Expression Pattern and Gene Characterization ofAsporin

We have discovered a new member of the class I small leucine-rich repeat proteoglycan (SLRP) family which is distinct from the other class I SLRPs since it possesses a unique stretch of aspartate residues at its N terminus. For this reason, we called the molecule asporin. The deduced amino acid sequence is about 50% identical (and 70% similar) to decorin and biglycan. However, asporin does not contain a serine/glycine dipeptide sequence required for the assembly of O-linked glycosaminoglycans and is probably not a proteoglycan. The tissue expression ofasporin partially overlaps with the expression ofdecorin and biglycan. During mouse embryonic development, asporin mRNA expression was detected primarily in the skeleton and other specialized connective tissues; very little asporin message was detected in the major parenchymal organs. The mouse asporin gene structure is similar to that of biglycan and decorin with 8 exons. The asporin gene is localized to human chromosome 9q22–9q21.3 where asporin is part of a SLRP gene cluster that includes extracellular matrix protein 2,osteoadherin, and osteoglycin. Further analysis shows that, with the exception of biglycan, all known SLRP genes reside in three gene clusters.

The small leucine-rich repeat proteoglycans (or SLRPs) 1 are a group of extracellular proteins (ECM) that belong to the leucine-rich repeat (LRR) superfamily of proteins (1,2). The LRR is a protein folding motif composed of 20 -30 amino acids with leucines in conserved positions. LRR-containing proteins are present in a broad spectrum of organisms and possess diverse cellular functions and localization (3). The members of the SLRP subfamily have core proteins of similar size (about 40 kilodaltons) that are dominated by a central domain composed of 6 -10 tandemly repeated LRRs. This domain is flanked by smaller, less conserved N-terminal and C-terminal regions containing cysteines in characteristic positions.
Most of the SLRP proteins are proteoglycans, and the SLRP gene family has been subdivided into 3 classes based on similarities in overall amino acid sequence, spacing of cysteine residues in the N terminus, and gene structure. The previously identified class I members, decorin (4) and biglycan (5), are the most closely related SLRPs based on amino acid sequences; the human sequences are 57% identical. The core proteins contain 10 LRRs, and the N-terminal regions of decorin and biglycan are substituted with one and two chondroitin/dermatan sulfate chains, respectively. The cysteine-rich cluster in the N terminus of class I SLRPs has an amino acid spacing of CX 3 CXCX 6 C. The mouse decorin (6) and biglycan genes (7) contain 8 exons.
The class II members, fibromodulin (8), lumican (9), PRELP (10), keratocan (11), and osteoadherin (12), have a pairwise amino acid sequence identity between 37 and 55% and have a common gene structure composed of three exons. The cysteine spacing in the N-terminal region of class II SLRPs is identical (CX 3 CXCX 9 C) but different from the other SLRP classes. The core proteins of class II SLRPs contain 10 LRRs and (with the exception of PRELP) can be substituted with N-linked keratan sulfate glycosaminoglycan chain(s).
The class III members, epiphycan/PG-Lb (13,14), osteoglycin/mimecan (15,16), and opticin (17), have a pairwise amino acid sequence identity between 35 and 42% and have a common gene structure composed of either 7 or 8 exons. Class III SLRPs contain only 6 LRRs, and the cysteine spacing in the N-terminal region is identical (CX 2 CXCX 6 C) for all members of this class. The recently identified opticin is substituted with Olinked sialylated oligosaccharides, and consequently is a glycoprotein rather than a proteoglycan. On the other hand, osteoglycin/mimecan and epiphycan can be substituted with N-linked keratan sulfate glycosaminoglycan chain(s) and Olinked chondroitin/dermatan sulfate chain(s), respectively. Many of the SLRP proteoglycans can be isolated from tissues without attached glycosaminoglycans, suggesting that they are "part-time" proteoglycans (11,16,18).
Several SLRP proteins display potent effects in vitro. For example, recombinant decorin, biglycan, and fibromodulin bind to transforming growth factor-␤ in vitro (19), and decorin can interfere with transforming growth factor-␤-dependent proliferation of Chinese hamster ovary cells (20). Furthermore, injection of decorin into rats with experimental glomerulonephritis curtailed the abnormal deposition of matrix suggesting that decorin may affect transforming growth factor-␤ activity also in vivo (21,22). Recently, it has been shown that decorin can down-regulate epidermal growth factor receptor leading to growth suppression, and decorin may act as a natural inhibitor of the epidermal growth factor receptor signaling pathway (23).
The SLRPs have been shown to interact with a variety of extracellular matrix proteins, such as collagens (24), fibronectin (25), and thrombospondin (26), as well as serum proteins, heparin cofactor II (27) and C1q (28). Biochemical assays have demonstrated that decorin (29), fibromodulin (30), and lumican (31) bind to collagens in vitro and modulate collagen fibril formation. Morphological analysis of mouse "knockouts" demonstrates that decorin (32), fibromodulin (33), and lumican (34), respectively, are necessary for normal collagen fibril formation in specialized connective tissues of skin, tendon, and cornea. Therefore, a role for SLRPs in collagen fiber formation is clearly established both in vivo and in vitro. Biglycan-null mice exhibit a mild osteoporosis-like phenotype (35). However, it is not known if this phenotype is the consequence of a primary defect in collagen fiber formation. Recently, patients with cornea plana 2 (MIM 217300) were shown to have mutations in the keratocan gene, a class II SLRP family member (36).
Nucleotide sequencing of a human bacterial artificial chromosome (BAC, RPCI11-917O5), and contigs of overlapping BAC clones revealed that four SLRPs genes (decorin, lumican, keratocan, and epiphycan/PG-Lb) are physically linked on human chromosome 12q (36). Previous genetic linkage studies in the mouse suggested that decorin, lumican, and epiphycan map together in a cluster in close proximity to the Mgf gene on mouse chromosome 10, and these genes are deleted in mice that have large deletion mutations at the Steel locus (37).
Chromosomal localization of three other SLRPs, fibromodulin (38), PRELP (39), and opticin (40,41) to human chromosome 1q32 by fluorescent in situ hybridization analysis and/or radiation hybrid mapping raised the possibility that these SLRP genes may also be physically linked. A computer homology search of the genome data bases was, therefore, initiated to look for additional, unidentified SLRP family members that might be associated with these clusters or a yet unidentified cluster.
In this study, we identify a novel SLRP family member that belongs to the class I subfamily and is closely related to biglycan and decorin. We have named this new SLRP asporin due to the unique aspartate stretch at the N terminus of the translated open reading frame. We report the molecular cloning of the full-length mouse and partial human cDNA and investigate asporin mRNA expression in mouse embryonic development. In addition, we have determined the mouse and human asporin gene structure and discovered that the human asporin gene is part of a SLRP gene cluster on human chromosome 9q21.3-9q22 that also contains osteoadherin, osteoglycin/mimecan, and a gene encoding another LRR-containing protein, ECM2 (42).

EXPERIMENTAL PROCEDURES
Materials-Chemicals and supplies were purchased from Sigma, Fisher, and Intermountain Scientific. Total RNA was extracted from confluent mouse ATDC5 cells (43) by using the QIAshredder kit and purified with the RNAeasy mini kit (Qiagen, Santa Clarita, CA). Total human heart RNA was obtained from Ambion (Austin, TX). Reverse transcriptase used was SuperScript II (Life Technologies, Inc., Rockville, MD), and first strand cDNA was synthesized by 5Ј and 3Ј-rapid amplification of cDNA ends (5Ј and 3Ј RACE) using SMART TM (Switch Mechanism At the 5Ј end of RNA Templates) technology (CLONTECH, Palo Alto, CA). The QIAPREP spin miniprep kit and the QIAEX II gel extraction kit (Qiagen) were used to purify DNA. The plasmid vector used in subcloning was pBluescriptKS (Stratagene, La Jolla, CA). Nucleotide sequencing reactions were performed at a University of Texas sequencing core facility. Oligonucleotides were purchased from Sigma-Genosys (Woodlands, TX). Polymerase chain reactions (PCR) were performed with one of the following polymerases: Taq polymerase (Life Technologies, Inc.), Advantage 2 Polymerase Mix (CLONTECH), Pfu Polymerase (Stratagene), or Takara LA Taq TM polymerase (Takara Biomedicals, Japan). PCR products were ligated into a TA-cloning vector (44) or the pGEM TM T-easy vector system (Promega). The mouse poly(A) ϩ multiple tissue Northern blot was from OriGene Technologies, Inc. The mouse BAC library was from Research Genetics (Huntsville, AL). Radioisotopes [␣-32 P]dATP and [␣-32 P]dCTP were purchased from PerkinElmer Life Sciences. Random labeling kits, T7 QuickPrime kit (Amersham Pharmacia Biotech), and DNA Strip-EZ (Ambion, Austin, TX) kit were used. Hybridization fluids used were either Rapid-Hyb (Amersham Pharmacia Biotech) or UltraHyb (Ambion). Imaging film and emulsion were purchased from Kodak (X-Omat AR film and NTB-2 emulsion).
Cloning Full-length Mouse cDNA-The full-length mouse cDNA was obtained by aligning nucleotide sequences of overlapping PCR products. RNA was extracted from mouse ATDC5 cells and first-strand cDNA was synthesized by reverse transcriptase SuperScript II (Life Technologies, Inc.) using the reagents provided in the SMART TM RACE cDNA amplification kit (CLONTECH). The gene-specific primers for mouse asporin were designed from nucleotide sequences contained in expressed sequence tags (ESTs) that were publicly available in GenBank TM data bases. For 5Ј RACE reactions, reverse oligonucleotide primers were designed against mouse EST GenBank TM accession number AI 006670 (MS ASP RV 406, 5Ј-AGGCTTCACTGGCTCTTTCGTAGGAAAAAG; and MS ASP RV 343, 5Ј-CGTCATCATCTGTGTCTTCCATATCCTTC). For 3Ј RACE reactions, forward oligonucleotides were designed against mouse EST GenBank TM accession number AA980962 (MS ASP FW 983, 5Ј-CTTGAAGATCTTAAACGGTACAGGGAACTGC; and MS ASP FW 1077, 5Ј-CCACGTGTGAGAGAGATACACTTGGAACAC). First round PCR conditions for 5Ј and 3Ј RACE were as follows: the template-RACE ready cDNA; gene specific oligonucleotides, MS ASP RV 406 (5Ј RACE) and MS ASP FW 983 (3Ј RACE), 25 cycles (5 s 94°C, 10 s 60°C, 2 min 72°C). First round PCR products were diluted and used as template in a second round "nested" PCR as recommended by the instructions provided with the kit. Nested PCR products for 14 clones harboring 5Ј RACE products were resolved electrophoretically on an ethidium bromide-stained 1% agarose gel. Five of the plasmids containing 5Ј RACE products of different sizes were sequenced in both directions using the T3 and T7 primers. The largest fragment was called p329, and was used in subsequent Northern, Southern, and in situ hybridization experiments. Analysis of this sequence in GenBank TM failed to reveal any homology to known cDNA or genomic sequences. Also, two clones harboring 3Ј RACE products of identical size were sequenced in both directions using the T3 and T7 primers.
Since the complete open reading frame (ORF) for mouse asporin could not be determined by alignment of overlapping mouse ESTs, two primers (Ms Start FW, 5Ј-CGCGGATCCAAACCCTTCTTTAGCCCTT-CCCAC; Ms Stop RV, 5Ј-CGCGGATCCTTATTTTCCAACATTCCCAA-GCTG) were designed to amplify by PCR the mouse asporin ORF (template of mouse RACE-ready cDNA, Pfu polymerase, 20 cycles (20 s 94°C, 30 s 60°C, 2 min 72°C). The amplified mouse asporin ORF was digested with BamHI restriction enzyme and ligated to BamHI-cleaved pBluescript KSϩ. The resulting subcloned ORF plasmid was sequenced with three primers: T3, T7, and MS FW 775 (5Ј-GGACACGTTCAAGG-GAATGAATGC) to determine the open reading frame of mouse asporin.
Human Partial cDNA-Human heart RNA (Ambion) was reverse transcribed to first strand cDNA by using the SMART TM RACE cDNA amplification kit (CLONTECH). A partial human cDNA was obtained that contained the open reading frame and the 5Ј-untranslated region. The gene-specific primers for human asporin were designed from nucleotide sequences contained in human ESTs, AK000136, FLJ20129, and AI539334. PCR conditions were as follows: template-human RACE ready cDNA; gene-specific oligonucleotides, HU ASP RV STOP (5Ј-CCGCTCGAGTTACATTCCAAAGTTCCCAAGCTGAAC) and HU ASP RV 1503 (5Ј-ACTGCAATAGATGCTTGTTTCTCTCAACCC), 30 cycles (5 s 94°C, 10 s 60°C, 2 min 72°C). PCR-amplified products from first round 5Ј RACE reactions were sequenced.
Northern Hybridization-Three consecutive Northern hybridizations were performed on a single mouse multi-tissue poly(A) ϩ RNA blot (Origene). DNA fragments of mouse asporin, biglycan, and decorin cDNAs were random-labeled in separate Northern hybridizations. The asporin probe (p329) is a 478-base pair (bp) PCR-amplified 5Ј RACE product that encodes for the 5Ј end of the mouse asporin cDNA that includes the 5Ј-untranslated region (region of cDNA that is encoded by exon I) and a portion of the open reading frame (a fragment of the cDNA that is encoded by exon 2). The biglycan probe (p368) is a 731-bp PCR-amplified fragment that encodes for a portion of the 3Ј-untranslated region of mouse biglycan (7). PCR parameters are as follows: primers are MS BGN3, 5Ј-CCTGAGACCCTGAACGAACTTCACCTGG, and MS BGN4, 5Ј CGGTGGCAGTGTGCTCTATCCATCTTTCC; template is mouse RACE-ready cDNA as described previously, 30 cycles (20 s 94°C, 20 s 60°C, 1 min 72°C). The decorin probe (p280) is a 399-bp XbaI/HindIII fragment from the 3Ј end of the mouse decorin open reading frame (6).
DNA probes were random-labeled by using the Strip-EZ TM kit (Ambion). Following an overnight hybridization at 42°C, the blot was washed under high stringency (1% SDS, 2 ϫ SSC) at 65°C. The same blot was subjected to 3 separate Northern hybridizations in this order: asporin hybridization, strip blot of probe, decorin hybridization, strip blot of probe, and biglycan hybridization. Radiolabeled probes were removed using Ambion's Strip-EZ technology between consecutive hybridizations. The wash conditions following each hybrization are as follows: asporin, 2 washes of 5 min, film exposure 16 h; decorin, 2 washes of 30 min, film exposure 2 h; biglycan, 3 washes of 10 min, film exposure 7 h. Radioactive Northern blots were exposed to Kodak film (X-Omat AR).
Mouse Gene Structure-A PCR-amplified 5Ј RACE product (p329) described earlier was used as a radiolabeled probe in a Southern hybridization to screen a mouse genomic BAC library (Research Genetics). After an overnight hybridization at 65°C, the blots were washed under high stringency (1% SDS, 2 ϫ SSC) at 65°C (3 ϫ 15 min) and exposed to x-ray film. Two BAC clones corresponding to positive signals seen on the developed film were purchased from Research Genetics.
After annotation of the genomic nucleotide sequence from BAC number AL137848, the exon/intron boundaries of the human asporin gene were determined by aligning homologous regions of this sequence with sequence from available human ESTs. Assuming that the mouse gene structure is similar to the human gene structure, the regions in the mouse cDNA that encoded for exons in the mouse gene were predicted. Forward and reverse primers were designed from regions in the mouse cDNA that were predicted to encode for consecutive exons (i.e. forward primer in exon 1, reverse primer in exon 2). With purified mouse BAC DNA as template, such primer pairs were used in long distance PCR reactions to amplify the introns of the mouse asporin gene. Amplified fragments were separated by electrophoresis on an ethidium bromidestained 0.8% agarose gel to judge intron size and were subcloned using the pGEM TM T-easy vector system. Subcloned fragments were sequenced with the T7 and SP6 primers to determine the sequence of the mouse exon/intron boundaries. The primer pairs used to amplify the introns of the mouse asporin gene are as follows: intron 1: Asp Ex1 Human Gene Structure of Asporin-During annotation of the nucleotide sequence from BAC number AL137848, the exon/intron boundaries of the human asporin gene were established by aligning homologous regions of the genomic sequence with available human ESTs (i.e. AK000136, FLJ20129, and AI539334), and determining the regions in the human cDNA that encoded for exons in the human gene. The ENSEMBL web site on the Sanger Center server confirmed the location of an open reading frame (ENST00000026531) in BAC number AL137848 that we have named asporin.
RNA in Situ Hybridization-In situ hybridizations were performed on sections from different stages of mouse embryos. Sections were hybridized with [ 35 S]UTP-labeled antisense or sense RNA probes generated from the plasmid p329 that contains the extreme 5Ј end of the mouse asporin cDNA (1-478 bp).
Pregnant C57Bl mice were sacrificed on various days post-coitus (dpc), embryos were harvested, rinsed in phosphate-buffered saline/ diethyl pyrocarbonate, and fixed in 10% (v/v) formalin in phosphatebuffered saline for 2-25 h. The fixed tissues were dehydrated through a series of increasing ethanol concentrations and then cleared in xylene before being embedded in paraffin. Sections of 7-m in thickness were mounted onto Superfrost Plus slides (Fisher Scientific, Pittsburgh, PA). Dimineralization was performed by placing the tissue into a solution of 0.1 M Na phosphate, pH 6.5, containing 0.26 M EDTA for 2-3 days at room temperature with several changes in between. The tissue was rinsed in diethyl pyrocarbonate/H 2 O, then dehydrated through a graded series of ethanol concentrations, and embedded and sectioned as described for embryonic tissue.
In situ hybridization was performed essentially as described previously (45). Hybridization was carried out at 50°C for 16 -17 h. Two high stringency washes were performed at 55°C in 50% formamide, 2 ϫ SSC for 20 min each. Autoradiography was carried out using NTB-2 Kodak emulsion. The slides were exposed for 16 h to 7 days at 4°C. Photomicrographs were taken using both bright and dark-field optics.

RESULTS
Human and Mouse Asporin cDNA-A novel member of the class I SLRP gene family has been identified and named asporin. The cDNA sequence of human decorin (4) was used as a query to search the human dbEST data base of GenBank TM using the BLAST-N algorithm. At the time, the nucleotide sequence of several human expressed sequence tags (ESTs AK000136, FLJ20129, and AI539334) exhibited strong homology to the nucleotide sequence of the class I SLRPs, but individual ESTs did not encode a full-length ORF. Furthermore, alignment of overlapping human ESTs did not establish a complete, "in-frame" ORF, so human genomic sequence from BAC AL137848 was used to correct the sequencing errors present in the human ESTs and to fill any gaps that were missing from the alignment of overlapping ESTs. These results revealed an open reading frame of 380 amino acids and were later confirmed experimentally by sequencing PCR products generated from 5Ј RACE reactions that used reverse transcribed human heart RNA (first-strand cDNA) as template. The transcription start site of human asporin, as well as the open reading frame and 5Ј-untranslated region, were determined by nucleotide sequencing of 5Ј RACE products obtained with the SMART cDNA amplification kit. Several 5Ј RACE products were resolved electrophoretically on an ethidium bromidestained 1% agarose gel (data not shown). No attempts were made to clone the 3Ј-untranslated region of human asporin.
An open reading frame of mouse asporin could not be obtained from overlapping mouse ESTs thus leaving a central gap in the computer-derived sequence. Furthermore, the genomic sequence of mouse asporin was not available in the public data bases. Therefore, oligonucleotide primers were designed from nucleotide sequences present in available 5Ј and 3Ј mouse ESTs, and a conventional PCR strategy was used to amplify a PCR fragment that bridged the gap in the open reading frame. The 5Ј and 3Ј cDNA ends of mouse asporin cDNA were determined by sequencing 5Ј and 3Ј RACE PCR products, and the transcriptional start site of the mouse asporin message was determined by sequencing the largest PCR-amplified 5Ј RACE product. This RACE product was subcloned and used as a probe for Northern hybridization, Southern hybridization, and RNA in situ hybridization experiments.
A full-length mouse cDNA of 2357 nucleotides was generated by aligning the nucleotide sequences of overlapping PCR reactions and is shown in Fig. 1. The mouse asporin ORF (including the stop codon) is 1122 bp. The 5Ј-untranslated and 3Ј-untranslated regions of the mouse cDNA are 305 and 930 bp, respectively. A noncanonical polyadenylation signal sequence of AATAA (cDNA 2326 -2331) is present very close to the end of the 3Ј-untranslated region of the mouse cDNA. The translated open reading frame encodes a protein of 373 amino acids that contains a putative signal peptide sequence of 15 amino acids assigned using the Signal P V1.1 program (46). The central domain is composed of an array of 10 LRRs, and each LRR contains 24 amino acids with the central consensus sequence The LRR domain is flanked by smaller cysteine containing N-and C-terminal regions. A cluster of four cysteines (C) in the N-terminal region conform to the amino acid spacing of CX 3 CXCX 6 C that is also found in the other class I SLRP members. The C-terminal region contains two cysteines with 32 intervening amino acids, and this exact spacing is also found in decorin and biglycan.
In contrast to decorin and biglycan, a serine/glycine dipeptide consensus sequence for O-linked glycosaminoglycan substitution is not present in the translated open reading frame of asporin. One putative N-linked oligosaccharide attachment site, located between LRRs 8 and 9, can be found in the asporin ORF. A stretch of 14 amino acids N-terminal to the first cysteine cluster contains 10 aspartic acid residues. A similar stretch of acidic residues is not present in the other class I SLRPs, and hence the new member was named asporin.
Comparison of Human and Mouse Asporin ORFs with Decorin and Biglycan-The translated human asporin ORF was aligned with mouse asporin, mouse biglycan (7), and mouse decorin (6) ORFs using the CLUSTAL program (Identity) contained in the MacIntosh MacVector software package (version 6.0.1) and is shown in Fig. 2. The mouse and human asporin ORFs are 91% identical. The major difference is located in an acidic stretch in the N-terminal region of the translated human ORF (380 aa) that is 7 amino acids longer than the corresponding acidic stretch in the mouse ORF (373 aa), and hence accounts for the size difference between the two ORFs. The human ORF, like the mouse, also lacks a dipeptide serine/ glycine consensus sequence for glycosaminoglycan substitution, but contains the same potential N-linked glycosylation substitution site located between the eighth and ninth LRR.
The three mouse class I SLRPs have remarkably similar amino acid sequences. The translated ORF of mouse asporin (373 aa) is most homologous to that of mouse biglycan (369 aa) with 52% identical and an additional 17% similar residues. The amino acid identity between mouse asporin (373 aa) and mouse decorin (354 aa) is slightly less at 49% with an additional 19% similar residues. The amino acid identity between mouse biglycan (369 aa) and mouse decorin (354 aa) is 54% with an additional 14% similar residues. The region of lowest homology among the three mouse translated ORFs is N-terminal to the first cysteine cluster. The aspartate-rich stretch of asporin is Tissue Distribution of Class I SLRPS-The mRNA expression for the class I SLRPs is broadly distributed in mammalian tissues (see Fig. 3). Three separate Northern hybridizations of a single mouse multitissue Northern blot using radiolabeled cDNA fragments of decorin (Fig. 3, top panel), biglycan (Fig. 3, center panel), and asporin (Fig. 3, bottom panel) were performed. The Northern results for mouse biglycan and mouse decorin confirm previously published work (6,7). The asporin probe recognized a single mRNA of 2.4 kb in the tissues tested. The asporin mRNA is comparable in size to the biglycan message of 2.4 kb and slightly larger than the decorin message of 1.8 kb.
For the 12 organs that were represented in the mouse adult multiple-tissue Northern blot, asporin message was most prominent in the heart. Asporin message was also detected in kidney, stomach, testes, and skin but only weakly in lung, skeletal muscle, small intestine, and thymus. However, asporin message in brain, liver, and spleen was virtually undetectable at the longest exposure tested.
Similarities and differences in the relative RNA expression pattern of the three genes were found. A message for all three genes was detected in heart, kidney, skin, testes, and small intestine, but the message in brain was extremely weak for all the genes. Biglycan message in spleen and lung was fairly robust, yet asporin and decorin message were very weak in these organs. Asporin message was virtually undetectable in the liver, yet expression of biglycan and decorin mRNA was observed in this organ.
Expression of Asporin in Mouse Development-To obtain a more complete picture of asporin expression, particularly dur-ing mammalian embryonic development, we used RNA in situ hybridization analysis of sagittally sectioned mouse embryos at different stages of development. No asporin mRNA was detected at the two earliest time points tested, 9.5 and 10.5 days dpc of mouse embryonic (ME) development (data not shown). Asporin mRNA was detected at 12.5 dpc in the maxilla and mandible (Fig. 4). At ME 12.5 dpc, a groove forms between the lower surface of the anterior tip of the tongue and the mandibular component of the first brachial arch. At this stage, asporin message is absent from the tongue, but is present in the mandibular (shown as an arrow in the 12.5 dpc panel) as well as maxillary components of the first branchial arch. A signal is also detected in the thoracic body wall adjacent to the heart. At ME 13.5 dpc, Meckel's cartilage is recognizable, and asporin mRNA expression is detected in the mesenchyme lateral to Meckel's cartilage, but not in Meckel's cartilage. Pronounced expression of asporin is observed in the perichondrium of the humerus, ribs, and scapula. At ME 14.5 dpc, the mesenchymal condensations lateral to Meckel's cartilage are clearly positive, and the asporin expression pattern appears as a "cusp" surrounding Meckel's cartilage. This "cusp-like" area will eventually ossify and give rise to intramembranous alveolar bone of the mandible. Asporin expression is also found in the perichondrium surrounding the central cartilaginous elements of the vertebrae. Weak asporin expression is detected in dermal mesenchyme.
At ME 15.5 dpc (Fig. 5), sagittal sections reveal a robust expression of asporin in the perichondrium/periosteum of the long bones (i.e femur, tibia, and fibula), some of the flat bones at the base of the skull (i.e sphenoid bone), ribs, clavicle, and vertebrae. The intramembranous bones of the maxilla and mandible (alveolar bone) are also positive for asporin. A strong signal was observed in sagittal sections of the subcutaneous The human asporin ORF of 380 amino acids is 91% identical to the mouse asporin ORF of 373 amino acids. The acidic stretch at the N terminus of human asporin (amino acids 33-53 of ORF) is 7 amino acids longer than the corresponding stretch in mouse asporin, and hence, accounts for the size difference between the two open reading frames. Neither the human nor the mouse asporin ORF contains a potential O-linked glycosaminoglycan substation site. However, both the human (asn number 282) and mouse (asn number 275) contain one potential N-linked glycosylation site. The amino acid identity between mouse asporin (373 aa) and mouse biglycan (369 aa) is 52% with an additional 17% similar residues. The amino acid identity between mouse asporin (373 aa) and mouse decorin (354 aa) is 49% with an additional 19% similar residues. The amino acid identity between mouse biglycan (369 aa) and mouse decorin (354 aa) is 54% with an additional 14% similar residues. Using the Clustal W (1.4) multiple alignment program, the number of identical amino acids shared by the three mouse aligned sequences is 141 amino acids. muscles or panniculus carnosus of the thorax, trunk, and head/ neck (platysma muscle) region and are shown as arrows in Fig.  5. Very little asporin message was detected in the major parenchymal organs (with the exception of the large bronchi of the lung). The strong expression of asporin in the perichondrium is underscored in a sagittal section of the digits of a 15.5 dpc forelimb (Fig. 5, panel C). Comparisons of dark and bright field micrographs of the distal end of the third digit (magnification of ϫ20, panel D) suggests that asporin signal is prominent throughout the perichondrium but not in differentiated cartilage.
Asporin RNA expression is prominent in the developing mouse skeleton, particularly in the perichondrium/periosteum of cartilage/bone, and is also found in other specialized connective tissues such as tendon, sclera, the connective tissue sheath surrounding muscle and dermis. Tendon expression of asporin at 15.5 dpc is shown in Fig. 6 (panel A). Expression of asporin in the sclera of the eye was first detected at 15.5 dpc and stronger expression was detected at 17.5 dpc (see Fig. 6, panel  B). The section of the eye shown in Fig. 6, panel B, is from an albino BALB/c mouse (the sagittal section of the 15.5 dpc embryo shown in Fig. 5 is from a C57Bl/6J mouse). Parasagittal sections of the tongue at ME 18.5 dpc reveals that the connective tissue layer of the lamina propria and the lingual fascia ensheathing the skeletal muscle bundles of the tongue are positive for asporin RNA expression (Fig. 6, panel C). The positive signal for the fascia surrounding the skeletal muscle bundles appears as parallel striations in the center of the section, and the myofibers are negative. Prior to embedding the 18.5 dpc embryos during the in situ hybridization protocol, the skin is peeled away from the embryo to allow for adequate penetration of fixatives. Fortuitously, some skin shavings remained on the slide during the procedure, and positive signal for asporin was observed in the dermis but not in the epidermis (Fig. 6, panel D).
Gene Structure of Asporin-A mouse BAC was screened with  (top panel, decorin; center panel, biglycan;  bottom panel, asporin). The blot was commercially prepared so that it contains about 2 g of poly(A) ϩ RNA per lane, and the tissues were taken from 9 -10-month-old Swiss Webster mice (thymus, 8 -12 weeks old). The RNA was loaded in 12 lanes (left to right) from the following tissues: brain (1), heart (2), kidney (3), liver (4), lung (5), muscle (6), skin (7), small intestine (8), spleen (9), stomach (10), testis (11), and thymus (12). Markers on the left side of the blot (dots representing Ambion RNA Millenium marker) from bottom to top are 0.5, 1, 1.5, 2.0, 2.5, 3.0, 4, 5, 6, and 9 kb. The RNA message size of 1.8 kb for mouse decorin (6) and 2.4 kb for mouse biglycan (7)  Dark field micrographs are shown to the right of the bright field images (magnification ϫ1). Top panels (A), at 12.5 dpc, asporin RNA is detected in the maxillary (Mx) and mandibular (Mn) components (arrow) of the first branchial arch and the thoracic body wall (Bw) adjacent to the heart. Middle panels (B), at 13.5 dpc, asporin is detected in the perichondrium of the scapula (Sp), ribs (Ri), and humerus (Hu). Asporin mRNA is not detected in Meckel's cartilage, but instead the mesenchymal cells lateral to Meckel's cartilage (shown with arrow). Bottom panels (C), at 14.5 dpc, asporin expression is detected in the perichondrium of the vertebrae (Ve). Condensing mesenchymal cells in the mandible surrounding Meckel's cartilage are positive for asporin RNA, and this cusp-like expression pattern is highlighted by the arrow. Strong expression of asporin is maintained in the mandible and maxilla, at future sites of intramembranous bone formation. Weak expression is also detected in the dermal mesenchyme (Dm) at 14.5 dpc.
an asporin probe and two unique BAC clones were obtained. Purified BAC DNA was used as template to PCR amplify the introns of the mouse asporin gene in a long distance PCR strategy. The amplified introns were resolved electrophoretically on an ethidium bromide-stained 0.8% agarose gel (data not shown), and their sizes were estimated (see Table I). The seven introns were subcloned and the exon/intron junctions were sequenced (see Table I). The mouse asporin gene spans about 23 kilobases and contains 8 exons.
The gene structure of human asporin (see Table I) was determined by alignment of annotated nucleotide sequence from genomic BAC clone AL137848 with overlapping human ESTs. The human gene spans at least 25 kb and also contains 8 exons. The size of the first exon was determined by nucleotide sequencing the largest PCR-amplified product from 5Ј RACE reactions. The size of the last exon was determined by comparing all the overlapping asporin 3Ј ESTs available on the public data bases and choosing the ones that were the longest in the 3Ј direction.
The gene size and structure of asporin in both species examined are very similar. The largest intron for both genes is the first, whereas the smallest intron is the sixth. The codon phas-ing at the exon/intron boundaries of the mouse and human genes is identical, and the intron sizes are similar (Table I).
Asporin Is Part of a SLRP Gene Cluster on Human Chromosome 9 -The nucleotide sequence of three overlapping human BAC clones (AL157827, AL137848, and AL354924) localized to chromosome 9q22-9q21.3 was subsequently annotated. A diagram depicting a 188-kb region of three overlapping BAC clones is shown in Fig. 7 and reveals a cluster of 4 genes that code for LRR-containing proteins: ECM2, asporin, osteoadherin, and osteoglycin. It was necessary to determine the location and orientation of a 16-kb contig located in the center of BAC AL137848, between asporin and osteoglycin, by performing bridging PCR reactions to neighboring contigs. Three gaps within intronic sequences could not be annotated and are shown as vertical dotted lines in Fig. 7. Within the 188-kb region, the 4 genes are arranged in a head-to-tail fashion with the same transcriptional orientation. The three SLRP genes are physically linked and include one member from each SLRP class: asporin (class I)-osteoadherin (class II)-osteoglycin/ mimecan (class III). The 5Ј-and 3Ј-untranslated region of each gene was estimated by comparison with nucleotide sequences contained in human ESTs that exhibited homology specifically to the extreme 5Ј and 3Ј ends of the genes. The four genes are also physically linked in the mouse.  (Panel A), and a serial section that was hybridized with a "control" sense probe is shown at the top right (Panel B). Asporin RNA is detected in the perichondrium/periosteum of the long bones such as the tibia (Ti), fibula (Fi), femur (Fe), iliac bone (Il), the flat bones at the base of the skull such as the sphenoid bone (Sh), ribs (Ri), clavicle (Cl), and vertebrae (Ve). Some of the intramembranous bones of the maxilla (Mx) and mandible (Mn) are also positive for asporin. A positive signal for asporin is detected in the region of the subcutaneous muscles of the thorax, trunk, and head (platysmal muscle), and these muscles are delineated with arrows. Very little asporin mRNA is detected in the major parenchymal organs, with the exception of the lung bronchi (arrow). A nonspecific signal is evident in the major parenchymal organs of heart (He), lung (Lu), and liver (Li) perhaps due to the nonspecific binding of the probe to the erythrocytes. A sagittal section of the digits from a forelimb at 15.5 dpc is shown in panel C (bright field to left; dark field to right; magnification of ϫ4). The tip of the third digit from panel C is shown in panel D (bright field to left, dark field to right; magnification of ϫ20). Asporin has a prominent expression in the fibroblast (Fb) layer of the perichondrium.

DISCUSSION
We have discovered a new member of the class I subfamily of SLRPs that we have named asporin. The size and amino acid sequence of the asporin protein are remarkably similar to those of the core proteins of the other members of the class I subfamily, decorin and biglycan. Almost 70% of the residues in these proteins are identical or conserved. Furthermore, they all contain 10 highly conserved LRRs in the central region, and the number and amino acid spacing of the cysteine residues in the N-and C-terminal domains are conserved. The region of the class I proteins that is least similar lies N-terminal to the first cysteine cluster. For the proteoglycans decorin and biglycan, the serine/glycine dipeptide sequence(s) required for xylosyl transfer and glycosoaminoglycan assembly are located in this region. Asporin does not contain this dipeptide, thus asporin is probably not a proteoglycan. Instead, asporin contains a stretch of aspartic acid residues in this region. This acidic motif in human asporin is composed of 18 residues, and in mouse asporin the acidic motif is 7 residues shorter. Two other identified SLRPs, osteoadherin and epiphycan have acidic regions. In epiphycan, this stretch is composed of glutamic acid residues and interestingly the acidic motif in human and bovine sequences (14,47) is longer than the corresponding motif in the mouse sequence (48). The C-terminal region of osteoadherin is rich in both aspartic and glutamic acid residues (49). The importance of these acidic motifs is unclear.
A dendrogram of the SLRP gene family is shown in Fig. 8.
With the introduction of asporin, 11 members are contained in the three SLRP gene family classes. Class I members include asporin, biglycan, and decorin; class II includes osteoadherin, lumican, fibromodulin, PRELP, and keratocan; class III includes osteoglycin/mimecan, opticin, and epiphycan/PG-Lb. Although chondroadherin (50,51) and the recently identified nyctalopin (52,53) have been granted membership to the SLRP gene family, they may have diverged from the other three SLRP classes early in evolution because their structures are significantly different from the conventional SLRPs. Human ECM2 (42) also has a LRR domain that is 34% identical to the corresponding domain in human decorin, but ECM2 is much larger and structurally different from the conventional SLRPs. However, ECM2 is physically linked to asporin on human chromosome 9 so it has been included in the dendrogram. Northern blot analysis suggests that the members of the class I SLRP subfamily appear to have a relatively broad tissue distribution in the adult mouse, but the distribution for asporin seems to be the most restricted. An earlier study of the RNA expression pattern of biglycan and decorin during human fetal development showed that biglycan and decorin expression patterns were "substantially divergent and sometimes mutually exclusive" (54). In mouse embryonic development at 14.5 dpc, biglycan is expressed in the perichondrium of the vertebrae, ribs, and large bones of the hind limbs (55), but decorin is not expressed in cartilage and bone (6,55). At this stage of mouse embryonic development, the RNA expression of asporin in the  7. Annotation of the nucleic acid sequences of three BAC clones to give the genomic organization of ECM2, asporin, osteoadherin, and osteoglycin on human chromosome 9q21.3-9q22. Wherever possible, comparison was made between overlapping clones. However, it was necessary to determine the location and direction of 16 kb of one contig in the center of AL137848, between asporin and osteoglycin, by performing bridging PCR reactions to neighboring contigs. Three gaps within intronic sequences could not be annotated.
skeleton is similar to that of biglycan and is specifically localized to the perichondrium. Asporin mRNA was observed in the periosteum of the long bones at ME 18.5 dpc (data not shown) and biglycan is clearly expressed in the periosteum at 2 days of postnatal development (55). The strong RNA expression of asporin observed in the fascia surrounding the muscle bundles of the tongue, and presumably the fascia surrounding subcutaneous muscles as well, coincides with a similar connective tissue expression pattern observed for mouse and human decorin. During human fetal development, decorin was localized to the connective tissue sheathes surrounding skeletal myofibers or fascia, whereas biglycan was localized within the connective tissue sheathes (endomysium) of the skeletal myofibers (54). Likewise, in the mouse, decorin was localized to the fascia (perimysium) of the subcutaneous muscle (32). We now report that asporin RNA expression, during mouse embryogenesis, partially overlaps with biglycan expression in the skeleton and with decorin expression in the fascia of skeletal muscle.
The close structural similarity and overlapping tissue distribution suggest that the SLRPs could represent a family of molecules with redundant functions. This hypothesis is supported by the observation that despite the potent in vitro effects of individual SLRPs, analyses of mice with inactivated or deleted genes reveal suprisingly mild phenotypes. The RNA expression pattern of decorin and biglycan appears to be divergent and in some cases completely nonoverlapping. Thus, it appears unlikely that these two molecules can have completely redundant functions. Our identification of a novel class I subfamily member may impact on our understanding of redundancy among class I members. Asporin has a partially overlapping RNA expression pattern with decorin and biglycan in mouse embryonic development, and consequently asporin must be considered as a candidate for functional redundancy with the class I SLRPs.
A certain degree of compensation has already been observed in targeted mutations of the SLRP class II genes in mice. Recently, morphological analysis of early tendon development in mice for the double "knockout" of lumican and fibromodulin revealed an additive phenotypic effect for the double mutant as compared with the single mutants (56). An increased deposition of lumican protein was observed in whole protein extracts of tails from fibromodulin-null mice suggesting that lumican and fibromodulin may share a binding site on the collagen fibril (33) which was subsequently demonstrated (57).
Some functional redundancy may be present between SLRPs of different subfamily classes. Abnormal collagen fibril formation was observed in the targeted mutations of both class I SLRPs and class II SLRPs, and in some cases, similar phenotypes were seen in the same tissue. For example, the decorinnull (32) and fibromodulin-null mice (33) exhibit a collagen fibril defect in tendon, and abnormal collagen fibril formation in the skin of decorin-null (32) and lumican-null mice (34) leads to skin fragility. In this study, asporin RNA was detected in the dermal mesenchyme, and this expression pattern has also been observed for certain members of the class I, class II, and class III (osteoglycin) 2 SLRPs. It is unclear whether it is possible for SLRP proteins of different classes to have interchangeable functions in certain tissues. The importance and extent of compensation may become more apparent as more SLRP double (and multiple) knockouts are generated and subsequently analyzed.
In this report, we describe a novel SLRP cluster in mammals and a schematic depiction of three SLRP clusters is shown in Fig. 9. In this figure, the horizontal distance between genes in a cluster is not to scale. SLRP class I gene members always lie 5Ј to class II members in a cluster. Likewise, class III members always lie 3Ј to class II members in a cluster. The transcrip- FIG. 8. A dendrogram showing predicted relationships between SLRP family members and other LRR proteins of the ECM. Horizontal distances of bars are proportional to evolutionary distance and are based on human protein sequences. The SLRP family is subdivided into 3 classes: class I contains 3 members; class II contains 5 members; and class III contains 3 members. Asporin is a class I member, and biglycan and decorin are more related. Although chondroadherin (50,51) and the recently identified nyctalopin (52,53) have been granted membership to the SLRP gene family, they may have diverged from the other three SLRP classes early in evolution because their structures are significantly different from the conventional SLRPs. Likewise, ECM2 is structurally different from the conventional SLRPs, but has a LRR domain that shows some homology with the SLRPs. Since ECM2 is physically linked to asporin on human chromosome 9, it has been included in the dendrogram. This analysis was done with public software using ClustalW version 1.81 and the output was generated with TreeViewer. FIG. 9. A diagram depicting the chromosomal organization of the SLRP genes. Three clusters of SLRP genes are represented as horizontal lines with their respective human chromosomal localization shown to the extreme left of the cluster. The genes are depicted as boxes, and the class designation for each SLRP gene (refer to dendrogram, Fig.  8) is shown as a number inside the box. Paralogous genes among the clusters are aligned vertically, and the horizontal distance between genes within a cluster are not to scale. If the transcriptional orientation of the genes in the cluster is known, it is shown as as arrow above the box. Upon alignment of the paralogous genes, one can speculate that biglycan may have been part of the cluster that resides on chromosome 1 at an early point in evolution and later migrated to the X-chromosome. If one compares the genetic distance of SLRP class members predicted by the dendrogram (see Fig. 8) with the paralogous genes arranged in the three clusters, it appears that the clusters on chromosomes 1 and 12 are more related and may have arisen from a duplication event of a "primordial" cluster. tional orientation of the genes located on chromosome 12 follows published reports (36). The relationship among human fibromodulin, PRELP, and opticin on human chromosome 1q32 was derived from the annotation of sequences present in Gen-Bank TM (NT_004523). Biglycan resides on human chromosome Xq28 (58).
The class members can be aligned vertically among paralogous genes in the three clusters (see Fig. 9). Upon examination of additional overlapping BACs from human chromosome 1q32, we have failed to discover a SLRP class I member upstream or downstream of the fibromodulin gene. Additionally, the biglycan gene on the X-chromosome does not appear to be physically linked to other genes that encode for leucine-rich repeating proteins. With the identification of asporin as a class I member on human chromosome 9, we propose that biglycan may have been previously linked with the SLRP cluster on chromosome 1, but early in evolution the gene migrated from this cluster and came to reside on the X-chromosome. If the three clusters are aligned based upon the evolutionary distances depicted in the dendrogram tree (see Fig. 8), it appears that the paralogous genes on chromosomes 1 and 12 are the most similar. Therefore, the SLRP genes clustered on human chromosome 9 may have arisen independently from the clusters on chromosome 1 and 12. Perhaps, the clusters on chromosome 1 and 12 arose from a second duplication of a common cluster.
The significance of the SLRP gene clusters is unclear. Since several of the SLRP genes have been "retained" in the clusters during evolution, it is tempting to speculate that a degree of functional redundancy has also been retained. We speculate that asporin, like the other class I molecules, plays a structural and/or signaling function in the extracellular matrix of skeletal tissues and other specialized connective tissues.