|Home | About | Journals | Submit | Contact Us | Français|
The chondroitin sulfate proteoglycan core protein aggrecan is the major protein constituent of cartilage aside from collagen, and is largely responsible for its distinctive mechanical properties. Aggrecan is required both for proper cartilage formation in development and maintenance of mature cartilage. Prominent Acan transcription is a conserved feature of vertebrate cartilage, although little is known about its specific transcriptional regulation. We examined the genomic interval containing human ACAN for transcriptional enhancers directing expression to cartilage, using a functional assay in transgenic zebrafish. We tested 24 conserved non-coding sequences, representing ~6% of the total sequence in the interval, and identified eleven independently capable of regulating reporter gene expression in cartilage. These enhancers were widely spaced, from >100 kb upstream of the gene to within the first intron. While the majority displayed broad cartilage expression in zebrafish larvae, several were restricted to a subset of cartilage cells in the craniofacial skeleton. In older fish, the enhancers displayed differential activity; some maintained expression, either in all cartilage or preferentially in articular cartilage at the joints, while others were not active. This remarkable degree of overlapping regulatory control has been highly conserved; we identified clear orthologues of six enhancers at the chicken ACAN locus, arranged in the same order relative to the gene. These were also functional in directing expression to cartilage in transgenic zebrafish. Several enhancers contain potential binding sites for Sox9, consistent with its described role as an upstream regulator of Acan expression. However, others lacked Sox9 consensus binding sites, implicating additional pathways and transcription factors as regulators of Acan expression in cartilage, either in development or adult tissue. Our identification of these enhancer sequences is the necessary first step in detailed examination of the upstream regulators of Acan expression.
Acan encodes aggrecan, the major chondroitin sulfate proteoglycan (CSPG) core protein of cartilage, a highly abundant component of the cartilage extracellular matrix. Although not exclusive to cartilage, aggrecan is found most prominently in cartilage and its deposition is considered the hallmark of chondrogenesis. Together, glycosylated aggrecan, the associated link protein, and hyaluronan lend cartilage its important mechanical properties of resistance to compression and viscosity. Not only is aggrecan essential for cartilage formation during development, but its maintained expression throughout life is likely to be essential to cartilage integrity.
The importance of aggrecan in development of cartilage is evident from the phenotypes of chicken and mouse mutants. The cartilage matrix deficiency (cmd) mouse mutant (Rittenhouse et al., 1978) is a naturally occurring mutation in Acan that is predicted to produce a truncated protein (Watanabe et al., 1994), although no aggrecan is detected in the cartilage matrix of homozygous mutant mice. Mutants display dwarfism, short snout, and cleft palate, and die shortly after birth. The chicken nanomelia mutant (Landauer, 1965) is also caused by a mutation in ACAN (Li et al., 1993; Primorac et al., 1994), and similarly results in a homozygous lethal chondrodysplasia. In humans, a recessive mutation in ACAN causes severe chondrodysplasia, but is not lethal; the mutant protein displays altered interaction with the extracellular matrix protein Tenascin, and does not seem to represent a null mutation (Tompson et al., 2009).
Aggrecan is also likely to play a central role in osteoarthritis and other degenerative processes affecting cartilage and intervertebral discs. Mice heterozygous for the cmd mutation appear normal at birth. However, they develop mild dwarfism, apparent at one month, and by one year develop severe curvature and misalignment of the spine, eventually leading to neurological impairment (Watanabe et al., 1997). A heterozygous mutation in ACAN has similarly been associated with spondyloepiphyseal dysplasia type Kimberley (SDK), which is clinically characterized by short stature and severe premature osteoarthritis (Gleghorn et al., 2005). Several studies have linked polymorphisms in the ACAN coding sequence with osteoarthritis (Horton et al., 1998; Kamarainen et al., 2006; Kirk et al., 2003) and lumbar disk degeneration (Mashayekhi et al., 2010), suggesting that even subtle differences in aggrecan function or levels could contribute to these common, multi-factorial diseases.
The expression of aggrecan in cartilage is highly conserved in evolution, and has been described in Xenopus and zebrafish as well as chicken, mouse, and other mammals. While presumably its similar expression in such disparate species involves common regulatory mechanisms, little is known about the regulation of Acan transcription. It is a target for positive regulation by the related factors Sox9, Sox5, and Sox6, acting through an upstream enhancer element (Han and Lefebvre, 2008). The homeobox transcription factor SHOX also regulates Acan expression, acting cooperatively with Sox5 and Sox6 at the same enhancer (Aza-Carmona et al., 2011). Other genes and signaling pathways are proposed to be upstream (Li et al., 2003; Pei et al., 2009; Pfander et al., 2003; Sen et al., 2004; Tim Yoon et al., 2003; Tsai et al., 2003), but no direct evidence has been shown for interactions with specific cis-regulatory sequences. Nonetheless, it is likely that variations in expression levels of the gene are linked to important pathogenic processes. Therefore, we undertook a comprehensive survey of the human ACAN locus to identify additional cis-regulatory elements capable of recapitulating, wholly or in part, the endogenous expression of ACAN in cartilage in vivo. Using moderate evolutionary sequence conservation as a guide to potential enhancer function, we tested 24 conserved sequences, from an interval of 220 kb containing ACAN. We assayed the sequences for their ability to direct reporter gene expression to cartilage in transgenic zebrafish. We uncovered a remarkable degree of overlap in regulatory control at the locus; in all, we identified eleven sequences independently capable of regulating expression in cartilage. Six of these are conserved at the chicken locus, in the same positions relative to the gene, indicating deep conservation of the regulatory machinery. Five of the six chicken enhancers also display conserved function, directing marker gene expression to cartilage in transgenic zebrafish.
Our identification of multiple enhancers associated with ACAN is the first necessary step in the investigation of specific upstream genes and signaling pathways for their effect on gene transcription. The sequences are also candidates for non-coding polymorphisms or mutations associated with changes in expression levels and human disease.
We took an unbiased approach to identify cis-regulatory sequences associated with ACAN, in addition to the proximal upstream enhancer that was previously described (Han and Lefebvre, 2008). We interrogated a 222 kb window of genomic sequence containing the human ACAN gene, extending to the genes on either side, using the PhastCons track on the UCSC genome browser. Through comparison of 28 vertebrate genomes, we identified 24 conserved non-coding sequences, representing the most conserved ~6% of the interval excluding coding sequence, for functional analysis (Table 1). These included sequences encompassing the 5′UTR (−1ACAN) and the 3′UTR (+71ACAN) of ACAN itself. The sequences are named according to their position in kilobases relative to the transcription start site of ACAN, and are broadly distributed across the interval, including seven from introns of the ACAN gene itself (Fig. 1).
We cloned the 24 potential regulatory sequences into a reporter plasmid containing egfp coding sequence in conjunction with the heterologous cFos minimal promoter, based on the Tol2 transposable element (Hu et al., 2011). We injected each of these constructs, together with RNA encoding the Tol2 transposase, into zebrafish embryos at the 1-cell stage. Because of the highly efficient mosaic expression of Tol2 vectors, we could examine the injected embryos directly for EGFP expression in skeletal structures from 1-6 days post fertilization (dpf).
Thirteen of the tested sequences did not regulate specific expression consistent with endogenous Acan expression, and were not further analyzed. For the remaining eleven constructs, the injected fish were raised to adulthood, and EGFP expression in their progeny was examined in detail after germline transmission.
Strikingly, we identified a large number of enhancers directing expression to cartilage. Eight of these regulate broad expression in craniofacial cartilages (Fig. 2; Fig. 3I, J). These vary in strength and exact timing of expression, with some active very early, at 2 dpf, while others are not detected until 3 dpf. Six were expressed robustly (Fig. 2), while two, −45ACAN and −55ACAN, showed significantly lower levels of expression (Fig. 3I, J)
Transgenics for −11ACAN, in addition to broad cartilage expression, showed strong expression in skeletal muscle, inconsistent with endogenous gene expression (Fig. 3A, B). This enhancer is homologous to the known regulatory element from the mouse gene, which does not display similar ectopic expression in transgenic mice (Han and Lefebvre, 2008). The sequence we initially analyzed was larger than the mouse enhancer. Therefore, we cloned the smaller equivalent human sequence for analysis. In transient transgenics with the shorter enhancer, we still observed prominent expression in skeletal muscle (data not shown), indicating that the ectopic expression was not due to the inadvertent cloning of an inappropriate activating sequence outside the native enhancer.
The remaining two enhancers were active only in a subset of cartilage cells. The −64ACAN enhancer was active at 2 dpf in the earliest forming cartilage, but by 3 dpf was active mosaically in only some cells (Fig. 3C-E). The expression was observed in multiple independent lines, so was not a result of mosaic inactivation of a particular transgene insertion. At 6 dpf, the expression was largely localized to medial cartilage, but was also still seen in scattered cells throughout the cartilage. The −107ACAN enhancer was broadly active in cartilage at 2 and 3 dpf. However, by 6 dpf, EGFP was largely confined to cells in the midline ethmoid plate and the medial portions of the ceratobranchial cartilages (Fig. 3G), similar to the expression regulated by −64ACAN. In addition, EGFP was also seen in bone, including the membranous bones opercle and cleithrum (Fig. 3H), and in perichondral cells surrounding the craniofacial cartilages.
Our initial screen was for enhancer activity in cartilage, but many of the identified enhancers had additional activity before the formation of morphologically apparent cartilage. The craniofacial cartilages, which strongly express Acan later in development, are derived from cranial neural crest cells. Interestingly, one enhancer regulated early and broad expression in neural crest, including in the trunk, while several additional ones regulated expression in the branchial arches consistent with neural crest-derived mesenchyme (Fig. 4A-E). The notochord has many similarities with cartilage in gene expression and composition of extracellular matrix, including abundant aggrecan. Consistent with this, we found three enhancers that regulate expression of EGFP in the notochord at 1 dpf (Fig. 4F-H). Two enhancers that regulated robust expression in branchial arches at 1 dpf (−55ACAN and −45ACAN) were weakly active in differentiated cartilage, suggesting that their primary activity is in early cartilage precursors.
By one month, perichondral ossification has replaced the bulk of most cartilage elements in the craniofacial skeleton with bone, surrounding the remaining cartilage core. There has also been substantial joint maturation and formation of articular cartilages. Both types of cartilage can be visualized with a universal cartilage reporter transgenic line −1.4col1a1:egfp (Fig. 5A). A gene trap into the trps1 gene (Talbot et al., 2010), an early patterning gene for joints, expresses EGFP preferentially in joints (Fig. 5B). Several ACAN enhancers were broadly active in cartilage of the juvenile fish, and regulated expression patterns similar to −1.4col1a1:egfp (Fig. 5E, H), while −108ACAN regulated expression more strongly in joints, in a pattern similar to trps1 (Fig. D, G).
The first exon of human ACAN is non-coding, and because of limited conservation is not annotated in many other species. Thus enhancers lying within the first intron in human are seemingly located upstream of the gene in many other species. Conversely, an additional upstream non-coding exon has been annotated for the mouse gene (Walcz et al., 1994), making the homologues of the −11ACAN and −34ACAN enhancers fall within the first intron. Interestingly, the +30ACAN enhancer overlaps with an expressed sequence tag (EST) from a chondrosarcoma line, and the orthologous chicken sequence overlaps with two ESTs isolated from cartilage, suggesting that this conserved enhancer sequence is normally transcribed in cartilage.
The expression of Acan in cartilage is highly conserved among vertebrates, suggesting that the regulatory sequences might also be highly conserved. As expected, all eleven enhancers were highly conserved among placental mammals, but for a subset of these, the conservation extended to other vertebrates, including marsupials and birds. The six enhancers with clear homologues at the chicken locus are arranged in the same order and orientation relative to the coding sequence (Fig. 6A), demonstrating deep conservation of the regulatory architecture at the Acan locus.
To test the significance of the sequence conservation observed at six of the enhancers, we cloned the homologous chicken sequences (Table 2) into our enhancer analysis vector. In transient transgenic zebrafish, five displayed conserved transcriptional activity in cartilage (Fig. 6B-F). The enhancers homologous to +4ACAN and −11ACAN regulated strong and specific expression in craniofacial cartilage in many (~80% and 30%, respectively) of the injected embryos at 3 dpf (not shown), and the expression was maintained at 6 dpf (Fig. 6D, E). While the enhancer homologous to −113ACAN also regulated prominent cartilage expression at 3 dpf in ~30% of injected embryos (Fig. 6B), expression had been largely down-regulated by 6 dpf (not shown). The homologue of +30ACAN also regulated expression in cartilage, visible by 6 dpf (Fig. 6F). However, the expression was not as strong or widespread, and present in a smaller percentage of embryos. The homologue of −107ACAN regulated expression prominently in heart at 3 dpf (~65% of injected embryos), and also less strongly in the cartilage in a subset of embryos (Fig. 6C). By 6 dpf, expression was also evident in perichondral cells associated with cartilage (data not shown), similar to the expression regulated by the homologous human enhancer.
We hypothesized that important regulatory inputs into Acan transcription will be conserved, and therefore took advantage of multi-species alignments to define potential transcription factor binding sites (TFBSs). For the six most highly conserved enhancers, we aligned sequences from human, mouse, opossum, and chicken, and simultaneously searched for predicted TFBSs in aligned regions. One known upstream regulator of ACAN is the transcription factor Sox9. In addition to the previously described Sox9 binding site in the −11ACAN enhancer, we found predicted sites in the −113ACAN and +30ACAN enhancer sequences (Fig. 7A, E). Two of the remaining three sequences contained no conserved predicted binding sites for Sox9 or related factors, but instead showed evidence for regulation by a variety of other transcription factors (Fig. 7B, D). One highly conserved sequence, −34ACAN, has multiple regions of alignment across species, and also regulates broad cartilage expression in transgenic embryos, but we found no predicted TFBSs in aligned regions.
Enhancers are frequently not associated closely with their corresponding genes, and can be located over a megabase away either upstream or downstream. In order to comprehensively survey a large locus for enhancers, an efficient functional assay is critical. A small subset of enhancers are highly conserved, and these are likely to be associated with developmentally important genes with tissue-specific expression (Pennacchio et al., 2006; Visel et al., 2008; Woolfe et al., 2005). Not surprisingly, these highly conserved sequences are often equivalently functional across divergent species. However, for most enhancers, there is no readily detectable sequence conservation extending beyond mammals. Despite this apparent lack of sequence conservation, we and others have previously shown that enhancer function can be deeply conserved, with mammalian enhancers regulating gene appropriate expression in transgenic zebrafish (Fisher et al., 2006a; Rada-Iglesias et al., 2011). Here we have exploited this convenient assay system to identify multiple enhancers associated with the ACAN locus. Importantly, the fact that these enhancers function across species strongly suggests that the important trans factors acting upon the enhancers are conserved, and also likely their binding sites.
Interestingly, we found little correlation between physical and functional characteristics among the tested sequences. For example, proximity to the gene, or to each other, did not correlate in any obvious way with activity. LOD score was also a poor predictor of enhancer activity; none of the positive enhancers had as high a score as the −64ACAN sequence, which had no activity in our assay. −64ACAN and similar sequences may have other important functions subjecting them to negative selection, that would not be apparent in an assay for positive transcriptional regulation. The negative results may also reflect limitations of our experimental system. We based our sequence selection for analysis on the most conserved core for each element, and could have failed to include important functional sequence in the less conserved flanks. A negative result could also reflect a real functional difference between human and zebrafish. The necessary activating signals for an individual enhancer may not all be present in zebrafish, or the human enhancer may simply have diverged sufficiently to no longer be recognized by zebrafish transcription factors. Therefore, a positive result in our assay is likely to represent deeply conserved enhancer function, while a negative result should be interpreted more cautiously.
Previously, one upstream enhancer had been identified for mouse Acan, and shown to bind several likely upstream regulators (Han and Lefebvre, 2008). We found the homologous human sequence to regulate, in addition to the expected cartilage expression, robust ectopic expression in skeletal muscle of transgenic zebrafish, not a normal site of endogenous Acan expression. Given the high homology between the human and mouse sequences, it is likely that this difference is due to the environment (mouse vs. zebrafish) rather than an intrinsic difference in the human sequence. The ectopic activity is consistent with the expression in zebrafish either of an activating trans factor in skeletal muscle that is not present in mouse at equivalent stages, or the absence in zebrafish of a negative regulator normally present in skeletal muscle of mouse. However, despite the ectopic expression, the enhancer also regulates appropriate cartilage expression.
We describe the identification of sequences independently capable of regulating expression in cartilage, consistent with Acan expression. Based on their proximity to the ACAN coding sequence, we presume that they are normally involved in its transcriptional regulation. One of the neighboring genes, ISG20, encodes a ribonuclease strongly induced in lymphatic cells by interferon (Gongora et al., 1997; Gongora et al., 2000). The other flanking gene, HAPLN3, is physically closely associated with ACAN, and it has been speculated that its ancestral gene contributed part of the coding sequence of the ancestral CSPG gene when it arose through domain shuffling (Kawashima et al., 2009). Hapln3 is expressed broadly, including prominently in the cardiovascular system, but is not apparently expressed in cartilage (Ogawa et al., 2004). Therefore, it is unlikely that that the enhancers we identified function in the transcriptional regulation of these flanking genes. A rigorous demonstration of their function in ACAN transcription will depend on mutating them in the context of the intact locus. Such experiments would also directly address the question of how functionally redundant the individual enhancers are.
Apparently redundant transcriptional regulation has been noted previously, in particular for several developmentally important genes in Drosophila (Hobert, 2010). It has been proposed that one important benefit to the organism of such seemingly unnecessary redundancy is that it lends robustness to gene expression in stressful situations (Frankel et al., 2010; Perry et al., 2010). In the case of Acan, mutations in mouse and human demonstrate the importance not just for its expression, but for maintaining its expression level within narrow parameters. Its conserved expression in the cartilage of all extant vertebrates further supports its critical importance. Therefore, we speculate that the large number of independently regulated Acan enhancers has evolved as a mechanism to ensure consistent, high levels of gene expression throughout life.
The Acan gene evolved by domain shuffling, and arose in the vertebrate lineage (Kawashima et al., 2009). Its appearance was likely a key event in the evolution of cellular cartilage in the vertebrate lineage. CSPG (as assayed by Alcian blue staining) is a prominent feature of cartilage in all extant vertebrates, and is presumed to correspond to Acan expression. Therefore, we speculated that the cis-regulatory mechanisms governing Acan might also be evolutionarily ancient. Through comparison of multiple vertebrate sequences, we identified conservation of six enhancers at the chicken Acan locus, in the same order relative to the gene. Several of these were also conserved at the locus in Xenopus tropicalis (data not shown).
In teleost fish with sequenced genomes, including zebrafish, there are two annotated acan orthologues. We directly compared the sequences upstream of the two zebrafish genes with other vertebrates. Using relaxed alignment parameters and an algorithm designed to detect alignments despite small scale rearrangements, we nonetheless only identified one possible homologous enhancer, a sequence aligned with +30ACAN at the acanb locus. It has been proposed that following the whole genome duplication in the teleost lineage, retained pairs of duplicates have evolved to have different expression patterns through differential retention of regulatory sequences (Cresko et al., 2003; Kleinjan et al., 2008; Kluver et al., 2005). However, if the Acan locus in the last common ancestor before the genome duplication had a similarly complex and redundant regulatory apparatus, we might predict that both teleost duplicates would still be expressed in cartilage, given the large number of independent enhancers regulating that expression. In zebrafish, acana on chromosome 7 is expressed prominently in cartilage (Kang et al., 2004), as is acanb on chromosome 25 (our unpubl. obs.), consistent with this prediction.
To define potential upstream regulators of the identified enhancers, we analyzed them with one of many available algorithms used to predict transcription factor binding sites. We imposed the additional restriction that the binding sites be conserved from mammals to avians. There will certainly be functional binding sites, less well conserved, that are not predicted by our analysis. In addition, alternative algorithms can utilize quite different starting data sets and criteria for matches to consensus sequences, frequently resulting in somewhat different sets of predicted binding sites. Overall, our analysis for potential TFBSs points to a number of possible upstream regulators of ACAN transcription in cartilage, and further suggests that enhancers mediating very similar expression patterns are subject to quite different regulation. However, as with any computational prediction of TFBSs, our analysis should serve as a guide for future functional validation.
Some of the enhancers we identified regulate broad and sustained expression in cartilage, although even these did not all regulate identical expression patterns. In addition, some of the sequences displayed strikingly different activity, for example regulating expression in only a small subset of cartilage cells, such as −107ACAN and −64ACAN. We would predict that these differences in expression patterns are mediated by different upstream regulators for each enhancer, and in fact we found quite different patterns of predicted TFBSs among the enhancer sequences. Previous characterization of one enhancer, equivalent to the human −11ACAN sequence, showed that it is bound by Sox9 (Han and Lefebvre, 2008). Consistent with this, there is a predicted binding site for an SRY-related protein, conserved to chicken. We also find similar conserved sites in two other enhancers, −113ACAN and +30ACAN; both of these regulate broad and specific cartilage expression, although at 1 dpf −113ACAN is active in the notochord while +30ACAN is active in the branchial arches. A sequence overlapping with one of these, +30ACAN, was also tested for potential enhancer activity by Han and Lefebvre (2008). They detected no activity; however, the sequence they tested was significantly shorter (320 bp) than the one we analyzed, and did not contain the complete predicted Sox9 binding site.
There is evidence for several cartilage-expressed genes that Sox9 regulates their transcription through binding as a dimer, to an inverted repeat of two binding sites separate by 3 or 4 bases (Bernard et al., 2003; Bridgewater et al., 2003; Coustry et al., 2010; Genzer and Bridgewater, 2007; Jenkins et al., 2005; Sock et al., 2003). Our approach to TFBS prediction, requiring both a high match to the consensus sequence and a high degree of conservation, revealed an inverted repeat sequence meeting these criteria in −11ACAN. We also examined all enhancers using relaxed criteria to define individual Sox9 binding sites and looking manually for inverted repeats with the correct spacing. Through this approach, we identified two additional potential sites for Sox9 regulation, in the −113ACAN and −107ACAN enhancers (data not shown). The first of these was not conserved in chicken, and did not overlap with the Sox9 binding site predicted by our other approach. An important complement to these TFBS predictions will be experimental validation through direct demonstration of binding, especially in the case of Sox9, which may have additional binding requirements for cartilage genes.
The remaining highly conserved enhancers do not contain predicted SRY/Sox binding sites, and presumably these are regulated independent of Sox9. We do find predicted binding sites for other transcription factor families, whose members include potential regulators such as Hox proteins, Gli transcription factors, and Ets-related transcription factors. Importantly, each sequence functions independently to regulate almost identical expression in developing cartilage, despite the fact that they are likely regulated by quite different factors. We would argue that this system of regulatory control, with multiple enhancers having different input but equivalent output, insures a maintained high level of Acan expression in a variety of environmental conditions.
A region encompassing ACAN was analyzed, starting from the upstream gene ISG20 and ending at the downstream gene HAPLN3, corresponding to genome coordinates chr15:86,993,001-87,225,000 on the NCBI 36/hg18 assembly from March, 2006. Conserved non-coding sequences (CNSs) were selected using the PhastCons algorithm (Siepel et al., 2005) as implemented on the UCSC genome browser, incorporating 28 vertebrate genome sequences.
Twenty-four sequences were chosen for analysis, representing ~6% of the non-coding sequence in the interval. Primers for PCR amplification were designed using Primer 3 software, based on each conserved sequence plus 200 base pairs on either side. The genome coordinates of the sequences cloned for analysis are given in Table 1. All CNSs were amplified from genomic DNA with high fidelity DNA polymerase, cloned into the Tol2 transposon based vector pGWcfosEGFP (Fisher et al., 2006a; Fisher et al., 2006b) which was altered with the addition of recombination sites for phiC31 as described (Hu et al., 2011), and analyzed for expression in transient transgenics.
Fish were cared for following standard protocols (Westerfield, 2007), and in accordance with animal care regulations at the University of Pennsylvania. Each construct was injected in at least two separate experiments, and mosaic GFP expression analyzed in a minimum of 200 embryos. Embryos were screened from 1 to 6 dpf using an Olympus MVX10 epifluorescence microscope. For those constructs regulating a consistent expression pattern, embryos were raised to adulthood and their progeny examined for expression after germline transmission. Unless otherwise noted, all constructs were examined in at least three independent transgenic lines, and among these each showed consistent expression.
Homologous sequences in other genomes were defined using the MultiZ alignment on the UCSC browser. In several cases for the chicken enhancer sequences, there was no apparent alignment with the human sequence as the base genome, but alignment was detected when another species (mouse or opossum) was used as the base. The genome coordinates of all sequences used for alignments and TFBS predictions are listed in Table S1.
For the six most conserved enhancers, the sequences from human, mouse, opossum, and chicken were aligned using DiAlign on the Genomatix Software Suite, and analyzed for potential TFBSs in the aligned regions by Matinspector using the program defaults, comparing with the vertebrate subsection from the latest matrix library (Release 8.3, Oct. 2010). We performed additional analyses with the entire upstream intervals (to the closest gene) from the above species, compared separately to the two zebrafish genes, with the glocal alignment program Shuffle-LAGAN implemented through the VISTA web server. For pairwise alignments other than human to mouse, the parameters were relaxed to 50% identity over 50 bases.
The authors thank Paula Roy and Laura Del Collo for excellent fish care. The work was supported by a grant from NHGRI to S. F.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.