Mispair extension fidelity of human immunodeficiency virus type 1 reverse
transcriptases with amino acid substitutions affecting Tyr115
Mispair extension fidelity of human immunodeficiency virus type 1 reverse transcriptases with amino acid substitutions affecting Tyr115
Ana M.
Martín-Hernández
+
,
Mónica
Gutiérrez-Rivas
,
Esteban
Domingo*
and
Luis
Menéndez-Arias
Centro de Biología Molecular `Severo Ochoa', Consejo Superior de Investigaciones Científicas-Universidad Autónoma de Madrid, 28049 Cantoblanco,
Madrid
,
Spain
Received December 5, 1996;
Revised and Accepted February 19, 1997
ABSTRACT
The role of Tyr115 of human immunodeficiency virus type 1 reverse transcriptase
(HIV-1 RT) in the mispair extension fidelity of DNA dependent DNA synthesis was
analysed by using a series of 15 mutant enzymes with substitutions at Tyr115.
Their kinetic parameters for elongation using homopolymeric RNA-DNA and heteropolymeric DNA-DNA complexes showed major effects of the amino acid substitutions
on the
K
m value for dNTP. Enzymes with large hydrophobic residues at position 115
displayed lower
K
m values than enzymes with small and charged amino acids at this position. The
influence of all these amino acid replacements in mispair extension fidelity
assays was analyzed using three different mismatches (A:C, A:G and A:A) at the
3
'
-terminal position of the primer DNA. For the A:C mispair, a 2.6-33.4-fold increase in mispair extension efficiency (
f
ext) was observed as compared with the wild-type enzyme. Unexpectedly, all the mutants tested as well as the wild-type RT were very efficient in extending the A:G and A:A
transversion mispairs. This effect was due to the template-primer sequence context and not to the buffer conditions of the assay.
The data support a role of Tyr115 in accommodating the complementary nucleotide
into the nascent DNA while polymerization takes place.
INTRODUCTION
During the retrovirus life cycle, the reverse transcriptase (RT) replicates the
viral genomic RNA to synthesize a double-stranded DNA which integrates into the host genome. Reverse transcription is error prone and contributes to the high genetic variability of
retroviruses. Mutation rates in a single cycle of retrotranscription are in the
range 10
-4
-10
-5
misincorporations per nucleotide (
1
). One of the consequences of the high mutation rates has been the emergence of
drug resistant HIV variants, which has become an important obstacle in the
control of AIDS. The HIV-1 RT is a heterodimeric enzyme composed of two subunits of 66 and 51 kDa
respectively (
2
,
3
). The catalytic properties of the enzyme reside within the 66 kDa subunit.
Studies with purified HIV-1 RT have revealed an unusually high error rate in copying DNA or RNA
templates (
4
-
9
; reviewed in
10
). Errors can be generated either by direct misinsertion of an incorrect
nucleotide or by transient primer slippage (
9
,
11
-
13
). However, the molecular mechanisms governing fidelity of DNA synthesis are
largely unknown. Site-directed mutagenesis studies on the HIV-1 RT have provided some clues on the role of different amino acids
in fidelity of DNA synthesis. The substitution of Gly262 or Trp266 by Ala
renders enzymes with decreased frameshift fidelity (
14
), probably because these amino acids are involved in interactions with the
template-primer (
15
). Both residues are located within [alpha]-helix H (residues 253-271) that together with [alpha]-helix I (residues 277-287) form the `helix clamp' at the thumb
subdomain of the 66 kDa subunit. Substitution of the other amino acids found at
these two helical regions did not have significant effects on fidelity of DNA
synthesis (
15
,
16
). Base substitution errors could occur during DNA polymerization by a simple
two-step mechanism. It would involve the misinsertion of a non-complementary base into the nascent DNA, followed by extension of
the 3'-terminal mismatch. The later step would fix the incorporated
mismatched nucleotide into the nascent DNA. Nucleoside analog resistant mutants
of HIV-1 RT, such as M184V (
17
,
18
) or E89G (
19
) displayed a 1.4-17-fold increase in insertion fidelity compared to the wild-type RT. A similar effect was also reported for a variant RT
with Leu instead of Met at position 184 (
20
). This substitution, as well as the replacement of Tyr183 by Phe, rendered
enzymes with enhanced fidelity of mispaired extension relative to wild-type RT (
20
). In contrast, other amino acid changes led to enzymes whose fidelity of
mispair extension was either similar to that of wild-type RT, as in mutants M184V, Y181I or Y188L (
18
,
21
), or somewhat reduced as observed in the case of M184A (
18
) or the double mutant C38S/C280S (
22
).
In a previous study, we described variant RTs with substitutions of Tyr115 by
Phe, Trp, Ala, Ser, Asp or Lys (
23
). While Y115F showed a wild-type phenotype, the other RTs had either an impaired dNTP binding function
or were almost inactive. We also showed that Tyr115 plays a role in
misinsertion fidelity of DNA synthesis, as judged by a nucleotide misinsertion
assay (
23
). In this report, we describe the preparation and purification of RTs, and
report kinetic properties of dTTP binding for 16 RT variants. The effects on
the fidelity of mispair extension are also shown. The comparison of frequencies
of nucleotide misinsertion and mismatch extension indicates that the mutator
phenotype of most of these enzymes is determined by the influence of Tyr115 in
dNTP recognition.
MATERIALS AND METHODS
Mutagenesis
Site-directed mutagenesis was carried out with the Altered Sites
in vitro
mutagenesis system kit from Promega following the manufacturer's instructions.
The single-stranded DNA (ssDNA) template used in the mutagenesis reaction was
obtained from
Escherichia coli
DH5[alpha]F' cultures harbouring a pALTER-derived construct containing the coding sequence of the 66
kDa subunit of HIV-1 RT (
23
). The RT mutations and the oligodeoxynucleotides used in the mutagenesis reaction are shown in Table
1
. Synthetic oligonucleotides were obtained from Isogen Bioscience (Maarssen,
Holland). The introduced mutations were confirmed by digestion with
Nsi
I and by DNA sequencing. DNA fragments containing the desired mutations were
cloned in the p66(RT) and pT51H expression vectors, as previously described (
23
).
.
Synthetic oligonucleotides used in the mutagenesis of DNA encoding HIV-1 RT
Mutations
a
Oligonucleotide
b
Y115V, Y115L
5'-GGGAACTGAAAAA
AS
TGCATCACCCACATC-3'
Y115M, Y115I
5'-GGGAACTGAAAA
SAT
TGCATCACCCACATC-3'
Y115N, Y115H
5'-GGGAACTGAAAAAT
K
TGCATCACCCAC-3'
Y115G, Y115C
5'-GGGAACTGAAAAA
CM
TGCATCACCCACATC-3'
Y115P
5'-GGGAACTGAAAAA
GG
TGCATCACCCACATC-3'
a
Mutations are identified by the corresponding amino acid position in HIV-1 RT, followed by the substituted amino acid. Amino acids are denoted by
the single-letter code.
b
Underlined nucleotides correspond to mutations introduced in the RT coding
region. Several mutants were obtained with oligonucleotide mixtures: S stands
for C+G, K for T+G and M for A+C. All the introduced mutations except for Y115C
lead to the loss of an
Nsi
I restriction site.
Expression and purification of HIV-1 RT variants
Purification of mutant and wild-type RTs was carried out after independent expression of their subunits, by following a previously described procedure (
23
). The 51 kDa subunit was obtained with an extension of 14 amino acid residues
at its N-terminal end, which includes six consecutive histidines to facilitate its purification by metal chelate affinity chromatography. In this study, amino acid
substitutions were introduced in both subunits of the RT. The purity of the
enzymes was assessed by SDS-PAGE. All enzymes were >= 95% pure. RT concentrations were determined using the BioRad protein
assay.
DNA polymerase activity assays
DNA polymerase activity of the purified RTs and their steady state kinetic
parameters were obtained as previously described, assuming that 50% of the enzyme was active as determined by active site
titration (
23
). The assay solution contained 50 mM Tris-HCl, pH 8.0, 20 mM NaCl, 10 mM MgCl
2
, 8 mM dithiothreitol, 3-5 [mu]Ci/ml [
3
H]dTTP and 1 [mu]M poly(rA)[middot] oligo(dT)
20
(concentration expressed as 3'-hydroxyl primer termini). For the determination of kinetic
parameters, the dTTP concentration was adjusted with non-radioactive nucleotide, and ranged from 1 [mu]M to 3.6 mM depending on the enzyme tested. Reactions (30 [mu]l) were initiated by the addition of 0.8-10 pmol enzyme, incubated at 37oC for 10-30 min and terminated by adding 20 [mu]l 0.5 M EDTA. After addition of 6 [mu]l 0.5 mg/ml salmon sperm DNA and 600 [mu]l cold 10% trichloroacetic acid (TCA) in 20 mM sodium
pyrophosphate, samples were kept on ice for 20-30 min. The TCA-precipitable materials were collected on Whatman GF/A filters and
counted for radioactivity in a liquid scintillation counter.
RNase H activity assays
Assays were done in buffer containing 25 mM Tris-HCl, pH 8.5, 5 mM MgCl
2
, 1.5% glycerol, 50 [mu]g/ml bovine serum albumin, 0.01% Nonidet P-40 and 2 [mu]Ci/ml [
3
H]poly(rA)[middot]poly(dT) (
24
). The RNase H substrate was prepared by mixing 10 [mu]Ci [
3
H]poly(rA) in 0.5 ml distilled water, with 0.53 ml (dT)
221
containing 0.5 A
260
U/ml. The mixture was incubated at 70oC for 5 min, slowly cooled at room temperature, and stored in 100 [mu]l aliquots at -20oC until use. The enzyme concentration in these assays was
around 100-150 nM. Samples (100-120 [mu]l) were incubated at 37oC for 5-30 min. At different times, 25 [mu]l aliquots were taken and reactions were
terminated by addition of 5 [mu]l salmon sperm DNA (0.5 mg/ml) and 85 [mu]l cold 10% TCA in 20 mM sodium pyrophosphate. Samples were kept on ice for 10 min and centrifuged 10 min at 12000 r.p.m. The supernatants (80 [mu]l) were diluted in 3 ml Optiphase `Hisafe' scintillation fluid
(Wallac, Turku, Finland) and counted for radioactivity in a liquid
scintillation counter.
Mispair extension fidelity assays
Assays were performed essentially as described (
25
) with the modifications introduced by Martín-Hernández
et al.
(
23
). Template-primers used in these assays are shown in Figure
1
. The template D2 and the complementary 16mer primers were taken from Ricchetti
and Buc (
26
). D2 is a 38mer mimicking the HIV-1
gag
sequence and includes nucleotides 1137 (5' end)-1174 (3' end), according to the sequence numbering of Ratner
et al.
(
27
). M13mp2 was grown in the
E.coli
NR9099 strain and template M13 ssDNA was prepared as described (
28
). The oligonucleotides used for mispair extension assays in which M13 ssDNA was
used as template were those described by Mendelman
et al
. (
29
) and correspond to positions 5386-5405 of the M13 genome (
30
). The oligonucleotides used in these experiments were from Pharmacia (D2 and
PG5), Isogen Bioscience (PG5C and PG5G) and Gibco BRL (PG5A, pT, pC and pG).
Primer 5' termini were labelled with [[gamma]-
32
P]ATP (10 mCi/ml, Amersham) and T4 polynucleotide kinase (Boehringer). The
templates and the corresponding
32
P-labelled primers were annealed in 150 mM NaCl and 150 mM magnesium
aspartate for 3 min at 90oC. Samples were then cooled slowly to room temperature. The template-primer concentration ratio was adjusted to 1:1, equivalent to a 3 [mu]M final concentration in the hybridization solution. Prior to
the elongation reaction assay, the DNA duplexes were diluted 10-fold in 500 mM HEPES, pH 7.0, 150 mM NaCl and 150 mM magnesium aspartate
(final concentrations). Steady state kinetics were performed in 50 mM HEPES, pH
7.0, 15 mM NaCl, 15 mM magnesium aspartate, 130 mM KCH
3
COO, 1 mM dithiothreitol and 5% polyethylene glycol 6000. The reaction volume was 20 [mu]l and the enzyme concentration in these assays was ~6 nM. The molar ratio of template-primer to enzyme in the reaction mixture was estimated to be 5:1
in assays with the D2 template and the PG5 primers, and 2.5:1 in those assays
performed using M13 ssDNA. The reaction was initiated by first equilibrating
the RT with the annealed template-primer in the absence of dNTPs (10 min at 37oC), followed by the addition of appropriate dNTPs at various
concentrations. The reaction was carried out for 30 s at 37oC, and then stopped by addition of 8 [mu]l 10 mM EDTA in 90% formamide containing 3 mg/ml xylene cyanol FF and 3
mg/ml bromophenol blue. Samples were denatured at 80oC for 5 min, cooled on ice and 4 [mu]l aliquots were loaded on a 20% polyacrylamide gel (35 * 42 * 0.04 cm), containing 8 M urea. The samples were
electrophoresed for 4-6 h at 65 W (~2000 V) to obtain good resolution of extended primers.
Autoradiography of samples labelled with
32
P was performed by exposing gels to photostimulable imaging plates (Fujifilm BAS-MP 2040S). Radioactive band intensities were measured in a Fujifilm Bio-imaging analyser BAS-1500, using the program Tina version 2.09 (Raytest Isotopenmessgerate GmbH, Staubenhardt, Germany). Elongation measurements were fitted to the Michaelis-Menten equation using the UltraFit Macintosh program
(version 1.03; Biosoft). Primer degradation was estimated to be very small in
control reactions performed in the absence of dNTP, indicating that the
purified RTs were free of contaminating nuclease activity.
.
Kinetic parameters for dTTP binding of wild-type and mutant RTs
a
Enzymes
K
m
k
cat
k
cat
/
K
m
([mu]M)
(s
-1
)
(mM
-1
s
-1
)
WT
b
6.7 +- 1.7
0.47 +- 0.16
70.1 +- 10.8
Y115F
b
3.0 +- 1.0
0.23 +- 0.10
76.7 +- 17.8
Y115I
52.8 +- 7.2
1.28 +- 0.22
24.2 +- 1.0
Y115V
62.7 +- 14.8
1.12 +- 0.17
18.3 +- 3.9
Y115W
b
44.8 +- 7.7
0.51 +- 0.02
11.4 +- 2.0
Y115M
92.9 +- 11.6
0.94 +- 0.25
10.0 +- 1.8
Y115N
116.2 +- 21.0
0.70 +- 0.09
6.2 +- 1.8
Y115C
133.9 +- 28.3
0.71 +- 0.11
5.5 +- 1.6
Y115L
177.9 +- 32.4
0.60 +- 0.17
3.6 +- 1.5
Y115A
b
156.7 +- 19.9
0.50 +- 0.05
3.2 +- 0.9
Y115S
b
235.2 +- 26.4
0.65 +- 0.05
2.8 +- 0.4
Y115H
166.8 +- 26.7
0.43 +- 0.11
2.0 +- 0.5
Y115G
566.0 +- 75.0
0.65 +- 0.23
0.9 +- 0.2
a
Poly(rA)
484
[middot]oligo(dT)
20
was used as substrate. The template-primer nucleotide ratio was 10:1 (approximate molar ratio 1:2.5). Mutants Y115D, Y115K and Y115P showed negligible activity (
k
cat
< 0.005 s
-1
). Data shown are the mean values +- standard deviation, obtained from a non-linear least squares fit of the kinetics data to the Michaelis-Menten equation. Each of the experiments was performed
independently at least twice.
b
Reported data for this enzyme were taken from ref. 23.
RESULTS
Effect of amino acid substitutions on dTTP binding
Steady state kinetic analysis of polymerization by wild-type and mutant RTs was performed with poly(rA)[middot]oligo(dT)
20
and dTTP. As shown in Table
2
, substitution of Tyr115 often renders a variant enzyme with lower affinity for dTTP than the wild-type enzyme, in agreement with our previously published observations (
23
).
k
cat
values were not largely affected by replacements at Tyr115. The
K
m
values for dTTP binding ranged from 3.0 [mu]M (as in Y115F, the only variant enzyme tested which displayed a similar or
slightly increased affinity for dTTP) to 566 [mu]M, as observed for Y115G. The size and hydrophobicity of the amino acid
occupying position 115 are apparently important to maintain the low
K
m
value. The substitution of Tyr115 by small non-hydrophobic amino acids resulted in a dramatic increase of the
K
m
for dTTP. The
k
cat
/
K
m
values were generally higher for variants with hydrophobic residues at position
115. RNase H activity assays performed with RTs having Ile, Val, Met, Asn, Cys,
Ala, His, Gly or Pro at position 115 failed to reveal any significant
differences in specific activity among them and the wild-type RT. Their average RNase H specific activity was 324.4 +- 57.4 U/mg (1 unit is defined as 1 nmol [
3
H]adenylate produced in 1 h at 37oC). Interestingly, mutant Y115P, which was devoid of DNA polymerase activity (Table
2
), showed nevertheless normal RNase H activity (data not shown).
Mispair extension fidelity
DISCUSSION
Tyr115 is located in the vicinity of the triad of aspartyl groups forming the
catalytic site of HIV-1 RT (
2
,
3
). The importance of this residue for polymerase activity was suggested from
assays performed using bacterial extracts containing variant enzymes with
substitutions at Tyr115. Thus, the replacement of Tyr by Phe or Val rendered
enzymes with similar activity to the wild-type RT (
31
-
33
). In contrast, the polymerase activity of enzymes having Asn or His instead of
Tyr115 was estimated as <15% of that reported for wild-type RT (
31
). Our published results (
23
) and those described in this paper are consistent with these observations.
Amino acid changes involving Tyr115 have a dramatic effect on the binding
affinity of dNTP. For example, the
K
m
for dTTP was 84.4 times higher for Y115G than for the wild-type RT when poly(rA)[middot]oligo(dT)
20
was used as template-primer. Other amino acid replacements affecting residues of the putative
dNTP binding site of HIV-1 RT (e.g. the substitution of Gln151 or Met184 by Ala) also produce an
increase in the
K
m
values for dTTP when poly(rA)[middot]oligo(dT) is used as template-primer (
18
,
34
,
35
). Our previously reported data showed that all mutants including the poorly
active Y115D and Y115K had normal DNA binding affinity (
23
). We have now described another mutant (Y115P) which was devoid of polymerase
activity, although it showed the same RNase H specific activity as the wild-type RT. Taken together, the results of DNA binding experiments and RNase
H assays suggest that the RT variants are correctly folded. The kinetic
parameters governing incorporation of nucleotides into correctly matched DNA-DNA template-primers correlate well with those obtained with homopolymeric RNA-DNA complexes. In both cases, HIV-1 RT variants with bulky hydrophobic residues at
position 115 show higher affinity for dTTP than those having small or charged
amino acids. Results of mispair extension efficiency for mismatch A:C revealed
a similar trend to the kinetic measurements. Although all mutants extend the
mismatch A:C at a higher rate than the wild-type RT, the effects were more pronounced when smaller and less
hydrophobic residues were found at position 115.
Retroviral reverse transcriptases exhibit a higher mismatch extension (
f
ext
)/nucleotide insertion (f
ins
) ratio than eukaryotic DNA polymerases (
29
). Extension of mismatched 3' termini of DNA has been shown to be a major determinant of the
infidelity of HIV-1 RT (
11
). Our data on mispair extension fidelity are in agreement with those
observations. In our assays, the
f
ext
value for the wild-type enzyme ranges from 8 * 10
-3
to 0.154, while f
ins
was reported to be 1.54 * 10
-5
(
23
). Substitutions involving Tyr115 have a major effect on the misinsertion
fidelity of DNA synthesis. Thus, the nucleotide misinsertion efficiency (f
ins
) of mutants Y115S, Y115A and Y115W is 590.9, 188.3 and 64.9 times higher
respectively compared with wild-type RT. However, the efficiency of mismatch extension (
f
ext
) for the same mutants is only 22.5, 4.4 and 3.4 times higher for mismatch A:C,
and roughly similar to wild-type RT, when mispairs A:G or A:A are considered. Therefore, Tyr115 appears to play a more determinant role in recognition of the correct nucleotide than in its further
extension, in agreement with our views on its role in dNTP binding (
23
). Patel
et al.
(
36
) suggested that Tyr115 may interact directly by hydrophobic forces with the
base of the incoming nucleotide. Molecular modeling of an incoming dNTP
suggests that the triad of aspartic acid residues of the active site interact
with the phosphates, while several amino acids forming the fingers, thumb and
palm subdomains of the 66 kDa subunit, would position the template-primer in an appropriate manner for catalysis. In this scenario, the
enzyme suffers a conformational change that positions the dNTP in the right
conformation for the nucleophilic attack by the 3' OH of the primer (
36
,
37
). If the nucleotide at the 3' OH of the primer is not correctly paired with the corresponding base in
the template, the 3' OH may not be correctly positioned, and the conformational change may
not facilitate the correct alignment for the interaction between the [alpha]-phosphate of the incoming dNTP and the 3' OH of the primer. The consequence would be that the enzyme
would lose affinity for the next nucleotide. Mutants of Tyr115 could, in some
way, accommodate better than the wild-type the bulged mismatch, and do so more easily when the residue at
position 115 is less hydrophobic and smaller than Tyr. In this case, the mutant
RTs could lose less affinity for the next dNTP than the wild-type RT, thus extending the mismatch more easily with the result of a
decreased fidelity.
In our assays, purine-purine mismatches were extended very efficiently even by wild-type RT (
f
ext
[approx] 0.1-0.2). This is surprising in view of previous studies with HIV-1 RT and other polymerases, which indicated that these kinds of
mismatches are poorly extended, with relative extension frequencies around 10
-4
-10
-5
(
11
,
13
,
20
,
22
,
29
,
38
). Base context appears to be important for the relative stability of base
mispairs (
29
,
39
). When the template-primer complexes used by Mendelman
et al.
(
29
) were assayed with the wild-type HIV-1 RT in the buffer conditions of our assay, we obtained a relative extension frequency of 2.9 * 10
-5
for the A:G mispair. This result further confirms the extreme dependence of
fidelity on sequence context (
12
). Interestingly, in the case of the A:C mispair, the extension efficiency of
the wild-type HIV-1 RT was similar with the M13-based duplexes than with the D2-containing template-primers. Furthermore, the effects of
substitutions involving Tyr115 seemed to follow a similar trend with both
template-primers, suggesting that the results obtained are representative of the
behaviour of these mutants.
Resistance mutations affecting Tyr115 are rarely found. Only Phe has been
reported to appear at this position after passage of the virus in the presence
of the nucleoside analog RT inhibitor 1592U89 (
40
). Our data indicate that this substitution does not have a significant
influence on fidelity of DNA-dependent DNA synthesis. The mutator phenotype as well as the low
polymerase activity of the other HIV-1 RT variants could explain why these mutations are not found
in vivo
. High mutation rates caused by inaccurate reverse transcription may interfere
with the coding ability of the genome and ultimately cause deterioration of the
quasispecies (
41
,
42
). In this context, mutator RTs can be useful tools to study the molecular
mechanisms of fidelity of DNA synthesis, and an aid leading to the design of
antiretroviral drugs targeting the fidelity properties of the RT.
ACKNOWLEDGEMENTS
We thank S. H. Hughes and P. Boyer for providing us with their RT expression
plasmid p66(RT), and T. A. Kunkel for a gift of M13mp2 and the
E.coli
strain NR9099. This work was supported by grants from Fondo de Investigaciones
Sanitarias (95/0034-1), Comunidad Autónoma de Madrid, Fundación Ramón Areces, and Fundación Rodríguez Pascual.
REFERENCES
1 Pathak,V.K. and Temin,H.M. (1990) Proc. Natl. Acad. Sci. USA 87, 6019-6023.MEDLINE Abstract
9 Boyer,J.C., Bebenek,K. and Kunkel,T.A. (1992) Proc. Natl. Acad. Sci. USA. 89, 6919-6923.MEDLINE Abstract
10 Bebenek,K. and Kunkel,T.A. (1993) In Skalka,A.M. and Goff,S.P. (eds) Reverse Transcriptase. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 85-102.
11 Perrino,F.W., Preston,B.D., Sandell,L.L. and Loeb,L.A. (1989) Proc. Natl. Acad. Sci. USA 86, 8343-8347.MEDLINE Abstract
37 Zinnen,S., Hsieh,J.-C. and Modrich,P. (1994) J. Biol. Chem. 269, 24195-24202.MEDLINE Abstract
38 Perrino,F.W. and Loeb,L.A. (1989) J. Biol. Chem. 264, 2898-2905.MEDLINE Abstract
39 Joyce,C.M., Sun,X.C. and Grindley,N.D.F. (1992) J. Biol. Chem. 267, 24485-24500.MEDLINE Abstract
40 Tisdale,M., Parry,N.R., Cousens,D., St. Clair,M.H. and Boone,L.R. (1994) Abstracts of the 34th InterScience Conference on Antimicrobial Agents and Chemotherapy, p.92, Orlando, FL.
41 Eigen,M. and Schuster,P. (1979) The Hypercycle. Springer-Verlag, Berlin-Heidelberg-New York.
42 Domingo,E. and Holland,J.J. (1994) In Morse,S.S. (ed.) The Evolutionary Biology of Viruses, Raven Press, New York, pp. 161-184.
