This study examined gene transcription at the exon level on a genome wide scale in DD patient fibroblasts. Each exon array provided extensive whole genome transcript coverage and allowed robust data acquisition for gene expression analysis superior to gene expression platforms used in earlier publications 
. Of the more than 15,000 genes tested, 302 had significantly higher transcript levels (>2 fold) in DD and 1276 had lower transcript levels (>−2 fold) in DD when compared with controls.
Most of the previous studies have compared expression levels in tissue biopsies between diseased regions and unaffected regions from the same patients 
and a few of these studies also included comparisons with tissue samples from carpal tunnel release patients 
, another soft tissue hand disease associated with diabetes 
. Few previous studies have compared expression levels in primary fibroblasts derived from the affected and unaffected regions of DD patients 
or from the two patient types 
. One compared the tissue between DD patients and hand trauma patients 
. Variation in the expression levels found can be accounted for by the differences in the cells contained in the tissue samples that were compared. In the tissue biopsies, the cell population consists of other cell types as well as fibroblasts. In addition, fibroblasts present in the DD tissue samples are in the diseased environment and expression levels may be influenced by the extensive cellular matrix and cell density. Differences in gene expression levels of primary fibroblast cells from diseased and non-diseased areas have been shown to decrease after 4 to 6 passages 
. These different comparisons provide information about different aspects of the disease. The primary fibroblast cells used as controls in our study were derived from skin punch biopsies taken from the thigh and may elicit some fibroblast specific expression differences due to cell derivation from skin of a different region 
, but these cells are derived from cancer patients with no history of Dupuytren's disease or other soft tissue diseases and had no major scarring after radiation therapy. Variation between our findings and others may reflect genetic susceptibility to DD in addition to the disease state.
The major collagen found in the palmar fascia tissue is type I but in Dupuytren's nodules it has been reported that there is an increase in collagen and in particular, a higher proportion of type III compared to type I 
. However, Murrell et al., 1991 found that the fibroblast cells (passage 3) from DD tissue and carpal tunnel control tissue did not show any noticeable difference in their collagen production and suggested that the increase in collagen type III to type I ratio found in the tissue samples was due to inhibition of collagen type I production in the fibroblasts growing in higher density as found in the DD tissue 
. They demonstrated that an increase in fibroblast density resulted in an increase in the ratio of collagen type III to type I due to a decrease in collagen type I production. The fibroblasts used in the present study were grown to approximately the same density to avoid issues in expression differences associated with cell density.
Compelling evidence shows that the collagen-associated transcripts are a key component of progression of DD 
. Satish et al (2008) 
transcripts were lower in DD samples which is the opposite to our results. However, Satish et al (2008) were comparing DD and carpal tunnel syndrome derived fibroblasts and the differences between our studies may be due to the differences in our controls. Some past studies have also shown a higher level of expression of various collagen genes in DD 
. Our analysis indicates a number of other collagens that showed higher levels in DD samples compared to controls. However, it is possible that some of the highly expressed collagen transcripts (e.g. COL15A1
) are binding to the similar PSR sets for other collagen genes which would then also manifest as increased in DD patients.
Modulated expression levels in matrix metalloproteinase (MMP) genes is a common finding in previous studies although results vary depending on the experimental design 
. Various MMP genes have been shown to have a higher expression level in DD and a few others have been shown to have lower expression. For example, MMP2
has previously been shown to have a higher expression in DD 
. We found a decrease in the expression level of MMP16
in DD. MMP16 protein activates MMP2 protein which in turn degrades type III collagen. We also found a substantial decrease in MMP1
gene expression in DD compared to the gene expression of control fibroblasts obtained from patients with no signs of DD. However, Johnston et. al. (2007) 
found a higher expression level of MMP1
in DD compared to carpal tunnel syndrome tissue samples. MMP1 protein functions as an interstitial collagenase to break down interstitial collagen types I, II and III. The expression levels of the MMP3
(stromelysin) gene, which codes for a protein that is able to activate the MMP1 protein 
was also down in DD cells. Rehmen et al (2008) 
who compared DD and carpal fascia tissue (from patients with carpal tunnel syndrome), also found a decrease in expression level of MMP3 in DD. We speculate that a low level of activated MMP1 proteins in DD may cause an accumulation of type I, II, and III collagens in the ECM due to an inability to break them down. In addition, low levels of MMP16 protein may decrease the activation of MMP2 and increase the build-up of collagen type III.
These findings provide compelling evidence that the development and progression of DD is closely associated with significant up-regulation of a broad group of collagen genes and down-regulation of matrix metalloproteinase and other collagenase genes which are required in remodelling the ECM. They also extend the understanding of the likely genetic origins of DD and provided the experimental rationale for the recent use of injectible collagenase from Clostridium histolyticum
in the non-surgical treatment of DD which has been found to be effective in controlling DD despite the associated pain tolerated by patients 
. A twelve month follow-up study of this treatment indicated that some patients had debilitating pain and deep tissue scarring and adhesion 
. Longer term studies are now required with this treatment to examine its effectiveness in preventing recurrence of DD and also to assess any negative consequences or non-specific effects of the treatment. More specific collagenases such as active MMP1 and MMP2 proteins may be better candidates for therapeutic treatment.
Other ECM components may also be involved in DD. Our findings show that transcripts of cathepsin K (CTSK
), a lysosomal cysteine proteinase involved in bone and possibly ECM remodelling and resorption, are also lower in DD samples. Other matrix remodelling genes such as plasmin-mediated matrix remodelling protein, tissue factor pathway inhibitor 2 (TFPI2
) and TFPI
transcripts were also expressed at significantly lower levels in DD samples. BMP4, which has increased transcription levels in DD fibroblasts, induces cartilage and bone formation but also has been shown to regulate tissue remodelling and fibrosis 
. The proteoglycan gene PRG4
was found to have higher expression levels in DD compared to controls which is consistent with other studies that have investigated fibroblasts from DD patients 
. This proteoglycan prevents protein deposition on to cartilage but its function in DD remains unclear.
Not all gene transcripts associated with the ECM showed higher levels in DD. For example, two fibronectin genes, fibronectin type III domain containing 3A (FNDC3A
), and fibronectin leucine rich transmembrane protein (FLRT2)
showed lower levels in DD fibroblasts. There was also a down-regulation of transcripts from laminin 4 alpha gene (LAMA4
). Laminin 4 alpha is part of laminin 411 which is found in endothelial basal laminae and is believed to up-regulate insulin gene expression 
. This gene may reflect the high incidence of diabetes in DD patients 
. Another gene that has been associated with type II diabetes and shows higher gene transcript levels in DD, is angiopoietin-like 4 (ANGPTL4
), a glycosylated, secreted protein with a fibrinogen C-terminal domain involved in glucose homeostasis, lipid metabolism and insulin sensitivity 
. It inhibits proliferation, migration and tubule formation in endothelial cells and is induced and accumulates in the ECM in response to hypoxia.
Our data indicate an increase in cell-to-cell interaction and dysfunction in the regulation of cytoskeletal structure. Many transcripts from genes involved in cell adhesion are found at a higher level in DD. For example, the vascular cell adhesion protein, (VCAM1) cell adhesion molecule 1 (CADM1), chitinase 3 like 1 (CHI3L1), neuronal cell adhesion molecule (NRCAM), and thrombospondin 4 (THBS4). As vascular cell adhesion gene (VCAM1) has an important function in cell-cell recognition and thrombospondin 4 is an adhesive glycoprotein that can bind fibrogen, fibronectin, laminin and type V collagen, an increase in these proteins would increase the amount of adhesion between the cells as well as with the ECM. There are also lower transcription levels in DD of the podocalyxin-like gene (PODXL), a negative regulator of the cell adhesion. These finding suggest that genes promoting cell adhesion are increased in the development and progression of DD.
ADAMs proteins are active metalloproteinases with gelatinolytic and collagenolytic activity. They inhibit beta-1 integrin mediated cell adhesion and migration. The ADAMs suppress cell mobility, cleave E-cadherin in response to growth factor depletion and may be active in cartilage remodelling. We found gene transcripts of these proteins, which increase cell adhesion and decrease cell mobility, are increased in DD as in previous studies 
. Integrins are also involved in cell adhesion and participate in cell-surface mediated signalling 
. The integrins (ITG) gene transcripts were found to be both higher and lower in DD. For example, ITGA11
transcripts were found at higher levels in DD whereas ITGA2
, and ITGA4
gene transcripts were lower in DD. Integrin alpha 11 cell surface adhesion receptor is involved in cell adhesion to the ECM and to other cells. The levels of this gene are increased in DD possibly increasing adhesion of cells and ECM in DD. Integrin alpha-2/beta-1 is a receptor for laminin, collagen, fibronectin and E-cadherin and is responsible for adhesion of cells to collagen, modulation of collagen and collagen gene expression, and organization of newly synthesized ECM. The levels of ITGA2
transcript, which encodes for integrin alpha 2, are down in DD which may cause a disorganisation of collagen.
Many of the gene transcripts that are lower in DD are involved in cytoskeletal structures (cytoskeleton associated protein 2 (CKAP2)), microtubule-based movement (KIF family), spindle formation (e.g., TPX2), centromere proteins such as kinetochores (NUF2, CENPF), and chromosome condensing (NCAPG). Lower levels of transcripts in these genes may reflect differences in proliferation between the two sample groups, however, there is also a lower level of Rho-associated genes that are involved in cytoskeletal rearrangement and cell motility. This includes SEMA3A, a protein possibly involved in cytoskeletal organisation, and indicates a possible association between DD and cytoskeletal structure.
We were interested to determine if follistatin and activin were involved in DD disease. Follistatin has been shown to antagonise fibrosis by complexing with activin 
and to modulate the proinflammatory and profibrotic actions of activin during wound healing, tumourigenesis 
and in rats treated with bleomycin, an agent that causes DNA double-strand breaks 
. When we analysed the differences in follistatin and activin subunit gene expression, we observed that levels of follistatin were much lower in the DD samples, which is consistent with the proposed anti-fibrotic actions of follistatin 
. Our study found unexpectedly that the levels of INHBA
, which codes for the βA subunit of activin and is known to be involved with fibrosis, were also down-regulated. In contrast, expression levels of INHBB
, which encodes for the activin βB subunit, were elevated in DD fibroblasts when compare with controls. However, the levels of INHBB
transcript varied greatly between individuals, for both DD and control patients. Further, there is little data on the role of activin B in the modulation of fibrosis as assays for this protein have only become available recently 
. In vitro
studies are required now to further explore the relationship between follistatin, activins and collagen synthesis in DD fibroblasts because of the potential for follistatin to be used as a novel treatment for DD.
Two fibroblast growth factor genes (FGF9 and FGF11)
were significantly up-regulated in DD fibroblasts. The FGF family of genes encode for mitogens and proteins involved in cell survival and various cell processes. Up-regulation of these growth factors in DD fibroblasts links with the increase in fibroblast proliferation and fibromatosis in DD. The proteins encoded by these genes are members of the fibroblast growth factor (FGF) family and are implicated in the stimulation of cell growth and tissue repair. The protein encoded by FGF9
was isolated as a secreted factor that stimulates growth in cultured glial cell but the exact functions of both FGF9
on fibroblasts, and particularly those from DD patients, have yet to be determined (http://www.ncbi.nlm.nih.gov/gene/2256
). Platelet-derived growth factors have been recently implicated in DD 
and have a specific effect on angiogenesis. However none of the three PDGF genes (PDGFB, PDGFC, PDGFD
) examined in this study were up-regulated in DD fibroblasts suggesting that enhanced angiogenesis is not a critical factor in the establishment of DD.
Other genes up-regulated in DD are also involved in inflammatory diseases. For example, tumour necrosis factor, alpha-induced protein 6 (TNFAIP6), which is found in the synovial fluid of patients with osteoarthritis and rheumatoid arthritis was up-regulated in DD. Vascular cell adhesion gene (VCAM1), which may play a role in atherosclerosis and rheumatoid arthritis, showed a 30-fold increase in expression in DD. VCAM1 also has an important function in cell-cell recognition. However another gene that codes for a protein involved in the innate immune response, STAT1, had a lower expression level in DD patients which may reflect the lack of inflammation observed in DD.
There is also up-regulation of a suite of keratin genes, particularly KRT34, in DD fibroblasts. Keratin is a protein usually involved in the formation stratified squamous epithelium and hair and is particularly associated with keratinocytes in the skin. Currently the nature of this relationship between up-regulated expression of KRT genes in DD fibroblast and DD is unclear.
Application of an alternative splicing algorithm to our exon array data revealed a number of gene isoforms that appear to have different levels between DD and control fibroblasts. These included THBS4
. There is approximately a 12 fold increase in the THBS4
gene transcript in DD patients but only for part of the gene. The data suggests an increase in a transcript that codes for a protein that is missing approximately the first 92 aa of the N-terminus. As the N-terminus of this protein binds to heparin (possible binding site between aa 102 and 105) 
; this protein may have an altered heparin binding capability but would probably still be involved in cell-matrix interactions. As there is an increase in KRT34 gene in all but the first PSR (first exon) in DD, the increased KRT34
transcript codes for a protein missing approximately the first 43 aa of the N-terminus. The TIRAP
transcript increased in DD patients is missing the 5′UTR of the gene. Although the resulting protein would be unchanged, the stability of the transcript may be altered. The expression profiles of both KIF14
indicate that the regions corresponding to exon 1 in both genes are unchanged, but all the following PSRs are decreased in DD fibroblasts. Transcripts in these unchanged regions may be protected from RNA degradation that is occurring in DD or premature termination of transcription is occurring at this point. A similar profile of decreased transcript expression with protection of the first exon was found 4 hours following 10 Gy ionizing radiation of fibroblasts in both KIF14
. Recently published papers examining the response of fibroblasts to ionizing radiation found that a common general response mechanism to that stress is the use of alternative start sites 
. The disease state in DD could reflect a defect in a stress response leading to alternative isoforms that have substantial effects on the normal production of the ECM. Alternatively, lower oxygen levels may be present in the zone of the affected region of the hand, which may induce alternative transcripts.
Recently, a large study that looked at SNPs in 1365 DD patient bloods identified a number of WNT gene SNPs as being associated with DD patients 
. One of the SNPs was in SFRP4
, a frizzle-related gene which was at the top of our list for genes that are more highly expressed in DD compared to normal samples. However, this SNP was more closely associated with the gene EPDR1
, a gene involved in cell adhesion, which we found had an increased expression in DD. These results provide intriguing clues to the cause of DD disease.
An advantage of our study was the use of control samples from donors with no DD genetic background. This avoided the potential complication of associated genetics in control samples from DD-affected tissue donor which have been used as controls in some previous investigations. We also had the advantage that highly sensitive exon arrays were used to obtain quality results. In this study we have not only identified transcripts which are precursors to known fibrotic components, but also have identified a large number of potential DD treatment candidates, some of which have also been identified in other genomic studies. Some of our findings, however, also contrast with and contradict those of other studies and require further examination to discovery how they relate to the onset and progression of DD.
In conclusion, we have comprehensively characterized the transcription profile differences between DD and normal primary fibroblasts. Our data indicate that in DD there is an excess of collagen and other ECM that is not controlled due to a reduction in matrix metalloproteinases and other matrix remodelling proteins. A reduction in the fibrotic control protein, follistatin, may also contribute to DD. In addition, the fibroblasts lack expression of genes involved in cell movement and cytoskeletal organisation and an increase in genes involved in cell adhesion. These indicate a lack of organisation of both extra- and intra-cellular matrix as well as a lack of cellular movement in DD. Alternative transcripts have also been identified which are expressed at different levels in the DD patients compared to the controls and may reflect cell stress such as hypoxia. These conclusions will be the basis for future experimentation. Many of the identified genes are potential candidates for the treatment of DD. There was a close correlation between expression levels in some genes from our study and data from previous studies using DD tissue samples providing reason to pursue investigations into potential therapeutic development strategies using in vitro studies on DD fibroblasts. It is likely some of these candidate genes for treating DD will also be effective for fibrotic diseases in general, including injury-related and radiotherapy-induced fibrosis.