Microsatellite Allelic Homoplasy Due to Variable Flanking Sequences

Microsatellite DNA sequences have become the dominant source of nuclear genetic markers for most applications. It is important to investigate the basis of variation between alleles and to know if current assumptions about the mechanisms of microsatellite mutation (that is to say, variations involving simple changes in the number of repeat) are correct. We have characterized, by DNA sequencing, the human alleles of a new highly informative (CA)n repeat localized approximately 20 kb centromeric to the HLA-B gene. Although 12 alleles were identified based on conventional length criteria, sequencing of the alleles demonstrated that differences between alleles were found to be more complex than previously assumed: A high degree of microsatellite variability is due to variation in the region immediately flanking the repeat. These data indicate that the mutational process which generates polymorphism in this region has involved not only simple changes in the number of dinucleotide CA repeats but also perturbations in the nonrepeated 5′ and 3′ flanking sequences. Three families of alleles (not visible from the overall length of the alleles), with presumably separate evolutionary histories, exist and can yield to homoplasy of size. Effectively, we can observe alleles of the same size with different internal structures which are separated by a significant amount of variation. Although allelic homoplasy for noninterrupted microsatellite loci has been suggested between different species, it has not been unequivocally demonstrated within species. A strong association is noted between alleles defined at the sequence level and HLA-B alleles. The observation of several families of alleles at the population level provides information about the evolutionary history and mutation processes of microsatellites and may have implications for the use of these markers in phylogenetic, linkage disequilibrium studies, and gene mapping. Received: 14 May 1996 / Accepted: 9 September 1996     

Divergent Human Y-Chromosome Microsatellite Evolution Rates

In this work, we analyze several characteristics influencing the low variability of the microsatellite DYS19 in the major founder Amerindian Y chromosome lineage containing the point mutation DYS199-T. Variation of DYS19 was compared with that of five other Y-linked tetranucleotide repeat loci (DYS389A, DYS389B, DYS390, DYS391, and DYS393) in the DYS199-T lineage. All the other microsatellites showed significantly higher levels of variability than DYS19 as measured by gene diversity and repeat number variance. Moreover, we had previously shown that DYS19 had high diversity in Brazilians and in several other populations worldwide. Thus, the slow DYS19 evolution in the DYS199-T lineage seems to be both locus and allele specific. To understand the slow DYS19 evolutionary rate, the microsatellite loci were compared according to their mapping on the Y chromosome and also on the basis of structural aspects such as the base composition of the repeat motif and flanking regions and the degree of perfection and size (repeat number) of the variable blocks. The only observed difference that might be related to the low DYS19 variability is its small average number of repeats, a value expected to be closer to the founder DYS19 allele in the DYS199-T lineage. These data were also compared to other derived Y lineages. The Tat-C lineage displayed a lower DYS19 variability correlated to a small average repeat number, while in the DYS234-G lineage, a high DYS19 variability was found associated to a larger average repeat number. This approach reveals that evolution of Y microsatellites in lineages defined by slowly evolving markers, such as point mutations, can be greatly influenced by the size (number of repeats of the variable block) of the founder allele in each microsatellite locus. Thus lineage-dating methods using microsatellite variation should be practiced with great care. Received: 7 November 1998 / Accepted: 9 April 1999     

Microsatellite Evolution: Testing the Ascertainment Bias Hypothesis

Previous studies suggest the median allele length of microsatellites is longest in the species from which the markers were derived, suggesting that an ascertainment bias was operating. We have examined whether the size distribution of microsatellite alleles between sheep and cattle is source dependent using a set of 472 microsatellites that can be amplified in both species. For those markers that were polymorphic in both species we report a significantly greater number of markers (P < 0.001) with longer median allele sizes in sheep, regardless of microsatellite origin. This finding suggests that any ascertainment bias operating during microsatellite selection is only a minor contributor to the variation observed. Received: 6 January 1997 / Accepted: 19 May 1997     

A Phylogenetic Perspective on Sequence Evolution in Microsatellite Loci

We examined the evolution of the repeat regions of three noncoding microsatellite loci in 58 species of the Polistinae, a subfamily of wasps that diverged over 140 million years ago. A phylogenetic approach allows two new kinds of approaches to studying microsatellite evolution: character mapping and comparative analysis. The basic repeat structure of the loci was highly conserved, but was often punctuated with imperfections that appear to be phylogenetically informative. Repeat numbers evolved more rapidly than other changes in the repeat region. Changes in number of repeats among species seem consistent with the stepwise mutation model, which is based on slippage during replication as the main source of mutations. Changes in repeat numbers can occur even when there are very few tandem repeats but longer repeats, especially perfect repeats led to greater rates of evolutionary change. Species phylogenetically closer to the one from which we identified the loci had longer stretches of uninterrupted repeats and more different motifs, but not longer total repeat regions. The number of perfect repeats increased more often than it decreased. However, there was no evidence that some species have consistently greater numbers of repeats across loci than other species have, once ascertainment bias is eliminated. We also found no evidence for a population size effect posited by one form of the directionality hypothesis. Overall, phylogenetic variation in repeat regions can be explained by adding neutral evolution to what is already known about the mutation process. The life cycle of microsatellites appears to reflect a balance between growth by slippage and degradation by an essentially irreversible accumulation of imperfections. Received: 13 April 1999 / Accepted: 8 September 1999     

Variation of Microsatellite Size Homoplasy Across Electromorphs, Loci, and Populations in Three Invertebrate Species

Size homoplasy was analyzed at microsatellite loci by sequencing electromorphs, that is, variants of the same size (base pairs). This study was conducted using five interrupted and/or compound loci in three invertebrate species, the honey bee Apis mellifera, the bumble bee Bombus terrestris, and the freshwater snail Bulinus truncatus. The 15 electromorphs sequenced turned out to hide 31 alleles (i.e., variants identical in sequence). Variation in the amount of size homoplasy was detected among electromorphs and loci. From one to seven alleles were detected per electromorph, and one locus did not show any size homoplasy in both bee species. The amount of size homoplasy was related to the sequencing effort, since the number of alleles was correlated with the number of copies of electromorphs sequenced, but also with the molecular structure of the core sequence at each locus. Size homoplasy within populations was detected only three times, meaning that size homoplasy was detected mostly among populations. We analyzed population structure, estimating F _st and a genetic distance, based on either electromorphs or alleles. Whereas little difference was found in A. mellifera, uncovering size homoplasy led to a more marked population structure in B. terrestris and B. truncatus. We also showed in A. mellifera that the detection of size homoplasy may alter phylogenetic reconstructions. Received: 21 July 1997 / Accepted: 29 January 1998     

Erratic Evolution of Glycerol-3-Phosphate Dehydrogenase in Drosophila,Chymomyza, and Ceratitis

We have studied the evolution of Gpdh in 18 fruitfly species by sequencing 1,077 nucleotides per species on average. The region sequenced includes four exons coding for 277 amino acids and three variable-length introns. Phylogenies derived by a variety of methods confirm that the nominal genus Zaprionus belongs within the genus Drosophila, whereas Scaptodrosophila and Chymomyza are outside. The rate of GPDH evolution is erratic. The rate of amino acid replacements in a lineage appears to be 1.0 × 10⁻¹⁰/site/year when Drosophila species are considered (diverged up to 55 million years ago), but becomes 2.3 × 10⁻¹⁰ when they are compared to Chymomyza species (divergence around 60 My ago), and 4.6 × 10⁻¹⁰ when species of those two genera are compared with the medfly Ceratitis capitata (divergence around 100 My ago). In order to account for these observations, the rate of amino acid replacement must have been 15 or more times greater in some lineages and at some times than in others. At the nucleotide level, however, Gpdh evolves in a fairly clockwise fashion. Received: 13 June 1996 / Accepted: 16 August 1996     

Molecular Evolution of P Transposable Elements in the Genus Drosophila. II. The obscura Species Group

A phylogenetic analysis of P transposable elements in the Drosophila obscura species group is described. Multiple P sequences from each of 10 species were obtained using PCR primers that flank a conserved region of exon 2 of the transposase gene. In general, the P element phylogeny is congruent with the species phylogeny, indicating that the dominant mode of transmission has been vertical, from generation to generation. One manifestation of this is the distinction of P elements from the Old World obscura and subobscura subgroups from those of the New World affinis subgroup. However, the overall distribution of elements within the obscura species group is not congruent with the phylogenetic relationships of the species themselves. There are at least four distinct subfamilies of P elements, which differ in sequence from each other by as much as 34%, and some individual species carry sequences belonging to different subfamilies. P sequences from D. bifasciata are particularly interesting. These sequences belong to two subfamilies and both are distinct from all other P elements identified in this survey. Several mechanisms are postulated to be involved in determining phylogenetic relationships among P elements in the obscura group. In addition to vertical transmission, these include retention of ancestral polymorphisms and horizontal transfer by an unknown mating-independent mechanism.     

Evolution of Orthologous Intronless and Intron-Bearing Globin Genes in Two Insect Species

While globin genes ctt-2β and ctt-9.1 in Chironomus thummi thummi each have a single intron, all of the other insect globin genes reported so far are intronless. We analyzed four globin genes linked to the two intron-bearing genes in C. th. thummi. Three have a single intron at the same position as ctt-2β and ctt-9.1; the fourth is intronless and lies between intron bearing genes. Finally, in addition to its intron, one gene (ctt-13RT) was recently interrupted by retrotransposition. Phylogenetic analyses show that the six genes in C. th. thummi share common ancestry with five globin genes in the distantly related species C. tentans, and that a 5-gene ancestral cluster predates the divergence of the two species. One gene in the ancestral cluster gave rise to ctn-ORFB in C. tentans, and duplicated in C. th. thummi to create ctt-11 and ctt-12. From parsimonious calculations of evolutionary distances since speciation, ctt-11, ctt-12, and ctn-ORFB evolved rapidly, while ctn-ORFE in C. tentans evolved slowly compared to other globin genes in the clusters. While these four globins are under selective pressure, we suggest that most chironomid globin genes were not selected for their unique function. Instead, we propose that high gene copy number itself was selected because conditions favored organisms that could synthesize more hemoglobin. High gene copy number selection to produce more of a useful product may be the basis of forming multigene families, all of whose members initially accumulate neutral substitutions while retaining essential function. Maintenance of a large family of globin genes not only ensured high levels of hemoglobin production, but may have facilitated the extensive divergence of chironomids into as many as 5000 species. Received: 31 December 1996 / Accepted: 16 May 1997     

Gene Conversion and Evolution of Daphniid Hemoglobins (Crustacea, Cladocera)

The extracellular hemoglobins of cladocerans derive from the aggregation of 12 two-domain globin subunits that are apparently encoded by four genes. This study establishes that at least some of these genes occur as a tandem array in both Daphnia magna and Daphnia exilis. The genes share a uniform structure; a bridge intron separates two globin domains which each include three exons and two introns. Introns are small, averaging just 77 bp, but a longer sequence (2.2–3.2 kb) separates adjacent globin genes. A survey of structural diversity in globin genes from other daphniids revealed three independent cases of intron loss, but exon lengths were identical, excepting a 3-bp insertion in exon 5 of Simocephalus. Heterogeneity in the extent of nucleotide divergence was marked among exons, largely as a result of the pronounced diversification of the terminal exon. This variation reflected, in part, varying exposure to concerted evolution. Conversion events were frequent in exons 1–4 but were absent from exons 5 and 6. Because of this difference, the results of phylogenetic analyses were strongly affected by the sequences employed in this construction. Phylogenies based on total nucleotide divergence in exons 1–4 revealed affinities among all genes isolated from a single species, reflecting the impact of gene conversion events. In contrast, phylogenies based on total nucleotide divergence in exons 5 and 6 revealed affinities among orthologous genes from different taxa. Received: 8 March 1999 / Accepted: 14 July 1999     

Rapid Evolution of the Ribonuclease A Superfamily: Adaptive Expansion of Independent Gene Clusters in Rats and Mice

The two eosinophil ribonucleases, eosinophil-derived neurotoxin (EDN/RNase 2) and eosinophil cationic protein (ECP/RNase 3), are among the most rapidly evolving coding sequences known among primates. The eight mouse genes identified as orthologs of EDN and ECP form a highly divergent, species-limited cluster. We present here the rat ribonuclease cluster, a group of eight distinct ribonuclease A superfamily genes that are more closely related to one another than they are to their murine counterparts. The existence of independent gene clusters suggests that numerous duplications and diversification events have occurred at these loci recently, sometime after the divergence of these two rodent species (∼10–15 million years ago). Nonsynonymous substitutions per site (d _N) calculated for the 64 mouse/rat gene pairs indicate that these ribonucleases are incorporating nonsilent mutations at accelerated rates, and comparisons of nonsynonymous to synonymous substitution (d _N / d _S) suggest that diversity in the mouse ribonuclease cluster is promoted by positive (Darwinian) selection. Although the pressures promoting similar but clearly independent styles of rapid diversification among these primate and rodent genes remain uncertain, our recent findings regarding the function of human EDN suggest a role for these ribonucleases in antiviral host defense. Received: 8 April 1999 / Accepted: 22 June 1999     

Evolution of Microsatellites in the Yeast Saccharomyces cerevisiae: Role of Length and Number of Repeated Units

The observed and expected frequencies of occurrence of microsatellites in the yeast Saccharomyces cerevisiae were investigated. In all cases, the observed frequencies exceeded the expected ones. In contrast to predictions by Messier et al. (1996), there is no critical number of repeats beyond which the observed frequencies of microsatellites significantly exceed the frequencies expected in a random DNA sequence of the same size. Rather, the degree of deviation from expectation was found to be dependent on the length of the microsatellite. That is, a fourfold concatemeric repeat of 3 bp was found to deviate from expectation as much as threefold concatemeric repeat of 4 bp, unlike the deviation of a fourfold concatemeric repeat of 4 bp. These findings suggest that microsatellites evolve through strand-slippage events, rather than recombination events. This, in turn, suggests that the chances of erroneous hybridizations leading to strand-slippage are length dependent. Received: 1 June 1998 / Accepted: 16 September 1998     

Evolution of Base Composition and Codon Usage Bias in the Genus Flavivirus

The extent to which base composition and codon usage vary among RNA viruses, and the possible causes of this bias, is undetermined in most cases. A maximum-likelihood statistical method was used to test whether base composition and codon usage bias covary with arthropod association in the genus Flavivirus, a major source of disease in humans and animals. Flaviviruses are transmitted by mosquitoes, by ticks, or directly between vertebrate hosts. Those viruses associated with ticks were found to have a significantly lower G+C content than non-vector-borne flaviviruses and this difference was present throughout the genome at all amino acids and codon positions. In contrast, mosquito-borne viruses had an intermediate G+C content which was not significantly different from those of the other two groups. In addition, biases in dinucleotide and codon usage that were independent of base composition were detected in all flaviviruses, but these did not covary with arthropod association. However, the overall effect of these biases was slight, suggesting only weak selection at synonymous sites. A preliminary analysis of base composition, codon usage, and vector specificity in other RNA virus families also revealed a possible association between base composition and vector specificity, although with biases different from those seen in the Flavivirus genus. Received: 29 August 2000 / Accepted: 19 December 2000     

Complex Evolution of Tandem-Repetitive DNA in the Chironomus thummi Species Group

The subspecies Chironomus thummi thummi and C. t. piger display dramatic differences in the copy number and chromosomal localization of a tandemly repeated DNA family (Cla elements). In order to analyze the evolutionary dynamics of this repeat family, we studied the organization of Cla elements in the related outgroup species C. luridus. We find three different patterns of Cla element organization in C. luridus, showing that Cla elements may be either strictly tandem-repetitive or be an integral part of two higher-order tandem repeats (i.e., Hinf[lur] elements, Sal[lur] elements). All three types of Cla-related repeats are localized in the centromeres of C. luridus chromosomes. This suggests that the dispersed chromosomal localization of Cla elements in C. t. thummi may be the result of an amplification and transposition during evolution of this subspecies. Received: 22 May 1996 / Accepted: 8 October 1996     

Evolution of trophic types in emperor fishes ( Lethrinus,Lethrinidae, Percoidei) based on cytochrome B gene sequence variation

Three trophic categories exist within emperor fishes, genus Lethrinus, relating to body form and dentition type. One group contains low-bodied, high speed, stalking predators with conical teeth. Another group comprises high-bodied, slow speed carnivores with molariform teeth capable of crushing hard-shelled benthic prey. A third group is also high-bodied but with conical teeth feeding mostly on small or soft-shelled benthic prey. Inferring the evolution of these trophic types within Lethrinus using morphology is problematic since these characters are typically correlated with feeding mode and are potentially homoplasious. We use mitochondrial DNA sequences, to independently determine a phylogenetic hypothesis for Lethrinus, which are not dependent on morphological characters relating to trophic categories. We analyzed complete cytochrome b gene sequences (1140 bp) for 20 species of Lethrinus, representing the three trophic types, and for 13 outgroup species, including four other representatives of the Lethrinidae. A monophyletic Lethrinidae did not resolve, but the monophyly of Lethrinus is well supported. In addition, two major clades within Lethrinus are well supported. One of these clades exclusively contains low-bodied species with conical teeth while the other clade only comprises the high-bodied species with molariform teeth. A high-bodied species with conical teeth, Lethrinus miniatus, appears most ancestral and sister to all other Lethrinus species. We hypothesize that this generalist trophic type was the evolutionary precursor to both of the other primary trophic types.     

Concordant Evolution of a Symbiont with Its Host Insect Species: Molecular Phylogeny of Genus Glossina and Its Bacteriome-Associated Endosymbiont, Wigglesworthia glossinidia

Many arthropods with restricted diets rely on symbiotic associations for full nutrition and fecundity. Tsetse flies (Diptera: Glossinidae) harbor three symbiotic organisms in addition to the parasitic African trypanosomes they transmit. Two of these microorganisms reside in different gut cells, while the third organism is harbored in reproductive tissues and belongs to the genus Wolbachia. The primary symbiont (genus Wigglesworthia glossinidia) lives in differentiated epithelial cells (bacteriocytes) which form an organ (bacteriome) in the anterior gut, while the secondary (S) symbionts are present in midgut cells. Here we have characterized the phylogeny of Wigglesworthia based on their 16S rDNA sequence analysis from eight species representing the three subgenera of Glossina: Austenina (=fusca group), Nemorhina (=palpalis group), and Glossina (=morsitans group). Independently, the ribosomal DNA internal transcribed spacer-2 (ITS-2) regions from these species were analyzed. The analysis of Wigglesworthia indicated that they form a distinct lineage in the γ subdivision of Proteobacteria and display concordance with their host insect species. The trees generated by parsimony confirmed the monophyletic taxonomic placement of Glossina, where fusca group species formed the deepest branch followed by morsitans and palpalis groups, respectively. The placement of the species Glossina austeni by both the traditional morphological and biochemical criteria has been controversial. Results presented here, based on both the ITS-2 and the symbiont 16S rDNA sequence analysis, suggest that Glossina austeni should be placed into a separate fourth subgenus, Machadomyia, which forms a sister-group relationship with the morsitans group species. Received: 17 March 1998 / Accepted: 1 May 1998     

Episodic Evolution of Protein Hormones: Molecular Evolution of Pituitary Prolactin

Previous studies have shown that pituitary growth hormone displays an episodic pattern of evolution, with a slow underlying evolutionary rate and occasional sustained bursts of rapid change. The present study establishes that pituitary prolactin shows a similar pattern. During much of tetrapod evolution the sequence of prolactin has been strongly conserved, showing a slow basal rate of change (approx 0.27 × 10⁹ substitutions/amino acid site/year). This rate has increased substantially (∼12- to 38-fold) on at least four occasions during eutherian evolution, during the evolution of primates, artiodactyls, rodents, and elephants. That these increases are real and not a consequence of inadvertant comparison of paralogous genes is shown (for at least the first three groups) by the fact that they are confined to mature protein coding sequence and not apparent in sequences coding for signal peptides or when synonymous substitutions are examined. Sequences of teleost prolactins differ markedly from those of tetrapods and lungfish, but during the course of teleost evolution the rate of change of prolactin has been less variable than that of growth hormone. It is concluded that the evolutionary pattern seen for prolactin shows long periods of near-stasis interrupted by occasional bursts of rapid change, resembling the pattern seen for growth hormone in general but not in detail. The most likely basis for these bursts appears to be adaptive evolution though the biological changes involved are relatively small. Received: 31 August 1999 / Accepted: 9 February 2000     

Episodic Evolution of Protein Hormones in Mammals

Pituitary growth hormone (GH) and prolactin have been shown previously to display a pattern of evolution in which episodes of rapid change are imposed on a low underlying basal rate (near-stasis). This study was designed to explore whether a similar pattern is seen in the evolution of other protein hormones in mammals. Seven protein hormones were examined (with the common α-subunit of the glycoprotein hormones providing an additional polypeptide for analysis)—those for which sequences from at least four eutherian orders are available with a suitable non-eutherian outgroup. Six of these (GH, prolactin, insulin, parathyroid hormone, glycoprotein hormone α-subunit, and luteinizing hormone β-subunit) showed markedly variable evolutionary rates in each case with a pattern of a slow basal rate and bursts of rapid change, the precise positions of the bursts varying from protein to protein. Two protein hormones (follicle-stimulating hormone β-subunit and thyroid-stimulating hormone β-subunit) showed no significant rate variation. Based on the sequences currently available, and pooling data from all eight proteins, the phase of slow basal change occupied about 85% of the sampled evolutionary time, but most evolutionary change (about 62% of the substitutions accepted) occurred during the episodes of rapid change. It is concluded that, in mammals at least, a pattern of prolonged periods of near-stasis with occasional episodes of rapid change provides a better model of evolutionary change for protein hormones than the one of constant evolutionary rates that is commonly favored. The mechanisms underlying this episodic evolution are not yet clear, and it may be that they vary from one group to another; in some cases, positive selection appears to underlie bursts of rapid change. Where gene duplication is associated with a period of accelerated evolution this often occurs at the end rather than the beginning of the episode. To what extent the type of pattern seen for protein hormones can be extended to other proteins remains to be established. Received: 10 October 2000 / Accepted: 18 December 2000     

The Length Distribution of Perfect Dimer Repetitive DNA Is Consistent with Its Evolution by an Unbiased Single-Step Mutation Process

We have examined the length distribution of perfect dimer repeats, where perfect means uninterrupted by any other base, using data from GenBank on primates and rodents. Virtually no lengths greater than 30 repeats are found, except for rodent AG repeats, which extend to 35. Comparable numbers of long AC and AG repeats suggest that they have not been selected for special functions or DNA structures. We have compared the data with predictions of two models: (1) a Bernoulli Model in which bases are assumed equally likely and distributed at random and (2) an Unbiased Random Walk Model (URWM) in which repeats are permitted to change length by plus or minus one unit, with equal probabilities, and in which base substitutions are allowed to destroy long perfect repeats, producing two shorter perfect repeats. The source of repeats is assumed to be from single base substutions from neighboring sequences, i.e., those differing from the perfect repeat by a single base. Mutation rates either independent of repeat length or proportional to length were considered. An upper limit to the lengths L≈ 30 is assumed and isolated dimers are assumed unable to expand, so that there are absorbing barriers to the random walk at lengths 1 and L+ 1, and a steady state of lengths is reached. With these assumptions and estimated values for the rates of length mutation and base substitution, reasonable agreement is found with the data for lengths > 5 repeats. Shorter repeats, of lengths ≤ 3 are in general agreement with the Bernoulli Model. By reducing the rate of length mutations for n≤ 5, it is possible to obtain reasonable agreement with the full range of data. For these reduced rates, the times between length mutations become comparable to those suggested for a bottleneck in the evolution of Homo sapiens, which may be the reason for low heterozygosity of short repeats.     

Sociality and the Rate of rDNA Sequence Evolution in Wasps (Vespidae) and Honeybees (Apis)

Sequence data of mitochondrial 16S ribosomal DNA (mt-rDNA) and nuclear 28S ribosomal DNA (nuc-rDNA) were compared in two honeybee species (Apis mellifera and Apis dorsata) and a selection of 22 wasp species (Vespidae) with different levels of sociality. The averge substitution rates in mt-rDNA and nuc-rDNA were almost-equal in solitary species. In species with larger nests, however, the difference between the nuclear and the mitochondrial substitution rate significantly increased. The average substitution ratio, ψ (nucleotide substitutions in mt-rDNA/nucleotide substitutions in nuc-rDNA) was 1.48 ± 0.12 (SE) among the solitary Eumeninae, 3.70 ± 0.15 among five primitive social Stenogastrinae species, 3.24 ± 0.20 among five Polistinae species, 5.76 ± 0.33 among nine highly eusocial Vespinae, and 12.7 in the two Apis species. The high egg-laying rate and the effective population size skew between the sexes may contribute to the rise of the substitution ratio in the highly eusocial species. Drift and bottleneck effects in the mitochondrial DNA pool during speciation events as well as polyandry may further enhance this phenomenon. Received: 12 January 1998 / Accepted: 28 April 1998