Frequency of Missense Mutations in the Coding Region of a Eukaryotic Gene Transferred by Retroviral Vectors

Abstract

A relatively high mutation rate is probably a major factor in the evolutionary success of retroviruses, because it generates the genetic diversity that helps them to cope with changes in the environment. When using recombinant retroviruses as vectors for gene transfer and gene therapy, it is important to consider the implications of this biological characteristic. Until now, the mutation rate has been studied by using noneukaryotic genes as reporters. Here we report point mutations in the human glucose-6-phosphate dehydrogenase (hG6PD) gene transferred by Moloney murine leukemia virus-based vectors into murine bone marrow cells and NIH 3T3 murine fibroblasts. After bone marrow transplantation, we observed an hG6PD with abnormal electrophoretic mobility for 2 out of 34 mice. Next, we studied this phenomenon quantitatively and found 1 electrophoretically abnormal hG6PD variant among 93 independently isolated NIH 3T3 clones, from which we estimate a mutation rate of 1.4 × 10⁻⁵ per base pair per replication cycle. Mutations in the transferred gene can thus contribute to the impairment of the effectiveness of retrovirus-mediated gene transfer.

A relatively low fidelity of replication (7) helps to generate diversity in retroviruses and thus helps them to cope with environmental changes and to evolve rapidly (3, 9). It is important to take into account this high mutation rate when we exploit the permanent integration of the retroviral genome into the host cell genome in order to transfer a gene for the ultimate purpose of clinical gene therapy. So far, the mutation rate of retroviruses has been studied in viral or bacterial target sequences. We report here that in the coding region of a eukaryotic gene transferred by a retrovirus, mutations are not infrequent.

We have used for this study the gene encoding glucose-6-phosphate dehydrogenase (G6PD), the deficiency of which causes hemolytic anemia, which, for a subset of patients, can be severe (reviewed in reference 11). We have previously constructed a set of Moloney murine leukemia virus (MMLV)-based vectors harboring the human G6PD (hG6PD) cDNA (Fig. 1) (5, 22). In a preclinical study, bone marrow cells from male donor mice were transduced with these vectors (obtained from stable transfected producer cell lines, viz., ecotropic ψ-CRE cells [4] or pantropic 293GPG cells producing vesicular stomatitis virus G protein-pseudotyped virions [16]) and transplanted into lethally irradiated syngeneic recipient females. Transgene expression on cell lysates was assessed each month by cellulose acetate gel electrophoresis (CAGE) followed by specific staining for G6PD activity, a technique that resolves the murine G6PD from the hG6PD (19) (Fig. 2). Expression of the hG6PD transgene was observed in the blood cells of 34 mice (reference 22 and unpublished results). To our surprise, for 2 of these 34 mice, the nonmurine G6PD band was shifted relative to the position of the expected hG6PD band (Fig. 2). Southern blot analysis did not show any gross rearrangement of the proviral fragment in blood cell DNA from either of these mice (data not shown). From this DNA, a specific proviral fragment that included the hG6PD cDNA was amplified, cloned, and sequenced. In mouse 3, we found a G-to-A base change at nucleotide 580 of the hG6PD cDNA, which causes a neutral Asn to replace the acidic Asp 194. In mouse 7, we found a G-to-A base change at nucleotide 1480, which causes a basic Lys to replace the acidic Glu 494. The changes, with charges of +1 and +2, respectively, are consistent with the observed changes in electrophoretic mobility. The presence of these mutations was confirmed by digestion with the appropriate restriction enzymes of a PCR-amplified fragment of DNA (data not shown). These mutations were not detected by PCR analysis in the producer cells.

FIG. 1.

Retroviral vectors harboring the hG6PD cDNA. The vectors we used (see references 5 and 22) include the cDNA coding region of the hG6PD gene of 1,545 bp plus 110 bp of the 3" untranslated region. The schematic proviral structure is shown. LTR, long terminal repeat; SD, splice donor; SA, splice acceptor; ψ, packaging sequence; ±B2, with or without the B2 mutation (1). Arrows indicate the origin and direction of transcription. NheI indicates the restriction sites used for Southern blot analysis of the number of proviral copies (bone marrow transplantation and NIH 3T3 experiments) (22); NdeI indicates the restriction site used for Southern blot analysis of clonal proviral integration (NIH 3T3 experiment) (5). The hG6PD cDNA used as a probe for Southern blot analysis is shown. Different shadings indicate different LTRs as follows. (a) MMLV-G6PD is based on the SFG plasmid backbone harboring the MMLV LTR (21). (b) Myeloproliferative sarcoma virus (MPSV)-G6PD is based on the MPSV-ADA plasmid backbone (20) which contains the 3" LTR derived from MPSV. (c) The GRU5-G6PD vector was produced by replacing most of the MMLV-G6PD 3" LTR with a fragment of the hG6PD promoter (5).

FIG. 2.

hG6PD mutations observed by electrophoretic analysis after transduction of mouse cells with normal hG6PD. The two left panels show results for peripheral blood from four mice that have been transplanted with bone marrow cells transduced with hG6PD. The right panel shows results for individual NIH 3T3 clones. The hG6PD and murine G6PD were resolved by CAGE, followed by specific staining for G6PD activity, as described previously (19). By this method, the hG6PD transgene (h) is visualized as a band running slower than that of the endogenous mouse G6PD (m). A band with intermediate electrophoretic mobility, consisting of a human-mouse heterodimer (hm), is also visible whenever both genes are coexpressed within the same cell. Double arrowhead, hG6PD variant; single arrowhead, presumptive heterodimer of the hG6PD variant and mouse G6PD; 3T3, untransduced NIH 3T3 cells. For clone IV/4, two fast-moving bands in addition to the murine G6PD band, are visible; it is likely that the slower of these two derives from the homodimeric mutated G6PD and that the other derives from the heterodimer. Mouse 16, mouse 18, and clone IV/1 show the expected G6PD patterns in murine cells transduced with wild-type hG6PD. Mouse 3, mouse 7, and clone IV/4 show three different abnormal patterns, arising from three different de novo mutations in hG6PD (see text). For mouse 7, in addition to the hm band, a heterodimer of the normal and the variant hG6PD (←) is visible. The abnormal patterns in both of these mice were observed throughout their lives.

We found two mutated proviral sequences in vivo after bone marrow transplantation in 34 mice, for which we have estimated by Southern blot analysis an average of 2.4 proviral copies per donor cell (reference 22 and unpublished results). Thus, in a complex experimental setting very similar to a clinical gene therapy protocol, we found a significant rate of mutations in the transferred gene. However, this system is not suitable for measuring the rate of mutations per viral replication cycle for two reasons. (i) The bone marrow of these mice contains a mixture, in different proportions, of cells clonally derived from a number of the originally transduced hematopoietic stem cells (each one with different proviral integrations which may vary in number) as well as untransduced cells of donor and recipient origin. (ii) For most of the mice, we used vesicular stomatitis virus G protein-pseudotyped retroviral particles whose production did not strictly confine viral replication to one cycle. In order to measure properly the true mutation rate, we transduced NIH 3T3 cell lines with recombinant retroviral vectors produced by stable transfected ecotropic ψ-CRE cell lines (a system in which viral replication is confined to one cycle), and we isolated by limiting dilution a total of 482 cell clones, 95 of which expressed hG6PD.

Of the 387 clones that did not express hG6PD, we tested 150 by PCR or Southern blotting and did not find proviral sequences in any of them. In each of the 95 clones expressing hG6PD, the number of copies was measured by Southern blotting. This analysis revealed gross rearrangements of the proviral sequence in two clones. One of these clones had a ∼250-bp insertion and the other had a ∼300-bp deletion. Both clones had multiple proviral copies, and therefore they were not investigated further. The 93 remaining hG6PD-expressing clones (52 single-copy and 41 multiple-copy clones) reflected a total of 171 different integration events (5). CAGE analysis revealed that in one of the single-copy clones expressing hG6PD, the wild-type hG6PD band was replaced by a band that migrated faster (charge change of −2) (Fig. 2). Sequence analysis revealed an A-to-G base change at nucleotide 289 (which causes an acidic Glu to replace the basic Lys 96), which was confirmed by PCR analysis (data not shown). This mutation was not detected by PCR in the producer cell line. The size of the hG6PD cDNA is 1,545 bp; as we found 1 point mutation in a total of 171 integration events, this means that there was 1 mutated nucleotide per 264,195 nucleotides (171 × 1,545). From the sequence of the hG6PD cDNA, we can predict that of the 4,635 possible point mutations, 22.4% are synonymous, 3.7% are nonsense, 47.6% are missense but electrophoretically silent, and 26.3% are missense and detectable by electrophoresis (these values are close to the previous estimates for proteins in general reported by Li and Sadler [10] and Marshall and Brown [14]). Therefore, the figure of 1 electrophoretically visible mutation per 264,195 bp must be corrected to about 1 mutation per 69,483 bp per replication cycle, i.e., μ = 1.4 × 10⁻⁵. This may still be an underestimate, because our technique misses null G6PD variants. However, these seem to be rare, because (i) we did not detect any proviral sequences in 150 clones that did not express hG6PD and (ii) 52 of the 93 clones expressing hG6PD contained only one proviral copy. Finally, there is the unlikely possibility that a missense mutation arose during the transfection; this is doubtful, because we did not detect the mutations in the producers by PCR analysis.

The replication cycle of retroviruses involves various steps. (i) Host DNA polymerase replicates the viral DNA integrated in the host genome. (ii) The host cell RNA polymerase II transcribes the provirus to produce viral RNA. (iii) Reverse transcriptase (RT) catalyzes (in two replication steps) the synthesis of viral double-stranded DNA from the RNA genome. Mutations can arise at any of these steps. Host DNA polymerase has a very low mutation rate (10⁻⁹ to 10⁻¹¹) thanks to efficient proofreading (6). In contrast, little is known about the fidelity of eukaryotic transcription. However, the evidence that different retroviruses have different rates of mutation (12, 13) and the recent finding that RNA polymerase II has some proofreading ability (8, 24) suggest that genetic variation in retroviruses can be attributed mainly to RT. Indeed, RT lacks proofreading ability, and its poor fidelity in vitro is well known (reviewed in reference 2).

The rates of different sorts of mutations have been estimated for both wild-type and replication-defective MMLVs (retroviral vectors) by using a variety of approaches and a number of noneukaryotic sequences as targets (Table 1). With respect to single-base substitutions, Varela-Echavarría et al. (25) measured the reversion rate of an amber codon in the neomycin resistance gene (neo) and reported a μ value of 2 × 10⁻⁶. Monk et al. (15) gave a higher figure, of 2 × 10⁻⁵ for a wild-type MMLV (by RNase T₁-oligonucleotide fingerprinting analysis or direct RNA sequencing). This figure is similar to that of 4 × 10⁻⁵ estimated in a retroviral vector by inactivation of the thymidine kinase gene (tk) (18).

TABLE 1.

Rates of mutation in MMLV^a

Mutation type	Virusb	Producer cell type	Method of detection	Target gene(s)	Mutation rate (mutations/bp/ replication cycle)	Source or reference
Base substitutions
All	WT	NAc	Fingerprint-sequence	LTRd-gag-pol-env	2 × 10−5	15
A to G	Vector	GP+E-86	Amber codon reversion	neo (Am)	2 × 10−6	25
All	Vector	GP+E-86; PG13	Inactivation	tk	4 × 10−5	17, 18
All	Vector	ψ-CRE	Change of charge	g6pde	1.4 × 10−5	Present study
Frame shift (±1, ±2)	Vector	GP+E-86; PG13	Inactivation	tk	1 × 10−5	17, 18
Deletion or insertion	WT	NA	Inactivation	gag-pol-env	3 × 10−5	23
	Vector	GP+E-86; PG13	Inactivation	tk	2 × 10−5	17, 26
Hypermutationf (A to G)	Vector	GP+E-86; PG13	Inactivation	tk	9 × 10−2	17, 18

^aThis table includes data from reference 18.

^bWT, wild-type MMLV virus; vector, MMLV-based retroviral vector.

^cNA, not applicable.

^dLTR, long terminal repeat.

^eg6pd, the G6PD gene.

^fClustered mutations.

In our study, we have measured for the first time the rate of forward point mutations after retroviral gene transfer, not in a reporter gene but in a potentially therapeutic human gene. By using a sensitive approach (electrophoresis followed by a functional protein assay), we have found three mutated hG6PD sequences: two among the 34 mice that underwent bone marrow transplantation and one among the 93 individual NIH 3T3 clones. Considering that in transplantation experiments, multiple stem cells contribute to the reconstitution of hematopoiesis, these two figures are entirely comparable. Both experimental approaches suggest that the frequency of point mutations after retroviral gene transfer is small but not negligible. In fact, the finding of 1 point mutation in 171 proviral integration events in NIH 3T3 clones yields an estimated μ value of 1.4 × 10⁻⁵, similar to that reported for noneukaryotic target sequences. It is likely that most studies of retrovirus-mediated gene transfer have missed or overlooked this phenomenon, because no tests are usually carried out for such qualitative changes, particularly for point mutations. In most cases, the transfer of a mutated gene would be just one of the factors that contribute to reducing the effectiveness of gene transfer. However, in some cases and in certain genes, a point mutation might produce a gain-of-function or an immunogenic protein with potentially more deleterious consequences. We suggest that a qualitative analysis of the transferred gene should be included in the quality assessment of gene transfer protocols for human therapy.