|Home | About | Journals | Submit | Contact Us | Français|
Keloids are benign dermal tumors that form during wound healing in genetically susceptible individuals. The mechanism(s) of keloid formation is unknown and there is no satisfactory treatment. We have reported differences between fibroblasts cultured from normal scars and keloids that include a pattern of glucocorticoid resistance and altered regulation of genes in several signaling pathways associated with fibrosis, including Wnt and IGF/IGF-binding protein 5 (IGFBP5). As previously reported for glucocorticoid resistance, decreased expression of the Wnt inhibitor secreted frizzled-related protein 1 (SFRP1), matrix metalloproteinase 3 (MMP3) and dermatopontin (DPT), and increased expression of IGFBP5 and jagged 1 (JAG1) are seen only in fibroblasts cultured from the keloid nodule. In vivo, decreased expression of SFRP1 and SFRP2 and increased expression of IGFBP5 are observed only in proliferative keloid tissue. There is no consistent difference in the replicative lifespan of normal and keloid fibroblasts, and the altered response to hydrocortisone (HC) and differential regulation of a subset of genes in standard culture medium are maintained throughout at least 80% of the culture lifetime. Preliminary studies using ChIP-chip analysis, Trichostatin A (TSA) and 5-aza-2′-deoxycytidine (5-aza-dC) further support an epigenetically altered program in keloid fibroblasts that includes an altered pattern of DNA methylation and histone acetylation.
Keloids are fibrotic tumors of the dermis that form during a protracted wound healing process. The predisposition to form keloids is found predominantly in people of African, Asian, and Hispanic descent (Butler et al., 2008; Niessen et al., 1999). Keloids occur in ~1/30 African Americans and ~1/625 of the overall US population (Barrett, 1973). While this disfiguring and sometimes disabling disorder of wound healing significantly impairs the quality of life, it is understudied relative to other chronic skin disorders (Bock et al., 2006). The key alteration(s) responsible for the pathological process has not been identified and, as for other fibrotic disorders, there is no satisfactory treatment (Butler et al., 2008; Lupher and Gallatin, 2006; Niessen et al., 1999). Moreover, keloid formation is one of a group of fibroproliferative diseases characterized by an exaggerated response to injury that occur at higher frequency or with more severe manifestations in people of African ancestry (Smith et al., 2008). We have reported differences in expression of a broad spectrum of wound healing-related genes between normal and keloid fibroblasts under standard culture conditions in medium containing 10% fetal bovine serum (Meyer et al., 2000; Russell et al., 1995; Smith et al., 2008). We have also reported a pattern of differences in growth and synthesis of extracellular matrix induced by several regulators of wound healing. These include an altered growth response to hydrocortisone (HC) (Russell et al., 1978) and resistance of keloid fibroblasts to HC downregulation of types I, III and V collagen, elastin, connective tissue growth factor (CTGF) and IGF-binding protein 3 (IGFBP3) gene expression (Russell et al., 1978; Russell et al., 1995; Russell et al., 1989; Smith et al., 2008). HC resistance is observed only in fibroblasts from the keloid nodule; fibroblasts cultured from superficial dermis of keloids and from unaffected dermis of keloid patients behave like normal skin and scar fibroblasts. These findings suggest that fibroblasts from the keloid nodule are distinct from other dermal fibroblasts in the affected individual. Although some cases of keloid formation may be due to somatic mutation (Saed et al., 1998), multiple keloids in the same individual and evidence for a multicellular origin of keloids (Chevray and Manson, 2004) argue against somatic mutation as the primary event and suggest that an environmental factor present during wound healing triggers abnormal gene expression in genetically susceptible individuals. Various studies have implicated keratinocytes, Langerhans cells, mast cells, and T cells in skin fibrosis (Butler et al., 2008; Niessen et al., 1999; Phan et al., 2003). Thus, differences in gene expression in keloid fibroblasts may be due either to expression of the abnormal gene(s) by the fibroblasts themselves, or expression by another cell type, causing selection of an epigenetically distinct subpopulation of fibroblasts in normal skin.
Since first described (Hayflick and Moorhead, 1961), it has been well established that human diploid fibroblasts display a limited proliferative lifespan followed by replicative senescence (Cristofalo et al., 2004). Effects of cellular aging on gene expression have been reported, including increased expression of matrix metalloproteinase (MMP)-1 and -3, plasminogen activator inhibitor and cyclin-dependent kinase inhibitors, and both increased and decreased collagen synthesis (Cristofalo et al., 2004; Ravelojaona et al., 2008; Zeng et al., 1996). Decreased responsiveness to HC during in vitro cellular aging has also been reported (Cristofalo and Rosner, 1979). Because fibroblasts in keloids may have undergone more population doublings during tumor formation than normal dermis or scar fibroblasts we investigated whether normal and keloid cultures differ in their in vitro proliferative capacity, and whether the different pattern of gene expression in normal and keloid fibroblasts is stable throughout the in vitro lifetime, or as normal fibroblasts age they adopt the gene expression pattern of keloid fibroblasts. A stable pattern over the in vitro lifetime and similar growth capacity of normal and keloid fibroblasts would refute the view that the altered program in keloid fibroblasts is due to increased cellular age and would support the hypothesis that keloid fibroblasts represent an epigenetically distinct population of fibroblasts selected during wound healing in individuals predisposed to form keloids.
Having previously observed that the pattern of glucocorticoid resistance of keloid fibroblasts to downregulation of elastin and collagen is confined to the keloid nodule, we determined whether differential expression of other genes in our gene expression profiling study was confined to fibroblasts from the lesion. We used quantitative real time (QRT)-PCR to measure levels of expression in cultures from an abdominal keloid, superficial dermis of the same lesion, and normal abdominal skin excised at the same time as the keloid (Russell et al., 1978).
In the gene profiling studies we observed significantly increased expression of IGFBP5, jagged 1 (JAG1), and CTGF, and decreased expression of secreted frizzled-related protein 1 (SFRP1), MMP3, and dermatopontin (DPT) in keloid fibroblasts (Smith et al., 2008). As seen in Table 1, altered expression is observed only in cultures from the lesion. Differences in expression in fibroblasts from the lesion compared to superficial dermis or nearby normal skin were similar to those previously seen when fibroblasts from keloid lesions were compared to normal scars from unaffected individuals. The only exception was that CTGF was expressed at higher levels in fibroblasts from the keloid lesion in the presence and absence of HC, although the increase was greater in its presence. Immunohistochemical examination revealed decreased SFRP1 and SFRP2 (Figure 1a–b) and increased IGFBP5 protein (Figure 1c) only in active areas of keloid tissue. Expression in inactive areas did not differ from that observed in normal dermis from unaffected individuals (data not shown).
To determine whether differential patterns of gene expression reflect the greater age of keloid cells that may have undergone more population doublings during tumor formation, we assessed differences in proliferative potential and whether normal cells aged in vitro adopt an expression pattern similar to that of keloid cultures.
Included in our collection of fibroblasts are two strains each of normal (21 and 130) and keloid (33 and 50) fibroblasts that were cultured to senescence in the presence and absence of 1.5 μM HC and preserved in liquid nitrogen at different in vitro culture ages ranging from approximately 44 to 5 population doublings from senescence. These cultures would be roughly equivalent to passage numbers 1–22, assuming two doublings per passage. Because the number of generations undergone prior to establishing in vitro cultures can not be accurately assessed, culture age is expressed as generations from senescence or, where several strains are averaged, the approximate percentage of in vitro lifetime.
Proliferative histories of the four fibroblast strains are depicted in Figure 2(a–d). Age-related changes in growth characteristics were similar for all strains and were unaffected by continuous growth in HC (data not shown). Maximum cell density decreased in a roughly linear manner throughout the culture lifetime while population doubling time remained relatively constant until approximately 10 generations prior to senescence and then increased in an accelerating manner.
The total number of population doublings for each strain cultured with and without HC is summarized in Supplementary Table 1. Keloid strain 50 underwent fewer population doublings than the other strains; however, keloid strain 33 did not differ significantly in total population doublings from cells derived from normal skin or normal mature scar.
We previously reported that growth of early passage fibroblasts in 1.5 μM HC increased the maximum cell density of normal fibroblasts but lowered or had no effect on the maximum density of fibroblasts from keloid lesions (Russell et al., 1978). When the effect of HC was examined in normal strain 21 at different generations from senescence, the differential effect of HC persisted as maximum cell density decreased linearly over the culture lifetime (Figure 2e–f).
While we observed no differences in rate of collagen synthesis between multiple strains of normal and keloid fibroblasts grown in standard culture medium, we have reported a differential effect of 1.5 μM HC on percent collagen synthesis, mRNA levels for types I, III, and V collagen and prolyl hydroxylase activity (Russell et al., 1978; Russell et al., 1989; Trupin et al., 1983).
To determine whether the rate of collagen synthesis changes and whether the differential effect of HC is maintained throughout the culture lifetime, percent collagen synthesis was estimated from rates of incorporation of 3H-proline into collagen and noncollagen protein (Russell et al., 1978) at different generations from senescence. We observed little change in percent collagen synthesis as a function of cellular aging (Table 2), and the differential effect of HC on collagen synthesis in normal versus keloid fibroblasts was maintained throughout most of the culture lifetime.
To further examine the effect of cellular aging on the HC response, QRT-PCR was used to measure expression of α1(I) collagen, elastin, and CTGF genes in cell strains 21 and 33 grown with 0.28 mM ascorbic acid with and without HC. As seen in Table 2, HC downregulated the expression of all three genes in normal fibroblasts over many cell generations, but had little effect on collagen gene expression in keloid fibroblasts, and increased expression of elastin and CTGF.
To determine the effect of cellular aging on several additional genes whose differential expression was seen only in fibroblasts from the keloid nodule, we compared normal strains 21 and 130 to keloid strains 33 and 50 at three different in vitro ages, approximately 16%, 59%, and 79% of the culture lifetime (Table 3). We observed that decreased expression of SFRP1 and MMP3, and increased expression of IGFBP5 and JAG1 in keloid fibroblasts were maintained for many cell generations.
Epigenetic silencing of tumor promoters by hypermethylation and differential histone acetylation, including those of multiple SFRPs, often occurs in early stages of tumorigenesis (Jones and Baylin, 2007; Kawamoto et al., 2008; Suzuki et al., 2002). However, analysis of the SFRP1 promoter from 20 keloid and 10 normal strains using the Sequenom MassARRAY System failed to show detectable methylation in either cell type (data not shown). While preliminary genome-wide ChIP-chip assay of pooled DNA samples revealed differential methylation of multiple genes in keloid versus normal fibroblasts, it did not show differential methylation of the SFRP1 gene (Supplementary Figure 1). Hypermethylation of genes in the homeotic (HOX)A cluster correlated with decreased expression of HOXA9 and HOXA10 in keloid cells. Hypomethylation of the asporin, thrombin-like receptor and MMP3 promoters in the keloid sample also correlated with differential gene expression; asporin and the thrombin-like receptor are overexpressed whereas MMP3 is underexpressed (Smith et al., 2008).
Preliminary experiments using 1-day treatment with 0.33 μM Trichostatin A (TSA), an inhibitor of histone deacetylation and 4-day treatment with 2μM 5-aza-2′-deoxycytidine (5-aza-dC), an inhibitor of DNA methylation, revealed that expression of SFRP1 in keloid fibroblasts was increased almost 15-fold by TSA but not by 5-aza-dC (Table 4). TSA, but not 5-aza-dC, decreased expression of JAG1 whereas both TSA and 5-aza-dC decreased IGFBP5 expression more in keloid than in normal cells. HoxA10 expression was undetectable in keloid fibroblasts in the absence of inhibitors but was increased by both TSA and 5-aza-dC. Both inhibitors decreased expression of collagen and CTGF to a similar extent in normal and keloid fibroblasts. Experiments on two additional normal and keloid strains confirmed that TSA selectively increases expression of SFRP1 and decreases expression of IGFBP5 and JAG1 in keloid but not in normal cells (data not shown).
We previously reported that an altered growth response to HC and resistance of keloid fibroblasts to HC downregulation of collagen and elastin are observed only in fibroblasts from the keloid nodule, findings that support the hypothesis that keloid fibroblasts are an epigenetically distinct subpopulation (Russell et al., 1978; Russell et al., 1995; Russell et al., 1989). Here we provide further evidence for that hypothesis. The differential expression of several fibrosis-associated genes, including the Wnt inhibitor SFRP1, MMP3, DPT, JAG1, CTGF, and IGFBP5 is confined to fibroblasts cultured from the keloid nodule. Immunohistochemical measurements confirmed that decreased levels of SFRP1 and SFRP2 and increased levels of IGFBP5 are confined to active keloid tissue.
The hypothesis that differences between normal and keloid fibroblasts in culture are due to differences in vivo aging is not supported by our studies. While the number of in vitro population doublings may not accurately reflect the number of divisions undergone in vivo (Cristofalo et al., 2004; Maier and Westendorp, 2009), detailed analysis of the replicative lifespan of two keloid and two normal strains aged in the presence or absence of HC revealed no consistent differences to support the hypothesis that fibroblasts cultured from keloids underwent more population doublings than fibroblasts from normal dermis or scar in the formation of the tumor. HC has been reported to extend (Cristofalo and Rosner, 1979) or have no effect (Didinsky and Rheinwald, 1981) on the replicative lifespan of normal fibroblasts. We observed little effect of HC on the lifespan of either normal or keloid fibroblasts.
It is routine to compare patterns of gene expression in cultured cells at low passage number to minimize loss of an in vivo phenotype (Feghali and Wright, 1999; Smith et al., 2008; Tuan et al., 2008). It has been reported that simply culturing cells results in loss of a difference in α1β1 integrin collagen receptor expression between fibroblasts from keloids and normal skin (Szulgit et al., 2002). However, some characteristics of an altered program are retained for many generations in culture. An example is the persistent downregulation of Fli1, a suppressor of collagen transcription, in scleroderma fibroblasts in vivo and in vitro (Asano et al., 2007; Wang et al., 2006). We have found that the altered pattern of gene expression in keloid fibroblasts, including failure of HC to downregulate collagen, elastin, and CTGF, decreased expression of SFRP1 and MMP3, and increased expression of IGFBP5 and JAG1 in standard culture medium is not abolished for at least 80% of the replicative lifespan. While not identifying a mechanism, these findings support the hypothesis of an epigenetically regulated program of fibrosis. Furthermore, persistence of the stimulatory effect of HC on growth of normal fibroblasts throughout the culture lifetime contradicts the notion that as normal fibroblasts age they act like keloid fibroblasts.
Recent inhibitor studies have provided evidence that epigenetic alterations occur during activation of wound healing and fibrosis. TSA blocks transforming growth factor β-mediated myofibroblastic differentiation (Glenisson et al., 2007) and induction of collagen gene expression (Ghosh et al., 2007; Rombouts et al., 2002) in human skin fibroblasts. TSA also prevents accumulation of extracellular matrix in a mouse model of bleomycin-induced skin fibrosis (Huber et al., 2007). TSA and 5-aza-dC have been reported to reverse epigenetic repression of the Fli1 gene and to decrease collagen expression in scleroderma fibroblasts (Wang et al., 2006). Gene profiling studies have revealed no differences in expression of Fli1 in keloid fibroblasts; thus Fli1 does not appear to play a role in the keloid program of fibrosis.
Our findings support an altered program of DNA methylation and histone acetylation that could account for the stable pattern of differential gene expression in keloid fibroblasts in culture. These epigenetically distinct fibroblasts may have been produced or selected in the wound-healing environment of genetically predisposed individuals. While not irreversible, patterns of DNA methylation and histone modifications can be replicated over many cell generations in vivo and in vitro by complex albeit incompletely understood mechanisms involving chromatin architecture, long-range gene interactions and a complex network of trans-acting proteins and noncoding RNAs (Margueron and Reinberg, 2010). The observation that TSA-induced reversal of SFRP1 gene silencing is associated with decreased expression of profibrotic IGFBP5 and JAG1 supports a role for differential histone acetylation of the SFRP1 gene or of a gene(s) that regulates SFRP1 expression in keloids. SFRP1 is best known as an inhibitor of Wnt signaling and increased Wnt signaling has been reported to play a role in the pathogenesis of keloids (Sato, 2006) and several other fibrotic disorders including pulmonary and renal fibrosis (He et al., 2009; Morrisey, 2003). SFRP1 and SFRP2 have recently been reported to inhibit bone morphogenetic protein signaling (Misra and Matise, 2010). Increased BMP signaling has been implicated in fibrotic disorders such as fibrodysplasia ossificans progressiva (Kaplan et al., 2009). Therefore, silencing of SFRP1 may be important in the fibrosis signature displayed by keloid fibroblasts. While these inhibitor studies do not identify causal relationships between the expression of different genes, they provide additional evidence for an epigenetically altered program in keloid fibroblasts. Further elucidation of this program may be achieved by determining individual gene and genome-wide patterns of DNA methylation and histone modification. Manipulation of expression of specific epigenetically modified genes may identify causal relationships. Characterization of an epigenetically altered program in cultured fibroblasts may reveal mechanisms leading to keloid formation and suggest strategies to treat or prevent keloids and possibly other fibrotic disorders that disproportionately affect individuals of African ancestry.
Methods of isolation and propagation of fibroblasts from keloids, normal dermis and scars have been described (Russell et al., 1978; Smith et al., 2008). Sources of cell strains, all of which were obtained from African American patients, have been presented (Russell et al., 1978; Smith et al., 2008; Trupin et al., 1983). Cultures are grown at 37°C in an atmosphere of air and CO2 adjusted to maintain a pH of 7.4 with 100% humidity. The culture medium consists of 90% F-10 (Sigma) and 10% fetal bovine serum (Invitrogen, Grand Island, NY)). No antibiotics or antimycotics are used so that culture dishes inadvertently subjected to bacterial or fungal contamination may be quickly detected, thus minimizing spread of infection to other dishes and avoiding antibiotic selection for mycoplasma. Cells are subcultured using 0.025% trypsin (Invitrogen) and 0.004% NaEDTA. Cell counts for initiating experimental cultures and for quantitating cell growth are done using a Coulter Counter Z1 (Beckman Coulter, Inc. Brea, CA). Cultures are preserved in liquid nitrogen using a programmable freezer (Planer Products, Sudbury on Thames, England). Institutional Review Boards at Vanderbilt University, Tennessee Valley Health Care System and Meharry Medical College have approved all described studies. The study was conducted according to the Declaration of Helsinki Principles.
Total RNA is isolated using an RNeasy Mini Kit (Qiagen, Valencia, CA) and reverse-transcribed using a cDNA Archive kit (Applied Biosystems, Foster City, CA). Assays are performed using an iQ real-time PCR system (Biorad Laboratories, Hercules, CA). Specific gene expression is measured as described (Smith et al., 2008). Fold-differences in expression of selected genes are estimated using the comparative CT method described in User Bulletin #2 ABI PRISM 7700 Sequence Detection System (Applied Biosystems 2001). Fold difference ranges are determined by evaluating the expression: 2−ΔΔCT ± s, where s is the standard deviation of ΔΔCT.
Standard immunohistochemical procedures available in Vanderbilt’s Immunohistochemistry Core Laboratory were modified where necessary for particular protein probes. For tissue sections, keloids were fixed in paraformaldehyde for 24 hours, embedded in paraffin, and sectioned at 5 microns. Sections were immunostained with rabbit polyclonal antibodies directed against IGFBP5 (Upstate Biotechnology, Lake Placid, NY), SFRP1 (Abcam, Cambridge, MA) and SFRP2 (Sigma). Antigen retrieval methods were used as needed.
Methylation was quantified using the Sequenom MassARRAY System (Sequenom, Newton, MA). Genomic DNA was isolated from fibroblasts and bisulfite treated to convert nonmethylated cytosines to uracils (C to T in PCR amplification products). These C/T variations appear as G/A variations in cleavage products generated from the reverse strand by base-specific cleavage. The G/A variations result in a mass difference of 16Da per CpG site, detectable by the Mass ARRAY system. Relative amount of methylation was calculated by comparing the signal intensity between the mass signals of methylated and nonmethylated templates. PCR primers, selected to hybridize with sequences that do not contain CpG, are designed to yield amplification products of between 200 and 600 bases. In the case of the SFRP1 CpG island, six amplicons were sufficient to provide overlapping coverage of the entire island.
Pooled DNA from normal and keloid strains used in gene profiling studies was digested with Mse1 to produce small fragments (200–1000 bp) while keeping CpG islands intact, denatured, and subjected to methylated DNA immunoprecipitation (MeDIP). Amplified ChIP samples were labeled with different fluorophores and co-hybridized to a Nimblegen CpG-Island Plus promoter array (Roche NimbleGen, Inc., Madison, WI). A computer program developed to analyze data from NimbleGen-tiled microarrays (ACME) was used to identify signals or “peaks” in the array data using a simple sliding window and threshold strategy. A probability value was assigned to each probe on the array (Scacheri et al., 2006). Peak files (.gff) identifying regions of DNA methylation were generated from the p-value files, and peaks were mapped to the transcription start site of each gene and visualized using SignalMap, a software package provided by Nimblegen.
We thank Joel Trupin for his contributions to every aspect of the keloid program since its inception. We acknowledge the outstanding contributions of the Molecular Genetics and Immunohistochemistry cores of the Vanderbilt Skin Diseases Research Center, the Vanderbilt Microarray Shared Resource and the DNA Resources Core of the Center for Human Genetics Research. We also thank Will Bush for his help in preparing figures. This work was supported by NIH grants CA-17229 (JDR), AG-02046 (JDR), P30AR041943 (SMW) F33AR052241 (SBR) 1UL1RR024975 (SBR), and by resources of the VA Tennessee Valley Health Care Center, Vanderbilt University School of Medicine and Meharry Medical College
CONFLICT OF INTEREST
The authors state no conflict of interest.