1 Department of Obstetrics and Gynaecology, University of Schleswig-Holstein, Kiel, 2 Department of Anatomy, University of Freiburg, Freiburg, Germany and 3 Harris Birthright Research Centre For Fetal Medicine, Kings College School of Medicine and Dentistry, London, UK
4 To whom correspondence should be addressed at: University of Schleswig-Holstein, Campus Kiel, Department of Obstetrics and Gynaecology, Michaelisstr. 16, 24105 Kiel, Germany. e-mail: vkaisenberg{at}email.uni-kiel.de
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: aneuploidy/extracellular matrix/glycosaminoglycans/nuchal translucency/proteoglycans
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
This study investigates further the hypothesis that increased nuchal translucency in Turner syndrome or aneuploidies may be the consequence of altered composition of the skin, in particular of glycosaminoglycans or proteoglycans. The aim of this study was to establish distribution patterns for a number of other glycosaminoglycansdermatan, heparan and keratan sulphate, chondroitin-6-sulphate and chondroitin-4-sulphate proteoglycanin the nuchal skin of normal and chromosomally abnormal fetuses at 1114 weeks. Fibroblast cultures from postnatal Turner patients contained only half the amount of biglycan (BGN) mRNA and BGN protein compared with age-matched control cultures (Geerkens et al., 1995). Therefore we also determined BGN mRNA and protein levels in the skin of Turner syndrome fetuses. BGN maps to the X chromosome and undergoes inactivation. Normal males and females have only one active copy of BGN, and fibroblast cultures in postnatal normal males and females did not show any difference in BGN expression, while individuals with additional X or Y chromosomes showed substantial overexpression (Table I). This suggests that BGN expression does not correlate with the number of BGN copies, but is rather correlated with the number of sex chromosomes. A regulator on the sex chromosomes (X and/or Y) may drive the transcription rate of BGN (Figure 1, Table I). If a gene copy of such a regulator on the X chromosome escapes inactivation or is pseudoautosomally regulated, different transcription rates of BGN in Turner patients and healthy females could be explained. The gene copy of the regulator gene on the Y chromosome should be a homologue of the X chromosomal copy, explaining identical transcription rates in females and males (Hibi et al., 1993
). BGN belongs to the family of the small-sized leucine-rich proteoglycans (SLRP) and is post-translationally modified by the addition of two glycosaminoglycan side-chains. The core protein may be modified by the addition of one or two glycosaminoglycan chains.
|
|
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Monoclonal antibodies directed against chondroitin-4-sulphate proteoglycan and chondroitin-6-sulphate proteoglycan specifically recognize 4-sulphated and 6-sulphated chondroitin following chondroitinase ABC (ChABC) digestion of various proteoglycans. Self-association among the glycosaminoglycans (GAG) and interaction with other extracellular matrix components may be affected by the particular charge organizations of the GAG molecule endowed by the sulphate group. The two monoclonals react with the chondroitin sulphate stubs which remain after extensive ChABC digestion. The antibody specificity is according to the sulphated position (4S, 6S) of these stubs (ICN product description). Proteoglycans, for example one of the small-sized proteoglycans BGN or DCN, have one or more glycosaminoglycan (GAG) side-chains attached to a core protein. Cross-reactivity of monoclonal antibodies directed against chondroitin sulphate proteoglycans with small proteoglycans may exist, but seems unlikely, since the antibody specificity is according to the sulphated position. Therefore it is unlikely that the finding of intense fluorescence in the dermis of nuchal skin of Turner fetuses using chondroitin-6-sulphate antibodies is the result of cross-reactivity of this antibody with either BGN or DCN (Figures 2a, b, f, 8, 12 and 13). The BGN and DCN antibodies did not cross-react, neither with each other, nor with dermatan, heparan or keratan sulphate, because BGN and DCN antibodies are directed against a short peptide of the core protein. No detailed information was available for dermatan sulphate or heparan sulphate, but keratan sulphate did not cross-react with any of the other antigens (Boehringer Mannheim Biochemica and ICN product information).
|
|
|
|
In-situ hybridization
Digoxigenin-labelled cRNA antisense riboprobes were taken as probes for in-situ hybridization according to the published protocol (Schaeren-Wiemers et al., 1993). The BGN and DCN probes were obtained from Dr Larry Fisher, NIH, USA (Fisher et al., 1989
). The BGN probe P16 was a 1685 bp insert in pBluescript containing the complete coding sequence of human bone proteoglycan I PGI (BGN). The antisense strand was linearized with KpnI and transcribed from the T3 promoter. The sense strand linearized with XbaI and transcribed with T7 polymerase. The DCN probe P2 was a 1600 bp insert into pBluescript containing the complete coding sequence of human bone proteoglycan II PGII (DCN) (Fisher et al., 1989
). The antisense strand was linearized with BamHI and transcribed from the T7 promoter, the sense strand linearized with KpnI and transcribed from the T3 promoter.
Isolation of RNA and protein
Skin tissues, dissected from control and Turner syndrome fetuses, were frozen immediately in RNase-free polypropylene tubes in liquid nitrogen and stored at 70°C. Total RNA and protein were isolated from identical probes by using the Trizol-based isolation procedure according to the protocol provided by the manufacturer (Life Technologies, Germany).
Northern blot analysis
We loaded 10 µg of total RNA per lane in the absence of ethidium bromide. RNA was blotted, UV-cross-linked, stained in methylene blue, and hybridized (Prols et al., 2001). Human BGN and DCN probes were labelled with [32P]dCTP using a random primer labelling kit (Life Technologies). For evaluating the quality of the individual RNA and to confirm equal loading per lane, the methylene blue-stained 28S and 18S rRNA are given as references.
Western blot analysis
Before electrophoresis the pelleted protein probes were rinsed three times in guanidium chloride (0.3 mol/l) and dissolved in 1% sodium dodecylsulphate overnight at 4°C. Dissolved protein samples were centrifuged at 10 000 g for 10 min at 4°C. The supernatants were transferred into new tubes and protein concentration was determined as described before (Prols et al., 1998). We loaded 25 µg protein per lane onto 12% polyacrylamide gels in the presence of 0.1% 2-mercaptoethanol. After electrophoresis, proteins were transferred onto nitrocellulose membranes in a semi-dry blotting chamber. To evaluate equal loading of the lanes, the membranes were stained with Ponceau S reagent (Sigma Aldrich Chemical Co., Germany).
Immunostaining was performed as previously described (Prols et al., 1998). Membranes were blocked in 5% non-fat dry milk in the cold overnight and stained with the polyclonal BGN antibody (1:2000). As second antibody we used horseradish-coupled goat-anti-rabbit antibody (1: 20 000). Staining was visualized using the ECL detection system (Pierce Reagent; KMF-Laboratories, St Augustin, Germany).
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
|
|
|
|
In-situ hybridization
Staining with BGN showed weak fluorescence in the area underlying the epidermis with homogeneous distribution. BGN seemed to be underexpressed in the nuchal skin of fetuses with Turner syndrome compared with an age-matched control (Figure 9), whereas DCN staining showed a comparable staining pattern in the nuchal skin tissue of the Turner and normal fetuses (Figure 10).
|
|
|
A consistent finding is that the 37/42 kDa BGN protein band is largely reduced in age-matched Turner fetuses (fetal probes 7 and 8) when compared with normal fetuses (fetal probes 2 to 5). However, the fetal probe 6, which originated from a 16 week old Turner fetus, showed more of the 37/42 kDa BGN protein, the amount being similar to the one in the normal 19 week fetus (Figure 15).
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
Chondroitin sulphate proteoglycans are known to have an influence on cell migration, in particular the migration of neural crest cells and outgrowing axons (Landolt et al., 1995). Potentially, the migration of blood and lymphatic vessel-forming cells could also be modified by an abnormal extracellular matrix (Miyabara et al., 1989
). This might be an explanation for lymphatic vessel hypoplasia, in particular in the upper part of the skin in Turner syndrome fetuses (von Kaisenberg et al., 1999
), where large skin edemas in the nuchal region at 1114 weeks can frequently be observed.
The finding that in Turner syndrome fetuses there was underexpression for BGN is not surprising since this small proteoglycan is encoded by the X chromosome (Figure 9). The BGN gene was mapped to chromosome Xq13-qter (McBride et al., 1990), Xq27-ter (Fisher et al., 1991
), Xq27-q28 (Traupe et al., 1992
), and recently to the distal long arm of the X chromosome, band Xq28, near the second pseudo-autosomal region (Geerkens et al., 1995
).
None of the seven fetuses with Turner syndrome showed mosaicism and there were no structural abnormalities of the X chromosome. However, one Turner fetus showed a Robertsonian translocation 13/14. This is of interest, because mosaicism was reported in 58/91 (66.7%) of Turner syndrome fetuses and 11/91 (12.6%) showed non-mosaic structural aberrations of the X chromosome, for example ring formation (Held et al., 1992). The existence of a normal additional sex chromosome in a mosaic could have substantially influenced gene expression.
BGN is expressed in a range of tissues and cell types, including connective tissue, human skin at the cell surface of differentiating keratinocytes and growing bone (Bianco et al., 1990; Traupe et al., 1992
). A critical role particularly in long-bone growth was therefore postulatedTurner syndrome individuals (45X) are known to be small in stature and have a short femur. This phenotype has also been observed in BGN-deficient mice (Xu et al., 1998
), while Klinefelter individuals (47XXY) and triple-X women (47XXX) are tall (Table I, Figure 1). BGN-deficient mice demonstrated both increased apoptosis of bone marrow stromal cells and decreased proliferation, leading to a deficiency in the generation of mature osteoblasts resulting in clinical osteopenia (Chen et al., 2002
). BGN knockout mice are defective in the synthesis of type I collagen mRNA and protein and mimic EhlersDanlos-like changes in bone and other connective tissues (Chen et al., 2002
; Corsi et al., 2002
). BGN is also located in the aorta and fetuses with Turner syndrome often suffer from extreme narrowing of the aortic isthmus presenting with aortic isthmus stenosis at birth. Fibroblast cultures from postnatal Turner patients contained only 50% of BGN mRNA and protein as age-matched control cultures (Table I, Figure 1), whereas normal males, who also have only one X chromosome, show transcript levels of 100% (Geerkens et al., 1995
). Abnormal distribution of BGN in Turner syndrome, found using in-situ hybridization, could not be confirmed by immunohistochemical studies and there was some degree of signal variation (Figures 2a, 3a, e and 12). Northern and Western blot analyses were performed in the skin of three fetuses with Turner syndrome and age-matched controls to quantify BGN mRNA and protein levels (Figure 15). BGN RNA seemed to be at similar levels compared with normal, but the stability of BGN RNA seemed to be reduced, which becomes apparent through the appearance of smaller BGN hybridizing bands. BGN protein patterns in Turner syndrome fetuses were largely different from normal. These data confirm the observation that BGN protein levels are decreased already in first trimester fetuses with Turner syndrome compared with normal controls (Figure 15). Further experiments could not be carried out, because Turner syndrome is one of the rarest numerical chromosomal abnormalities and further fetal samples were not available.
DCN, which is encoded on chromosome 12 (McBride et al., 1990), was as expected unchanged in the nuchal skin of Turner syndrome fetuses by immunohistochemistry (Figures 2b, 4a and e), in-situ hybridization (Figures 10 and 11) and Northern blotting (Figure 15). DCN is associated with fibrillar collagen I and II and its distribution is mutually exclusive to BGN (Bianco et al., 1990
). Its distribution suggests that DCN is involved in regulating collagen fibril formation and organization. Moreover it has been shown that it can modulate the activity of growth factors by binding them (Yamaguchi et al., 1990
). By using a double staining technique, we could demonstrate that DCN and collagen type I are already co-localized in the fetal skin (Figure 11). It was shown in vitro that DCN of bovine tendon demonstrated a unique ability to inhibit fibrillogenesis of both type I and type II collagen from bovine tendon and cartilage respectively (Vogel et al., 1984
; Uldbjerg et al., 1988
; Brown et al., 1989
). Unlike BGN, DCN and collagen levels did not differ from control in Turner patients (Figure 11).
Keratan, heparan and dermatan sulphate did not show any obvious difference in distribution or expression in the nuchal skin of Turner syndrome fetuses, fetuses with trisomies or normal controls (Figures 2 and 57). The balance between proteases and inhibitors is critical for the steady state level of glycosaminoglycans, which involves the rate of protein synthesis, of post-translational modifications by glycosidases and sulphatases as well as proteases. The study of dermatan and heparan sulphate was performed because the enzyme iduronate-2-sulphatase (IDS), which metabolizes both molecules, is located on Xq28. We therefore assumed that the activity of this enzyme could be reduced in Turner syndrome resulting in an accumulation of its substrates. However, no changes have been observed at this early stage of pregnancy.
One of our most striking findings was the intense staining for chondroitin-6-sulphate in the dermis of nuchal skin of Turner fetuses, which extends down to a chain of vessels at the dermissubcutis junction (Figures 2f, 8 and 14). The area below this junction was practically devoid of staining. This is compatible with previous findings in the nuchal skin of fetuses with Turner syndrome, where aberrant distribution of lymphatic vessels in the upper dermis was found using a PTN63 antibody against 5'-nucleotidase (von Kaisenberg et al., 1999). The scarcity of lymphatics in the upper dermis and the congestion of large diameter vessels at the dermissubcutis junction was presumably the cause for the increase in nuchal translucency arising from failure of interstitial fluid drainage in this area.
Potential mechanisms for increased nuchal translucency are increased proteoglycans, which bind large amounts of water, thus creating swelling of the skin. This is true for chondroitin-6-sulphate in the dermis of nuchal skin of Turner fetuses found in this study and for hyaluronan in the dermis of nuchal skin of trisomy 21 fetuses (Böhlandt et al., 2000). Furthermore, adhesive glycoproteins influence cell behaviour by attachment and abnormal migration, potentially altering the migration of blood and lymphatic vessel-forming cells (Ayad et al., 1994
), leading to lymphatic vessel hypoplasia (von Kaisenberg et al., 1999
) or narrowing of the aortic isthmus (Hyett et al., 1997
). BGN is located in the aorta. In Turner syndrome fetuses BGN is underexpressed in the nuchal skin as shown in our study and is likely to be reduced in the aortic arch. Turner syndrome fetuses frequently suffer from narrowing of the aortic isthmus, potentially as a result of decreased BGN, or of abnormal migration of cells from the neural crest, which form the aortic arch. In Turner fetuses with narrowing of the aortic arch, there is increased impedance to flow in the aorta and potentially overperfusion of the head and neck, leading to the highest degree of increase in nuchal translucency. In fetuses with trisomies 21, 18 and 13, unlike in Turners syndrome, there is normal BGN expression, and the mechanism of increased nuchal translucency is different. There is however, as the result of a gene dosage effect, overexpression of insoluble fibrils as part of the framework of connective tissues, in particular collagen type VI, laminin and collagen type IV respectively (von Kaisenberg et al., 1998a
,b).
In conclusion, this study shows that fetuses with Turner syndrome show a distribution pattern for chondroitin-6-sulphate and BGN that is substantially different from normal. This indicates a genetic regulation of extracellular matrix components in first trimester fetuses and may help explain the pathophysiology of increased nuchal translucency.
![]() |
Acknowledgements |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bianco, P., Fisher, L.W., Young, M.F., Termine, J.D. and Robey, P.G. (1990) Expression and localization of the two small proteoglycans biglycan and decorin in developing human skeletal and non-skeletal tissues. J. Histochem. Cytochem., 38, 15491563.[Abstract]
Böhlandt, S., von Kaisenberg, C.S., Wewetzer, K., Christ, B., Nicolaides, K.H. and Brand-Saberi, B. (2000) Hyaluronan in the nuchal skin of chromosomally abnormal fetuses. Hum. Reprod., 15, 11551158.
Brand-Saberi, B., Epperlein, H.H., Romanos, G.E. and Christ, B. (1994a) Distribution of extracellular matrix components in nuchal skin from fetuses carrying trisomy 18 and trisomy 21. Cell. Tiss. Res., 277, 465475.[CrossRef][ISI][Medline]
Brand-Saberi, B., Floel, H., Christ, B., Schulte-Vallentin, M. and Schindler, H. (1994b) Alterations of the fetal extracellular matrix in the nuchal oedema of Downs syndrome. Anat. Anz., 176, 539547.[Medline]
Brown, D.C. and Vogel, K.G. (1989) Characteristics of the in vitro interaction of a small proteoglycan (PG II) of bovine tendon with type I collagen. Matrix, 9, 468478.[ISI][Medline]
Chen, X.D., Shi, S., Xu, T., Robey, P.G. and Young, M.F. (2002) Age-related osteoporosis in biglycan-deficient mice is related to defects in bone marrow stromal cells. J. Bone Miner. Res., 17, 331340.[ISI][Medline]
Corsi, A., Xu, T., Chen, X.D., Boyde, A., Liang, J., Mankani, M., Sommer, B., Iozzo, R.V., Eichstetter, I., Robey, P.G. et al. (2002) Phenotypic effects of biglycan deficiency are linked to collagen fibril abnormalities, are synergized by decorin deficiency, and mimic EhlersDanlos-like changes in bone and other connective tissues. J. Bone Miner. Res., 17, 11801189.[ISI][Medline]
Fisher, L.W., Termine, J.D. and Young, M.F. (1989) Deduced-protein sequence of bone small proteoglycan I (Biglycan) shows homology with proteoglycan II (Decorin) and several nonconnective tissue proteins in a variety of species. J. Biol. Chem., 246, 45714576.
Fisher, L.W., Heegaard, A.M., Vetter, U., Vogel, W., Just, W., Termine, J.D. and Young, M.F. (1991) Human biglycan gene. Putative promoter, intronexon junctions, and chromosomal localization. J. Biol. Chem., 266, 1437114377.
Geerkens, C., Vetter, U., Just, W., Fedarko, N.S., Fisher, L.W., Young, M.E., Termine, J.D., Robey, P.G., Wohrle, D. and Vogel, W. (1995) The X-chromosomal human biglycan gene BGN is subject to X inactivation but is transcribed like an X-Y homologous gene. Hum. Genet., 96, 4452.[ISI][Medline]
Gleizes, P.E. and Rifkin, D.B. (1999) Activation of latent TGF-beta. A required mechanism for vascular integrity. [In French.] Pathol. Biol., 47, 322329.[ISI][Medline]
Held, K.R., Kerber, S., Kaminsky, E., Singh, S., Goetz, P., Seemanova, E. and Goedde, H.W. (1992) Mosaicism in 45, X Turner syndrome: does survival in early pregnancy depend on the presence of two sex chromosomes? Hum. Genet., 88, 288294.[ISI][Medline]
Hibi, I. and Takano, K. (eds) (1993) Basic and Clinical Approach to Turner Syndrome. Excerpta Medica, Amsterdam, pp. 143146.
Hyett, J.A., Brizot, M.L., von Kaisenberg, C.S., McKie, A.T., Farzaneh, F. and Nicolaides, K.H. (1996) Cardiac gene expression of atrial natriuretic peptide and brain natriuretic peptide in trisomic fetuses. Obstet. Gynecol., 87, 506510.
Hyett, J.A., Moscoso, G. and Nicolaides, K.H. (1997) Abnormalites of the heart and great arteries in first trimester chromosomally abnormal fetuses. Am. J. Med. Genet., 69, 207216.[CrossRef][ISI][Medline]
Landolt, R.M., Vaughan, L., Winterhalter, K.H. and Zimmermann, D.R. (1995) Versican is selectively expressed in embryonic tissues that act as barriers to neural crest cell migration and axon outgrowth. Development, 121, 23032312.
Liu, N.F. and He, Q.L. (1997) The regulatory effects of cytokines on lymphatic angiogenesis. Lymphology, 30, 312.[ISI][Medline]
McBride, O.W., Fisher, L.W. and Young, M.F. (1990) Localization of PGI (biglycan, BGN) and PGII (decorin, DCN, PG-40) genes on human chromosomes Xq13-qter and 12q, respectively. Genomics, 6, 219225.[ISI][Medline]
Miyabara, S., Sugihara, H., Maehara, N., Shouno, H., Tasaki, H., Yoshida, K., Saito, N., Kayama, F., Ibara, S. and Suzumori, K. (1989) Significance of cardiovascular malformations in cystic hygroma: a new interpretation of the pathogenesis. Am. J. Med. Genet., 34, 489501.[ISI][Medline]
Nicolaides, K.H., Azar, G., Byrne, D., Mansur, C. and Marks, K. (1992) Fetal nuchal translucency: ultrasound screening for chromosomal defects in first trimester of pregnancy. Br. Med. J., 304, 867869.[ISI][Medline]
Polkinghorne, J. (1989) Review of the Guidance on the Research Use of Fetuses and Fetal Material. Her Majestys Stationery Office, London, Cm762.
Prols, F., Loser, B. and Marx, M. (1998) Differential expression of osteopontin, PC4, and CEC5, a novel mRNA species, during in vitro angiogenesis. Exp. Cell Res., 239, 110.[CrossRef][ISI][Medline]
Prols, F., Mayer, M.M., Renner, O., Czarnecki, P.G., Ast, M., Gässler, C., Wilting, J., Kurz, H. and Christ, B. (2001) Upregulation of the cochaperone Mdg1 in endothelial cells is induced by stress and during in vitro angiogenesis. Exp. Cell Res., 269, 4253.[CrossRef][ISI][Medline]
Schaeren-Wiemers, N. and Gerfin-Moser, A. (1993) A single protocol to detect transcripts of various types and expression levels in neural tissue and cultured cells: in situ hybridization using digoxigenin-labelled cRNA probes. Histochemistry, 100, 431440.[ISI][Medline]
Traupe, H., Van den Ouweland, A.M.W., Van Oost, B.A., Vogel, W., Vetter, U., Warren, S.T., Rocchi, M., Darlison, M.G. and Ropers, H.H. (1992) Fine mapping of the human biglycan (BGN) gene within the Xq28 region employing a hybrid cell panel. Genomics, 13, 481483.[ISI][Medline]
Uldbjerg, N. and Danielsen, C.C. (1988) A study of the interaction in vitro between type-I collagen and a small dermatan sulphate proteoglycan. Biochem. J., 251, 643648.[ISI][Medline]
Vogel, K.G., Paulsson, M. and Heinegard, D. (1984) Specific inhibition of type I and type II collagen fibrillogenesis by the small proteoglycan of tendon. Biochem. J., 223, 587597.[ISI][Medline]
von Kaisenberg, C.S., Huggon, I., Hyett, J.A., Farzaneh, F. and Nicolaides, K.H. (1997) Cardiac expression of sarcoplasmic reticulum calcium ATPase in fetuses with trisomy 21 and trisomy 18 presenting with nuchal translucency. Fetal Diagn. Ther., 12, 270273.[ISI][Medline]
von Kaisenberg, C.S., Brand-Saberi, B., Christ, B., Vallian, S., Farzaneh, F. and Nicolaides, K.H. (1998a) Collagen type VI gene expression in the skin of trisomy 21 fetuses. Obstet. Gynecol., 91, 319323.
von Kaisenberg, C.S., Krenn, V., Ludwig, M., Nicolaides, K.H. and Brand-Saberi, B. (1998b) Morphological classification of nuchal skin in fetuses with trisomy 21, 18 and 13 at 1218 weeks and in a trisomy 16 mouse. Anat. Embryol., 197, 105124.[CrossRef][ISI][Medline]
von Kaisenberg, C.S., Nicolaides, K.H. and Brand-Saberi, B. (1999) Lymphatic vessel hypoplasia in fetuses with Turner syndrome. Hum. Reprod., 14, 823826.
Xu, T., Bianco, P., Fisher, L.W., Longenecker, G., Smith, E., Goldstein, S., Bonadio, J., Boskey, A., Heegaard, A.M., Sommer, B. et al. (1998) Targeted disruption of the biglycan gene leads to an osteoporosis-like phenotype in mice. Nat. Genet., 20, 7882.[CrossRef][ISI][Medline]
Yamaguchi, Y., Mann, D.M. and Ruoslahti, E. (1990) Negative regulation of transforming growth factor- by the proteoglycan decorin. Nature, 346, 281284.[CrossRef][ISI][Medline]
Submitted on December 12, 2002; resubmitted on April 3, 2003; accepted on September 8, 2003.