1 Fertilitetscentrum, 402 29 Göteborg, Sweden and 2 Department of Obstetrics and Gynaecology and Reproductive Medicine Unit, University of Adelaide, Adelaide 5005, Australia
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: blastocyst/culture/embryo/GM-CSF/growth factor
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Studies in rodents and livestock species suggest that the growth and development of the preimplantation embryo is regulated by an array of cytokines and growth factors secreted from epithelial cells lining the oviduct and uterus (Pampfer et al., 1991; Robertson et al., 1994
). Their synthesis occurs in precise, differentially regulated spatial and temporal patterns driven predominantly by ovarian steroid hormones but also by factors in seminal plasma (Robertson et al., 1996
; Tremellen et al., 1998
). Embryos of many species, including human, have been shown to express receptors for many of the growth factors secreted by the tract (Pampfer et al., 1991
; Sharkey et al., 1995
), and the addition of growth factors and cytokines to embryo culture medium has been shown to promote blastocyst development in the human (Dunglison et al., 1996
; Lighten et al., 1998
; Martin et al., 1998
) and in a number of mammalian species (reviewed by Kane et al., 1997).
Granulocytemacrophage colony-stimulating factor (GM-CSF) is a cytokine originally identified as a product of activated T-lymphocytes involved in the proliferation and differentiation of myeloid haematopoietic cells (Ruef and Coleman, 1990). GM-CSF is produced by oestrogen-primed epithelial cells in the oviduct and uterus in mice (Robertson et al., 1992
), sheep (Imakawa et al., 1993
) and women (Zhao and Chegini, 1994
; Giacomini et al., 1995
).
Animal studies suggest that GM-CSF can act as a survival factor for the developing embryo. Murine preimplantation embryos express the -chain of the GM-CSF receptor, and culture in recombinant GM-CSF-containing media has beneficial effects on murine embryo development to blastocyst stage and subsequently on their capacity to hatch from the zona pellucida and attach to the culture dish (Robertson et al., 1991
, 2000
). Recent studies have shown that genetically GM-CSF-deficient mice have retarded blastocyst formation with a significantly lower number of blastomeres, principally due to a diminished inner cell mass size. The effects on the null mice also include decreased fetal size and increased rates of fetal resorption during late gestation and mortality during early post-natal life (Robertson et al., 1999). The development of in-vitro produced bovine embryos is improved by addition of GM-CSF to culture media (de Moraes and Hansen, 1997
). Exposure of ovine embryos to GM-CSF in vitro increases their implantation potential through enhanced expression of the anti-luteotrophic signal interferon (IFN)-
(oTP-1) in trophectoderm cells (Imakawa et al., 1993
) .
Together these findings suggest that GM-CSF is a potential regulator of human preimplantation embryo development. The purpose of this study was to explore this possibility by investigating the effects of recombinant human GM-CSF on human embryo development in vitro.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Media and embryo culture
All media were from Scandinavian IVF Science AB, Göteborg, Sweden. Embryo culture was performed in IVF-50, S2, G1.2 and G2.2, and micro-drops were overlaid with Ovoil-150. Gamete-100, comprised of HEPES-buffered IVF-50, was used for differential staining. IVF-50 is a low glucose and low phosphate medium modified from the original human tubal fluid (HTF) formulation (Quinn et al., 1985). Media S2, G1.2 and G2.2 are modifications of the original G1 and G2 (Gardner, 1994
; Barnes et al., 1995
). Modifications to the original culture media formulations include reduced glutamine, EDTA and phosphate concentrations, the inclusion of specific vitamins in G2.2, and the use of human serum albumin rather than bovine serum albumin. S2 medium is a modification of the G2 medium, supplemented with human serum albumin. Apart from the insulin in S2, the media do not contain any protein or growth factors other than pharmaceutical grade human serum albumin. The precise formulations of these media are withheld by Scandinavian IVF Science for commercial reasons.
Ovarian stimulation and in-vitro fertilization
Patients received 300 µg buserelin gonadotrophin-releasing hormone agonist (GnRHa, Suprecur; Hoechst, Frankfurt, Germany) three times daily intranasally, starting 1 week before expected menses and lasting for 2 weeks. Down-regulation was confirmed by a serum oestradiol content of <0.2 nmol/l. Patients were then given recombinant follicle stimulating hormone (r-FSH; Gonal-F; Serono Laboratories, Aubonne, Switzerland; 150225 IU/day s.c.). The starting dose was dependent on the patient's age and/or previous response during ovarian stimulation (Wikland et al., 1994). The ovarian response was monitored by ultrasound and serum oestradiol concentration as previously described (Bergh et al., 1997
). GnRHa and rFSH were administered until there was at least one follicle >18 mm in mean diameter and two others
16 mm. Finally, oocyte maturation was triggered by one s.c. injection of 10 000 IU of human chorionic gonadotrophin (HCG, Profasi; Serono Laboratories).
Oocytes were retrieved 3638 h after HCG administration, assessed morphologically and fertilized in vitro. The embryos were cultured in IVF-50 and frozen on day 2 using a three-step propanediol cryopreservation kit (Freeze Kit 1, Scandinavian IVF Science) according to the manufacturer's instructions.
Recombinant GM-CSF
Two different commercial sources of recombinant human (rh)GM-CSF were used in these experiments. A laboratory grade preparation was obtained from R&D Systems Europe Ltd, Abingdon, Oxon, UK, and a pharmaceutical grade preparation, Molgramostim (Leucomax) was obtained from Schering & Plough, Madison, NJ, USA. The biological activity of both recombinant cytokine preparations was measured in a bioassay employing a GM-CSF responsive cell line (human myeloid TF-1 cell line), essentially as previously described (Kitamura et al., 1989). Duplicate serial 1:2 dilutions of both preparations were incubated with 2000 TF-1 cells in 200 µl of RPMI-1640 (GIBCO/BRL, Grand Island, NY, USA) supplemented with 10% fetal calf serum (FCS; Commonwealth Serum Laboratories, Parkville, Victoria, Australia), 5x105 mol/l ß-mercaptoethanol and antibiotics. After 2 days, cultures were pulsed with 1 µCi of [3H]thymidine (Amersham, Arlington Heights, IL, USA) for 6 h, harvested onto glass fibre paper using a PHD automated cell harvester (Cambridge Technology Inc., Cambridge, MA, USA) and radioactivity measured in a liquid scintillation beta counter.
Embryo thawing, allocation and culture
Frozen 24-cell embryos were thawed in four steps using a propanediol method for embryo thawing (Thaw Kit 1, Scandinavian IVF Science) following instructions given by the manufacturer. The viable embryos were classified and graded according to criteria listed in Table I. To avoid bias the embryos were randomly allocated, with regard to patient and embryo grade, into the different culture groups (Table II
). The embryos were cultured in groups of five embryos per drop. To avoid the toxic effects of ammonium, released due to metabolism and breakdown of amino acids, the culture medium was renewed every 48 h until hatching occurred.
|
|
In a third experiment the embryos were transferred from IVF-50 into S2 medium at the 68 cell stage. Additions of GM-CSF and carrier were the same as in the two previous experiments. When blastocysts were observed they were transferred to Falcon 3037 dishes, coated 24 h previously with Biomatrix EHS (Boehringer Ingelheim Bioproducts, Heidelberg, Germany). Developmental stage was scored every 8 h from thawing until 2300 h on day 8 (200 h post-insemination).
In a fourth experiment, embryo culture was performed in two different sequential media systems using two different commercial sources of rhGM-CSF. After thawing, the embryos were cultured first in G1.2 or IVF-50. At 68-cell stage the embryos were transferred into G2.2 or S2. The experiment included six groups: (i) G1.2/G2.2 alone; (ii) G1.2/G2.2 containing 2 ng/ml rhGM-CSF (R&D Systems); (iii) G1.2/G2.2 containing 2 ng/ml Molgramostim (diluted 1:75 000 from stock material); (iv) IVF-50/S2 alone; (v) IVF-50/S2 containing 2 ng/ml rhGM-CSF (R&D Systems); (vi) IVF-50/S2 containing 2 ng/ml Molgramostim. Developmental rate was scored every eighth hour until expanded blastocyst stage. Blastocysts were scored on day 5 at 120 h post-insemination according to criteria described previously (Dokras et al., 1993). Briefly, grade A blastocysts exhibited an expanded cavity with a distinct trophectoderm (TE) and an eccentrically located inner cell mass (ICM); grade B blastocysts were not yet expanded but otherwise morphologically identical to A; and grade C blastocysts exhibited poor morphology characterized by a number of degenerative foci in the ICM and TE and a poorly developed blastocoel cavity. Embryo scoring in each of the experiments was performed by the same person (C.S.).
Statistical analysis was performed using Fisher's exact test and independent samples t-test (StatSoft, Inc., Tulsa, OK, USA). Differences in data were considered significant when P < 0.05.
Differential labelling of blastocysts
Differential labelling was performed using a modification of a protocol described previously (Handyside and Hunter, 1984). Human blastocysts were cultured from excess embryos, surplus to treatment and freezing. On day 5 of culture (120124 h post-insemination) the zona was removed in acid Tyrode's solution containing 4 mg/ml polyvinylpyrrolidone (PVP; 360 000 mol. wt) and embryos were washed once in Gamete-100 and three times in albumin-free S2 containing 4 mg/ml PVP (S2-PVP). The blastocysts were incubated in trinitro-benzene sulphonic acid (TNBS, Sigma Chemical Co., St Louis, MO, USA; 10 mmol/l in S2-PVP pH 8.5, 4°C/20 min in the dark) and washed three times in Gamete-100. TNBS-treated blastocysts were incubated in anti-dinitrophenyl antibody (anti-DNP; Sigma, 0.2 mg/ml diluted in Gamete-100; 37°C/30 min). Embryos were then washed and incubated in guinea-pig complement serum (Sigma; diluted 1:10 in Gamete-100; 37°C/30 min). Embryos were washed again and labelled with fluorochromes (Sigma; 0.05 mmol/l bisbenzimide and 10 µg/ml propidium iodide in Gamete-100, 37°C/30 min). After extensive washing embryos were fixed briefly in 1% paraformaldehyde and 0.5% glutaraldehyde in PBS, mounted under coverslips in 20% glycerol in PBS and examined by fluorescence microscopy using a 400 nm excitation filter. Nuclei stained pink were scored as lysed trophectoderm cells (TE) and blue nuclei were scored as viable inner cell mass cells (ICM).
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
Figure 2 illustrates the effect of rhGM-CSF on blastocyst formation according to embryo grade at thawing. GM-CSF exerted a comparable effect in all grades of 24-cell embryos, with similar increases in the proportion of poor quality compared with good quality embryos forming blastocysts.
|
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
These results concur with previous findings in rodents, where increased developmental potential afforded to embryos by GM-CSF is associated with promotion of inner cell mass viability. Blastocysts recovered from GM-CSF null mice have a decreased total cell number and inner cell mass size compared to blastocysts from wild-type littermates, and culture of embryos in the presence of recombinant murine GM-CSF can alleviate the effect of GM-CSF deficiency (Robertson et al., 2000). These data together with the findings in this study suggest that GM-CSF promotes blastomere survival and/or proliferation, and that the effect is preferentially exerted in the inner cell mass. Although the molecular events that culminate in these effects are not fully understood, our recent experiments using mouse blastocysts show that GM-CSF can promote glucose uptake through binding to trophectoderm cells via the
-chain of the GM-CSF receptor (Robertson et al., 1999). Increased glucose uptake is known to stimulate metabolic activity, and might therefore promote cell division. It has been shown in other cells that cytokine withdrawal can induce apoptosis (Hetts, 1998
) and the diminished survival rate of embryos cultured in growth factor-free medium could be due to apoptosis resulting from metabolic starvation. Our recent findings (unpublished data) have shown that human blastocysts express the
-chain but not the ß-chain of the heterodimeric GM-CSF receptor, but whether improved glucose transport is stimulated by GM-CSF in human blastocysts remains to be determined.
The ability of GM-CSF to preserve inner cell mass viability is important since at least in other species the number of inner cell mass cells in the blastocyst at the time of implantation appears to be a key parameter influencing implantation success, and subsequent fetal size and health. Depletion of inner cell mass cells is characteristic of blastocysts generated in vitro, and this defect is correlated with abnormal fetal size late in gestation (Leese et al., 1999). In mice, the beneficial effect of insulin in in-vitro culture is mediated through its capacity to promote the survival and/or proliferation of inner cell mass cells (Kaye et al., 1992
), whereas cytokines with embryotoxic properties such as tumour necrosis factor (TNF)
appear to constrain embryo development by inhibiting the survival and/or proliferation of inner cell mass cells (Pampfer et al., 1995
).
A likely role for GM-CSF in the physiological development of the preimplantation embryo in the human reproductive tract is suggested by the temporal pattern of expression of GM-CSF in the oviduct and uterus, which coincides with the time of fertilization and early embryo development and implantation. Expression of GM-CSF in the human Fallopian tube has been shown to be dependent on the stage of the menstrual cycle, with peak expression occurring during the midlate proliferative and earlymid secretory phases (Zhao and Chegini, 1994). A similar pattern of cycle-related synthesis of GM-CSF occurs in the uterus, where GM-CSF secreted principally from endometrial epithelial cells is present throughout the cycle with a moderate increase during the late proliferative and early secretory phases (Giacomini et al., 1995
).
The presence of a large repertoire of different cytokines and growth factors in the reproductive tract during the time of preimplantation embryo development and implantation indicates that many different growth regulators may act in concert to orchestrate optimal embryo development. Other factors known to be expressed in the female reproductive tract have been reported to promote human embryo development. Addition of leukaemia inhibitory factor (LIF) to a complex serum-free human embryo culture media increased the blastulation rate from 18 to 44% (Dunglison et al., 1996). The effect of LIF was limited to increased blastocyst formation, since no effects beyond blastulation were observed. Culture of human embryos in the presence of heparin-binding epidermal growth factor (HB-EGF) also improves the proportion of embryos developing through to the blastocyst stage as well as their developmental competence as assessed by hatching, adherence to extracellular matrix proteins and trophectoderm outgrowth (Martin et al., 1998
). The extent of the effect was comparable to that reported for GM-CSF in the current study, with blastulation rates increased from 41% in the control group to 71% in the presence of HB-EGF. In contrast to the GM-CSF, however, HB-EGF did not improve the developmental rate or cell number and its effect was limited to good quality embryos. Addition of insulin-like growth factor (IGF)-1 to in-vitro culture media has also been shown to benefit human embryo development (Lighten et al., 1998
), increasing the blastulation rate from 35% in the control to 60% in the treatment group. The effect of IGF-I also included an increase in blastocyst cell number, due entirely to an increase of cells in the inner cell mass. It is not known whether the mechanisms underlying the effects of all of these growth factors are related, but it could be speculated that each has influence on the metabolic activity of inner cell mass and/or trophectoderm cells, leading to an improved rate of cell division and/or protection from apoptosis.
In-vitro culture conditions for the culture of human embryos are generally considered to be suboptimal and this is believed to compromise the quality of embryos and may contribute to the high rates of implantation failure seen in human IVF. Culture to the blastocyst stage provides a means for selecting the most developmentally competent embryos for transfer to patients as well as achieving a better synchronization between the embryo stage and uterine development. Thus, the transfer of blastocysts to the uterus may lead to higher implantation rates and help to reduce the number of multiple births resulting from IVF. However, clinical application of these findings in human IVF must await the results of further experiments designed to elucidate the mechanism of action of GM-CSF in the embryo, and to evaluate the effect of culture with GM-CSF on the subsequent in-vitro developmental potential of blastocysts in animal models. Finally, any transfer to patients of blastocysts cultured in GM-CSF or other growth factors will be undertaken with caution, given that the events of early preimplantation development are now recognized to have long-term consequences for the health of the individual in childhood and in adult life (Seamark and Robinson, 1995).
![]() |
Acknowledgments |
---|
![]() |
Notes |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bergh, C., Howles, C.M., Borg, K. et al. (1997) Recombinant human follicle stimulating hormone (r-hFSH; Gonal-F) versus highly purified urinary FSH (Metrodin HP): results of a randomized comparative study in women undergoing assisted reproductive techniques. Hum. Reprod., 12, 21332139.[Abstract]
de Moraes, A.A. and Hansen, P.J. (1997) Granulocyte-macrophage colony-stimulating factor promotes development of in vitro produced bovine embryos. Biol. Reprod., 57, 10601065.[Abstract]
Dokras, A., Sargent, I.L. and Barlow, D.H. (1993) Human blastocyst grading: an indicator of developmental potential? Hum. Reprod., 8, 21192127.[Abstract]
Dunglison, G.F., Barlow, D.H. and Sargent, I.L. (1996) Leukaemia inhibitory factor significantly enhances the blastocyst formation rates of human embryos cultured in serum-free medium. Hum. Reprod., 11, 191196.[Abstract]
Gardner, D.K. (1994) Mammalian embryo culture in the absence of serum or somatic cell support. Cell Biol. Int., 18, 11631179.[ISI][Medline]
Gardner, D.K. and Lane, M. (1997) Culture and selection of viable blastocysts: a feasible proposition for human IVF? Hum. Reprod. Update, 3, 367382.
Gardner, D.K. and Lane, M. (1998) Culture of viable human blastocysts in defined sequential serum-free media. Hum. Reprod., 13 (Suppl. 3), 148159.[Medline]
Giacomini, G., Tabibzadeh, S.S., Satyaswaroop, P.G. et al. (1995) Epithelial cells are the major source of biologically active granulocyte macrophage colony-stimulating factor in human endometrium. Hum. Reprod., 10, 32593263.[Abstract]
Handyside, A.H. and Hunter, S. (1984) A rapid procedure for visualising the inner cell mass and trophectoderm nuclei of mouse blastocysts in situ using polynucleotide-specific fluorochromes. J. Exp. Zool., 231, 429434.[ISI][Medline]
Herrler, A., Krusche, C.A. and Beier, H.M. (1998) Insulin and insulin-like growth factor-I promote rabbit blastocyst development and prevent apoptosis. Biol. Reprod., 59, 13021310.
Hetts, S.W. (1998) To die or not to die: an overview of apoptosis and its role in disease [see comments]. J. Am. Med. Assoc., 279, 300307.
Imakawa, K., Helmer, S.D., Nephew, K.P. et al. (1993) A novel role for GM-CSF: enhancement of pregnancy specific interferon production, ovine trophoblast protein-1. Endocrinology, 132, 18691871.[Abstract]
Jones, G.M., Trounson, A.O., Gardner, D.K. et al. (1998) Evolution of a culture protocol for successful blastocyst development and pregnancy. Hum. Reprod., 13, 169177.[Abstract]
Kane, M.T., Morgan, P.M. and Coonan, C. (1997) Peptide growth factors and preimplantation development. Hum. Reprod. Update, 3, 137157.
Kaye, P.L., Bell, K.L., Beebe, L.F. et al. (1992) Insulin and the insulin-like growth factors (IGFs) in preimplantation development. Reprod. Fertil. Dev., 4, 373386.[ISI][Medline]
Kitamura, T., Tange, T., Terasawa, T. et al. (1989) Establishment and characterization of a unique human cell line that proliferates dependantly on GM-CSF, IL-3, or erythropoietin. J. Cell. Physiol., 140, 323334.[ISI][Medline]
Leese, H.J., Donnay, I. and Thompson, J.G. (1999) Human assisted conception: a cautionary tale. Lessons from domestic animals. Hum. Reprod., 13 (Suppl. 13), 184202.
Lighten, A.D., Moore, G.E., Winston, R.M. et al. (1998) Routine addition of human insulin-like growth factor-I ligand could benefit clinical in-vitro fertilization culture. Hum. Reprod., 13, 31443150.[Abstract]
Martin, K.L., Barlow, D.H. and Sargent, I.L. (1998) Heparin-binding epidermal growth factor significantly improves human blastocyst development and hatching in serum-free medium. Hum. Reprod., 13, 16451652.[Abstract]
McKiernan, S.H. and Bavister, B.D. (1994) Timing of development is a critical parameter for predicting successful embryogenesis. Hum. Reprod., 9, 21232129.[Abstract]
Ménézo, Y., Hazout, A., Dumont, M. et al. (1992) Coculture of embryos on Vero cells and transfer of blastocysts in humans. Hum. Reprod., 7 (Suppl. 1), 101106.[ISI][Medline]
Olivennes, F., Hazout, A., Lelaidier, C. et al. (1994) Four indications for embryo transfer at the blastocyst stage. Hum. Reprod., 9, 23672373.[Abstract]
Pampfer, S., Arceci, R.J. and Pollard, J.W. (1991) Role of colony stimulating factor-1 (CSF-1) and other lympho-hematopoietic growth factors in mouse pre-implantation development. Bioessays, 13, 535540.[ISI][Medline]
Pampfer, S., Vanderheyden, I., Vesela, J. et al. (1995) Neutralization of tumor necrosis factor alpha (TNF alpha) action on cell proliferation in rat blastocysts by antisense oligodeoxyribonucleotides directed against TNF alpha p60 receptor. Biol. Reprod., 52, 13161326.[Abstract]
Quinn, P., Kerin, J.F. and Warnes, G.M. (1985) Improved pregnancy rate in human in vitro fertilization with the use of a medium based on the composition of human tubal fluid. Fertil. Steril., 44, 493498.[ISI][Medline]
Robertson, S.A., Lavranos, T.C. and Seamark, R.F. (1991) In vitro models of the maternal fetal interface. In Wegmann, T.G., Nisbett-Brown, E. and Gill, T.G. (eds) The Molecular and Cellular Immunobiology of the Maternal-Fetal Interface. Oxford University Press, New York, pp. 191206.
Robertson, S.A., Mayrhofer, G. and Seamark, R.F. (1992) Uterine epithelial cells synthesize granulocyte-macrophage colony-stimulating factor and interleukin-6 in pregnant and nonpregnant mice. Biol. Reprod., 46, 10691079.[Abstract]
Robertson, S.A., Seamark, R.F., Guilbert, L.J. et al. (1994) The role of cytokines in gestation. Crit. Rev. Immunol., 14, 239292.[ISI][Medline]
Robertson, S.A., Mayrhofer, G. and Seamark, R.F. (1996) Ovarian steroid hormones regulate granulocyte-macrophage colony-stimulating factor synthesis by uterine epithelial cells in the mouse. Biol. Reprod., 54, 183196.[Abstract]
Robertson, S.A., Roberts, C.T., Farr, K.L. et al. (1999a) Fertility impairment in granulocyte-macrophage colony-stimulating factor-deficient mice. Biol. Reprod., 60, 251261.
Robertson, S.A., Sjöblom, C., Jasper, M.J. et al. (2000) Granulocyte-macrophage colony-stimulating factor promotes glucose transport and inner cell mass viability in murine blastocysts (in press).
Ruef, C. and Coleman, D.L. (1990) Granulocyte-macrophage colony-stimulating factor: pleiotropic cytokine with potential clinical usefulness. Rev. Infect. Dis., 12, 4162.[ISI][Medline]
Seamark, R.F. and Robinson, J.S. (1995) Potential health problems stemming from assisted reproduction programmes. Hum. Reprod., 10, 13211322.[ISI][Medline]
Sharkey, A.M., Dellow, K., Blayney, M. et al. (1995) Stage-specific expression of cytokine and receptor messenger ribonucleic acids in human preimplantation embryos. Biol. Reprod., 53, 974981.[Abstract]
Tremellen, K.P., Seamark, R.F. and Robertson, S.A. (1998) Seminal transforming growth factor beta1 stimulates granulocyte-macrophage colony-stimulating factor production and inflammatory cell recruitment in the murine uterus. Biol. Reprod., 58, 12171225.[Abstract]
Wikland, M., Borg, J., Hamberger, L. et al. (1994) Simplification of IVF: minimal monitoring and the use of subcutaneous highly purified FSH administration for ovulation induction. Hum. Reprod., 9, 14301436.[Abstract]
Zhao, Y. and Chegini, N. (1994) Human fallopian tube expresses granulocyte-macrophage colony stimulating factor (GM-CSF) and GM-CSF alpha and beta receptors and contain immunoreactive GM-CSF protein. J. Clin. Endocrinol. Metab., 79, 662665.[Abstract]
Submitted on June 7, 1999; accepted on September 20, 1999.