Technical Comments

HIV-1 RNA Editing, Hypermutation, and Error-Prone Reverse Transcription

See allHide authors and affiliations

Science  06 Apr 2001:
Vol. 292, Issue 5514, pp. 7
DOI: 10.1126/science.292.5514.7a

Bourara et al. (1) reported that transcripts of the human immunodeficiency virus–type 1 (HIV-1) were subject to RNA editing in chronically infected cells. They observed multiple guanine-to-adenine (G-to-A) and cytosine-to-uracil (C-to-U) changes in several regions of the HIV-1 RNA; commonly, a G-to-A change in the untranslated leader was present exclusively in spliced HIV-1 messenger RNA (mRNA), but not in the unspliced RNA and the proviral DNA genome. Changes in the viral protein R (vpr) gene were present in spliced and unspliced HIV-1 RNA extracted from the cell, but not in the unspliced RNA genome that is packaged in virion particles. Therefore, Bourara et al. proposed that post-transcriptional mRNA-editing events occur for a subset of viral RNAs. Known editing mechanisms, however, cannot easily explain these changes in the HIV-1 genome. We here propose an alternative mechanistic model based on HIV-1 reverse transcription to explain some of the nucleotide changes.

Bourara et al. argued that the chronically infected cells represent a clonal population with one proviral genome, based on the idea that reinfection of HIV-producing cells is restricted by a mechanism known as superinfection interference. There is convincing evidence, however, that viral interference is not complete in chronically infected cell cultures (2, 3). Thus, reverse transcribed viral genomes are likely to end up in the DNA of a subset of cells, thereby producing a heterogeneous cell population. Also, it is crucial to understand why a chronically infected cell could be established with this cytopathic virus. Viral latency is frequently associated with mutational inactivation of the viral transcription machinery, in particular the essential tat-TARaxis (4, 5).

These mechanisms can result in a very complex population of chronically infected cells. All cells may harbor the original proviral genome that is transcriptionally impaired, but there may be several subsets of cells with additional proviruses that underwent at least one round of reverse transcription. These minority proviruses will not be picked up by Southern blot analysis, but they may largely determine the HIV-1 RNA content of the cell pool. The proviruses can have the typical mutations due to reverse transcription errors, which—because of mutational inactivation of motifs that regulate either splicing (therev-RRE axis) or RNA packaging (Gag protein and the psi motif)—may not be distributed equally over the spliced versus unspliced RNA and the cellular versus virion HIV-1 RNA. Thus, the chronically infected cell system is far too complex to provide evidence for RNA editing based on sequence differences in the viral DNA and RNA.

There is also some experimental evidence that these typical mutations arise by means of reverse transcription. We have accumulated sequence data of spontaneous HIV-1 variants that evolve in long-term tissue culture infections that were started with molecular clones of known sequence. This experimental system reflects a natural infection in that beneficial errors introduced during error-prone reverse transcription will end up in the majority of proviruses by natural selection. These studies focused on the untranslated leader region of the HIV-1 RNA genome. We frequently observed the identical G-to-A change at position 181 that was also reported in the RNA-editing study (Fig. 1A). This residue is immediately upstream of the primer-binding site (PBS) that base-pairs with the tRNA primer for reverse transcription. Remarkably, we found many other G-to-A changes in this region upstream of the PBS, whereas few sequence changes were observed in the regions either further upstream or downstream of the PBS (Fig. 1, A and B).

Figure 1

(A) Distribution of spontaneous mutations surrounding the PBS signal of the HIV-1 RNA genome. The 105/247 region of the untranslated leader is plotted, and the PBS motif is highlighted in the gray box. The HIV-1 leader sequences were obtained in 87 independent virus evolution experiments that each lasted approximately three months (9–11). (B) Mutational bias for a total of 127 leader mutations. (C) The 5′ and 3′ nucleotide context of all G's involved in a G-to-A mutational event. Data have been corrected for the occurrence of the different NGN triplets in the HIV-1 leader segment.

Because the mutations cluster in the region that is copied first during reverse transcription, we propose that G-to-A mutation is a typical property of the reverse transcriptase (RT) complex that executes the initial stages of reverse transcription. This initial phase of reverse transcription may be hindered by a low deoxycytidine 5′-triphosphate (dCTP) concentration in the virion particle, which can trigger G-T mispairing that is the likely cause of biased G-to-A mutation in the mechanism of hypermutation (6). The subsequent phases of reverse transcription will occur in the cytoplasm of infected cells, and the surplus of deoxynucleotide triphosphate (dNTP) building blocks may thus effectively turn off this typical mutational bias. The PBS motif itself is almost invariable, which is not unexpected because this sequence is copied from the tRNAlys3 primer during the process of reverse transcription. The position immediately 3′ of the PBS is highly mutable (T200N), because this position corresponds to the extension point after the second strand transfer of reverse transcription. The mechanisms of error-prone initiation of reverse transcription and hypermutation may be very similar, but hypermutation cannot explain the clustering of mutations upstream of the PBS. We also analyzed the sequence context of the residues that undergo the G-to-A change. A preference for mutation of G residues in the NGT trinucleotide sequence is apparent (Fig. 1C), whereas hypermutable G residues are usually in the NGA sequence context (7, 8).

These combined results are consistent with the hypothesis that reverse transcription, rather than RNA editing, is responsible for the acquired leader mutations. Similarly, the preference for G-to-A and C-to-U changes in the vpr gene is a hallmark of error-prone reverse transcription (Fig. 1B).


Response: Berkhout et al. propose an interesting alternative to explain the nucleotide changes in HIV-1 transcripts we observed (1) in chronically infected cells: they suggest that changes are produced via the error-prone reverse transcription and hypermutation steps (2, 3). Their hypothesis rests on five assumptions. (i) All cells may harbor the original provirus; (ii) the original provirus may be transcriptionally impaired—for example, via thetat-TAR axis (4, 5); (iii) a minority of cells are superinfected and contain additional proviral genomes generated by reverse transcription; (iv) the viral transcripts found in infected cells are largely derived from the few cells containing the mutated provirus generated after superinfection; and (v) mutations affect viral gene expression such as splicing (rev-RRE axis) and RNA packaging (Gag and the psi motif).

Is it necessary to propose such sophisticated situations to conclude that the main cause of HIV-1 variability is the reverse transcription process? Whether the answer is yes or no, mutation events are what ultimately lie at the root of the model of Berkhout et al., and the modifications thus produced are transferred to the offspring, a process that gives rise to a heterogeneous population of virus by natural selection.

The differences between our model and that of Berkhout et al. can be, and indeed were, experimentally tested by analyzing the viral genome produced by infected cells. The provirus sequence will determine whether virions are generated by the main provirus or by the “minority” of mutated proviruses issued from superinfection. The hypothesis of Berkhout et al. requires that all virions should bear mutated genomes, because the main provirus is transcriptionally impaired. Some mutants may be impaired in packaging and thus may not be represented in the viral population, whereas those that are affected in the rev-RRE axis should be. Additionally, under this model, one can invoke other complications, such as the notion that mutations may occur simultaneously in therev-RRE axis and the Gag-psi motif.

The main problem with this idea is that it requires multiple events—inactivation of the original provirus, superinfection of some cells with viruses that went through “one round of reverse transcription,” and mutations affecting either splicing or RNA packaging. Each of these possibilities in itself is feasible, but the low probability of the simultaneous emergence of several random, low-frequency events makes this explanation difficult to accept. In very long-term cell cultures, of course, some of these random events may lead to such cell populations; if all these mutations take place, they can be tested by RT polymerase chain reaction (PCR). The “minority” of mutated proviruses will be spotted using powerful PCR technology, with the appropriate primers able to amplify mutant proviruses, even if they are present at very low levels.

Our study (1) was not based on the idea of a total absence of superinfection, but on the fact that the cellular model used in our experiments drastically reduced this possibility. Moreover, we used short-term cultures from frozen cells, not multiple passes lasting several months. Southern blot analysis allowed us to verify that chronically infected cells were unchanged during the study period. Berkhout et al. maintain that the chronic infection state is “far too complex” to be used as a model; their argument is based on the observation of superinfection in long-term cell culture studies (6) with durations of 3 months to 1½ years. It should be noted that these studies described as superinfection the differences found by Southern blot analysis of genomic DNA using a viral probe; genetic rearrangements that cells should undergo in long-term cultures were not considered.

Thus, to the questions raised by Berkhout et al., we would answer (i) that the “main” provirus does not show modifications in the tat-TAR axis; (ii) that no traces of mutants were detected in viruses produced by chronically infected cells (i.e., all carried the “main” proviral sequence); (iii) that, based on this observation, we conclude that the “main” provirus was not mutationally inactivated for transcription; and (iv) that no mutant sequences were found from the analysis of proviral DNA.

Berkhout et al. argue that the emergence of the mutation adjacent to the PBS is a hallmark of error-prone reverse transcription (7–9). Other authors (10–12), however, have offered a different explanation of this event—and no matter which turns out to be the best explanation, some of these putative mutants should have been found when the viral RNA genome was sequenced. In the experimental evidence that we described (1), mutational events were not detected. Because transcripts generated by the transcription-competent provirus were specifically modified without changing the proviral sequence, we are convinced that these events were generated by RNA editing.


Navigate This Article