Download Failure to infect embryos after virus injection in mouse zygotes

Human Reproduction Vol.17, No.3 pp. 760–764, 2002
Failure to infect embryos after virus injection in mouse
zygotes
L.Tebourbi1, J.Testart1,3, I.Cerutti2, J.P.Moussu2, A.Loeuillet2 and A-M.Courtot1,4
1Institut
National de la Santé et de la Recherche Médicale, Clamart, 2Service d’Expérimentation Animale et de Transgénèse,
Centre National de la Recherche Scientifique, Villejuif and 3AMP Center, American Hospital of Paris, Neuilly, France
4To
whom correspondence should be addressed at: INSERM U 355, 32 rue des Carnets, 92140 Clamart, France.
E-mail: [email protected]
BACKGROUND: The intracytoplasmic injection of sperm raises the problem that viral elements may be transported
into the oocyte by the spermatozoon or the surrounding medium. It also raises questions about how the developing
zygote will behave. METHODS: We used the murine model to microinject murine cytomegalovirus (MCMV) into
the zygote ooplasm and followed the changes in these microinjected zygotes in vivo and in vitro over time. RESULTS:
80% of zygotes microinjected with viral suspension, and 80% injected with medium alone, survived. Although
MCMV DNA was detected in 56% of injected embryos, up until the blastocyst stage, the mice born from these
injected zygotes developed normally and did not contain MCMV DNA. When embryonic stem cells were coincubated with MCMV and then transferred into healthy blastocysts, the offspring were normal and did not contain
any MCMV DNA. CONCLUSIONS: Our observations suggest that even if MCMV DNA persists from the zygote
to the blastocyst stage, its presence has no detrimental effect on pre-implantation or post-implantation development.
Key words: ES cells/ICSI/MCMV DNA/mouse/zygote
Introduction
The intracytoplasmic injection of sperm has been particularly
well developed recently as a means of medically assisted
procreation (MAP). This technique raises the question of
whether viral elements can be transported via the spermatozoon or the surrounding medium to the oocyte at
fertilization and what effects this would have during development.
It is known that ejaculated semen can contain viral particles,
such as human immunodeficiency virus (HIV: Baccetti et al.,
1994, 1999), cytomegalovirus (HCMV: Mansat et al., 1997)
papillomavirus (HPV: Lai et al., 1996) herpes simplex virus
type 2 (HSV2: Moore et al., 1989) and hepatitis C virus (HCV:
Leruez-Ville et al., 2000). Percoll gradients are generally used
to select motile sperm from the ejaculate and this reduces the
number of viral elements in the pool of selected sperm (Levy
et al., 1997). Accordingly, the probability of selecting a viral
particle with the motile spermatozoon when using ICSI pipettes
is low. However, sperm may carry viral particles. This is
particularly true of morphologically abnormal sperm (Huang
et al., 1986; Dussaix et al., 1993; Baccetti et al., 1994), which
are frequent in the case of male infertility. It is not always
possible to select only motile sperm in oligoasthenospermic
semen. In this case, the problem of transporting viral particles
to the oocyte is more crucial. Conversely, ICSI is often
recommended in azoospermia, and sperm are collected from
the epididymis or testis biopsies. These biopsies consist of
760
germinal, interstitial and blood tissues, which means that sperm
may come into contact artificially with any viral particles that
are present in interstitial or blood tissues. Furthermore, in
certain ICSI cases, sperm are directly injected into the oocyte
without being washed (De Croo et al., 2000).
The outcome of viral elements carried by the injected
sperm or the surrounding medium is more of a problem with
ICSI than with conventional IVF. The main studies on viruses
and fertilization in mammals have concentrated on the coincubation of viral particles with oocytes or embryos that are
surrounded by a zona pellucida, which is known to be an
efficient barrier against viruses (Neighbour and Fraser, 1978;
Heggie and Gaddis, 1979; Tsutsui, 1995; Tsuboi and Imada,
1997; Baccetti et al., 1999; Vanroose et al., 1997, 2000).
However, non-infectious viruses have been observed in the
perivitelline space and are thought to have been carried by the
sperm that were co-incubated with murine cytomegalovirus
(MCMV) (Neighbour and Fraser, 1978). Furthermore the
exposure of two-cell mouse embryos to the Coxsackie virus
B4 and MCMV inhibited blastocyst formation, although only
the Coxsackie virus B4 was recovered from the embryonic
cells (Heggie and Gaddis, 1979). Some studies have been
carried out on zona-free oocytes or embryos: Neighbour
showed that exposing zona-free pre-implantation mouse
embryos to MCMV does not affect development up to the
blastocyst stage (Neighbour, 1978). Conversely, Vanroose et
al. found bovine blastomeres expressing viral antigens in
© European Society of Human Reproduction and Embryology
Virus injection and ICSI
morulae and blastocysts after the co-incubation of zonafree embryos with bovine herpes simplex virus-1 (Vanroose
et al., 1997).
Purified viral RNA (Gamarnik et al., 2000) and DNA
(Baskar et al., 1993) have been injected into the oocytes.
When mengovirus RNA is microinjected into Xenopus oocytes
it leads to the complete cycle of viral replication, yielding a
high viral titre (107 pfu/oocyte, 30 h after microinjection;
Gamarnik et al., 2000). Baskar et al. injected MCMV DNA
into the male pronucleus of the zygote by use of the same
technique as that used for transgenesis (Baskar et al., 1993).
This resulted in smaller litter sizes, developmental retardation
and fetal abnormalities in mice. MCMV DNA was demonstrated in fetuses by PCR and DNA–DNA cytohybridization.
DNA viruses introduced into the ooplasm during ICSI are
predicted to have similar effects to viral DNA microinjected
into the male zygote pronuclei. However, studies concerning
the introduction of viral elements into the oocyte cytoplasm
are scarce. Ivanova et al. discussed experiments with adenoviruses and concluded that the injection of adenoviruses into
the ooplasm of mice and rabbits has no effect on development
(Ivanova et al., 1999).
We introduced MCMV into the cytoplasm of mice zygotes
at the pronucleus stage to evaluate the outcome of introducing
viral elements into the ooplasm during fertilization.
We chose MCMV as it is very similar to human CMV. This
virus can occur in complete or incomplete forms or even as
DNA and is known to induce severe fetal abnomalies. By
injecting the virus directly into the zygote, we were certain of
its presence in the zygote, which is not the case when it
is simply co-incubated with sperm. We also evaluated the
consequence of the virus in later stages of development by
coincubating embryonic stem (ES) cells with MCMV before
transferring them into healthy blastocysts.
Materials and methods
Virus preparation
The MCMV used in this study was the Smith strain obtained from
American Type Culture Collection (ATCC, Maryland, Rockville,
USA; VR-1399). It was maintained in the laboratory by serial passage
in FVB/N mouse embryo fibroblast (MEF) cultures. The infected
MEF cells were incubated for 3–5 days at 37°C to obtain characteristic
MCMV cytopathic effects (CPE). The infected MEF were then
collected and purified by use of a sucrose gradient [15% in phosphatebuffered saline (PBS)] and centrifuged as previously described
(Tebourbi et al., 2001). The concentration of the purified virus stock
(MCMVp) was 108 pfu/ml.
Collection of eggs and blastocysts from mice
Collection of ova
Four to six week old female FVB/N mice from our pathogenfree breeding colony [Service d’Expérimentation Animale et de
Transgénèse (SEAT), Institut André Lwolf, CNRS, Villejuif, France]
were induced to ovulate by i.p. injection with 5 IU of pregnant mare’s
serum gonadotrophin (PMSG) between 1300 and 1400 h, followed
by 5 IU of HCG 48 h later. They were immediately placed with
FVB/N males.
The following morning, the females with a vaginal plug were
killed, the oviducts were removed and placed in M2 medium (Quinn
et al., 1982) containing 3% hyaluronidase. The cumuli were discarded
from the ampulla, the cumulus cells were removed by the enzymatic
treatment and washed in M2 medium. After removing the defective oocytes, the intact oocytes were stored in M2 medium until
microinjection.
Collection of blastocysts
Female FVB/N mice which had been induced to ovulate were mated
with male FVB/N. The following morning, females with a vaginal
plug were separated from the males. Four days later, they were killed,
the uterine horns were recovered, and flushed with M2 medium. The
recovered blastocysts were transferred into a fresh microdrop of M2
medium in a Petri dish and stored at 37°C until injection of ES cells.
MCMV infection
Microinjection of MCMVp into zygote ooplasm
Micromanipulations were carried out in microdrops of M2 medium
covered with light paraffin oil in Petri dishes, to create a microinjection chamber.
The oocytes were maintained with a holding pipette. The injection
pipette was filled by capillary action with MCMVp that had been
centrifuged at 20 800 g for 5 min. The microinjection was performed
into the ooplasm of the zygote, leading to a transient swelling of the
oocyte. After ten washes with M2 (200 µl/drop), the zygotes were
transferred into a fresh M2 microdrop before further manipulations.
The controls consisted of injecting the zygote ooplasm with medium
alone. To ensure that the virus was present in each injected volume,
an equal volume was placed in the bottom of tubes containing lysis
buffer for PCR analysis.
MCMVp infection of ES cells
Cloned E14 was derived from murine ES cells from the ‘OLA 129’
strain of agouti mice (a gift from Doctor Boucheix, Laboratoire de
Différenciation Hématopoièse normale et leucémique, INSERM;
U268). The ES cells were cultured on gelatine-coated tissue culture
dishes containing a confluent monolayer of mitomycin C-treated
MEF with 1000 IU/ml leukaemia inhibitory factor (LIF). ES cells
and MEF were incubated at 37°C with high-glucose Dulbecco’s
modified Eagle’s medium (DMEM), supplemented with 20% fetal
calf serum (FCS), 2 mmol/l glutamine, 1 mmol/l 2-mercaptoethanol
and 1 mmol/l sodium pyruvate. Every 2 days, the medium was
removed and the cells were trypsinized. After centrifugation, the cell
suspension was reseeded in a gelatine-coated Petri dish for 1 h and
non-adherent cells, enriched with ES cells, were recovered. They
were incubated for 1 h in the presence of MCMVp (1⫻107 pfu/ml
for 1⫻107 cells). After washing in M2 medium, the infected ES cells
were centrifuged (2 ml at 500 g for 5 min). The pellet was resuspended
in 100 µl fresh M2, and 20 µl of the suspension was placed into the
microinjection chamber containing the blastocysts. Twelve to fourteen
infected ES cells were injected into each blastocyst. The control
group consisted of injecting non-infected ES cells into the blastocysts.
These blastocysts were then transferred into pseudopregnant females.
Embryo transfer
C57BL/6J recipient mice were placed with vasectomized males to
induce pseudopregnancy. One- or two-cell embryos that had been
microinjected with MCMVp were then transferred into the recipient
females the following day. The blastocysts which received ES cells
were transferred into recipient females 72 h later.
Detection of MCMV DNA by PCR
The presence of MCMV DNA was tested in injected medium, twocell embryos, blastocysts, fetuses, newborn mice and in the placentas
and target organs of the recipient females.
761
L.Tebourbi et al.
A pair of 20 bp oligonucleotides based on the BamH1-B fragment
of the Smith strain MCMV genome (Klotman et al., 1990) was used
to amplify the sequence. The primer sequences were as follows:
(bp 35708 to 35728), 5⬘ CGTCGCGAAGGAACACCTCA 3⬘; and
(bp 35946 to 35966, antisense), 5⬘ GTCGCCATGCGGGAGGTTTC 3⬘ (Genset, Paris, France). This primer pair amplified a 258 bp
fragment.
Samples were incubated in lysis buffer (Tris/EDTA 10/1; pH 7. 8;
proteinase K) for 2 h at 55°C or overnight at 55°C depending on
the sample.
One microlitre of each lysate was amplified by use of Ready-toGo® beads (Amersham, Pharmacia Biotech, Orsay, France). The PCR
conditions were as follows: 94°C 5 min, 65°C 2 min and 70°C 4 min
for the first cycle, followed by 30 cycles of 94°C 1 min, 65°C 1 min
and 70°C 1min. PCR products were separated on a 1.4% agarose gel
and stained with ethidium bromide.
Results
Microinjection of zygotes with MCMVp
Effect of microinjection on zygotes
We studied the development in vivo and in vitro of microinjected zygotes. One hundred zygotes were microinjected
with the viral suspension and 80 survived (80%). Twenty five
zygotes were microinjected with the medium alone and 20
survived (80%).
Development of microinjected oocytes
Twenty zygotes were microinjected with the viral suspension
and cultured for 4 days giving rise to 17 blastocysts (85%).
The control zygotes resulted in a similar proportion of blastocysts (80%).
Of the 60 infected zygotes (30 of which were transferred at
the two pronuclei stage and 30 at the two-cell stage), 40
developed normally and gave rise to apparently normal young
(67%). This result was similar to that obtained in the control
group—7/15, i.e. 47%, (Table 1).
Table I. In-vitro and in-vivo development of MCMVp microinjected
zygotes
MCMVp microinjected
Controls
In vitro
blastocyst stage
In vivo
newborn mice
17/20a
4/5
40/60b
7/15
aNumber
of blastocysts obtained/number of microinjected zygotes.
of newborn mice/number of zygotes (two pronuclei) or embryos
(two-cell stage) transferred.
bNumber
In another experiment, following transfer to pseudopregnant
recipient mice, 34 infected zygotes developed into 19 normal
fetuses (56%) and five regressed at day 13.
Detection of MCMV DNA in injected medium, embryos, fetuses
and newborn mice
The medium injected with MCMVp always contained viral
DNA. Moreover MCMV DNA was detected in the medium
containing the MCMVp microinjected oocytes and in the
washing medium.
MCMV DNA was found after in-vitro culture in 22 of the
49 two-cell embryos analysed (45%) and in 53 of the 94
blastocysts (56%). However, no characteristic MCMV DNA
band was found in normal or abnormal embryos at day 13 of
pregnancy or in newborn mice (Table II). Similarly, MCMV
DNA was not detected in the placenta or target organs (salivary
glands) of the recipient mice.
ES cells co-incubated with MCMVp
Nine blastocysts that had been injected with 12–14 infected
ES cells were transferred into two pseudopregnant mice. They
gave birth to seven normal newborn, four of which where
chimerae, as shown by their hair colour which was a mixture
between FVB/N (white) and OLA 129 (agouti). In the control
group, five of the seven offspring were chimerae, a similar
result to that obtained with infected ES cells.
The organs (salivary glands, lung, liver and testis) of the
seven mice born after the injection of infected ES cells were
analysed on day 14 following birth and no specific MCMV
DNA was recovered.
Discussion
The aim of this study was to evaluate the risks of viruses
being transported by oocytes during ICSI fertilization
in humans.
We found that injecting MCMV into fertilizing mice
ooplasm had no effect on the pre-implantation development
in vitro. The embryos obtained from these injected zygotes
also developed normally in vivo and did not present
anomalies like those described in cases of congenital CMV
infection of human (Hansaw, 1966) or murine (Baskar et al.,
1983) fetuses. We followed the development of the injected
virus by testing for the presence of viral DNA at different
stages of development (two-cell stage, blastocyst, fetus,
newborn). The viral DNA, which was present in all the
tested microinjection droplets, persisted in half of the two-
Table II. Presence of viral DNA during embryonic development in vitro and in vivo after microinjection of zygotes with MCMVp
In vitro
Proportion (%) of samples
containing viral DNA
Microinjected
Washing
medium
Two-cell
stage
Blastocyst
stage
Normal
embryos
Abnormal
embryos
Newborn
medium
⫹
⫹
22/49
(45.0)
53/94
(56.0)
0/19
0/5
0/40
⫹ ⫽ presence of DNA–MCMVp in samples.
762
In vivo
Virus injection and ICSI
cell embryos and blastocysts, but disappeared in fetuses and
newborn. As we were unable to eliminate viral DNA by
repeatedly washing MCMV-injected embryos, we assume
that PCR analysis detected DNA adherent to the
outer layer of zona pellucidae as a result of contamination
during microinjection procedures. However, viral DNA might
correspond also to the microinjected DNA inside the
cytoplasm of the zygote. This DNA might be stored either
inside the embryo, or outside the embryo, in the perivitelline
space or attached to the inner side of the zona pellucidae.
Moreover it was not possible to determine whether the virus
persists in its complete form or as DNA.
We also transferred ES cells previously co-incubated with
MCMV into healthy blastocysts. No anomalies or viral
DNA were observed in the embryos or newborn resulting
from these blastocysts. In BDF1 mice, Tsutsui did not
observe any anomalies or detect any virus in 11 day
embryos after the injection of MCMV into blastocysts and
transfer into pseudopregnant mice (Tsutsui, 1995).
It was reported that sperm can be viral DNA vectors
(Bracket et al., 1971; Habrova et al., 1996), delivering
imported replicating DNA to the oocyte and the embryo.
The discrepancy between these reports and our results could
be due to the fact that the above authors used purified
forms of the viral DNA or modified DNA. Purified DNA
(or plasmids) are known to fix and even to be internalized
in the sperm, and are then expressed in the embryo as
shown after IVF (Bracket et al., 1971: rabbit; Lavitrano
et al., 1989: mouse) or ICSI (Chan et al., 2000: monkey).
Concerning MCMV, it was shown that microinjecting MCMV
DNA into mouse male pronuclei causes malformations and
that viral protein is subsequently found in the embryos
(Baskar et al., 1993). However, to obtain such results, the
authors injected purified MCMV DNA and used transgenic
procedures as they injected DNA into the male pronucleus
through the pronucleus membrane. The DNA integrated
randomly, and in some cases viral proteins were expressed.
In our experiments, which were similar to the ICSI procedure,
purified DNA was not used. We injected either the complete,
or incomplete form of the virus or unpurified DNA present
in the viral suspension as shown at the ultrastructural level
(Tebourbi et al., 2001).
These results only concern MCMV and cannot be extended
to other viruses, even other DNA viruses. Ivanova et al.
(1999) reported that the injection of adenoviruses into the
cytoplasm of rabbit and mouse zygotes had no effect on
subsequent development; however, these results have not
been substantiated. Heggie and Gaddis recovered the
Coxsackie virus B4 from murine blastomere extracts after
the co-incubation of the virus with two-cell embryos (Heggie
and Gaddis, 1979). However, the Coxsackie virus B 4 is a
very small RNA virus that is able to cross the zona
pellucida.
Our result clearly show that even if complete and infectable
MCMV persists in the mouse embryo until the blastocyst
stage, it has no impact on the further development of the fetus
or the newborn.
Acknowledgements
We thank Dr Boucheix, Laboratoire de Differentiation, Hématopoièse
normale et leucémique, for his generous gift of ES cells, Michel
Prenant, Catherine Landreau and Rene Duchateau (SEAT) for
technical assistance.
References
Baccetti, B., Benedetto, A., Collodel, G. et al. (1994). HIV-1 particles in
spermatozoa of patients with AIDS and their transfer into the oocytes. J.
Cell Biol., 127, 903–914.
Baccetti, B., Benedetto, A., Collodel, G. et al. (1999). The debate on the
presence of HIV-1 in human gametes. J. Reprod. Immunol., 41, 41–67.
Baskar, J.F., Stanat, S.C., Sulik, K.K. et al. (1983) Murine cytomegalovirusinduced congenital defects and fetal maldevelopment. J. Infect. Dis., 148,
836–843.
Baskar, J.F., Furnari, B. and Huang, E.S. (1993) Demonstration of
developmental anomalies in mouse fetuses by transfer of murine
cytomegalovirus DNA-injected eggs to surrogate mothers. J. Infect. Dis.,
167, 1288–1295.
Bracket, B.G., Baranska, W., Sawicki, W. et al. (1971) Uptake of heterologous
genome by mammalian spermatozoa and its transfer to ova through
fertilization. Proc. Natl Acad. Sci. USA, 68, 353–357.
Chan, A.W.S., Luetjens, C.M., Dominko, T. et al. (2000) Foreign DNA
transmission by ICSI: injection of spermatozoa bound with exogenous DNA
results in embryonic GFP expression and live Rhesus monkey births. Mol.
Hum. Reprod., 6, 26–33.
De Croo, I., Van der Elst, J., Everaert, K. et al. (2000) Fertilization, pregnancy
and embryo implantation rates after ICSI in cases of obstructive and nonobstructive azoospermia. Hum. Reprod., 15, 1383–1388.
Dussaix, E., Guetard, D., Dauguet, C. et al. (1993) Spermatozoa as potential
carriers of HIV. Res. Virol., 6, 12–14.
Gamarnik, A.V., Boddeker, N. and Andino, R. (2000) Translation and
replication of human rhinovirus type 14 and mengovirus in Xenopus
oocytes. J. Virol., 74, 11983–11987.
Habrova, V., Takac, M., Navratil, J. et al. (1996) Association of Rous sarcoma
virus DNA with Xenopus laevis sperrmatozoa and its transfer to ova through
fertilization. Mol. Reprod. Dev., 44, 332–342.
Hansaw, J.B. (1966) Congenital and acquired cytomegalovirus infection.
Pediatr. Clin. N. Am., 13, 279–293.
Heggie, A.D. and Gaddis, L. (1979) Effects of viral exposure of the two-cell
mouse embryo on cleavage and blastocyst formation in vitro. Pediatr. Res.,
13, 937–941.
Huang, E.S., Davis, M.G., Baskar, J.F. et al. (1986) Molecular epidemiology
and oncogenicity of human cytomegalovirus. In Harris, C.C. (eds)
Biochemical and molecular epidemiology of cancer. Liss, NewYork,
pp. 323–344.
Ivanova, M.M., Rosenkranz, A.A., Smirnova, O.A. et al. (1999) Receptormediated transport of foreign DNA into pre-implantation mammalian
embryos. Mol. Reprod. Dev., 54, 112–120.
Klotman, M.E., Henry, S.C., Greene, R.C. et al. (1990). Detection of mouse
cytomegalovirus nucleic acid in latently infected mice by in vitro enzymatic
amplification. J. Infect. Dis., 161, 220–225.
Lai, Y.M., Yang, F.P. and Pao, C.C. (1996) Human papillomavirus
desoxyribonucleic acid and ribonucleic acid in seminél plasma and sperm
cells. Fertil. Steril., 65, 1026–1030.
Lavitrano, M., Camaioni, A., Fazio, V.M. et al. (1989) Sperm cells as vectors
for introducing foreign DNA into eggs: genetic transformation of mice.
Cell, 57, 717–723.
Leruez-Ville, M., Kunstman, J.M., De Almeida, M. et al. (2000) Detection of
hepatitis C virus in the semen of infected men. Lancet, 356, 42–43.
Levy, R., Najioullah, F., Keppi, B. et al. (1997) Detection of cytomegalovirus
in semen from a population of men seeking infertility evaluation. Fertil.
Steril., 68, 820–825.
Mansat, A., Mengelle, C., Chalet, M. et al. (1997) Cytomegalovirus detection in
cryopreserved semen samples collected for therapeutic donor insemination.
Hum. Reprod., 12, 1663–1666.
Moore, D.E., Ashley, R.L., Zarutskie, P.W. et al. (1989) Transmission of
genital herpes by donor insemination. J. Am. Med. Assoc., 261, 3441–3443.
Neighbour, P.A. and Fraser, L.R. (1978) Murine cytomegalovirus and fertility:
potential sexual transmission and effect of this virus on fertilization in vitro.
Fertil. Steril., 30, 216–222.
763
L.Tebourbi et al.
Neighbour, P.A. (1978) Studies on the susceptibility of the mouse preimplantation embryo to infection with cytomegalovirus. J. Reprod. Fertil.,
54, 15–20.
Quinn, P., Barros, C. and Whittingham, D.G. (1982) Preservation of hamster
oocytes to assay the fertilizing capacity of human spermatozoa. J. Reprod.
Fertil., 66, 161–168.
Tebourbi, L., Courtot, A.M., Duchateau, R. et al. (2001) Experimental
inoculation of male mice with murine cytomegalovirus and effect on
offspring. Hum. Reprod., 16, 2041–2049.
Tsuboi, T. and Imada, T. (1997) Effect of bovine herpes virus-1, bluetongue
virus and akabane virus on the in vitro development of bovine embryons.
Vet. Micobiol., 57, 135–142.
764
Tsutsui, Y. (1995) Developmental disorders of the mouse brain induced by
murine cytomegalovirus: animal models for congenital cytomegalovirus
infection. Pathol. Inter., 45, 91–102.
Vanroose, G., Nauwynck, A.,Van Soom, E. et al. (1997) Susceptibility of
zona-intact and zona-free in vitro-produced bovine embryos at different
stages of development to infection with bovine herpesvirus1.Theriogenology, 47, 1389–1402.
Vanroose, G., Nauwynck, A.,Van Soom, E. et al. (2000) Structural aspects of
the zona pellucida of in vitro-produced bovine embryos: a scanning electron
and confocal laser scanning microscopic study. Biol. Reprod., 62, 463–469.
Submitted on August 29, 2001; accepted on November 11, 2001