MRC Human Genetics Unit, Western General Hospital, Crewe Road, Edinburgh EH4 2XU, UK
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: AZF/Dazl1/Rbm/spermatogenesis/Y chromosome
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Several genes have been mapped into each of the AZF intervals (Lahn and Page, 1997) but for most of them their individual contribution to the AZF phenotype remains to be determined. However, biochemical characterization of some gene products and functional studies of different animal homologues led to the assignment of at least one strong candidate gene for each microdeletion interval, namely DFFRY (Brown et al., 1998
) for AZFa, RBM (Ma et al., 1993
; Elliott et al., 1997
) for AZFb, and DAZ (Reijo et al., 1995
) for AZFc, all of which may cause azoospermia if deleted.
Despite a growing number of screening programmes for deletions in infertile men involving the AZF regions, the link between RBM and DAZ deletions and impaired spermatogenesis is still formally unproven. All microdeletions described so far encompass a substantial amount of DNA and there is evidence that genes other than RBM or DAZ might also be affected by the deletions (Lahn and Page, 1997; Wong et al., 1999
). Moreover, deletions within the AZFc region that apparently do not include the DAZ gene family have been reported in azoospermic individuals (Najmabadi et al., 1996
; Stuppia et al., 1997
; Foresta et al., 1997
).
Proof of a causative link between RBM and DAZ genes and impaired spermatogenesis would be the detection of point mutations, small intragenic deletions or rearrangements in these genes, which should give rise to the same phenotype as observed in men with microdeletions of Yq11. This goal has not been achieved to date but is of importance as concerns arise about the functionality of the DAZ genes in general (Vereb et al., 1997; Agulnik et al., 1998
).
The individual features of these genes have hampered the investigation of the controversial role of RBM and DAZ as AZF: RBM is a multigene family of at least 2040 micro-heterogeneous members on the Y chromosome. While a minimal critical region for RBM expression in AZFb has been determined (Elliott et al., 1997), it is difficult to assess whether deletions or sequence variations affect `the' essential gene/genes.
The situation for DAZ is even more complicated as the micro-heterogeneous multicopy organization on the Y comes together with a highly related homologous gene on chromosome 3, DAZL1 (Saxena et al., 1996), whose function in relation to DAZ is poorly understood.
Two approaches have been used to make genotype-phenotype correlations of the candidate genes for AZF: one is to investigate the biochemical properties of RBM and DAZ, which both seem to be involved in RNA metabolism (Houston et al., 1998; Elliott et al., 1998
), the other is to address this question in model organisms that are more accessible to study defined alterations of the genes. DAZ homologues of Drosophila (boule) (Eberhart et al., 1996
), Xenopus (Xdazl) (Houston et al., 1998
), and C. elegans (dazl1) (Karashima et al., 1997
) have all been shown to function in male and/or female germline-specific processes. The mouse is, however, the only animal model in which both genes have been investigated. The conserved Y chromosomal location of Rbm (Elliott et al., 1996
) has hampered a simple knock-out experiment, but asymmetrical pairing and exchange of Yp-Sxra has produced a sex reversed mouse line with defined deletions of the 2550 Rbm copies (Capel et al., 1993
; Laval et al., 1995
). In one of these lines, namely Yd1, the number of residual Rbm genes was estimated as 24. The introduction of an Sry transgene into this line produced males with only low level transcription of Rbm paired with an increased proportion of abnormal sperm heads. The sperm count and the overall fertility of these animals was unaffected (Mahadevaiah et al., 1998
). These authors attribute the observed sperm head abnormalities to the lack of Rbm function. However, since the deletion interval of the Yd1 chromosome is estimated to encompass 34 Mb, disruption of genes other than Rbm may also be involved in the Yd1 phenotype.
The human DAZ genes are a recent acquisition to the Y chromosome and are only found in humans and old world monkeys. The murine homologue Dazl1 (Reijo et al., 1996; Cooke et al., 1996
) however is located on chromosome 17, and readily accessible for knock-out experiments. Heterozygote DazlTm1Hgu/+ animals have a reduced sperm count and show minor sperm tail defects, but are fertile. The reproductive organs of homozygous males and females are devoid of meiotic germ cells and DazlTm1Hgu/DazlTm1Hgu animals are therefore infertile (Ruggiu et al., 1997
).
Since both DAZ and RBM are multigene families in man and since deletions of the Y chromosome can include both DAZ and RBM genes simultaneously and to varying extents (Pryor et al., 1997; Elliott et al., 1997
), we reasoned that a part of the variability in phenotype might be due to interaction between the effects of deleting the two genes and/or other genes located in the vicinity. To mimic this in the mouse we crossed the Yd1 chromosome on a DazlTm1Hgu/+ background to investigate the resulting phenotype. Since a Y chromosomal localization of RBM is conserved in mammals, it is likely that genes interspersing the RBM family in humans might also be found in similar location in mouse. Hence, although the Yd1 deletion is not fully characterized with respect to affected genes other than Rbm, the same genes might be removed by human AZF deletions. If Dazl1 and Rbm (or other gene products) are acting in different molecular pathways these mutations might be expected to have additive or synergistic effects which might be expected to lead to infertility, whereas if acting in the same pathway the effect of combining the mutations might be minimal.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Sperm analysis
Sperm counts per epididymis were carried out as previously described (Searle and Beechey, 1974). Spermatozoa for morphology analysis were released from the epididymis into phosphate-buffered saline, the suspension smeared onto clean slides and allowed to dry. Slides were then fixed in methanol for 10 min at room temperature (RT) and air-dried. Preparations were silver-stained (Howell and Black, 1980
) and counter stained with 2% Giemsa (pH 6.8) for 5 min at RT. Slides were coded and scored blind to genotype. Various schemes for classifying sperm head morphology have been reported (Krzanowska, 1976
; Mahadevaiah et al., 1998
). We most closely followed that of Mahadevaiah et al. (1998). Several additional minor categories observed in DazlTm1Hgu/± and Yd1 mice have been included (Table I
). Only complete spermatozoa with head and tail attached were scored to ensure consistency between scored numbers of different animals and to rule out preparation artefacts. Tail abnormalities were scored according to our own classification.
|
Statistical analysis
The results were analysed using Minitab12. General linear model (GLM) analysis of variance assumed a 2x2 cross-classification with the mean squares adjusted to allow for unequal replications of each genotype. Standard errors appropriate to each class shown in Tables I and II were calculated from the mean square errors in the analyses of variance and are employed in t-tests (with 7 degrees of freedom) of contrasts between the four classes. Homogeneity of variance tests (Bartlett's and Levine's) indicated that data transformation was only necessary for the `grossly abnormal' percentages of sperm heads. These were analysed using a natural log transformation, hence standard errors are not shown in Table I
.
|
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Although the DazlTm1Hgu/+ males are fertile (Ruggiu et al., 1997), the sperm count was considerably reduced, being only 45.9% (n = 3) of normal. Sperm counts for the Yd1 males were slightly impaired (Mahadevaiah et al., 1998
) having a mean count that was 81.5% of normal (n = 3). The combination of heterozygous Dazl1 and Yd1 produced the most marked effect, the mean sperm count only being 16.6% of normal (n = 4). There was variation in this group, one animal resembled the Dazl1 heterozygotes, two were extremely low at 5.2% of normal (n = 2), and one was intermediate between them.
The numbers of spermatozoa from the DazlTm1Hgu/+ Yd1 mice were significantly less than both Dazl+/+ Yd1 and wild type Y (P < 0.0001) and also than the DazlTm1Hgu/+ with wild type Y (P < 0.01). Statistical analysis of variance showed that the interaction between the effects of deletion and heterozygosity were negligible (P = 0.485). The depression in the sperm count for heterozygous Y-deleted males could be adequately explained by a main effect of heterozygosity added to a smaller main effect of Y deletion. In biological terms, this means that the effects of the two mutations are not synergistic.
Sperm morphology
Heads
Two features are important for assessing head morphology. Firstly the insertion of the mid-piece into the head under normal circumstances should have a clear stepped appearance (Table I). Secondly the overall shape of the sperm head must be considered. Table I
shows two examples of each of the varying types of sperm head identified. It is apparent that within such subjective categories minor variations are common. The grossly abnormal sperm heads in categories 26, where both assessment factors are changed, present few problems in classification. The changes in categories 11d are more subtle, and can be described as:
1: Normal; 1a: Mid-piece insertion normal, head showing a flattened frontal curvature; 1b: Mid-piece insertion normal, head with a rear projection above the mid-piece; 1c: Mid-piece insertion normal, curved hook of spermatozoa malformed; 1d: Mid-piece insertion abnormal, head shortened with very angular outline, hook not fully formed.
The numbers observed in each category for each genotype are shown in Table I.
Categories 1c and 1d were observed predominantly in Yd1 and DazlTm1Hgu/+ Yd1 males.
It has previously been shown (Bruce et al., 1974) that sperm abnormalities can vary between 1 and 15% depending on strain. The figure of 2.5% for our control was at the lower limit. Analysis of variance showed very strong evidence (P = 0.00002) that there were fewer normal heads in the Yd1 groups than in the wild type Y-carrying groups (Table I
). This was also the case for heterozygous Dazl1 males compared to males with two wild type Dazl1 alleles (P = 0.001). However, there is also considerable non-additivity since the two wild type Y-carrying groups do not differ significantly where there is strong evidence (P = 0.00004) that DazlTm1Hgu/+ Yd1 has far fewer normal heads than +/+ Yd1 animals. The DazlTm1Hgu/+ males exhibit a mean total of 11.2% abnormalities, Yd1 males 30.8% and DazlTm1Hgu/+ Yd1 males 73.8%. The increased level in Yd1 over DazlTm1Hgu/+ is due to the proportion of slight head abnormalities, the class 1a and 1b being far more frequent in the former. The elevated levels in the DazlTm1Hgu/+ Yd1 males reflect increased numbers of all classes of grossly abnormal spermatozoa, particularly the class 4 (tulip-headed spermatozoa).
Tails
Sperm tails can be divided into two regions, the mid-piece and the flagellum. Primary defects in each region such as multiple tails and bent or coiled tails may be due to defects in the structure of organelles such as the internal fibres or the outer sheath, and such spermatozoa have generally been considered incapable of fertilization. Multiple tails or mid-pieces (Figure 1ac) were not seen in the control, they occurred at low levels in both the DazlTm1Hgu/+ (0.7%) and Yd1 (0.2%) males, and while still at low levels were most frequent in the DazlTm1Hgu/+ Yd1 males (2.1%) (Table III
). Bent and coiled tails (Figure 1df
) were most frequent in the DazlTm1Hgu/+ (41.1%, n = 3). They were also present in the Yd1 males but about four-fold lower in frequency (10.1%, n = 3). Unlike sperm head abnormalities the frequency of tail defects in the DazlTm1Hgu/+ Yd1 was not significantly increased in comparison with DazlTm1Hgu/+ males (31.3%, n = 4) (Table III
). However in these males the defects were qualitatively more severe with extreme coiling and an apparent self-adhesion of the mid-piece and flagella (Figure 1gl
). Minor structural defects were also seen in areas of both the mid-piece and flagella (Figure 1jl
), these often splitting into two sub-components.
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We have shown that the Dazl1 heterozygote phenotype is a reduction of sperm output and tail abnormalities mainly affecting the mid-piece. The replacement of a normal Y with the Yd1 chromosome, accompanied by the induction of sex reversal through an Sry transgene, leads to an increased production of spermatozoa with abnormal heads. Although our numbers for Yd1 related head abnormalities differed slightly from the figures reported by Mahadevaiah et al. (1998), the overall picture is the same. Mahadevaiah and colleagues' report is based on spermatozoa entering the epididymis. In contrast, we isolated spermatozoa from the entire epididymis. These different sources of spermatozoa might account for the differing values of abnormal spermatozoa obtained rather than age dependent variations.
Apart from the assigned predominant phenotypes we observed a small overlap in the phenotypes of the mutations, i.e. the number of abnormal sperm heads was slightly increased in DazlTm1Hgu/+ compared to wild type animals, and a small proportion of spermatozoa in Yd1-carrying animals exhibited tail defects. Yd1 animals have also a lower sperm count than wild type animals.
These slightly overlapping phenotypes were enhanced when the Yd1 chromosome was crossed onto a DazlTm1Hgu/+ background resulting in a significant drop of the sperm count, an increase of abnormal sperm heads and more prominent mid-piece defects of the tails. These data suggest a model in which Rbm genes (or in the case of the Yd1 chromosome all genes in the deletion), and Dazl1, affect at least three different phenotypic pathways in mouse spermatogenesis. It is difficult to assess whether both genes act in the same or in different molecular pathways. However, since all observed phenotypes seem to be more severe in DazlTm1Hgu/+ Yd1 animals, this suggests an additive or synergistic effect, and we assume that Rbm and Dazl1 are mainly acting in different molecular pathways. Since the phenotypes of both mutations on their own overlap only slightly, we conclude that the contribution of each gene to the main pathway of the other is only small.
Sperm head abnormalities and more strikingly sperm output vary between individual animals in DazlTm1Hgu/+ Yd1 mice, whereas little variation is seen in all other genotypes studied here. It has been reported that Y chromosomes from different genetic backgrounds can affect the total percentage of sperm head abnormalities in inbred mouse strains (Krzanowska, 1976). An increase in specific types of abnormality has not been observed in the study of Krzanowska, in contrast to the situation observed in DazlTm1Hgu/+ Yd1 animals. In these mice, the increase of specific abnormalities, e.g. the tulip-shaped class 4, is therefore not an effect of different genetic backgrounds, but is caused by reduced activity of the affected Y chromosomal genes and Dazl1.
A variant penetration of the phenotype in DazlTm1Hgu/+ Yd1 was also observed in testis sections. Some animals, i.e. number 8, were only slightly affected by the reduction of functional Dazl1 and Rbm genes and exhibited a mild phenotype that resembled a DazlTm1Hgu/+ phenotype. The tubules of these animals had germ cells of all stages and a complete depletion of germ cells was never observed. In contrast, very badly affected animals (numbers 10 and 11) showed only few tubules with complete spermatogenesis but predominantly tubules with arrested pachytene stages or Sertoli cells only. Fertilization ability of these most severely affected animals has not been investigated.
The variation in severity of the double mutant phenotype could also be caused by strain-dependent segregation of different modifier genes. Although initially generated by modifying a C129-derived allele, the Dazl1 null allele was maintained on a random bred MF1 strain. Crossing of this null allele in different inbred strains did not result in a variable phenotype (data not shown). Thus the Dazl1 null phenotype seems not to be susceptible to the action of modifying genes. Variations of sperm output in different animals has not been observed for the Yd1 chromosome, which was also crossed onto the MF1 background (Mahadevaiah et al., 1998). However, involvement of modifying genes cannot be ruled out to contribute to the observed individual variations in DazlTm1Hgu/+Yd1 males.
The variable penetration of the AZF phenotype is also apparent in human patients with deletions in the AZFc, ranging from Sertoli cell only to meiotic or spermatid maturation arrests that lead to absence of mature (azoospermia) or only few spermatozoa (oligozoospermia). In some cases increased rates of dysmorphic sperm heads (oligoasthenoteratozoospermia) have also been observed (Vogt et al., 1996).
This variability might point to an involvement of other factors apart from the single candidate gene DAZ, influencing the severity of the AZFc phenotype in humans. Those factors can be environmental, epigenetic or genetic in nature.
The data presented in this study, however, show that the variable AZFc phenotype in humans is mimicked in the mouse by copy number reduction of Dazl1 and Rbm, and/or additional Y chromosomal genes. Assuming similar functions during spermatogenesis for Rbm and Dazl1 in humans and mouse, this observation provokes the question as to whether RBM is also involved in the AZFc phenotype. According to the data presented by Elliott and colleagues (1997) RBM expression from the minimal critical region in AZFb is unaffected in the AZFc patients studied, excluding a direct involvement of RBM itself. Another possible explanation would be the involvement of other genes that are localized in the vicinity of or in the murine Rbm cluster itself. Hence, the similar phenotype observed in this study would not reflect an effect mediated by the copy number reduction of Rbm but by the removal or silencing of those different genes, which might on the other hand be localized in the AZFc region on the human Y. Deletions in AZFc also reduce the distance between the heterochromatic region of the Y chromosome and active genes. Hence, position effects evoking silencing of active genes to various extents might contribute to the variability of the AZFc phenotype in some individuals.
In summary, this study has shown firstly that Dazl1 and Rbm and/or other genes localized in the Yd1 deletion interval have multiple targets that contribute to different areas of the establishment and progression of spermatogenesis. Secondly, that a reduction of active gene product of those genes gives rise to a phenotype similar to the heterogeneous AZFc phenotype observed in humans.
![]() |
Acknowledgments |
---|
![]() |
Notes |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Brown, G.M., Furlong, R.A., Sargent, C.A. et al. (1998) Characterisation of the coding sequence and fine mapping of the human DFFRY gene and comparative expression analysis and manning to the Sxr(b) interval of the mouse Y chromosome of the Dffry gene. Hum. Mol. Genet., 7, 97107.
Bruce, W.R., Furrer, R. and Wyrobek, A.J. (1974) Abnormalities in the shape of murine sperm after acute testicular x-irradiation. Mutat. Res., 23, 381386.[ISI][Medline]
Capel, B., Rasberry, C., Dyson, J. et al. (1993) Deletion of Y chromosome sequences located outside the testis determining region can cause XY female sex reversal. Nature Genet., 5, 301307.[ISI][Medline]
Cooke, H.J., Lee, M., Kerr, S. et al. (1996) A murine homolog of the human DAZ gene is autosomal and expressed only in male and female gonads. Hum. Mol. Genet., 5, 513516.
Eberhart, C.G., Maines, J.Z. and Wasserman, S.A. (1996) Meiotic cell-cycle requirement for a fly homolog of human deleted in azoospermia. Nature, 381, 783785.[ISI][Medline]
Elliott, D.J., Ma, K., Kerr, S.M. et al. (1996) An RBM homolog maps to the mouse Y-chromosome and is expressed in germ-cells. Hum. Mol. Genet., 5, 869874.
Elliott, D.J., Millar, M.R., Oghene, K. et al. (1997) Expression of RBM in the nuclei of human germ cells is dependent on a critical region of the Y chromosome long arm. Proc. Natl Acad. Sci. USA, 94, 38483853.
Elliott, D.J., Oghene, K., Makarov, G. et al. (1998) Dynamic changes in the subnuclear organisation of pre-mRNA splicing proteins and RBM during human germ cell development. J. Cell Science, 111, 12551265.
Foresta, C., Ferlin, A., Garolla, A. et al. (1997) Y-chromosome deletions in idiopathic severe testiculopathies. J. Clin. Endocrinol. Metabol., 82, 10751080.
Houston, D.W., Zhang, J., Maines, J.Z. et al. (1998) A Xenopus DAZ-like gene encodes an RNA component of germ plasm and is a functional homologue of Drosophila boule. Development, 125, 171180.
Howell, W.M. and Black, D.A. (1980) Controlled silver-staining of nucleolus organizer regions with a protective colloidal developer: a 1-step method. Experientia, 36, 10141015.[ISI][Medline]
Karashima, T., Sugimoto, A. and Yamamoto, M.A. (1997) C.elegans homologue of DAZ/boule is involved in progression through meiosis during oogenesis. Worm Breeders Gazette, 15, 6566.
Krausz, C. and McElreavey, K. (1999) Y chromosome and male infertility. Front. Biosci., 4, E1-E8.[Medline]
Krzanowska, H. (1976) Inheritance of sperm head abnormality types in mice the role of the Y chromosome. Genet. Res., 28, 189198.[ISI][Medline]
Lahn, B.T. and Page, D.C. (1997) Functional coherence of the human Y chromosome. Science, 278, 675680.
Laval, S.H., Glenister, P.H., Rasberry, C. et al. (1995) Y chromosome short arm-Sxr recombination in XSxr/Y males causes deletion of Rbm and XY female sex reversal. Proc. Natl Acad. Sci. USA, 92, 1040310407.[Abstract]
Ma, K., Inglis, J.D., Sharkey, A. et al. (1993) A Y-chromosome gene family with RNA-binding protein homology candidates for the azoospermia factor AZF controlling human spermatogenesis. Cell, 75, 12871295.[ISI][Medline]
MacGregor, G.R., Russell, L.D., Van Beek, M.E. et al. (1990) Symplastic spermatids (sys): a recessive insertional mutation in mice causing a defect in spermatogenesis. Proc. Natl Acad. Sci. USA, 87, 50165020.[Abstract]
Mahadevaiah, S.K., Odorisio, T., Elliott, D.J. et al. (1998) Mouse homologues of the human AZF candidate gene RBM are expressed in spermatogonia and spermatids, and map to a Y chromosome deletion interval associated with a high incidence of sperm abnormalities. Hum. Mol. Genet., 7, 715727.
Najmabadi, H., Huang, V., Yen, P. et al. (1996) Substantial prevalence of microdeletions of the Y-chromosome in infertile men with idiopathic azoospermia and oligozoospermia detected using a sequence-tagged site-based mapping strategy. J. Clin. Endocrinol. Metab., 81, 13471352.[Abstract]
Odorisio, T., Rodriguez, T.A., Evans, E.P. et al. (1998) The meiotic checkpoint monitoring synapsis eliminates spermatocytes via p53-independent apoptosis. Nature Genet., 18, 257261.[ISI][Medline]
Pryor, J.L., Kent-First, M., Muallem, A. et al. (1997) Microdeletions in the Y chromosome of infertile men. N. Engl. J. Med., 336, 534539.
Reijo, R., Lee, T.Y., Salo, P. et al. (1995) Diverse spermatogenic defects in humans caused by Y chromosome deletions encompassing a novel RNA-binding protein gene. Nature Genet., 10, 383393.[ISI][Medline]
Reijo, R., Seligman, J., Dinulos, M.B. et al. (1996) Mouse autosomal homolog of DAZ, a candidate male-sterility gene in humans, is expressed in male germ-cells before and after puberty. Genomics, 35, 346352.[ISI][Medline]
Ross, A.J., Waymire, K.G., Moss, J.E. et al. (1998) Testicular degeneration in Bclw-deficient mice. Nature Genet., 18, 251256.[ISI][Medline]
Ruggiu, M., Speed, R., Taggart, M. et al. (1997) The mouse Dazla gene encodes a cytoplasmic protein essential for gametogenesis. Nature, 389, 7377.[ISI][Medline]
Russell, L.D. et al. (1990) Histological and Histopathological Evaluation of the Testis. Cache River Press, Clearwater, USA.
Saxena, R., Brown, L.G., Hawkins, T. et al. (1996) The DAZ gene cluster on the human Y chromosome arose from an autosomal gene that was transposed, repeatedly amplified and pruned. Nature Genet., 14, 292298.[ISI][Medline]
Searle, A.G. and Beechey, C.V. (1974) Sperm-count, egg-fertilization and dominant lethality after X-irradiation of mice. Mutat. Res., 22, 6372.[ISI][Medline]
Stuppia, L., Gatta, V., Mastroprimiano, G. et al. (1997) Clustering of Y chromosome deletions in subinterval E of interval 6 supports the existence of an oligozoospermia critical region outside the DAZ gene. J. Med. Genet., 34, 881883.[Abstract]
Tiepolo, L. and Zuffardi, O. (1976) Localisation of factors controlling spermatogenesis in the non fluorescent portion of the human Y chromosome long arm. Hum. Genet., 34, 119124.[ISI][Medline]
Vereb, M., Agulnik, A.I., Houston, J.T. et al. (1997) Absence of DAZ gene mutations in cases of non-obstructed azoospermia. Mol. Hum. Reprod., 3, 5559.[Abstract]
Vogt, P.H. (1998) Human chromosome deletions in Yq11, AZF candidate genes and male infertility: history and update. Mol. Hum. Reprod., 4, 739744.[Abstract]
Vogt, P.H., Edelmann, A., Kirsch, S. et al. (1996) Human Y-chromosome azoospermia factors (AZF) mapped to different subregions in Yq11. Hum. Mol. Genet., 5, 933943.
Wong, J., Blanco, P. and Affara, N.A. (1999) An exon map of the AZFc male infertility region of the human Y chromosome. Mamm. Genome, 10, 5761.[ISI][Medline]
Submitted on May 17, 1999; accepted on September 10, 1999.