In the present work, we have performed large-scale EST sequencing and analysis in order to generate sufficient sequence information to study fracture healing at a genomic level in the sheep for the first time. The results of our study are of interest not only in the context of fracture healing and bone development, but will also provide useful information for all fields of research that use the sheep as a model organism. The ESTs generated in the present work represent an increase in the number of available ESTs by a factor of over 4.5, and provide at least partial sequences for over 5800 sheep genes never before sequenced. The sequences can be downloaded from the online supplemental site and are also available in the GeneNest Sheep database.
Genomic approaches to the sheep have great potential to accelerate progress in research areas using the sheep as a model organism. For instance, the first study using an ovine cDNA microarray prepared from unpublished EST libraries was very recently published. A number of genes differentially expressed in sheep resistant and susceptible to gastrointestinal nematode infection were identified .
Although we now have initial concepts of the molecular processes involved in fracture healing, many questions remain to be answered before a comprehensive picture of the genetic programs involved in fracture healing can emerge. An in-depth understanding of the genes involved in fracture healing and the regulatory mechanisms that control them will be prerequisite for developing optimal treatments for disturbances of fracture healing and will have relevance to other disorders such as osteoporosis. At present, we still do not have a comprehensive list of key genes and gene products involved in bone development and fracture healing. In this study, we have performed the first large-scale EST sequencing project using a model for fracture healing and have identified numerous differentially regulated genes not previously known to be involved in fracture healing. EST sequencing gives a nearly unbiased view of highly expressed sequences in a tissue (assuming that the process of producing the ESTs does not introduce a significant amount of bias), and therefore offers some advantages over candidate gene approaches and even over microarray-based approaches that may not cover the entire genome.
In this work, we have generated a profile of postfracture day 7 and day 10 genetic expression using EST libraries produced by automated plasmid purification. No normalization procedures were applied to select ESTs for sequencing. Therefore, we believe that the distribution of ESTs in these libraries gives a nearly unbiased view of genetic expression at these stages of fracture healing. In addition, we generated a third EST library from a pool of later stages of fracture healing and used RT-PCR to examine differential expression of 30 of the 39 genes found to be enriched in the PF7 and PF10 libraries and 48 other genes (Tables 2 and 3).
Previous genomic approaches to fracture healing have involved analysis of a rat femur fracture model based mainly on microarray expression analysis, but different experimental and statistical approaches were used [11, 17, 18]. These studies have collectively identified thousands of genes that are differentially expressed in fracture healing. The genes were grouped by the authors into a number of functional classes, which overlap to a large extent with the classes identified in the present investigation, including extracellular matrix proteins, ribosomal proteins, resorption/remodeling, transcriptional regulation and signal transduction. However, the specific genes belonging to the functional classes are often different among the three studies cited above as well as in comparison to our results. For instance, Hadjiargyrou and coworkers (2002) identified 20 differentially expressed ribosomal genes; in the present study, 16 differentially expressed ribosomal genes were identified. With the exception of ribosomal protein SA, there is no overlap between the two lists. This suggests that we are just beginning to identify to the full complement of genes involved in fracture healing.
Genes differentially expressed in fracture healing
Many of the differentially expressed genes we identified were previously known to be involved in various biological processes involved in bone development or fracture healing (see Tables 2 and 3 for references), suggesting that our approach has captured some essential aspects of fracture healing. Additionally, we have demonstrated differential expression for many genes not previously known to be involved in fracture healing.
The differentially expressed genes can be divided into several different classes, as will be discussed below. Three of these classes, extracellular matrix proteins, resorption/remodeling/inflammation, and transcriptional regulation/signal transduction, have been well studied in the context of fracture healing, and our results have identified both previously known and novel differentially regulated genes with these functional roles.
In the following, we describe three other classes with numerous differentially regulated genes, angiogenesis, free-radical control, and ribosomal genes. These functional classes have received less attention to date, although they are known to be involved in the process of fracture healing. Fracture-related differential expression had not previously been shown for the great majority of genes in these classes (cf. Tables 2 and 3). These observations thus represent starting points for future studies on the role of processes such as angiogenesis and free-radical control in fracture healing.
The single largest group of putatively upregulated genes in our analysis was made up of ribosomal proteins. There are over 75 ribosomal proteins , which are not necessarily expressed in concert . We found 9 genes coding for ribosomal proteins to be significantly overexpressed in our analysis of EST distribution compared to normal human bone. All of these genes, as well as seven additional genes coding for ribosomal proteins (Table 3) showed differential expression in the RT-PCR analysis, with expression levels peaking at postfracture days 7 and 10 (Fig. 7). We additionally identified four translation factors as being overexpressed in our EST analysis (Table 2). RT-PCR for two of these genes confirmed differential expression. To the best of our knowledge, a specific role in fracture healing has not been identified for all but two of these genes to date, although some of them are known to be differentially expressed in certain developmental processes (Tables 2 and 3), and a number of other ribosomal genes are known to be differentially expressed in fracture healing .
Cluster B (Fig. 7) contained 36 genes. Interestingly, 13 of the 15 ribosomal genes analyzed by RT-PCR, and both of the translation factors analyzed by RT-PCR were in this cluster, whose expression levels peaked during the early phases of fracture healing at days 7 and 10 postfracture, suggesting that there is an especially high level of protein synthesis at these stages.
Extracellular matrix proteins
Several matrix proteins were found to be differentially regulated. The gene with the highest overall expression, Collagen Iα1, is the main protein component of bone, and the fact that it is one of the most highly upregulated genes is not surprising. Two other collagens and osteonectin were also found to be upregulated by EST analysis (Table 2). We investigated these genes and four other collagen genes and two proteoglycans by RT-PCR and showed differential expression for all genes. In general, the expression of these matrix genes was highest at postfracture day 7 or 10 but was still relatively high at day 14. IBSP is an osteoblast marker that is a major noncollagenous structural protein of the bone matrix. IBSP showed an expression pattern different from that of the other matrix genes, with significant expression beginning at day 10, peaking at day 14, and continuing until day 42.
The lysyl-tRNA synthetase KARS has no known involvement in fracture healing. Given the important role of lysine crosslinks in the biosynthesis of collagen, it is interesting that lysyl-tRNA synthetase was the only tRNA synthetase gene that was overexpressed in our data.
SERPINH1 was also found to be significantly overexpressed in the PF7 and PF10 ESTs. It is a molecular chaperone involved in the maturation of collagen molecules . This is presumably important for bone development because SERPINH1 is localized to regions of type I collagen production in developing murine femurs and tibiae [22, 23].
It has long been known that vascular invasion is necessary for bone differentiation , and angiogenesis is a key process for fracture healing . A number of genes related to angiogenesis have previously been shown to be upregulated during bone repair, such as vascular endothelial growth factor .
COL3A1 and COL4A1 were shown to be overexpressed in the PF7 and PF10 ESTs (Table 2). Both of these collagens have known roles in angiogenesis [27, 28].
Six of the eight genes in our dataset with prominent roles in angiogenesis (CSRP1, COL4A1, COL4A2, Perlecan, HDGF, Endoglin) were in cluster A. The smooth muscle marker CSRP1  and the TGFβ-family auxiliary receptor endoglin are required for angiogenesis and heart development [30, 31]. HDGF is a highly expressed vascular endothelial cell protein .
The genes of this cluster show especially strong expression on day 7 postfracture with lesser amounts of expression on days 10 and 14, which is not surprising given that new blood vessel formation is a prerequisite for the formation of new bone from callus tissue. The other two genes, COL3A1 and TAGLN2, were in cluster B, which like cluster A is characterized by high early expression (PF7 and PF10), but shows a stronger drop in expression at the 14 day timepoint.
Resorption, remodeling and inflammation
Several genes that probably play a role in the inflammatory response characteristic of early stages of fracture healing (FCGRT, Basigin, HLA-A, and CD74) were found to be overexpressed in the PF7 and PF10 ESTs (Table 2).
Resorption of cartilage and bone mediated by osteoclasts is an important part of the remodeling processes involved in new bone formation in fracture healing. The main enzymes involved are thought to be the cathepsins and the matrix metalloproteinases (MMPs), which act in a concerted fashion. Although not all details are clear, our current understanding of this process suggests that osteoclasts create an acidic environment in their resorption lacunae, which results in dissolution of the mineral, and then secrete cysteine proteinases that are active at a low pH, to initiate proteolysis of the proteinaceous bone matrix. Finally, MMPs exert their activity, once the pH has increased sufficiently . Our data has demonstrated differential expression for several enzymes from both classes.
Interestingly, four of the five genes in RT-PCR cluster E encode enzymes with roles in osteogenesis (ACP5, CTSK, MMP9, and MMP13). The first three genes are expressed by osteoclasts. The expression of genes in this cluster was minimal at day 7 but peaked at day 10 postfracture, suggesting that these enzymes might be particularly important for the processes involved in the bone formation that becomes visible by day 10.
We showed differential expression for MMP2, MMP9, MMP13, MMP14, and MMP19. All of these MMPs had previously known roles in bone development except MMP19, providing further evidence for the similarity of bone development and fracture healing. MMP2, MMP9, and MMP14 are differentially regulated in scarless fetal wound healing  and during osteogenesis , which is interesting because bone is one of the few adult tissues that can heal without scarring. MMP19 previously had no known role in skeletal development or fracture healing, but it is highly expressed in dermal wounds, suggesting a role in wound repair . The highest expression of MMP19 was observed in the day 7 and 10 stages.
Cluster F contained only two genes, cystatin C and TPM2. These genes were the only ones with an expression profile that peaked at the last time point (42 days postfracture). Cystatin C (CST3) can inhibit bone resorption and osteoclast formation . Osteoclast activity begins early and increases over the course of fracture healing , and one may speculate that CST3 may be important for regulation of osteoclasts at this time point.
Control of free radicals
Vascular invasion of ossified cartilage during enchondral ossification is associated with breakdown of tissue by the release of lytic enzymes by invading cells such as macrophages and endothelial cells. This process is likely to involve the production of reactive oxygen species (ROS) . At present, little is known about the mechanisms by which unwanted tissue damage by ROS is controlled during fracture healing. The glutathione peroxidases (GPX) are enzymes that catalyze the reduction of hydrogen peroxide, organic hydroperoxide, and lipid peroxides by reduced glutathione. They are thus involved in the protection of cells against oxidative damage. Although none of the four GPX genes we tested in RT-PCR were individually statistically significantly overexpressed in the EST analysis, there were no ESTs for any of these GPX genes in normal bone and a total of 85 ESTs for the four GPX genes in the PF7 and PF10 EST libraries. We showed significant differential expression patterns for each of these genes by RT-PCR, although they did not show a uniform expression pattern (Fig. 7).
HMOX1 was also found to be significantly overexpressed in the PF7 and PF10 EST libraries, as has been previously described for postfracture day 3 in a rat femur fracture model . HMOX1 is an inducible heme oxygenase that could conceivably be involved in heme catabolism during resorption of the initial fracture hematoma, although other roles in development have been postulated for this gene .
Transcription regulation and signal transduction
Two membrane-bound proteins with roles in signal transduction were found to be significantly overexpressed in the PF7 and PF10 libraries. ITM2C is highly expressed during chondro-osteogenic differentiation , but its specific function and its role in fracture healing remain unknown. The G protein GNAI2 was also found to be overexpressed, but no clues are available as to its particular role in fracture healing.
Signal transduction and modification of transcriptional programs by transcription factors represent important phenomena in development, and presumably are essential for fracture healing. Nevertheless, the EST counts for the transcription factors examined in this study were too low to reach statistical significance, even though in many cases ESTs for transcription factors were found only in the PF7 and PF10 libraries but not in the human normal bone libraries. For instance, we identified only one EST each for the transcription factor runx-2 and osterix, which are both key players in skeletogenesis [43, 44].
We performed RT-PCR analysis on a number of transcription factors with relatively large numbers of ESTs in our libraries (Table 3). We were able to demonstrate differential expression with the highest expression at days 7 and 10, for JunB, which among other roles is a positive regulator controlling primarily osteoblast as well as osteoclast activity . MORF4L1, which may play a role in mechanotransduction , and TCF4, which may mediate Wnt signaling during limb development  also showed highest expression levels in the first two time points. Additionally, we showed differential expression for NSEP1, PPIB, and RARA, for which to the best of our knowledge no previous information about differential expression during bone development or fracture healing was available.