A novel gene, tendin, is strongly expressed in tendons and ligaments and shows high homology with chondromodulin-I

Abstract

Abstract

Chondromodulin-I (CHM1) was identified recently as an angiogenesis inhibitor in cartilage. It is highly expressed in the avascular zones of cartilage but is absent in the late hypertrophic region, which is invaded by blood vessels during enchondral ossification. Blast searches with the C-terminal part of CHM1 in available databases led to the identification of human and mouse cDNAs encoding a new protein, Tendin, that shares high homology with CHM1. Based on computer predictions, Tendin is a type II transmembrane protein containing a putative proteinase cleavage and two glycosylation sites. Northern assays with mouse RNAs demonstrated strong expression of a 1.5-kb tendin transcript in the diaphragm, skeletal muscle, and the eye and low levels of expression in all other tissues investigated. In 17.5-day-old mouse embryos, in situ hybridization revealed high levels of tendin transcript in tendons and ligaments. Additional signals were detected in brain and spinal cord, liver, lung, bowels, thymus, and eye. Cartilage, where CHM1 is found, revealed low levels of tendin m-RNA. In adult mice, tendin is expressed in neurons of all brain regions and the spinal cord. The tendin gene is localized in the human Xq22 region, to which several human diseases have been mapped. © 2001 Wiley-Liss, Inc.

INTRODUCTION

Chondromodulin-I (CHM1) is a type II transmembrane glycoprotein whose extracellular part can be cleaved and is then called mature CHM1. It has been isolated from bovine epiphyseal and nasal cartilage (Hiraki et al., 1991, 1997a; Neame et al., 1990). Recombinant mature CHM1 protein induced proliferation, proteoglycan synthesis, and colony formation of chondrocytes in vitro (Hiraki et al., 1991, 1997a; Inoue et al., 1997). The expression pattern of CHM1, in vivo and in vitro experiments suggest that it can exert an inhibitory role during angiogenesis (Hiraki et al., 1997b; Hiraki and Shukunami, 2000; Shukunami et al., 1999).

Most of the skeletal elements form by means of an enchondral process. During enchondral ossification, CHM1 is already expressed in committed mesenchymal cells, which differentiate into completely avascular cartilage moulds of the developing bones (Aszódi et al., 2000). Later, chondrocytes undergo a terminal differentiation sequence characterized by cell proliferation, hypertrophy, and cell death. Mineralization and vascularization of the hypertrophic matrix, cartilage resorption, and deposition of bone matrix accompany these cellular events (Aszódi et al., 2000). One of the most critical steps during enchondral ossification is the switch from the avascular to the vascular cartilage phenotype that is regulated by a delicate balance of angiogenic and anti-angiogenic cues (Harper and Klagsbrun, 1999). The mouse CHM1 gene is highly expressed in avascular cartilage regions, in resting and proliferating chondrocytes of the growth plate of long bones, epiphyseal cartilage, and the cartilage of the skull. Its expression is abolished in the area of late hypertrophic chondrocytes where blood vessel invasion takes place (Shukunami et al., 1999). A direct involvement of CHM1 in angiogenesis was shown both in vitro and in vivo. First, recombinant CHM1 inhibited proliferation and tube morphogenesis of endothelial cells in cell culture experiments (Hiraki et al., 1997b). Second, injection of CHM1 into growing mouse osteoblastomas inhibited angiogenesis and resulted in profound inhibition of tumor growth (Hayami et al., 1999).

CHM1 is almost exclusively expressed in cartilaginous tissues. However, inhibitors are also important for the maintenance of other vessel-free or poorly vascularized tissues like tendons, vitreous body, or the retina. To identify new molecules with anti-angiogenic properties, we performed a database search with the mature CHM1 protein. Homologous sequences were found on the human PAC dJ479J7 from Xq22 and on several human and mouse ESTs. We have sequenced the mouse and human cDNAs and investigated the mRNAs expression pattern by Northern and in situ hybridization. The new gene was named tendin, reflecting its high expression in tendons and ligaments.

RESULTS

Cloning of the Mouse and Human tendin cDNAs

A blast search in the mouse EST database with the mature CHM1 protein yielded two overlapping ESTs (IMAGE Clone 463876 and IMAGE Clone 2811628) for tendin. The corresponding cDNA clones revealed a sequence consisting of 202-bp 5′-untranslated sequence, 951bp open reading frame (ORF), and 184-bp 3′-untranslated sequence (Fig. 1A). 5′RACE experiments from a whole mouse embryo marathon cDNA library (Clontech) did not result in clones extending the sequenced cDNA (data not shown). Alternative internal splicing was investigated by reverse transcriptase-polymerase chain reaction (RT-PCR) experiments amplifying the complete coding region of the cDNA from several mouse tissues such as heart, liver, brain, cartilage, skin, muscle, and kidney. All experiments resulted in a single band with no evidence for alternative splicing (data not shown). The homologous human cDNA with a 951-bp open reading frame (ORF) was identified by sequencing the IMAGE Clone 1657819 (Fig. 1B). Mouse and human cDNAs have a homology of 89%. Because this new gene is highly expressed in tendons and ligaments (see below), we named it tendin.

Figure 1. Nucleotide and deduced amino acid sequences of mouse (A) and human (B) tendin. Exon-intron boundaries in the human sequence are marked by arrows. Predicted transmembrane regions are underlined, the homologous cysteine-rich extracellular part is boxed.

The predicted Tendin protein, deduced from the cDNA sequence, consists of 317 amino acids. Protein prediction programs (Tmpred, TMAP, Phd-TM, DAS) proposed for Tendin a structure similar to CHM1, with a type II transmembrane topology, a transmembrane region from amino acid (aa) 31 to aa 49 and no signal peptide (Fig. 2A). An alignment of Tendin and CHM-I reveals high overall homology to CHM1 (similarity 54%, identity 31%), especially in the C-terminal cysteine-rich region (77%/66%) (Fig. 2B). The C-terminal extracellular part contains 10 cysteines, which are also present in CHM-I. Eight of these make the cysteine pattern of the cleaved mature CHM1. For the mature CHM1, disulfide bonds have been experimentally shown between the fourth and seventh and fifth and sixth cysteine. The eighth cysteine has a disulfide bond with either cystein1 or 2. The remaining two cysteines, first or second and third establish a fourth disulfide bond (Neame et al., 1990). In Tendin, the sixth cysteine at position 292 of the C-terminal domain is shifted by 4 aa N-terminally compared with CHM1. Two N-glycosylation sites and one O-glycosylation site have been described in the mature bovine CHM1 (Neame et al., 1990). One N-glycosylation site present at position 243 is conserved in human and mouse CHM1. The other sites are present neither in human nor mouse protein sequences. All three possible glycosylation sites are absent in Tendin. Instead, the PROSITE program (Bairoch et al., 1997) predicts two N-glycosylation sites for position 94 and 180. Although the endoprotease cleavage site RERR present in CHM1 is not conserved in Tendin, a potential protease cleavage motif RXXR could be found at position 233-236 that might allow cleavage by endoproteases (Barr, 1991). Sequences homologous to CHM1 or Tendin have only been found in vertebrates and are absent in Drosophila and C. elegans genomes.

Figure 2. Genomic organization of human tendin and alignment of the predicted amino acid sequences of chondromodulin-I (CHM1) and Tendin. A: The gene has seven exons depicted as black boxes and extends over 15-kb genomic sequence. Cytoplasmic part and transmembrane domain are encoded by the first two exons, the membrane bound extracellular part is encoded by exon 3-7 and the cleaved extracellular domain (C-terminal cysteine-rich) is encoded only by exon 7. B: Alignment of amino acid sequences. h and m denote the human and mouse sequences, respectively. Putative transmembrane regions are boxed; the highly conserved C-terminal cysteine-rich extracellular part is underlined. Glycosylation sites are marked below the alignment with N. Conserved cysteines are marked under the alignment by C. Putative protease cleavage sites are indicated by scissors.

Genomic Structure of the Human tendin

The PAC dJ479J7 that contains the human tendin sequences was mapped by FISH to Xq22 (Maeda et al., 1998). For the human CHM1 and tendin genes, genomic sequences are available from the Human Genome Project. Based on these sequences, the exon intron boundaries and the exon and intron sizes were analyzed by using the NIX platform at the Human Genome Mapping Project Resource Centre (URL: http://www.hgmp.mrc.ac.uk/Registered/Menu/). The CHM1 gene (unfinished sequence RP11-16F6, acc. AC012239; RP11-93H24, acc. AL139089; RP11-431O22, acc. AL139085) covers more than 43 kb, whereas tendin spans approximately 15 kb [Fig. 2A]). Both genes consist of seven exons, which all follow the AG-GT splicing consensus. The cytoplasmic N-termini of both genes are encoded by exon 1. Exon 2 encodes the transmembrane region. The extracellular part of the protein is encoded by exons 3-7. Exon 7 contains the complete C-terminal domain with eight conserved cysteine residues. With the exception of intron 5, which is in phase1 in the CHM1 as well as in the tendin gene, all introns are in phase 0. The exon sizes are well conserved, whereas the intron sizes differ between the two genes (Table 1).

Table 1. Genomic Organization of CHM1 and Tendina

	Splice acceptor	Splice donor	Exon size (bp)	Intron size (bp)	Intron phase
a CHM1, chondromodulin-I.
CHM1 exon
1	CCCCGgtgagtaccgcc	72	505	0
Pro
2	cccggcccgcagGCGTA	GTCACgtaagtccagag	141	5698	0
Ala	His
3	tccactttcgagATTTA	AGAATgtaagtatattt	141	9113	0
Ile	Asn
4	tttgtttcatagGGCAT	AACTGgtgggtaccaac	114	11140	0
Gly	leu
5	ttttactttcagGAAGG	AAAAGgtaacattttaa	154	>14 kb	1
Glu	LysG
6	aaatttttgcagAAATC	ATCATgtgcgttccggt	167	>3.8 kb	0
luIle	His
7	tttatttcacagCAGCA
Gln
Tendin exon
1	TAAATgtaagttgattc	163	146	0
Asn
2	ttctttggttagGCAGA	AAAAAgtaagtaaatac	132	8532	0
Ala	Lys
3	ctgttctcccagGCCTA	AAAACgtaagttggatg	140	225	0
Ala	Asn
4	tctgttttatagGGATA	ATGAGgtatgtaagaag	102	3140	0
Gly	Glu
5	ttcttctttcagAATGA	ATCAGgtatgacattct	154	1357	1
Asn	SerV
6	ttattatcttagTTTCT	ACTATgtgagttatgtt	167	326	0
alSer	Tyr
7	ttcttctttcagACTGA	210
Thr

Expression Analysis

Analysis of the available sequences in the EST database revealed 11 ESTs from nine cDNA clones for the human form of tendin. Among the nine cDNA clones, four are from heart cDNA libraries; four from pooled libraries derived from heart, melanocyte, and lung tissue; and one clone is from a metastasizing colon tumor library.

Despite the significantly lower number of EST sequences in the mouse database, there are 33 sequences for tendin present, which were derived from several tissues, including 1 from bone, 4 from mammary cancer, 1 from skin, 1 from medulla oblongata, 1 from corpora quadrigemina, and 25 from whole embryos from stages E12.5 to neonate.

Northern blot analysis of newborn mouse tissues showed high expression of tendin in skeletal muscle, diaphragm, and eye. Weak signals were visible in almost all other tissues investigated, including brain, liver, lung, kidney, heart, skin, thymus, and perichondrium- and periosteum-free rib cartilage (Fig. 3). After probing the same blot with a CHM1 cDNA probe, signals were visible in cartilage, thymus, eye, and skeletal muscle. The latter signal was most probably due to cartilage contamination of the skeletal muscle RNA preparation (data not shown). To determine the temporal expression profile of tendin during development, we analyzed total RNAs isolated from embryos by Northern hybridization at various time points. tendin expression was already detected at E9.5, the earliest embryonic stage that has been analyzed to date (data not shown).

Figure 3. Northern blot analysis of mouse total RNA from various newborn tissues. The same filter was sequentially hybridized with cDNA probes specific for tendin, chondromodulin-I (CHM1), and glyceraldehyde phosphodehydrogenase (GAPDH).

To localize the expression at the cellular level, in situ hybridization of mouse embryos and adult tissues was performed with sense and antisense tendin riboprobes. At E17.5, strong expression was detected in the tendons and ligaments of the skeletomuscular system, including the knee joint (Fig. 4A), the upper limb (Fig. 4B), and the intercostal ligaments (Fig. 4C). In the knee, tendons inserting into the patella, tendons of muscles, and all ligaments, including the cruciate ligaments, gave very high signals. A weaker signal was also detected in the developing medial meniscus (Fig. 4A). In the forelimb, high expression was seen in all tendons investigated, including the tendons of the flexor muscles (Fig. 4B). A high level of tendin expression was observed in the tendinous part of the diaphragm (Fig. 4D). Low-level expression of tendin was detected in resting and proliferative chondrocytes of long bones (Fig. 4E) and vertebral bodies (Fig. 4E), and in the cartilaginous part of the intervertebral disks (Fig. 4F). In hypertrophic chondrocytes, no tendin signals were observed (Fig. 4E).

Figure 4. In situ hybridization of tendin antisense riboprobe in the appendicular and axial skeleton of embryonic day (E) 17.5 mouse embryo. A: Hindlimb: strong signals in the tendons (small arrows) and in cruciate ligament (large arrow) and weak signals in the medial meniscus (asterisk). pa, patella; fe, femur; ti, tibia; fi, fibula. B: Forelimb: tendons of the musculi flexor digitorum (arrows) inserting to the ulna (ul) give hybridization signals. C: Rib: the ligament intercostale (arrow) shows positive signals. Ri, rib cartilage. D: The diaphragm shows strong signals in the tendinous parts (arrow). E: Metatarsal bone: weak tendin signals in the chondrocytes of epiphyseal cartilage (ec) and the proliferative zone of growth plate (p). Note the lack of expression in the hypertrophic region (h). F: Vertebral column: low level of tendin expression can be seen in the chondrocytes of vertebral bodies (vb) and the inner, cartilaginous intervertebral disks (ivd).

Outside the skeletomuscular system signals were further detected in cortex and medulla of the thymus (Fig. 5A). In the liver, only single cells throughout the parenchyme express tendin (Fig. 5B). Signals in the lung were weak and mainly detected in the bronchial regions (Fig. 5C). In heart tissues, cells in the aortal and pulmonary outflow tract, as well as in the aortic and pulmonary valves showed weak positive signals (Fig. 5D). Epithelial cells at the basis of crypts and in villi in duodenum showed also evidence of tendin expression (Fig. 5E). In the eye, signal was visible in the tendons of the ocular muscles and within single cells in the retina (data not shown). tendin expression was also detected in cells of the olfactory epithelium (data not shown) and in the spleen (Fig. 5F).

Figure 5. In situ hybridization of tendin antisense riboprobe in extraskeletal tissues of embryonic day (E) 17.5 embryos. In extracellular tissues, signals were detected in the cortex and medulla of the thymus (A), cells in the liver (B), in cells (arrow) surrounding bronchioli in the lung (C), and in the aortal and pulmonary outflow tract of the heart aorta (ao), aortal valve (av), pulmonary artery (pa), right atrium (ra) (D). E: In duodenum, cells at the basis of the villi stain for tendin expression. F: In spleen, tendin expression can also be shown.

In brain, tendin expression was detected in neuronal cells of all brain regions at E17.5. Highest levels of expression were found in the developing cerebellum, the neocortex, the olfactory bulb, and the spinal cord (data not shown). tendin expression was studied in more detail on adult mouse brain where it was found in all regions analyzed. Expression was observed in all hippocampal fields and in a dentate gyrus (Fig. 6A). Neuronal cells in the corpus striatum and thalamus also expressed tendin (Fig. 6B). In the cerebellum, Purkinje cells and neurons in the cerebellar nuclei expressed high levels of tendin mRNA (Fig. 6C). In the cervical, thoracic, and sacral regions of the spinal cord, neuronal cells in the anterior and posterior horns were positive for tendin (Fig. 6D). In the cortex, neuronal cells of all layers were positive for tendin mRNA (Fig. 6E). Expression in glial cells was not detected by in situ hybridization. However, very low level of expression cannot be excluded.

Figure 6. In situ hybridization of tendin antisense riboprobe in adult mouse brain. A: High expression of tendin was detected in the dentate gyrus (DG,) and the CA regions of the hippocampus. B: Neurons in the cerebral nuclei stain positive for tendin. pu, putamen; th, thalamus; V3, third ventricle. C: In the cerebellum, Purkinje cells (arrow) and neuronal cells in the cerebellar nucleus (asterisk) show strong signals. D: Neurons in the anterior and the posterior horn of the spinal cord show high tendin expression. E: In the cerebral cortex, neurons from all cell layers (1-6b) show clear signals of tendin expression.

DISCUSSION

In this study, we describe the identification and analysis of a new cDNA, tendin, with high homology to chondromodulin-I (CHM1), which modulates chondrocyte differentiation and angiogenesis (Hiraki et al., 1991, 1997a,b; Hiraki and Shukunami, 2000). We have sequenced the cDNAs for the human tendin and the mouse tendin homologues. Computer-based analysis of the deduced protein sequence predicted a type II transmembrane protein with a cleaved extracellular C-terminal cysteine-rich domain. Mouse tendin is widely expressed, but Northern blot and in situ hybridization analyses revealed the strongest signals in tendons and ligaments.

CHM1 is cell membrane anchored by its transmembrane domain. Endoproteolytic cleavage at position 218 releases the C-terminal part into the extracellular matrix (Hiraki and Shukunami, 2000). Whether this cleavage occurs in the cell or after the incorporation into the cell membrane is so far unclear. In Tendin, no cleavage site corresponding to position 218 in CHM1 was found. Protein sequencing of CHM1 revealed an even shorter form which is cleaved at position 252 and carries no glycosylation sites (Neame et al., 1990). Proteolytic cleavage at the predicted site for Tendin (position 236) would release a nonglycosylated cysteine-rich fragment (mature Tendin) similar to the shorter mature CHM1.

In situ hybridization analysis revealed the dominant expression of mouse tendin in tendons, ligaments, and the adult central nervous system. However, a low level of tendin expression was also observed in almost every mouse tissue, including cartilage, where CHM1 is expressed. Interestingly, although mouse and human tendin have a very high homology, they seem to have different expression patterns. Although the mouse tendin mRNA shows only low level expression in heart, the human homologue has been found to date almost exclusively in ESTs from heart cDNA libraries. Therefore, it is possible that there is one further member of this family in the mouse strongly expressed in the heart and vice versa one in humans, strongly expressed in tendons and ligaments. However, so far no sequences are available in the relevant human and rodent databases to support this hypothesis.

The function of Tendin is unknown. However, its close similarity to CHM1 might suggest an anti-angiogenic function in ligaments and tendons where tendin is highly expressed. Ligaments and tendons are poorly vascularized tissues. A finely balanced regulation of angiogenesis is necessary to impede sprouting of vessels into the tendon (Benjamin and Ralphs, 2000; Brenchley, 2000). tendin might be involved in this process and the regulation of angiogenesis in ligaments and tendons during development, trauma, or inflammatory processes. Besides the high expression in tendons and ligaments, tendin is found in neuronal cells and a wide range of tissues. Although the C-terminal part of CHM1 and Tendin show striking homology, the presumption of an angiogenesis inhibiting function for the C-terminal domain of Tendin needs to be confirmed and its function in extraskeletal tissues should be investigated.

Arts syndrome has been mapped to Xq21.33-q24 where tendin is also located (Arts et al., 1993; Kremer et al., 1996). Male newborns with Arts syndrome show an early-onset floppiness and the development of ataxia, areflexia, flaccid tetraplegia, and myelin loss in the posterior columns of the spinal cord. They often die in early childhood due to susceptibility to infection, especially infection of the upper respiratory tract. Some patients also suffer from deafness and optic atrophy. Female carriers show only hearing loss (Arts et al., 1993). Although the function of Tendin in the central nervous system is not known, the disease symptoms fit well with the expression pattern of tendin and make it a candidate gene for Arts syndrome.

EXPERIMENTAL PROCEDURES

Cloning of the tendin cDNA

A blast search with the CHM1 complete cDNA using the advanced blast server at NCBI (Altschul et al., 1990) against the nonredundant database identified highly homologous human genomic sequences in PAC dJ479J7 acc. AL035608 and mouse ESTs. The IMAGE clone 463876 from a mouse cDNA library, which showed homology to the human sequences was sequenced with T3 and T7 primers by using the ABI Prism Dye terminator kit (Applied Biosystems). Sequences were analyzed on an ABI 373A automatic sequencer (Applied Biosystems). The 5′part of the cDNA was amplified with the Advantage cDNA PCR kit (Clontech) from mouse whole embryo Marathon-ready cDNA (Clontech) by nested PCR. The following primers were used: first-round gene-specific primer 5′-ATCCAGTACATGGTCACATTATCG; AP1 primer (Clontech) 5′-CCATCCTAATACGACTCACTATAGGGC; second-round gene-specific primer 5′-TTGAAGACCTACAAAGTAGATGCC; AP2 primer (Clontech) 5′-ACTCACTATCGGGCTCGAGCGGC. Amplified products were subcloned into pBluescript (Stratagene) and sequenced as described above. The complete ORF was amplified by RT-PCR from cartilage cDNA with the primers TendinF (5′-CTTAAGCTTACCATGGCAAAGAATCCTCC) and TendinR (5′-GGTACCGGTGACTCTCCCAAGCATGCG). The amplification products were subcloned into pBluescript and sequenced. For protein prediction the PIX platform provided by the HGMP (URL: http://www.hgmp.mrc.ac.uk/Registered/Menu/) was used.

Northern Analysis

Total RNA was extracted from various tissues of newborn C57B6 mice by the guanidium-thiocyanate method (Chomczynski and Sacchi, 1987). Five- or 15-μg aliquots of total RNA were electrophoretically separated on a 1% agarose-2.2 M formaldehyde gel, and transferred onto Hybond N+ membrane (Amersham). The blot was consecutively hybridized with ³²P-labeled cDNA probes specific for tendin (nt 531-1310), CHM1 (nt 529-1044), and GAPDH. The hybridizations were performed in Church buffer at 65°C overnight. Blots were washed twice for 20 min in Church washing buffer (1% SDS, 40 mM sodium pyrophosphate) and exposed to x-ray film for 5 days.

In Situ Hybridization

Digoxigenin-UTP (Boehringer Mannheim) labeled sense and antisense riboprobes were generated from a plasmid containing a fragment of the tendin cDNA (nt 74-628). In situ hybridization of alkaline fixed cryosections of embryonic and adult tissues was performed as described by Basyuk et al. (2000).

Acknowledgements

The mouse and human tendin cDNA sequences (acc. AF291655 and acc. AF291655, respectively) have been submitted to Genbank. We thank Dr. Heide Hellebrandt for experimental help and Dr. Mike Dictor for discussion and reading of the manuscript.