In this study, maspin was down regulated during the conversion of human corneal stromal keratocytes to phenotypically distinct fibroblast and myofibroblast cells and cells in the presence of SFDM. Down regulation of gene expression and protein synthesis was progressive with assumption of the typical fibroblastic morphology. Maspin was observed in the nucleus and in the cytoplasm of P0 cells, some in both locations in P1 stromal cells treated with FGF-2 and only nuclear maspin in P1 cells treated with FBS. The role of cytoplasmic and nuclear maspin is not clearly understood, however some hints can be obtained from the effect of maspin interaction with proteins in these two parts of cells. In the nucleus, maspin binds to and inhibits histone deacetylase (Li et al., 2006
) and binds to interferon regulatory protein-6 (Bailey et al., 2005
). The presence of maspin prevents interferon regulatory protein-6 induced increases in N-cadhedrin expression in the invasive MDA-MB-231 mammary carcinoma cells. In the cytoplasm, maspin interacts with glutathione-S-transferase and enhances its activity thus playing a regulatory role in oxidative stress (Yin et al., 2005
). The presence of maspin in the nucleus of some carcinoma cells indicates a better prognosis than if maspin is only found in the cytoplasm (Sood et al., 2002
). The implications of maspin in the nucleus vs the cytoplasm await further research.
In the present study, we demonstrated that the overall methylation state of the mapsin promoter is altered in primary human corneal stromal cells in culture. This is consistent with the function of DNA methylation which is important in X-chromosome-inactivation, development, and tumorigenesis (Holliday and Pugh, 1975
; Riggs, 1975
; Tycko, 2000
). Recently, promoter methylation of maspin was shown to regulate tissue-specific gene expression and changes in gene expression upon transformation of normal cells to carcinoma cells (Domann et al., 2000
; Domann and Futscher, 2003
; Fujisawa et al., 2005
; Murakami et al., 2004
; Ohike et al., 2003
; Sato et al., 2004
; Yatabe et al., 2004
). Conversion of the maspin promoter from the hypomethylated state observed in freshly isolated (non-cultured) stromal cells to a hypermethylated state occurred between the time corneal keratocytes were released from the matrix of the cornea and when the non-dividing P0 stromal cells in SFDM were harvested at three days or the P0 cells cultured in the presence of 5% FBS became confluent. This de novo methylation of the maspin gene in the P0 cells probably is carried out by DNMT-3a and b which do not require a methylation template and do not require cell division for activity (Okano et al., 1999
). The phenotype of these P0 cells cultured in the presence of FBS retained some of the characteristics of keratocytes in that they have not yet assumed the characteristic elongated shape of fibroblasts (Ngamkitidechakul et al., 2001
). Hypermethylation of the maspin promoter also occurred early in the conversion of mammary epithelial cells to carcinoma cells (Futscher et al., 2004
). The maspin promoter in ductal carcinoma in situ cells (DCIS), an intermediate cell type between normal mammary epithelial cells and carcinoma cells, is often hypermethylated when maspin mRNA and protein are still observed. Maspin mRNA and protein were also observed at high levels for the P0 corneal stromal cells cultured cells in the presence of FBS and at low concentrations for the P1 corneal stromal cells cultured in the presence of FBS, FGF-2 or TGF-β1 despite promoter hypermethylation. This may reflect a long half life for the message and protein, or residual transcription of the maspin gene. Alternatively, since RNA polymerase II can bind and initiate transcription of methylated DNA (D’Alessio et al., 2007
), the observed mRNA may be newly transcribed.
Hypomethylation of the maspin promoter was concordant with high levels of maspin in the in situ cells and hypermethylation of the promoter was concordant with low levels of maspin expression in the P0 cells cultured in SFDM and in the P1 cells. However, this did not hold for the P0 stromal cells cultured in the presence of 5% serum which were hypermethylated yet expressed maspin mRNA at the same levels as that observed for the hypomethylated in situ stromal cells. DNA methylation is only one component in the downregulation of gene transcription. Methyl CpG binding proteins can bind methylated DNA and repress transcription (Bird and Wolffe, 1999
; Yoon et al. 2003
; Kuzmichev et al. 2004). In addition, specific modification of histones and binding of proteins that recognize these modifications can repress gene expression (Reviewed by Ruthenburg et al. 2007
and Taverna et al. 2007
). The present study would predict that the rate of downregulation of maspin expression through these mechanisms is prolonged by the presence of FBS in the stromal cell cultures. The interaction of proteins with the methylated maspin DNA or associated histones has not been studied.
Analysis of the patterns of methylation in the region of −247 to 155 of the promoter relative to the mRNA levels in the P0 and P1 cells showed total methylation in non-expressing SFDM cells at positions 12 and 32 while at least one of the clones from the maspin expressing P0 and P1 cells were demethylated at these sites. This pattern did not hold for the P4 cells. The non-expressing cells produced clones that were demethylated at one or more of these sites. Position 194 was methylated in the maspin non-expressing P4 cells (untreated and TSA treated) and was demethylated in at least one clone from the maspin expressing 5-Aza-dC and 5-Aza-dC plus TSA treated cells. This difference in methylation was true for the maspin non-expressing P0 but not the P1 cells cultured in SFDM. Analysis of the methylation patterns at the individual sites between the low and non-expressing cells by the Kruskal-Wallis One Way Analysis of Variance on Ranks showed no significant differences between these two groups. Thus maspin methylation patterns in the corneal stromal P0, P1 and P4 cells are probably random as previously found for the maspin promoter in other tissues (Futscher et al., 2002
The DNA methyl transferase inhibitor 5-Aza-dC at 500 nM induced low levels of maspin reexpression without significant demethylation of the maspin promoter. However, this treatment was able to demethylate the H3 histones associated with the maspin promoter. This probably resulted in alteration of the chromatin allowing a very low level of transcription by RNA polymerase II, which binds and initiates transcription of methylated DNA (D’Alessio et al., 2007
). DNA demethylation of the maspin gene probably follows this low level of transcription as suggested by D’Alessio et al. 2007
The observation that 5-Aza-dC at 500 nM induced low levels of maspin reexpression without significant demethylation of the maspin promoter is not unique to the corneal stromal cells. Low levels of maspin reexpression were observed with 5-Aza-dC in mammary carcinoma cell lines without demethylation of the maspin promoter (Wozniak et al., 2007
). Use of small interfering RNA to the DNA methyltransferase, DNMT1 increased the expression of maspin in MDA-MB-231 breast mammary carcinoma cells supporting DNA methylation in the regulation of maspin expression.
Even though the primary function of 5-Aza-dC at 10 μM or less is to inhibit the DNA methyltransferase, DNMT-1, it could have another role as a histone methyltransferase inhibitor (Kondo et al., 2003
; Murakami et al., 2004
; Wozniak et al., 2007
; Zhu et al., 2001
). This role has been demonstrated for histone H3 associated with the maspin gene. Treatment of P4 corneal stromal cells with 5-Aza-dC demethylated histone H3 at K9 in the same manner as was observed in mammary carcinoma cells (Wozniak et al., 2007
Reexpression of maspin may depend on factors other than demethylation of histone H3. Despite demethylation of this histone by TSA, maspin transcription was not stimulated by this HDAC1 inhibitor in the P4 corneal stromal cells. Demethylation of histone H3 associated with the genes for Ikaros and FGFR2 occurs upon TSA treatment of pituitary cells (Zhu et al., 2007a
; Zhu et al., 2007b
). In contrast to the maspin gene in corneal stromal fibroblasts, FGFR2 and Ikaros were re-expresed in pituitary cells with TSA treatment alone. The inability of TSA to induce the reexpression of maspin is a characteristic of the human corneal stromal fibroblasts since this reagent can also induce maspin transcription in hepato-billiary tract carcinomas and in normal pancreatic cells (Fujisawa et al., 2005
; Ohike et al., 2003
As suggested by Wozniak et al. (Wozniak et al., 2007
), the lack of demethylation of the maspin promoter when maspin is re-expressed in response to 5-Aza-dC could be due to gene reactivation from a few alleles that become demethylated in the cell population. This might account for the really low maspin levels (mRNA and protein) observed in the presence of 5-Aza-dC. Another possibility is that a non-epigenetic factor is at play here. There may be a transcription factor necessary for maspin gene transcription or a RNA regulatory molecule required for stability that is down-regulated during the phenotypic change from human stromal keratocytes to fibroblastic and myofibroblastic cells that is up regulated at low levels in the presence of 5-Aza-dC but not TSA.
The role of methylation of the maspin promoter across species is not known. The only species besides humans for which the distribution of maspin synthesis across tissues has been reported are rat and mouse (Zhang et al., 1997c
; Umekita et al., 1997
). The distribution is similar for all three species. This would suggest that regulation of expression, including regulation by DNA methylation is conserved. Using the Vista Genome Browser (pipeline.lbl.gov/), multiple areas of high identity were found when 50 kb 5′ and 3′ of the maspin transcription start site were examined for human vs chimpanzee, dog and mouse (, arrows). These areas represent potential important binding sites for transcription factors that may regulate maspin synthesis. When any one of these areas was queried with the transcription binding site programs, TFBIND (http://tfbind.hgc.jp/
) and PATCH (http://www.gene-regulation.com/pub/programs.html#patch
) hundreds of potential sites were identified. Future studies will identify the importance of these sites.
Figure 9 In silico analysis of the maspin gene across species. Areas of high identity between human and chimpanzee, dog and mouse maspin genes were calculated using the Vista Genome Browser (pipeline.lbl.gov/) for Ch18:59,244,500–59,344,500 and are given (more ...)
Because some transcriptional binding sites have been identified in the region of the transcription start site, this area was selected for more in depth in silico
exploration. Comparison of multiple species using the Vertebrate Multiz Alignment of the UCSC Genome Browser (http://genome.ucsc.edu/cgi-bin/hgGateway
) revealed considerable identity with the human maspin promoter between −284 and +184 (Ch18: 59,294,792-59,295,260) (). The greatest identity is among human, chimpanzee and rhesus, less with dog, cow and rabbit and even less with mouse and rat. The chimpanzee and rabbit contain 20 and 24 potential CpG methylation sites, respectively, in comparison to the 22 CpG sites that are methylated in the human maspin promoter. Rhesus, cow, mouse and rat have 15 or less sites in this region. If all species use CpG methylation of this region for control, the most important methylation sites for silencing the gene may be −170, −103 and 32 which are conserved in 6 or 7 of the 8 species compared. There is no consistent pattern for these sites between the low expressing (0.1–5%) cells and those that do not express maspin. The importance of individual CpG sites awaits further experimentation. Use of reporter constructs is problematic to study regulation of gene transcription by CpG methylation because chromatin associated with plasmids is abnomal (Jeong et al., 1994
The identified transcription factor binding sites in this region experimentally elucidated by previous studies include the sequences in the negative regulator site HRE and the positive regulator binding sites, ETS, AP1, ATF-126 and ATF-2 sites (Maekawa et al., 2007
; Zhang et al., 1997 a
; Beltran et al., 2007
). These sites are 60–83% identical with human suggesting these transcription factors may be active across the different species (). Although p53 is known to stimulate maspin transcription, the exact sites that are important for p53 binding have not been determined. There are two p53 binding sites between −103 and −1 with at least 50% and 75% identity with human (). The first site has at least one CpG site associated with the exception of dog. The most homologous region within the −284 to 181 region is −119 to 23 (). Within this region, there are multiple potential transcription binding sites including SP1, CAP, GC, ELK 1, HSF 1, HSF2, E47, GATA1, CETSIP54, NRF 2, LYF 1, E2F, CP2, YY1, NYCMA X, STAF, OLF 1, CEBPB, SRY, NF 1, MYO D, AP2, PAX 5, XFD 3, AHRARNT, RFX1, MYB, AP4, SOX 5, T3R, CJUN, NRF, GFI, IRF 2, YM, P300, AP 1, HOX 13, OCT 1, SRF, MZF 1, USF, NFKB, EGR 1 and SEF 1 as determined by the TFBIND program. This region of the maspin gene is only one of many in the 50 kb on the 5′ and 3′ side of the transcription site that is highly conserved across several species ().
One of the open questions is whether DNA methylation regulates corneal wound healing in vivo
. For this analysis, an animal model is needed, preferably one that has CpG islands and CpG sites in similar regions to the human. Based on the presence of 24 CpGs () and an identified CpG island using MethPrimer (http://www.urogene.org/methprimer/
) in the same area as those for the human, the overall identity of the sequence and the size of the cornea, the rabbit model of corneal wound healing probably is the best. These studies are being initiated in our laboratory.
Genes other than maspin may be controlled by DNA methylation upon conversion of corneal stromal keratocytes to fibroblasts. Examples of molecules down-regulated to undetectable levels upon this phenotypic change include the keratin sulfate proteoglycans, lumican and keratocan, and the corneal crystallins, aldehyde dehydrogenase-1, aldehyde dehydrogenase-3 and α-transketolase (Saghizadeh et al., 2005
). CpG islands in the 5′ UTR and 1000 bp upstream of the start site were found in aldehyde dehydrogenase-3 and α-transketolase using MethPrimer. In addition, DNA demethylation may play a role in the keratocyte to fibroblast conversion. The promoter for tenascin-C, a gene that is not observed in keratocytes but is expressed by corneal fibroblasts, has two CpG islands. Thus, DNA methylation may be pertinent to the changes in gene expression observed upon cornea wound healing.
The elucidation of corneal wound healing mechanisms is important based on the number of refractive surgeries performed. Analysis of corneas following LASIK and PRK by non-invasive techniques and post mortem has revealed altered structure of the cornea even 7–10 years following these procedures (Kramer et al., 2005
). The role of maspin in corneal wound healing is an area of intense study in our laboratory. Our hypothesis is that downregulation of maspin synthesis upon phenotypic change of stromal keratocytes to wound healing fibroblasts and myofibroblasts allows migration of these cells into the wounded area. Upon reestablishment of the epithelial sheet, maspin is secreted from the epithelial cells and inhibits migration of the stromal cells. Exogenously added maspin inhibits migration of cultured human corneal stromal fibroblasts and the closure of epithelial wounds in organ culture (Unpublished Data). Maspin also upregulates the synthesis and secretion of plasminogen activators involved in extracellular matrix remodeling (Unpublished Results) indicating that maspin may stimulate the turnover of proteins that contribute to corneal opacities.
This is the first study to show down regulation of a gene is epigenetically controlled during the conversion of corneal keratocytes to fibroblasts. Down regulation of maspin synthesis was shown to involve methylation of the maspin promoter on CpG islands and dimethylation of K9 of the histones associated with the maspin gene. Future research will determine the universality of this mechanism for turning off genes during the phenotypic change of corneal keratocytes to fibroblasts.