The Evolution of MHC Diversity by Segmental Duplication and Transposition of Retroelements

Sequence analysis of a 237 kb genomic fragment from the central region of the MHC has revealed that the HLA-B and HLA-C genes are contained within duplicated segments peri-B (53 kb) and peri-C (48 kb), respectively, and separated by an intervening sequence (IF) of 30 kb. The peri-B and peri-C segments share at least 90% sequence homology except when interrupted by insertions/deletions including Alu, L1, an endogenous retrovirus, and pseudogenes. The sequences of peri-B, IF, and peri-C were searched for the presence of Alu elements to use as markers of evolution, chromosomal rearrangements, and polymorphism. Of 29 Alu elements, 14 were identified in peri-B, 11 in peri-C, and 4 in IF. The Alu elements in peri-B and peri-C clustered phylogenetically into two clades which were classified as ``preduplication' and ``postduplication' clades. Four Alu J elements that are shared by peri-B and peri-C and are flanked by homologous sequences in their paralogous locations, respectively, clustered into a ``preduplication' clade. By contrast, the majority of Alu elements, which are unique to either peri-B or peri-C, clustered into a postduplication clade together with the Alu consensus subfamily members ranging from platyrrhine-specific (Spqxcg) to catarrhine-specific Alu sequences (Y). The insertion of platyrrhine-specific Alu elements in postduplication locations of peri-B and peri-C implies that these two segments are the products of a duplication which occurred in primates prior to the divergence of the New World primate from the human lineage (35–44 mya). Examination of the paralogous Alu integration sites revealed that 9 of 14 postduplication Alu sequences have produced microsatellites of different length and sequence within the Alu 3′-poly A tail. The present analysis supports the hypothesis that HLA-B and HLA-C genes are products of an extended segmental duplication between 44 and 81 million years ago (mya), and that subsequent diversification of both genomic segments occurred because of the mobility and mutation of retroelements such as Alu repeats. Received: 21 May 1997 / Accepted: 9 July 1997     

Duplication and Diversification of the Apolipoprotein CI (APOCI) Genomic Segment in Association with Retroelements

We have previously shown that several multicopy gene families within the major histocompatibility complex (MHC) arose from a process of segmental duplication. It has also been observed that retroelements play a role in generating diversity within these duplicated segments. The objective of this study was to compare the genomic organization of a gene duplication within another multicopy gene family outside the MHC. Using new continuous genomic sequence encompassing the APOE-CII gene cluster, we show that APOCI and its pseudogene, APOCI′, are contained within large duplicated segments which include sequences from the hepatic control region (HCR). Flanking Alu sequences are observed at both ends of the duplicated unit, suggesting a possible role in the integration of these segments. As observed previously within the MHC, the major differences between the segments are the insertion of sequences (approximately 200–1000 bp in length), consisting predominantly of Alu sequences. Ancestral retroelements also contribute to the generation of sequence diversity between the segments, especially within the 3′ poly(A) tract of Alu sequences. The exonic and regulatory sequences of the APOCI and HCR loci show limited sequence diversity, with exon 3 being an exception. Finally, the typing of pre- and postduplication Alus from both segments indicates an estimated time of duplication of approximately 37 million years ago (mya), some time prior to the separation of Old and New World monkeys. Received: 17 July 1999 / Accepted: 6 November 1999     

Coevolution of HLA-B and PERB11.1 (MICA): Significance of Independent Triplet Expansion Within the Transmembrane Region of PERB11.1 (MICA)

Several highly polymorphic sequences are present in the beta block of the MHC, especially HLA-B, HLA-C, PERB11.1 (MICA), and PERB11.2 (MICB). It is now apparent that the polymorphism of PERB11.1 is of the same order as that of HLA-A, -B, and -C and it has been suggested that PERB11 could explain some of the disease associations previously attributed to HLA-B. Phylogenetic analysis of PERB11 α-domain sequences demonstrates relationships with HLA-B cross-reactive serogroups. In contrast, the transmembrane polymorphisms do not appear to be associated with either PERB11 or HLA-B. These data indicate that PERB11 and HLA-B have evolved in concert from their common ancestors and that the transmembrane polymorphisms have arisen independently and more recently. MHC disease associations will need to be reviewed in the light of mechanisms such as receptor binding and signaling. Received: 22 September 1999 / Accepted: 11 January 2000 (Revised)     

Genomic and Phylogenetic Analysis of the Human CD1 and HLA Class I Multicopy Genes

The human CD1 proteins belong to a lipid-glycolipid antigen-presenting gene family and are related in structure and function to the MHC class I molecules. Previous mapping and DNA hybridization studies have shown that five linked genes located within a cluster on human chromosome 1q22-23 encode the CD1 protein family. We have analyzed the complete genomic sequence of the human CD1 gene cluster and found that the five active genes are distributed over 175,600 nucleotides and separated by four expanded intervening genomic regions (IGRs) ranging in length between 20 and 68 kb. The IGRs are composed mostly of retroelements including five full-length L1 PA sequences and various pseudogenes. Some L1 sequences have acted as receptors for other subtypes or families of retroelements. Alu molecular clocks that have evolved during primate history are found distributed within the HLA class I duplicated segments (duplicons) but not within the duplicons of CD1. Phylogeny of the alpha3 domain of the class I-like superfamily of proteins shows that the CD1 cluster is well separated from HLA class I by a number of superfamily members including MIC (PERB11), HFE, Zn-alpha2-GP, FcRn, and MR1. Phylogenetically, the human CD1 sequences are interspersed by CD1 sequences from other mammalian species, whereas the human HLA class I sequences cluster together and are separated from the other mammalian sequences. Genomic and phylogenetic analyses support the view that the human CD1 gene copies were duplicated prior to the evolution of primates and the bulk of the HLA class I genes found in humans. In contrast to the HLA class I genomic structure, the human CD1 duplicons are smaller in size, they lack Alu clocks, and they are interrupted by IGRs at least 4 to 14 times longer than the CD1 genes themselves. The IGRs seem to have been created as "buffer zones" to protect the CD1 genes from disruption by transposable elements.     

Genomic Organization Around the Centromeric End of the HLA Class I Region: Large-Scale Sequence Analysis

We previously sequenced two regions around the centromeric end of HLA class I and the boundary between class I and class III. In this paper we analyze the two regions of about 385 kb and confirm, giving a new line of evidence, that the following two pairs of the genomic segments were duplicated in evolution: (i) a 43-kb genomic segment including the HLA-B gene showing the highest polymorphism among the classical HLA class I loci (class Ia) and a 40-kb segment including the HLA-C locus showing the lowest polymorphism and (ii) a 52-kb segment including the MIC (MHC class I chain related gene) B and a 35-kb segment including MICA. We also found that repetitive elements such as SINEs, LINEs, and LTRs occupy as much as 47% of nucleotides in this 385-kb region. This unusually high content of repetitive elements indicates that repeat-mediated rearrangements have frequently occurred in the evolutionary history of the HLA class Ia region. Analysis of LINE compositions within the two pairs of duplicated segments revealed that (i) LINEs in these regions had been dispersed prior to both the duplication of the HLA-B and -C loci and the duplication of the MICB and MICA loci, and (ii) the divergence of the HLA-B and -C loci occurred prior to the duplication of the MICA and MICB loci. To find novel genes responsible for HLA class I-associated or other diseases, we performed computer analysis applying GenScan and GRAIL to GenBank's dbEST. As a result, at least five as yet uncharacterized genes were newly mapped on the HLA class I centromeric region studied. These novel genes should be analyzed further to determine their relationships to diseases associated with this region. Received: 16 June 1998 / Accepted: 18 August 1998     

Using Alu J Elements as Molecular Clocks to Trace the Evolutionary Relationships Between Duplicated HLA Class I Genomic Segments

The class I region of the major histocompatibility complex contains two subgenomic blocks (250–350 kb each), known as the alpha and beta blocks. These blocks contain members of multicopy gene families including HLA class I, HERV-16 (previously called P5 sequences), and PERB11 (MIC). We have previously shown that each block consists of imperfect duplicated segments (duplicons) containing linked members of different gene families, retroelements and transposons that have coevolved as part of two separate evolutionary events. Another region provisionally designated here as the kappa block is located between the alpha and the beta blocks and contains HLA-E, -30, and -92, HERV-16 (P5.3), and PERB11.3 (MICC) within about 250 kb of sequence. Using Alu elements to trace the evolutionary relationships between different class I duplicons, we have found that (a) the kappa block contains paralogous (duplicated) Alu J sequences and other retroelement patterns more in common with the beta than the alpha block; (b) the retroelement pattern associated with the HLA-E duplicon is different from all other HLA class I duplicons, indicating a more complex evolution; (c) the HLA-92 duplicon, although substantially shorter, is closely related in sequence to the HLA-B and -C duplicons; (d) two of the six paralogous Alu J elements within the HLA-B and -C duplicons are associated with the HLA-X duplicon, confirming their evolutionary relationships within the beta block; and (e) the paralogous Alu J elements within the alpha block are distinctly different from those identified within the beta and kappa blocks. The sequence conservation and location of duplicated (paralogous) Alu J elements in the MHC class I region show that the beta and kappa blocks have evolved separately from the alpha block beginning at a time before or during the evolution of Alu J elements in primates. Received: 22 September 1999 / Accepted: 24 January 2000     

Duplication and Quadruplication of Arabidopsis thaliana Cysteinyl- and Asparaginyl-tRNA Synthetase Genes of Organellar Origin

Two cysteinyl-tRNA synthetases (CysRS) and four asparaginyl-tRNA synthetases (AsnRS) from Arabidopsis thaliana were characterized from genome sequence data, EST sequences, and RACE sequences. For one CysRS and one AsnRS, sequence alignments and prediction programs suggested the presence of an N-terminal organellar targeting peptide. Transient expression of these putative targeting sequences joined to jellyfish green fluorescent protein (GFP) demonstrated that both presequences can efficiently dual-target GFP to mitochondria and plastids. The other CysRS and AsnRSs lack targeting sequences and presumably aminoacylate cytosolic tRNAs. Phylogenetic analysis suggests that the four AsnRSs evolved by repeated duplication of a gene transferred from an ancestral plastid and that the CysRSs also arose by duplication of a transferred organelle gene (possibly mitochondrial). These case histories are the best examples to date of capture of organellar aminoacyl-tRNA synthetases by the cytosolic protein synthesis machinery. Received: 8 October 1999 / Accepted: 23 January 2000     

Characterization of Two Alcohol Dehydrogenase (Adh) Loci from the Olive Fruit Fly, Bactrocera (Dacus) oleae and Implications for Adh Duplication in Dipteran Insects

We report the cloning and structural characterization of two Adh loci of the olive fruit fly, Bactrocera oleae. Each of the two genes, named Adh1 and Adh2, consists of three exons and two introns for a total length of 1981 and 988 nucleotides, respectively. Their deduced amino acid sequences of 257 and 258 residues exhibit a 77% identity and display the characteristics of the insect ADH enzymes, which belong to the short-chain dehydrogenases/reductases family. The Adh genes of B. oleae are compared to the two genes of the Mediterranean fly, Ceratitis capitata, the only other species of the Tephritidae family in which the Adh genes have been studied. On the basis of amino acid divergence the four genes form two clusters each containing one gene from each species, as expected if there was one duplication event before speciation. On the basis of nucleotide sequence the four sequences form two clusters each containing the two sequences from the same species, as expected if there was a separate duplication event in each species. To help decide between the two alternatives, we compared at both the amino acid and DNA level the Adh genes of five Drosophila species that are known to carry two such genes and observed that, with only one exception at the amino acid level, conspecific loci cluster together. We conclude that the information we have at present does not allow a firm choice between the hypothesis of a single duplication event that occurred before the split of Bactrocera and Ceratitis from their common ancestor and the hypothesis of two independent duplication events, one in each of the two genera. Received: 30 May 2000 / Accepted: 17 August 2000     

Evidence for Gene Conversion Between Tandemly Duplicated Cytoplasmic Actin Genes of Helicoverpa armigera (Lepidoptera:Noctuidae)

Tandemly duplicated actin genes have been isolated from a Helicoverpa armigera genomic library. Sequence comparisons with actin genes from other species suggest they encode cytoplasmic actins, being most closely related to the Bombyx mori A3 actin gene. The duplicated H. armigera actin genes, termed A3a and A3b, share 98.3% nucleotide sequence identity over their entire putative coding region. Analysis of the distribution of nucleotide differences shows the first 763 bp are identical between the two coding regions, with the 18 nucleotide changes occurring in the remaining 366 bp. This observation suggests a gene conversion event has taken place between the duplicated H. armigera A3a and A3b actin genes. Translation of the open-reading frames indicates the products of these genes are identical, apart from a single amino acid difference at codon 273. Polymerase chain reaction and northern blot analysis have shown both H. armigera A3a and A3b genes are expressed during pupal development and in the brain of newly eclosed adults. A region 5′ of the H. armigera A3a actin gene start codon has been identified which contains regulatory sequences commonly found in the promoter region of actin genes, including TATA, CAAT, and CArG motifs. Received: 10 January 1996 / Accepted: 12 March 1996     

The Heat-Shock Gene hsp83 of Drosophila auraria: Genomic Organization, Nucleotide Sequence, and Long Antiparallel Coupled ORFs (LAC ORFs)

The genomic organization of the hsp83 gene of Drosophila auraria, a far-eastern endemic species belonging to the montium subgroup of the melanogaster species group, is presented here. Based on in situ hybridization on polytene chromosomes, cDNA and genomic clone mapping, nucleotide sequencing, and genomic Southern analysis, hsp83 is shown to be present as a single-copy gene at locus 64B on the 3L chromosome arm in D. auraria. This gene is organized into two exons separated by a 929-bp intron. The first exon represents the mRNA leader sequence and is not translated, while the coding region, having a length of 2,151 bp, is solely included in the second exon. Nucleotide sequence comparisons of D. auraria hsp83 with homologous sequences from other organisms show high conservation of the coding region (88–92% identity) in the genus Drosophila, in addition to the conserved genomic organization of two-exons–one-intron, of comparable size and arrangement. A phylogenetic tree based on the protein sequences of homologous genes from representative organisms is in accord with the accredited phylogenetic position of D. auraria. In the hsp83 gene region, a second case of long antiparallel coupled open reading frames (LAC ORFs) for this species was found. The antiparallel to the hsp83 gene ORF is 1,554 bases long, while the two ORFs overlap has a size of 1,548 bp. The anti-hsp83 ORF does not show significant homology to any known gene sequences. In addition, no similar LAC ORF structures were found in homologous gene regions of other organisms. Received: 18 April 1997 / Accepted: 1 August 1997     

Evidence for AT-Transversion Bias in Wasp (Hymenoptera: Symphyta) Mitochondrial Genes and Its Implications for the Origin of Parasitism

We inferred the incidence of nucleotide conversions in the COI and 16S rRNA mitochondrial genes of members of the Symphyta and basal Apocrita (Hymenoptera). Character-state reconstructions in both genes suggested that conversions between A and T (AT transversions) occurred much more frequently than any other type of change, although we cannot wholly discount an underlying transition bias. Parsimony analysis of COI nucleotide characters did not recover phylogeny; e.g., neither the Tenthredinoidea nor Apocrita were recovered as monophyletic. However, analysis of COI amino acid characters did recover these relationships, as well as others based on fossil and morphological evidence. Analysis of 16S rRNA characters also recovered these relationships providing conversions between A and T were down-weighted. Analysis of the combined data sets gave relatively strong support for various relationships, suggesting that both data sets supported similar topographies. These data sets, both separately and combined, suggested that the phytophagous Siricidae were more closely related to the predominantly parasitic Apocrita than were the ectoparasitic Orussoidea. This suggests that the wasp parasitic lifestyle did not have a single origin, unless the Siricidae have more recently reverted to phytophagy. Alternatively, parasitism evolved twice independently, once in the Orussoidea and again in the Apocrita. The latter scenario is supported by the observation that the evolution of parasitism was accompanied by a tendency for the larvae to develop inside plant tissues. Adaptations that accompanied the movement of wasps into a confined, wood-boring habitat may have preadapted them to becoming ectoparasitic. Received: 27 March 1996 / Accepted: 2 August 1996     

Mitochondrial DNA of Hydra attenuata (Cnidaria): A Sequence That Includes an End of One Linear Molecule and the Genes for l-rRNA, tRNAf-Met, tRNATrp, COII, and ATPase8

The 3231-nucleotide-pair (ntp) sequence of one end of one of the two linear mitochondrial (mt) DNA molecules of Hydra attenuata (phylum Cnidaria, class Hydrozoa, order Anthomedusae) has been determined. This segment contains complete genes for tRNA^f-Met, l-rRNA, tRNA^Trp, subunit 2 of cytochrome c oxidase (COII), subunit 8 of ATP synthetase (ATPase8), and the 5′ 136 ntp of ATPase6. These genes are arranged in the order given and are transcribed from the same strand of the molecule. As in two other cnidarians, the hexacorallian anthozoan Metridium senile and the octocorallian anthozoan Sarcophyton glaucum, the mt-genetic code of H. attenuata is near standard. The only modification appears to be that TGA specifies tryptophan rather than termination. Also as in M. senile and S. glaucum, the encoded H. attenuata mt-tRNA^f-Met has primary and secondary structural features resembling those of Escherichia coli initiator tRNA^t-Met. As the encoded mt-tRNA^Trp cannot be folded into a totally orthodox secondary structure, two alternative forms are suggested. The encoded H. attenuata mt-l-rRNA is 1738 nt, which is 451 nt shorter than the M. senile mt-l-rRNA. Comparisons of secondary structure models of these two mt-l-rRNAs indicate that most of the size difference results from loss of nucleotides in the H. attenuata molecule at a minimum of 46 locations, which includes elimination of six distinct helical elements. Received: 9 March 2000 / Accepted: 24 July 2000     

Updating Carbamoylphosphate Synthase (CPS) Phylogenies: Occurrence and Phylogenetic Identity of Archaeal CPS Genes

Among Bacteria the carA and carB genes encoding the small (CarA) and large (CarB) subunits of carbamoylphosphate synthase (CPS) have been lost in certain symbionts (Haemophylus influenzae) and in most obligate intracellular parasites (Chlamydiae, Spirochaetes, Mycoplasmatales, Rickettsiae) having genome sizes in the 0.7- to 1.1-Mb range. Compared to Bacteria, Archaea exhibit a more varied pattern of CPS gene losses and an unusual propensity to incorporate CPS genes derived from both Bacteria and other Archaea. Schematically they fall into three groups. Group 1 taxa (the crenarchaeon Aeropyrum pernix and the euryarchaea Pyrococcus horikoshi and Pyrococcus abyssii) lack CPS genes altogether. Group 2 taxa (comprising Halobacteriales, Thermoplasmales, Methanococcales, Methanomicrobiales, Archaeoglobales) harbor CPS genes whose encoded CarB and CarA subunit proteins are ostensibly bacterial in origin; that is, they are intermixed with bacterial homologues on a phylogeny of concatenated CarA and CarB sequences and are not distinguishable from bacterial sequences after searching for domain-specific amino acid residue positions. Group 3 taxa (the crenarchaea Pyrobaculum aerophilum, Sulfolobus solfataricus, and Sulfolobus tokodaii and the euryarchaeon Pyrococcus furiosus) harbor CPS genes whose encoded proteins appear to be archaeal: consistent with an archaeal origin, the CarA and CarB sequences in this group possess both unique signatures and signatures affiliating them to Eukarya. Based on the topology of the clade comprising the four Group 3 taxa, we argue that CPS genes of P. furiosus (a euryarchaeon) and those of the crenarchaea P. aerophilum, S. solfataricus, and S. tokodaii are of a single type, resulting from the two genes being laterally transferred from a crenarchaeon to P. furiosus.     

The Root of the Universal Tree of Life Inferred from Anciently Duplicated Genes Encoding Components of the Protein-Targeting Machinery

The key protein of the signal recognition particle (termed SRP54 for Eucarya and Ffh for Bacteria) and the protein (termed SRα for Eucarya and Ftsy for bacteria) involved in the recognition and binding of the ribosome SRP nascent polypeptide complex are the products of an ancient gene duplication that appears to predate the divergence of all extant taxa. The paralogy of the genes encoding the two proteins (both of which are GTP triphosphatases) is argued by obvious sequence similarities between the N-terminal half of SRP54(Ffh) and the C-terminal half of SRα(Ftsy). This enables a universal phylogeny based on either protein to be rooted using the second protein as an outgroup. Phylogenetic trees inferred by various methods from an alignment (220 amino acid positions) of the shared SRP54(Ffh) and SRα(Ftsy) regions generate two reciprocally rooted universal trees corresponding to the two genes. The root of both trees is firmly positioned between Bacteria and Archaea/Eucarya, thus providing strong support for the notion (Iwabe et al. 1989; Gogarten et al. 1989) that the first bifurcation in the tree of life separated the lineage leading to Bacteria from a common ancestor to Archaea and Eucarya. None of the gene trees inferred from the two paralogues support a paraphyletic Archaea with the crenarchaeota as a sister group to Eucarya. Received: 19 March 1998 / Accepted: 5 June 1998     

Multiple Recurrent Evolution of Trophic Types in Northeastern Atlantic and Mediterranean Seabreams (Sparidae, Percoidei)

Seabreams are among the most valuable fish, not only for small-scale and semiindustrial fisheries but also for aquaculture throughout the Mediterranean. Nevertheless, their phylogenetic relationships are not at all clear. The current taxonomy is based solely on trophic morphology and rests on the assumption that each trophic type evolved only once from a less specialized ancestral condition. We analyzed a 486-bp segment of the mitochondrial 16S rDNA of all 24 seabream species described for the northeastern Atlantic and the Mediterranean to elucidate their generic and subfamily-level relationships. Three major mitochondrial lineages, each comprising species of different feeding strategy and dentition, were found that do not agree with the present taxonomic assignments. Most of the investigated genera were resolved paraphyletically, indicating that the structure and arrangement of oral teeth must have repeatedly evolved from a less specialized ancestral condition. Further, the genus Sparus was resolved as distantly related to the genus Pagrus, in that it was assigned to a different major mitochondrial lineage. Oblada melanura was consistently placed within the Diplodus radiation as sister group to Diplodus puntazzo. Our phylogenetic hypothesis thus suggests multiple independent origins of similar trophic specializations within the Sparidae and indicates that the currently recognized three or four subfamilies need to be redefined. Received: 5 October 1999 / Accepted: 9 November 1999     

On the Evolution of Dopa decarboxylase (Ddc) and Drosophila Systematics

We have sequenced most of the coding region of the gene Dopa decarboxylase (Ddc) in 24 fruitfly species. The Ddc gene is quite informative about Drosophila phylogeny. Several outstanding issues in Drosophila phylogeny are resolved by analysis of the Ddc sequences alone or in combination with three other genes, Sod, Adh, and Gpdh. The three species groups, melanogaster, obscura, and willistoni, are each monophyletic and all three combined form a monophyletic group, which corresponds to the subgenus Sophophora. The Sophophora subgenus is the sister group to all other Drosophila subgenera (including some named genera, previously considered outside the Drosophila genus, namely, Scaptomyza and Zaprionus, which are therefore downgraded to the category of subgenus). The Hawaiian Drosophila and Scaptomyza are a monophyletic group, which is the sister clade to the virilis and repleta groups of the subgenus Drosophila. The subgenus Drosophila appears to be paraphyletic, although this is not definitely resolved. The two genera Scaptodrosophila and Chymomyza are older than the genus Drosophila. The data favor the hypothesis that Chymomyza is older than Scaptodrosophila, although this issue is not definitely resolved. Molecular evolution is erratic. The rates of nucleotide substitution in 3rd codon position relative to positions 1 + 2 vary from one species lineage to another and from gene to gene. Received: 2 June 1998 / Accepted: 3 September 1998     

Molecular evolution of calmodulin-like domain protein kinases (CDPKs) in plants and protists

Many genes for calmodulin-like domain protein kinases (CDPKs) have been identified in plants and Alveolate protists. To study the molecular evolution of the CDPK gene family, we performed a phylogenetic analysis of CDPK genomic sequences. Analysis of introns supports the phylogenetic analysis; CDPK genes with similar intron/exon structure are grouped together on the phylogenetic tree. Conserved introns support a monophyletic origin for plant CDPKs, CDPK-related kinases, and phosphoenolpyruvate carboxylase kinases. Plant CDPKs divide into two major branches. Plant CDPK genes on one branch share common intron positions with protist CDPK genes. The introns shared between protist and plant CDPKs presumably originated before the divergence of plants from Alveolates. Additionally, the calmodulin-like domains of protist CDPKs have intron positions in common with animal and fungal calmodulin genes. These results, together with the presence of a highly conserved phase zero intron located precisely at the beginning of the calmodulin-like domain, suggest that the ancestral CDPK gene could have originated from the fusion of protein kinase and calmodulin genes facilitated by recombination of ancient introns. Received: 11 July 2000 / Accepted: 18 April 2001