cytoskeleton genes, with the exception of
EPB41L1. Our findings therefore call for a detailed investigation of 4.1N expression and function in prostatic cells in future studies.
Investigation of
EPB41L3 methylation in a larger series of samples confirmed the presumed association between its downregulation and hypermethylation.
EPB41L3 hypermethylation tended to be more frequent in higher stage tumors and downregulation of gene expression tended to be associated with earlier recurrence. The two changes did not correlate significantly with the same clinical indicators of tumor progression, which may be due the limited number of samples studied, since the tendencies were the same. In any case, these findings concur with the idea from functional studies in mouse and cell line models that 4.1B may act as a metastasis suppressor in prostate cancer [
7]. Likewise, although we did not observe significant associations of
EPB41L4B expression with cancer stage or grade, increased mRNA expression was associated with earlier recurrence, in accord with the immunohistochemical study of Wang et al. [
10].
Moreover, the extended
EPB41L3 methylation analysis allowed a comparison with the well-studied hypermethylation of
GSTP1. Hypermethylation of
GSTP1 can be detected in a significant fraction of HG-PINs [
28], but more consistently after progression towards invasive carcinomas. It is thought to occur as part of an „epigenetic catastrophe" [
29] that involves hypermethylation of further genes, e.g.
APC,
RARB2 and
RASSF1A. In our tissue series, accordingly, cases with hypermethylation of
GSTP1 have a very high likelihood of displaying hypermethylation of these genes as well [
22].
EPB41L3 hypermethylation appears to occur slightly less frequently and was notably essentially restricted to cases with
GSTP1 hypermethylation. A straightforward interpretation of this finding is that
EPB41L3 hypermethylation occurs in many prostate cancers in conjunction with or following the hypermethylation of
GSTP1.
In the present study, we observed a good correlation between the presence of
EPB41L3 and
EPB41L4B expression changes on one hand and
ERG overexpression on the other hand. Simply stated, carcinoma tissues without
ERG overexpression showed no differences in expression of the 4.1 genes towards normal tissues (Fig.
4). Our data therefore suggest that the changes in the expression of the two cortical cytoskeleton genes might be a consequence of ERG overexpression. This hypothesis fits with the presumed sequence of events during prostate cancer development. Fusion genes causing ERG overexpression can be detected in some pre-neoplastic HG-PINs, but more consistently in invasive carcinomas [
21] supporting a role of ERG in promoting prostate cancer invasion and progression. As mentioned above, this transition is also associated with consistent hypermethylation of
GSTP1, which may coincide with or precede hypermethylation of
EPB41L3.
A relationship between ERG and the cortical cytoskeleton proteins is also plausible on functional grounds. Overexpression of ERG is thought to result in increased proliferation, invasiveness and motility of prostate cancer cells [
18‐
20,
30], whereas protein 4.1B has been shown to oppose invasiveness and metastasis of prostate cancer cells [
5,
7] and EHM2 to modulate adhesion of prostate cancer cells [
10]. Our data invite the interpretation that the changes in
EPB41 gene expression are part of a presumed invasion program directed by ERG and serve to implement the necessary changes in the cortical cytoskeleton. Whether cortical cytoskeleton genes are indeed - directly or indirectly - regulated by ERG, must now be investigated in cell line and animal models.
Along the same line of argument, we have previously observed changes in the expression of ECM genes encoding fibulin 1 (
FBLN1) and testican 1 (
SPOCK1). The causes of these changes are so far unclear, since we did not observe altered gene copy numbers in the majority of cases or obtain evidence for
FBLN1 hypermethylation [
25]. We have now become aware that upregulation of
SPOCK1 parallels that of
EPB41L4B and downregulation of
FBLN1 parallels that of
EPB41L3. Accordingly, the changes in the expression of the ECM genes were also associated with
ERG overexpression. They may therefore represent changes in the composition of the extracellular matrix brought about as a consequence of ERG oncogenic activation. The observation that EHM2 influences cell adhesion to basement membrane collagen is intriguing in that respect [
10]. Of note, the strikingly close correlation between
EPB41L4B and
SPOCK1 expression in the tissue samples may be enhanced by their common regulation by androgens [
10,
31]. It remains to be determined whether the ECM genes are indeed ERG targets.
In this study, we also report the first immunohistochemical investigation of 4.1B protein in prostatic tissues. Not unexpectedly [
2,
3], we observed the most prominent expression in the basal cells of the prostatic glands, while luminal cells contained less protein. Loss of the basal cell population is a well-established characteristic of prostate cancers. The downregulation of
EPB41L3 expression may therefore largely reflect the loss of basal cells during prostate cancer development. Basal cells in the prostate epithelium express
GSTP1, in contrast to luminal cells [
32]. Accordingly, loss of expression and hypermethylation of
GSTP1 may be related to the loss of the basal cell compartment [
33]. Consequently, it is tempting to speculate that the hypermethylation of
EPB41L3 in prostate cancer, too, might be associated with the loss of the basal cell compartment. Unlike GSTP1, however, 4.1B protein is expressed also in luminal cells of normal glands and the expression in cancerous prostatic glands remained overall comparable to that in the luminal cells, despite the presence of significant hypermethylation in a majority of the cases. Indeed, in prostate cancer cell lines with pronounced hypermethylation of the
EPB41L3 gene and low levels of mRNA [
6], a small amount of 4.1B protein remains detectable by blot techniques (Schulz et al., unpublished results). The residual expression of the gene in cancer cells could be due to a low level of transcription despite hypermethylation of the main promoter or from an alternative weaker promoter suspected in the gene [
34].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
WAS designed the study. ECD and CH performed RNA and DNA analyses, RE performed and evaluated immunohistochemistry. MI and JR conducted statistical analyses. WAS, MI, JR and RE interpreted the data. WAS, MI and RE wrote the manuscript. All authors have given final approval to the manuscript.