Our bioinformatics analysis of NGS and comparison of transcriptomes between RNAs of control and DSLD-affected horses provided a window into numerous factors and molecules potentially involved in DSLD. As skin biopsy would be preferred method as a source of a diagnostic marker, analysis of DEG in skin suggest the feasibility of using the results for development of a diagnostic assay for DSLD. This is comparable to the use of subcutaneous adipose-tissue derived fibroblasts rather than tendon-derived cells to study changes in gene expression in DSLD by Lu et al (26). As shown by Seidler et al. and Miyake et al. cultures of skin fibroblasts have been a useful tool for determination of genetic and biochemical causes of several less common forms of Ehlers-Danlos syndrome, a systemic disease similar to DSLD (27, 28). The good separation of control and DSLD transcriptomes into two clearly distinguishable groups suggests that a diagnostic assay based in these results could be developed and would validate our use of skin for RNA extraction. Moreover, DSLD-affected horses clustered together regardless of their breed, Peruvian Pasos or Warmbloods. We were not able to determine the role of age and sex on the DEG, in part because of the small number of horses in each group, and because of lack of more data on gene expression in horses in general. Peruvian Pasos are rather uncommon in the US (about 5,000 such horses are in the US) and it was not easy to convince owners to participate in such study. Unfortunately, because horses are fairly long living and the course of DSLD is unpredictable and so far not examined in a systematic matter (e.g., prospective study), it is difficult to estimate the impact and course of RNA changes in the natural course of the disease. In addition, though we know that DSLD is a progressive disease we do not know much about its pathogenesis and the nature of the progression. Ours is the first report examining DEG in DSLD.
We do recognize that this will have to be confirmed with a larger number of horses, both controls and with DSLD, and of other breeds as well. As expected, analysis of transcriptomes revealed differential expression in numerous genes, in 1567 to be exact, with more genes downregulated than upregulated. To no surprise differential gene expression affected many proteoglycans, growth factors and signaling molecules, and ECM constituents. Our results brought some expected outcomes, and many unanticipated results as well.
As part of our ongoing efforts to identify a factor(s) initiating and/or driving the proteoglycan accumulation in DSLD we reported on increased content of BMP2, a chondrogenic, osteogenic and tenogenic growth factor and a member of the TGFβ super-family (29, 30), in cellular foci in DSLD. These foci consists of active fibroblasts/tenocytes with small amounts of proteoglycans and high content of BMP2 in their cytoplasm (8). This finding correlates well with BMP2 overexpression described here. The fact that skin RNA rather than RNA from tendon or ligament was analyzed may account for the relatively low degree of overexpression of BMP2 in skin, and high overexpression of other, rather unexpected genes, such as those for keratins (see below). The BMP2 overexpression was not accompanied by an increase in transcription in genes encoding for Smads, the mediators of the canonical TGFβ/BMP signaling pathways (31). Instead, the observed overexpression of genes encoding for Fos and many mediators in the MAPK pathways indicates that MAPK pathway plays an important role in inappropriate expression and activity of BMP2 in DSLD. Crosstalk between TGFβ/BMP signaling and Ras/MAPK system has been noted in other systems as well (31-33).
The underexpression of TGFB1, TGFBR3, LTBP1 (encoding for latent TGFβ-binding protein 1), CHRDL1 (encoding for chordin-like 1 protein; antagonist of BMP4), and TGFBI (TGFβ induced), also known as INHBA (inhibin β A chain) genes correlates well with previously observed of no or only small changes in TGFβ content in DSLD (8).
Dysregulation of action of TGFβ and related molecules, such as BMPs (BMP2, BMP4 and BMP6) and CTGF (a mediator of BMP activity) in damaged tendons has been well documented in human and animal tendinopathies where the excessive presence of BMPs can lead to increased synthesis and deposition of proteoglycans in the tendon (29, 34). CTGF (encoding for CTGF/CCN2) is active in chondrocytes, and plays important roles in wound healing and fibrotic processes (35). Under normal conditions the regulation of terminal chondrocyte differentiation by CTGF/CCN2 is opposed by tsukushi, a member of the SLRP group, that affects proliferating and hypertrophic zones of the growth plate (36). Underexpression of TSK, gene for tsukushi, might thus contribute to the presence of not well organized and differentiated cartilage islands in DSLD tendons (1).
Interestingly, both FGF18 and FGF19 genes, encoding for chondrogenic growth factors, were underexpressed, perhaps as the result of BMP2 overexpression. FGF18 is an anabolic chondrogenic and osteogenic growth factor acting through FGFR3 (37, 38). We hypothesize that the underexpression of FGF18 and FGF19 may contribute to further underexpression of genes encoding for core proteins of many proteoglycans, especially those negatively regulated by BMP2 as well (Fig. 4). It is likely that FGF5 overexpression is associated with overexpression of keratin genes (Fig. 6) as FGF5 is involved in normal follicle structure and hair growth (39).
Rui et al. have observed that treatment of tendon-derived stem cells with BMP2 leads to decrease in deposition of several proteoglycans, such as decorin, biglycan and fibromodulin, though they noted overall increase in GAG production and increase in aggrecan as well (30). Obviously, the decrease in expression of many genes for core proteins of proteoglycans in DSLD tissues does remain somewhat mysterious as it is proteoglycans that accumulate in connective tissues in other organs besides tendons and ligaments in DSLD (1). ACAN, gene encoding for aggrecan core protein was downregulated, at least in skin, but ADAMTS4 which encodes for aggrecanase was upregulated. The increase in ADAMTS4 is in agreement with report by Plaas et al. (7). However, they found an increased presence of aggrecan in DSLD-affected tendons and concluded that accelerated degradation of aggrecan by aggrecanases led to DSLD as the result of accumulation of aggrecan degradation products. Our previous, unpublished data found no changes in aggrecan staining in DSLD tendons. By the way, the degradation of articular cartilage in osteoarthritis is thought to be the results of ADAMTS5 and likely also of ADAMTS4 activity (38), two enzymes thought to be involved in degradation of certain SLRPs, e.g., of fibromodulin as well (40). Though our previous work has demonstrated the presence of modified decorin in tendons with DSLD, it was clear from immunohistochemistry that the majority of the proteoglycan in these tissues was neither decorin nor aggrecan (1, 4).
The observed increased expression of hyaluronan synthase and binding protein genes may represent a compensatory mechanism of (attempted) increased hyaluronan synthesis which would offset the decrease in ACAN expression. This finding will have to be confirmed in other organs besides the skin. TSK, a gene encoding for tsukushi, a member of class IV SLRPs functionally related to class I SLRPs of which decorin and biglycan are also members (41) was underexpressed as well (see also above). Tsukushi, decorin and biglycan are known to inhibit TGFβ/BMP/Smad pathways (42). Several studies indicate that tsukushi modulates osteoblast differentiation through inhibition of BMP4 signaling, inhibits Wnt pathways and regulates hair follicle cycle, all features it shares with decorin and biglycan (41, 42).
Gene defects in several human enzymes participating in GAG synthesis, among them xylosyltransferases 1 and 2, and at least two galactosyltransferases, are held responsible for several uncommon disorders affecting skeletal and joint structures (43). A defect in B3GALT6 (encoding for β-1l3-Galactosyltransferase-II) is tied to the progeroid type of Ehlers-Danlos syndrome (44). We did report similarities between this type of Ehlers-Danlos syndrome and DSLD in our earlier work (4). Some underexpressed and overexpressed genes encoding for enzymes involved in synthesis and degradation of proteoglycans and glycoproteins are listed in the Results section. We did not find any changes in the expression of glucuronyl C5-epimerase (dermatan sulfate epimerase), a limiting enzyme in the synthesis of dermatan sulfate (45), however the possibility of a mutation cannot be excluded. Previously we hypothesized that this epimerase might play an important role in pathogenesis of DSLD (4). A complete list of genes for enzymes of interest can be found in the submitted data set.
The expression of several other growth factors was decreased (Fig. 3). The significance of GDF10 underexpression in DSLD is difficult to assess at this time. GDF10 encodes for BMP3B. Though BMP3B was characterized as a primarily growth factor stimulating axonal sprouting in the cerebral cortex (46) it has been described also as an inhibitor of osteoblastic differentiation (47). Similar phenomenon was observed with IGF1, and IGFBP4 and IGFBP5. IGF-1 and IGFBP-4 are involved in stimulation of osteogenic differentiation. IGF-1 and IGFBP-4 promote proliferation and maturation of chondrocytes using the Wnt/catenin signaling pathway (48-50) whereas IGFBP-5 promotes fibrosis, cell senescence, and it also promotes migration of macrophages, an inflammatory step preceding fibrosis (51, 52). Whether the decrease in expression of IGF-1, IGFBP-4 and -5 is the result of negative feedback by BMP2 or one of the other dysregulated growth factors or signaling molecules remains an open question. However, the lack of extensive calcifications in most cases of DSLD would be compatible with these results (1, 4, 53). Primary calcifying desmopathy in horses presents as extensive calcifications of tendons but it is encountered rather infrequently (54). The underexpression of several members of the PDGF/VEGF family (VEGFC, KDR - encoding for VEGF receptor-kdr-like protein, PDGFRB and PDGFRA) is more difficult to explain as their expression is enhanced in other systems by increased BMP2 presence (29), however, this corresponds to minimal presence of significant blood vessels in the DSLD affected tissues, including active foci producing BMP2 (1, 8).
Genes encoding α chains of numerous collagen types were underexpressed. This is indicative of profound disturbance in collagen metabolism, whether it is the consequence of altered expression of BMP2 or changes in proteoglycan synthesis remains to be established (29, 55).
Only genes for α chains of two collagen types were overexpressed, one for the α chain for type 17 collagen, the 2nd gene was for the α chain for type 26 collagen. Type 17 collagen coordinates cell proliferation in interfollicular epidermis (23). Its function was shown to be defective in human epidermolysis bullosa (24). It is possible that its overexpression in DSLD horses explains the presence of loose wrinkly skin, patches of grey hair, and bruises in some of these horses (personal communications). In addition, the overexpression of BMP2 may contribute to these changes as well. BMP2 plays a significant role in the embryonic development of skin and its appendages, including hair follicles, specifically in hair placode (56), whereas BMP4 directs the development in mesenchymal cells located beneath the hair placode (57). A more recent report has shown that overexpression of constitutively active BMP-receptor-IB (one of the receptors for BMP2) in transgenic mice leads to ichthyosis-vulgaris-like skin disorder characterized by hyperkeratosis (58). The overexpression of FGF5 in DSLD transcriptomes points to a possible involvement of FGF5 in impaired hair growth (39, 59)
Phenotypically, DSLD is clearly and unequivocally distinct from Hereditary Equine Regional Dermal Asthenia (HERDA) (60, 61) and Warmblood Fragile Foal Syndrome or WFFS with primary skin involvement, and only occasional presence of affected tendons and joints (62). A pinpoint mutation in the equine procollagen-lysine, 2-oxoglutarate 5-dioxygenase (PLOD1) gene is implicated as the cause of WFFS, an autosomal recessive condition. Horses affected with WFFS present shortly after birth with thin fragile skin, hyperextended joints, and poor wound healing (62). People with mutation in PLOD1 suffer from so called kyphoscoliotic Ehlers-Danlos syndrome (63), a disorder reminiscent of other, rare types of Ehlers-Danlos with mutations in carbonic sulfotransferase 14 or dermatan-sulfate epimerase (64)
The role and significance of type 26 collagen is unknown. Its expression appears to be limited to the testis and ovary (65).
Genes for MMPs 1,9,19, 23B and 25 and at least two of the tissue inhibitors of metalloproteinases (TIMPs 1 and 2) were underexpressed. TIMP1 inhibits the activity of MMP 9 (66), and it is thought that TIMP1 plays an important role in limiting inflammation following injury (67). TIMP2 inhibits the activity of MMP2, but it is also participatory in indirect activation of MMP2 through association with MMP14 that may promote cancer progression (68) and, more importantly in the context of DSLD, aortic aneurysm development (69). It might be of some significance that not only these MMPs are collagenases and/or gelatinases, but most of them degrade proteoglycans (e.g., aggrecan and versican) as well (66, 68).
Hofberger et al have associated idiopathic chronic degeneration of the SL, including DSLD, with pituitary pars intermedia dysfunction or PPID (70, 71). PPID is characterized by elevated free cortisol fraction levels accompanied by increased immunostaining for 11-β-dehydrogenase type I in SL and skin. We did not notice any changes in expression of HSD11B1 gene (which encodes for 11-β-dehydrogenase type I), however HSD11B2 gene encoding for 11-β-dehydrogenase type II was found to be underexpressed. Similar decrease in staining for 11-β-dehydrogenase type II was predicted, but not verified by Hofberger et al (71). Interestingly, they did find skin thinning in their PPID-affected horses. Whether the decrease in gene for 11-β-dehydrogenase type II, and SL and skin changes in horse with DSLD found by us are analogous to findings identified in horses with PPID by Hofberger remains to be determined. No clinical signs of PPID were observed by us, owners and any of the veterinarians who provided skin samples or horses for our study.
The lack of inflammatory cells in DSLD-affected tissues is rather conspicuous (1, 8). As noted in the Results section many genes for proinflammatory proteins and peptides, including chemokines, TNFα and TNFα-system related molecules were downregulated. The expression of genes for ADAM 9, 19 and 33 was decreased as well. In general, ADAM genes and their products are involved in a variety of pro-inflammatory processes. ADAM 9 and 19 are membrane-anchored enzymes activating cytokine precursors, including that for TNFα into active molecules (72, 73). ADAM 33, the third underexpressed gene of the ADAM family, has been identified as a susceptibility gene for asthma and chronic obstructive pulmonary disease, and it likely plays a role in stimulating immune function, and remodeling of extracellular matrix (74).
Though NGS is a powerful tool to evaluate level of expression of individual genes, or transcriptomes, it does not tell us much about the translation mRNAs into actual protein synthesis and function. Another drawback of NGS is that it does not identify the presence of mutations in individual genes that might be instrumental in pathogenesis of DSLD, more specifically, in the increased proteoglycan presence either due to a mutation in a core protein of a less characterized proteoglycan, or in an enzyme facilitating synthesis of GAGs attached to proteoglycans.