Comparison of Moloney Murine Leukemia Virus Mutation Rate with the Fidelity of Its Reverse Transcriptase in Vitro *

The role of Moloney murine leukemia virus (MoMLV) reverse transcriptase (RT) in the generation of base substitution mutations during retroviral replication was analyzed. To that effect, the in uitro fidelity of the MoMLV RT was compared to the rate of base substitution mutations occurring during the replication of an MoMLV-based retroviral vector. Using the vector in an amber reversion assay, the base substitution mutation rate at a single locus was found to be 2 X lo-'/base pair in one cycle of vector virus replication. Analysis of the fidelity of the purified RT using the same template sequence revealed that, of the two mispairs (A.C and T.G) that would lead to reversion of the amber codon during replication, A . C occurs at a rate of 4.0 X lo-', and T.G occurs at a rate of 0.7 X While the rate of formation of A . C is very similar to the vector mutation rate, the rate of formation of T . G is more than 30 times higher. This discrepancy in rates suggests that there are other elements in the infected cells that contribute to the fidelity of viral replication.

Mutations can be introduced during transcription by RNA pol 11, minus or plus strand DNA synthesis by RT, or provirus replication by cellular DNA polymerases. Due to the high fidelity of cellular DNA replication (lo-' to lo-'' substitutions/base pair) (5), it is unlikely that mutations occurring during cellular replication of the provirus contribute significantly to the high retroviral mutation rates (6). Therefore, most mutations are probably introduced during provirus transcription or reverse transcription.
Retroviral mutation rates or the mutation frequency in a single cycle of retroviral replication have been quantitated for some retroviruses (7)(8)(9). These determinations, which measure the combined contributions of RT and RNA pol I1 to the mutation rate, have been done for two retroviruses of avian origin, spleen necrosis virus (SNV) and Rous sarcoma virus. For SNV, retroviral vectors and helper cells were employed and the base substitution mutation rate at a defined locus was found to be 2 X 10-5/base pair/replication cycle (7). Using a different SNV-based vector, the base substitution mutation rate over a longer sequence was found to be 0.7 x lO+/base pair/replication cycle (9). Using a different experimental system, the rate of base pair substitution mutations of Rous sarcoma virus was calculated as 1.4 X mutations/base pair/replication cycle (8). Thus, the average base substitution mutation rate of these avian retroviruses is 6 x or about 0.5 base substitutions/retroviral genome/replication cycle.
The fidelity of different purified RTs has been determined using cell-free systems. Nucleotide misincorporation rates were found to be sequence-and polymerase-dependent. Rates, in general, varied from 2.5 X 1O"j to 6 X (10-17). Based on those studies, it has been proposed that the high mutation rate during retroviral replication is, a t least in part, a consequence of errors made by R T (10). However, RNA pol I1 could also contribute to retroviral mutation rates.
T o date, experiments have not been performed to directly compare the fidelity of purified RT to retroviral mutation rates a t identical genetic loci. This is particularly important in view of the established effects of template sequence on RT fidelity. T o gain insight into the relative contribution of R T to retroviral mutation rates, we compared the base substitution mutation rate a t a defined locus during replication of a Moloney murine leukemia virus (MoMLV)-based vector in cultured cells with the fidelity of the purified MoMLV R T measured in vitro at the same locus. We found that the fidelity of plus strand synthesis by MoMLV RT in vitro was about 30 times lower than the overall mutation rate of the virus. These results suggest that other factors increase the fidelity of viral replication in the cell.

Enzymes
The MoMLV RT was obtained from Pharmacia LKB Biotechnology Inc. and is a recombinant product (18)

DNA Constructions
Phmids-The vector pAV was constructed in two sequential steps as follows. The SNV promoter contained in the EcoRI-AuaI fragment of pJD214Hy (19) was cloned by blunt-end ligation into the SalI site of pJD220Hy (19) resulting in the plasmid pJD220SNVHy. From this vector, the fragment containing the SNV promoter linked to the hygromycin phosphotransferase gene (hygro) was excised with XbaI and ClaI and cloned into the XhoI site of pN2 (20) by blunt-end ligation yielding pAV. The vector pAVneoAm differs from pAV by a single base pair which creates an amber codon at the 15th amino acid of the neomycin phosphotransferase gene (neo). pAVneoAm was made by substituting the BclI-BstBI fragment of pAV with the corresponding fragment from pJD216NeoamHy (7) which contains the 5'-coding region of ne0 with the amber codon. Vector viruses derived from the plasmid pAV and pAVneoAm are referred to as AV and AVneoAm, respectively.
M13 Constructions-The 1.3-kilobase pair HindIII fragment of pJD215Am (map available upon request), containing the same ne0 gene with the amber codon (neoAm) that was used in the construction of pAVneoAm, was cloned into the HindIII site of M13 mp18 in both orientations to create M13neoAm+ and M13neoAm-. In M13neoAm+ the ne0 coding sequence is inserted in the same orientation as that of the phage transcription; in M13neoAm-it is in the opposite orientation. These constructs yield single-stranded phage DNA (21) containing neoAm template sequences corresponding to the plus and minus strands, respectively, of the MoMLV-based pAVneoAm vector.

CeZls
NIH/3T3 is a murine fibroblast cell line permissive for infection by MoMLV and MoMLV-based vectors. GP+E-86 is an NIH/3T3derived helper cell line which provides the MoMLV trans-acting functions required for propagation of replication-defective MoMLV vectors without producing replication competent virus (22). The MoMLV provirus clone used to establish the helper cell line was the same as that used for the production of the recombinant MoMLV RT used in these experiments (18).

Transfections and Infections
Vector plasmid DNA was transfected by the polybrene-dimethyl sulfoxide method (23) into GP+E-86 helper cells. Vector virus harvested from these cells was in turn used to infect fresh GP+E-86 cells as previously described (7). Transfection and infection were followed by selection with hygromycin B (see below). Infected helper cell clones harboring a single provirus were isolated and used as source of vector virus for the determination of mutation rates. Vector virus titers of these clones were determined by infecting 2 X lo5 NIH/3T3 cells in 60-mm dishes with 0.2 ml of 10-fold serial dilutions of virus stocks in the presence of polybrene (10 pg/ml). 24-h post infection, the cells were selected in parallel for either G418 (300 pg/ml) or hygromycin B (350 pg/ml) resistance. Titers are given as G418' or hygro' colony forming units/0.2 ml of virus stock, respectively.

DNA Analysis
Southern Blotting-Analysis of proviral DNA was done by standard Southern blot hybridization (21). Genomic DNA was digested with DdeI, electrophoresed in a 1.2% agarose gel, and blotted onto a nitrocellulose membrane by capillary transfer. The blotted DNA was then hybridized with a 32P-labeled neo-specific probe (EcoRI-XhoI fragment of pN2, Ref. 20) followed by autoradiography.
DNA Amplification and Sequencing-The 5"coding region of the ne0 gene was amplified from genomic DNA of infected G418' NIH/ 3T3 cell clones using Taq polymerase (Perkin-Elmer Cetus) and the polymerase chain reaction (24). The primers used were 20-mers which, upon amplification, yielded a specific fragment of 199 bp from nucleotide position 123 to position 321 of ne0 (25). The specific 199bp fragment was purified by electrophoretic separation in an 8% polyacrylamide gel followed by passive elution and used to synthesize single-stranded DNA (minus strand) with an internal primer in an asymmetric enzymatic amplification reaction (26). The primer for the asymmetric amplification was a 20-mer complementary to sequence 267-286 of neo. The single-stranded DNA was sequenced with the 5'-TCAAGAGACAGGATGAGGA-3' primer by the dideoxychain termination method using [ ( u -~~S I~A T P (~1 0 0 0 Ci/mmol, Amersham Corp.) and T7 DNA Polymerase with a commercial kit (Pharmacia LKB Biotechnology Inc.).

Primed DNA Templates
Oligonucleotides used as primers for RT in vitro were synthesized by Operon Technologies, Inc. (Alameda, CA). Each was radiolabeled at the 5' terminus using T4 polynucleotide kinase and [ T -~~P ] A T P prior to hybridization. Single-stranded DNA from the M13 vectors containing the neoAm gene in both orientations (see DNA Constructions) was used to assay RT fidelity on substrates simulating both plus and minus strand template intermediates of the AVneoAm vector in culture. A DNA template was used to model plus strand RNA in order to minimize heterogeneities in the template sequence (27) which could mask true misincorporation events.
For mispair formation experiments, templates were hybridized in 45 pl at a 2:3 template/primer molar ratio with 12.5 pmol of one of two 5' end-labeled 16-base complementary oligonucleotides in 100 mM HEPES (pH 7.3) and 300 mM KCl. Hybridization was carried out by sequential incubation at 100 'C for 5 min, room temperature for 30 min, and 0 'C for 30 min. The 3'-hydroxyl terminus of one 16base oligonucleotide hybridized immediately adjacent to the A residue of the ti"TAG-3' amber codon in the plus strand of the ne0 gene. The 3'-hydroxyl terminus of the second 16-base oligonucleotide hybridized adjacent to the T residue of the 5'-CTA-3' in the minus strand that is complementary to the amber codon (Fig. 3). Mispair extension substrates were prepared in the same manner with the exception that the oligonucleotide primers were 17 bases long, containing an additional base at the 3' end. Thus, the extra terminal nucleotide was positioned opposite the A or T residues of the amber sequence of the plus or minus strand DNA templates, respectively (Fig. 3).

Mispair Formation and Extension Assays
The primer extension assay used for kinetic analysis is basically that described by Boosalis et al. (28). The reactions contained MoMLV RT at 0.002-0.15 unitlpl (units as defined by the manufacturer). Reaction components, in addition to enzyme, were 20 mM Tris-HC1 (pH 8.0), 2 mM dithiothreitol, 10 mM MgC12, 0.1 mg/ml bovine serum albumin, 2.2 nM of hybridized DNA template, and dNTPs as indicated in a total volume of 10 pl. Reactions were initiated by the addition of dNTP, incubated at 30 "C for varying times, then terminated by the addition'of NazEDTA, and transferred to an icewater bath. Each reaction was eluted through a 0.5-ml Sephadex G-100 column to remove non-hybridized primer, electrophoresed in 22% polyacrylamide-urea gels, visualized by autoradiography, and quantitated by densitometric scanning (Hoeffer GS-300 Scanning Densitometer, Ref. 11). Reaction times and enzyme levels were experimentally predetermined to ensure steady-state conditions during the measurement of both mispair and correct pair formation (<20% of total primer extension).
Stocks of dCTP were treated with purified E. coli dUTPase (kindly provided by Marshall Williams, Ohio State University) to remove dUTP, the product of dCTP deamination. This ensured that correct A. U pairs would not be formed during measurement of A. C mispair formation. 10 mM stocks of dCTP were incubated with 0.001 unit/pl dUTPase at 37 "C for 60 min in 50 mM Tris-HC1 (pH 7 . 3 , 2 mM 2mercapto-ethanol, 1 mM MgClZ, and 1% (w/v) bovine serum albumin. The formation of true mispairs was confirmed by comparing the electrophoretic mobilities of the products with those of synthetic markers containing correct or incorrect residues at the target site. Each marker migrates with slightly different mobility, and thus comparison of bands corresponding to mispair products with marker bands insured accurate identification of misincorporation events and demonstrated that the dNTP stocks were not cross-contaminated.

Assay for MoMLV Mutation Rates during a Single Cycle of
Replication-To determine the base substitution mutation rate of MoMLV, an MoMLV-based vector system was developed that scored reversion mutations during a single cycle of viral replication (Fig. 1). This system is an extension of an assay previously used to measure SNV mutation rate (7).
A replication-defective MoMLV-based vector pAVneoAm The mutant titer divided by the total titer represents the mutation frequency. The spread of the vector virus among the helper cells is effectively blocked due to superinfection immunity (31), and in the target NIH/3T3 cells due to the absence of viral proteins required for replication. Thus, this system limits retroviral replication to a single cycle which starts with a provirus in a GP+E-86 helper cell and ends with a provirus in the target NIH/3T3 cell. Since the measurement of the mutation frequency is confined to a single cycle, it constitutes a measure of the mutation rate. It should be noted, however, that sequence analysis of the mutants is required before the final computation of the mutation rate (see below). The polymerization steps during replication of this vector system are identical to those in MoMLV replication involving one step of genomic RNA synthesis by transcription (catalyzed by cellular RNA pol 11 in the helper cells), and two steps of DNA polymerization (minus and plus strand synthesis catalyzed by MoMLV RT in the target cells).
Mutation Rate during Replication of MoMLV Vector Virus-Measurements of mutant frequency following a single cycle of AVneoAm vector virus replication revealed that neoAm reversion mutations occurred at a relatively low rate (Tables I and 11). In one experiment, eight independent GP+E-86 helper cell clones (Table I,

Helper cell clones
a To obtain virus titers, vector virus was harvested from the helper cell clones (GP+E-86) harboring an AVneoAm provirus and used to infect NIH/3T3 cells. The infected cells were in turn selected for G418-resistance (G418') or hygromycin B-resistance (hypo').

TABLE I1
Summary of the mutation rates The abbreviation used is: cfu. colonv formine unit. 1 2 2 3 2 15 X lo5 1.4 X lo-' 4 6.1 X lo5 6.6 X lo-' MoMLV Mutation Rate and R T Fidelity X lo5 total virus screened in this experiment, 12 G418' NIH-3T3-infected clones were detected, corresponding to a mutation rate of 4 X 10-6/cycle. T o obtain significant numbers of G418' reversion mutants and to examine the possible effects of different helper cell clones on mutation rate, four additional experiments were conducted (Table 11). All experiments yielded similar mutation rates (1.4 X to 6.8 X 10-6/cycle) with an average mutation rate from 16 independent proviral helper cell clones of 4 X 10-6/cycle.
For each experiment a control was performed using a mass population of helper cells established by infection with the wild-type neolhygro provirus AV which yielded very similar titers for both marker genes (the ratio of the (3418' titers to the hygro' titers ranged from 1.3 to 5.7 with an average value of 3.2; data not shown). Thus, the conditions for selection were such that allowed the wild-type ne0 and hygro genes to be scored with similar efficiencies. Therefore, the ratio of G418/hygror titers obtained with AVneoAm virus should accurately reflect the mutation rate.
Southern Blot Analysis of Proviral DNA from G418' Colonies-During the construction of the vector pAVneoAm, a DdeI site (CTNAG) was generated in the neo gene by the substitution of the amber codon TAG for the tryptophan codon TGG (Fig. LA). Any base change in the second or third position of the amber codon should result in the loss of the DdeI site. T o determine whether the DdeI site was lost in the (3418' mutant clones obtained during AVneoAm vector replication, their genomic DNA was digested with DdeI and analyzed by Southern blotting. Of 45 clones analyzed, 31 had multiple proviral copies integrated in the genome since the effective multiplicity of infection was usually greater than one. Of the 14 cell clones with one provirus/cell, seven lost the DdeI site and seven retained it. Two cases in which the DdeI site was lost and two in which it was retained are shown in Fig. 2. Sequence Analysis of G418' Mutant Proviral DNA-To determine the exact nature of the mutations in the G418' clones, the region of the provirus containing the amber codon was sequenced using asymmetric polymerase chain reaction and the dideoxy-chain termination method (26). Analysis of the 14 singly infected G418' mutant clones revealed that those in which the DdeI site was lost had an A.T to G. C base pair substitution in the second position of the amber codon (data not shown). This substitution restored the wild-type ne0 sequence coding for tryptophan (TGG). This A.T to G. C transition was the only mutation seen at the amber codon in a similar reversion assay employing an SNV-based vector (7), suggesting that only reversion to the tryptophan codon can restore neo activity. In the seven clones in which the DdeI site was still present, no change was found at the amber codon or in the surrounding 60 base pairs (30 base pairs upstream and downstream of the amber codon). We are currently testing if other mutations outside the sequenced region suppress the amber codon thereby yielding G418' clones. These unidentified mutations, however, can be excluded from the calculation of the base substitution mutation rate at the amber codon. Therefore, if half of the G418' clones result from a mutation at the amber codon, the mutation rate at this locus is (4 X 10-6)(0.5), that is 2 x lO"j/base pair/replication cycle. This constitutes a minimum mutation rate at this locus since of the three possible base substitutions at the second position of the amber codon only the A to G transitions were detected by our selection system.
Measurement of R T Fidelity-The base substitution mutations that occur during retroviral replication presumably arise from two sequential polymerization events at the target site. The first is incorporation of a noncomplementary nucleotide to generate a mispair at the 3' terminus of the nascent strand. The second step is extension of the terminal mispair, thereby permitting complete genome synthesis. Thus, mutagenesis depends upon the frequency of mispair formation as well as the efficiency of mispair extension. In order to explore the contribution of R T in the generation of the vector amber reversion mutations, we have measured the relative efficiencies of both mispair formation and extension at the amber reversion site by purified MoMLV R T using the kinetic assay depicted in Fig. 3 (28). Briefly, a radiolabeled oligonucleotide (16 nucleotides in length, 16-mer) is hybridized to a singlestranded neoAm template so that the 3' end of the primer is immediately adjacent to the second position of the amber codon. Purified MoMLV R T is incubated in the presence of excess template .primer and increasing levels of a single dNTP species, either complementary or noncomplementary to the template base. Reaction products are separated by polyacrylamide gel electrophoresis, and the relative amounts of product formed at each dNTP concentration are quantitated to obtain apparent Vm../Km values for incorporation of both correct and incorrect dNTPs at the amber site. The efficiency of base pair formation is described by the term V,../K,,, (28,32).
Thus, the relative frequency for the formation of each mispair is given by Vm.,/Km for that mispair divided by V,,,.,/K,,, for formation of the correct pair (28). Efficiencies of mispair extension are measured basically in the same manner, using template-primers with preformed 3'-terminal mispairs at the amber site.
Mispair Formation by MoMLV RT Panel B depicts the same process on the M13neoAm-template, which corresponds to plus strand DNA synthesis during vector virus replication. The amber reversion site is the second base pair of the amber codon.

MoMLV Mutation Rate and RT Fidelity
A. Minus-strand synthesis. Amber rmrsiw site 4

G C T T A G G T G ---( + ) C C A C w m e r
( -)

A A C C C A C ~-m e r
B. Pius-strand synthesis.

"-C G A A T C C A C " -5
Mispair Formation RT + dGTP J.

"-C G A A T C C A C -" 5'
Mispair Extension RT + dGTP formed by RT during viral minus strand synthesis or from T,.G mispairs formed either by RT during plus strand synthesis or by RNA pol I1 during proviral transcription. As a model for events that occur at this site on each strand during viral replication, we have measured the relative efficiency of formation of each mispair by purified MoMLV RT in a cellfree system containing purified template-primers. For minus strand DNA synthesis, incorporation was measured opposite the template A of the 5'-TAG-3' sequence of M13neoAm' which contains the plus strand sequence, thus simulating the amber codon found in the AVneoAm vector ( Figs. 3A and 4, A and B). The extent of base pair formation, as measured by the quantity of radiolabel in bands larger than 16 nucleotides, was clearly dependent upon the nature and amount of dNTP present in the reaction. Fig. 4A shows that yields of 17-mer product increased with increasing micromolar concentrations of the correct nucleotide, dTTP. In contrast, extension to 17-mer in the presence of millimolar concentrations of the incorrect nucleotide dCTP was negligible (Fig.   4B), requiring overexposure of the film for visualization (data not shown). Thus, purified MoMLV R T forms A,. C mispairs very poorly during minus strand synthesis of the amber codon reversion site. This is consistent with other assays on MoMLV R T showing relatively low rates of A,. C mispair formation on both DNA and RNA templates (16, 33, 34).
As a model for plus strand DNA synthesis, we examined the reaction products made by MoMLV R T on M13neoAmtemplates which contain the minus strand ne0 sequence found in the vector AVneoAm (Figs. 3B and 4, C and 0 ) . Incorporation was measured opposite the T in the 5'-CTA-3' sequence which is complementary to the amber codon. The correct substrate, dATP, readily supported polymerization (Fig. 4C) at PM concentrations, while mispaired product was only detectable at mM concentrations of dGTP (Fig. 40). The product containing the T,. G mispair was 19 nucleotides in length due to correct incorporation of dGTP opposite the two template C residues downstream from the target site. To confirm that the observed 19-mer resulted from dGMP incorporation at the template T, a correctly paired 19-mer was made by extension of 16-mer in the presence of 50 PM each of dATP and dGTP. Under these conditions, the product 19mer will contain the correct base, A, at the target site, since mispairs are not formed detectably at micromolar concentrations of dGTP. It is clear that this 19-mer marker migrates differently than the 19-mer formed in the presence of millimolar concentrations of dGTP (Fig. 4 0 , marker lane indicated with an *). From this we infer that the 19-mer formed in the kinetic fidelity assay contained an incorrectly incorporated G residue opposite the T of the 5'-CTA-3' sequence, and extension was not due to correct incorporation of contaminant dATP in the dGTP stocks. These data also provide evidence that the RT is able to extend the Tt. G mispair because only the 19-mer product was detectable on the gel. Apparent V,,,/K, values for formation of each nucleotide pair were determined, and the frequencies of mispair formation ( f i n = ) were calculated as described (Table 111, Ref. 28).
The fine for A,-C is 3.5 (k0.6) x and the fin, for T,.G is Thus, the error rate of T,. G mispair formation in vitro is over 30 times higher than the mutation rate at this site during AVneoAm vector virus replication.

Extension of Preformed Mispaired Termini by MoMLV
RT-Since extension of mispairs might influence mutagenesis during viral replication, we quantitated the propensity of purified MoMLV RT to extend preformed A,. C and T,. G mispairs. For these experiments, primers 17 bases in length were constructed. The primers contained C or G terminal bases so that upon hybridization, A,. C or T,. G mispaired termini were formed. For assays of correct pair extension, A,.T or T,.A, 17-mers terminating in bases complementary to the target site were used (Fig. 5 ) . Inspection of Fig. 5, A and C, shows that the 17-mers containing the correctly paired termini were extended with ~L M dNTPs to yield primarily 19mers representing the addition of two template-directed residues. Extension of both A,. C and T, + G mispaired termini were detected at millimolar nucleotide concentrations of the next nucleotide (Fig. 5 , B and D). Apparent V, , , . , /K, , , values and relative efficiencies of mispair extension ( fext) are listed in Table  IV. There was no significant difference between the efficiency of extension of the A,. C and Tt. G mispairs ( f e X t = 1.5 X for A,. C; fert = 0.8 X for Tt. G).

DISCUSS~ON
T o gain insight into the role of R T in the generation of retroviral mutations, we compared the fidelity of MoMLV R T  during polymerization in vitro of an amber codon in the ne0 gene to the mutation rate of an MoMLV-based viral vector at the same locus during replication in the cell. After a single cycle of MoMLV vector replication, G418-resistant reversion mutants arose at a rate of 4 X Sequence analysis of the mutant clones revealed that seven of 14 had undergone an A.T to G. C transition at the second base pair of the amber codon; the other seven clones showed no sequence changes in the amber codon and thus were not true revertants. Therefore, the base substitution mutation rate at the second base pair of the neo amber codon is 2 x substitutions/base pair/virus replication cycle.
The A. T to G. C transitions observed in the revertant proviruses could arise as an A,. C mispair formed during minus strand DNA synthesis by R T or as a T,. G mispair formed either during plus strand DNA synthesis by R T or during provirus transcription by RNA pol 11. In an attempt to assess which of those steps contributed most significantly to the mutation rate, we examined the fidelity of MoMLV R T during polymerization of both strands of the amber codon i n vitro. Our results showed that the A,.C mispair is formed at an average rate of 4 x which is comparable to the 2 x rate observed during MoMLV vector replication. However, the T,. G mispair is formed a t a rate of 7 x in vitro which is over 30 times higher than the vector virus mutation rate in the cell. This difference might be slightly greater since the G418' titers obtained with the control vector AV, which contains both wild-type neo and hygro genes, were on average three times higher than the corresponding hygro' titers, thus resulting in a small overestimation of the mutation rate during virus replication. During replication, viral RNA serves as the template for minus strand DNA synthesis, and minus strand DNA serves as template for plus strand DNA synthesis. In the in vitro system employed, DNA templates were used as models for both minus and plus strand DNA synthesis during vector replication. Several studies show that the average fidelity of MoMLV R T is comparable on both RNA and DNA templates (16,33,34), including ribo-and deoxyribo-templates of identical sequence (16). Thus, DNA appears to be a reasonable model for RNA in MoMLV R T fidelity assays. Furthermore, if the rate of misincorporation for the i n vitro model corresponding to minus strand synthesis was significantly higher than the mutation rate obtained during replication, it could be argued that the discrepancy is the result of the use of a DNA template instead of RNA, the normal template during replication. However, a significant difference was found only between the in vitro model corresponding to DNA-directed plus strand synthesis and the mutation rate measured during vector virus replication. Since DNA serves as template for plus strand synthesis both during virus replication and in our i n vitro model, it is difficult to argue that the nature of the templates accounts for that difference.
There are at least four possible reasons for the discrepancy in rates. First, conditions in the cellular environment may contribute to fidelity by decreasing the rate of mispair formation. During reverse transcription in the cell, RT is initially found in an RNA-protein complex, and by the end of reverse transcription, the product viral DNA is still associated with subviral particles that contain RT, gag proteins, and viral integrase (35-37). Such nucleoprotein complexes could enhance the fidelity of R T by affecting processivity, templateprimer recognition, or dNTP discrimination. Unidentified cellular factors might also contribute to R T fidelity.
Second, the rate of mispair formation could be the same both in vitro and during viral replication, but its contribution   (46).
Third, cellular DNA mismatch repair mechanisms might correct mispairs formed by RT. This would require a repair system that gains access to the replicative complex or integrated proviral DNA, discriminates between the template and nascent DNA strands, and selectively repairs incorrectly inserted bases on the nascent strand. DNA repair systems capable of strand-specific mismatch repair have been detected in nuclear and cytoplasmic extracts of HeLa cells (47)(48)(49) and nuclear extracts of Drosophila melanogaster cells (47). These systems might be capable of correcting replication errors during or after retroviral plus strand DNA synthesis. Studies of retroviral recombination suggest that mismatch repair might occur during replication sometime after reverse transcription and before cellular replication of integrated proviral DNA (50). Although strand discrimination for mismatch repair in retroviruses has not been examined, selective removal of RT errors in the nascent plus strand could be signaled by discontinuities often present in this strand (44). It should be noted that repair by the mammalian G .T to G C short patch repair system cannot account for the low rate of AVneoAm virus mutation in our experiments, since this mismatch repair system exhibits a strand bias that would actually increase the reversion rate during replication (51).
The fourth possibility is that differences between the MoMLV RT used by the vector virus during replication and the recombinant MoMLV RT used in vitro resulted in the observed discrepancies. The purified RT utilized in our in vitro studies is a fusion protein produced in E. coli which contains some amino acid residues of non-viral origin at both termini and lacks 7 residues normally present at the carboxyl end of the mature enzyme (18). Similar experiments to those described herein revealed that the recombinant RT used in our studies and another recombinant MoMLV RT (Bethesda Research Laboratories, Ref. 52) incorporate nucleotides with similar fidelity opposite a G on a @X174 single-stranded DNA template? The second enzyme has all the amino acids present in the authentic RT, but it also contains an extra methionine at the amino terminus as well as 6 extra residues present in the pol polyprotein (not found in the normally processed MoMLV RT) plus 6 residues encoded by a terminator linker at the carboxyl terminus. Since the accuracies of both recombinant proteins are the same, it suggests that the differences MoMLV Mutation Rate and RT Fidelity at their termini do not affect fidelity.
Of the four possible mechanisms discussed above, we favor the first, that is, accessory viral or cellular factors increase R T fidelity during plus strand synthesis in the cell. Replicative polymerases do not generally function as isolated enzymes and often require accessory factors to conduct efficient, faithful DNA synthesis (53). Several retroviral proteins exhibit biochemical properties that suggest accessory roles during reverse transcription (54), but contribution of these and other cellular factors to RT fidelity has not been addressed experimentally. Analyses of the fidelity of reverse transcription complexes in uitro involving other virion-associated proteins should provide valuable information concerning the possible roles of these proteins in retroviral DNA synthesis. Similar analyses including cellular extracts might prove useful in elucidating the role of mismatch repair systems in retroviral replication fidelity. In light of these results, caution should be exercised when extrapolating in vitro measurements of RT fidelity to the corresponding mutation rates during retroviral replication.
Our results also provide information about the fidelity of RNA pol 11. RNA pol I1 transcribes genomic RNA from proviral DNA and thus can contribute to retroviral mutagenesis. Our data indicate that RNA pol I1 forms T,. G mispairs in the neo amber codon at a rate no greater than 2 x ( i e . the overall mutation rate of the virus). This relative high fidelity is not predicted from teleological arguments nor from studies on other polymerases lacking exonucleolytic proofreading (11,55), including E. coli RNA polymerase which is relatively error prone (27). However, the fidelity of RNA pol 11, like other polymerases (55), is likely to be influenced by sequence context, and the neo amber locus studied here represents only a single template sequence. Establishment of in vitro fidelity assays employing RNA pol I1 should be useful in assessing its overall fidelity and its contribution to retroviral mutagenesis.