The genetic assays that measure cDNA recombination separately from transposition were used previously to screen a library of Tf1 elements for mutations that were capable of reverse transcription but not integration (1
). Although the entire transposon was mutagenized, the majority of the elements identified had mutations in IN. This observation coupled with other analyses of the mutants demonstrated that the genetic assays were able to identify transposons that were specifically defective for integration. These results suggested that the same assays used with mutagenesis of just RT could identify a specific class of mutations that inhibited specialized functions of RT. The sharp clustering of the current mutations in RNase H was another indication that the genetic assays could identify amino acids with a specific function not required for cDNA synthesis.
Tf1 with the mutations that clustered in RNase H produced cDNA that was indistinguishable in size and amount from that of wild-type Tf1. The use of a restriction digest and the location of the cut site in the cDNA was designed to provide a sensitive assay for intermediates of reverse transcription. The wild-type amounts of the BstXI fragment indicated that the mutants produced normal levels of full-length double stranded cDNA. The one exception was I784T, a mutation in the cluster that showed reduced cDNA. One possible explanation for why this mutant exhibited high levels of homologous recombination is that it produced single-stranded intermediates of heterogeneous length that did not resolve on the blot.
Immunoblots indicated that the mutations in the cluster did not change IN levels and except for one mutation, they did not change amounts of RT either. Interestingly, L783I caused a significant reduction in the 60-kDa species of RT without affecting the 72-kDa PR-RT species. Although it is not known which form of RT is responsible for reverse transcription in vivo, the dramatic reduction in the 60-kDa species suggests that the 72-kDa RT has sufficient activity to produce wild-type levels of cDNA (Fig. ). This possibility is supported by a study of a closely related transposon, Tf2, that found the PR-RT species was likely responsible for reverse transcription (7
The cDNA produced by the mutants was characterized further to identify defects that could inhibit integration. We examined the sequences at the 3′ ends of the cDNA because small changes could remove or reposition the CA dinucleotide that, for integration to occur, must be at the 3′ terminus. Our initial analysis of cDNA produced by wild-type Tf1-neo
AI revealed that the sequences at the 3′ end of the minus strands were surprisingly heterogeneous. The dominant species constituted just 20% of the cDNA. In comparison, analyses of cDNA produced by HIV-1, Moloney murine leukemia virus, and the LTR-retrotransposon Ty1 indicate that 50 to 60% of the cDNA terminates precisely at the consensus site (3
). In addition to the broad distribution of the Tf1 cDNA, another unusual feature was that 85% of the cDNAs terminated with 3′ untemplated additions. This is a significantly higher level than that of Ty1, where 25% of the minus-strand cDNA had nontemplated additions (22
). The nontemplated additions present at the 3′ ends of cDNAs are thought to result from RT because in purified form it has terminal transferase activity (24
The high proportion of untemplated additions at the end of cDNA produced by wild-type Tf1-neo
AI would in principle inhibit integration by placing the CA dinucleotide at an internal position. Although the INs of some retroviruses and LTR-retrotransposons have processing activities that remove nucleotides past the CA, the position of the minus-strand primer suggests that the IN of Tf1 should lack such an activity (17
). Nevertheless, recent work with recombinant protein demonstrated that Tf1IN does possess a processing activity capable of removing nontemplated additions (6a
). Thus, the products with termini that include an internal CA have the potential to be integrated.
The most prevalent species of cDNA produced by wild-type Tf1-neoAI terminated beyond the CA with two T's that likely were templated by the last two A's of the PPT. The resulting species of cDNA would have a two nucleotide extension that would have to be removed by the processing activity of IN to participate in integration. That this species is the dominant product of wild-type Tf1-neoAI is consistent the idea that it is the principal substrate for integration in vivo. Other evidence that this species ending at the −2 position was the major donor for integration in vivo came from analyses of cDNA produced by Tf1 with the mutations that clustered in RNase H. Of the six mutations in RNase H that were studied, five caused substantial reductions in the dominant species of cDNA ending at position −2. For R786H, R786C, and L783I, the major product was the only species to be significantly reduced. The correlation of reduced integration with the specific drop in the major cDNA indicates that, in vivo, this species ending at position −2 is likely the dominant substrate for integration.
The changes in the sequence profiles at the 3′ end of minus-strand cDNA revealed important information about how the mutations in RNase H altered the cDNA. Of the five mutations that caused significant reductions in the major species of cDNA, N782S, L783I, and R786H, exhibited corresponding increases in cDNAs with PPT sequence. In each case, the amount of the cDNA lost from the population of the major species matched quantitatively the increase in species that ended with the PPT sequence. This strong correlation indicates that the mutations we studied caused a severe defect in the ability of RNase H to either recognize or remove the PPT. Such a defect would cause the residual PPT sequence at the 5′ end of the plus strand to be templated into the 3′ end of the minus strand. The evidence that the mutations altered the processing of the PPT is supported by the observation that all six of the mutations we identified corresponded to residues associated with the RNase H primer grip, the domain of HIV-1 RNase H that in a crystal structure binds the DNA annealed to the PPT (30
). Of the three mutations that caused quantitative shifts of cDNA from the major class to species with PPT sequence, R786H and N782S correspond to amino acids of HIV-1 that interact directly with nucleotides opposite the PPT. The third mutation resulting in a quantitative shift of cDNA, L783I maps one amino acid from a contact with the nucleic acid at a residue that forms the base of a critical alpha helix (Fig. ). Thus, L783I is also likely to disrupt specific interactions with nucleic acid.
Two of the mutants, S749L and R786C, showed reductions in the major cDNA species without a corresponding increase in cDNA ending with PPT sequence. Neither of these mutants exhibited corresponding increases in cDNA anywhere in the 40 nucleotide windows (Fig. and ). Thus, S749L and R786C may have increased levels of cDNA that terminate prematurely in positions that were close to or downstream of the primers we used for ligation mediated PCR. Evidence supporting this possibility for S749L was that approximately 25% of the cDNA species we sequenced terminated hundreds of nucleotides before the 3′ of the minus strand. The only reason we were able to clone and sequence these species was that tandem ligations of Rag208 allowed the product to migrate within the window of the agarose gels that we analyzed. Such tandem ligations were not detected or expected with cDNA from other mutants because Rag208 was 3′ blocked. Apparently this block was incomplete in the ligation with cDNA from the S749L mutant.
Despite the complexities of the S749L and R786C substitutions, the other three mutations that reduced the amounts of the −2 species, N782S, L783I, and R786H, caused increases in the cDNA that ended with PPT sequences. Along with the crystal structure of HIV-1 RT, these results suggest the three amino acids in Tf1 RNase H recognize the PPT due to direct contacts they make with the nucleic acid. Damaging this ability did not cause any apparent defects in the other functions of RNase H required for reverse transcription. The function damaged by these mutations was likely the specific removal of the PPT RNA from the 5′ end of the plus strand. This is supported by the preponderance of full-length PPT associated with the cDNA of R786H and N782S. However, there was a more complex possibility that the mutations move the positions of the RNase H-mediated cleavages that “select” the PPT such that the plus-strand primer was 5′ of the PPT and as a result the PPT sequence in the plus strand would be copied into DNA. In DNA form, the PPT could not then be removed by RNase H. This scenario was ruled out by the primer extension experiment that showed there was no PPT DNA at the 5′ end of the plus strand.
The PPT sequence at the 3′ end of the minus-strand cDNA that resulted from the RNase H mutants was likely the reason transposition was reduced since the critical CA was no longer at the 3′ terminus. Even though the IN of Tf1 has processing activity, the processing activities of retroviral INs are unable to remove double-stranded sequences longer than 4 base pairs from the CA (4
). The processing activity of Tf1 IN has a similar inability to process long double-stranded sequences distal to the CA (6a
One of the most interesting results from these experiments was that reverse transcription could proceed through the strong stop transfers in the absence of primer degradation. This indicates that primer removal is not a necessary component of reverse transcription. A suggestion that this is true for another element is that the retrotransposon Ty1 proceeds through plus-strand transfer without first copying the tRNA primer into cDNA (12
Our study provides strong evidence that the RNase H primer grip of Tf1 makes contacts with DNA that are specifically required for the removal of the PPT. Other reports have suggested that the primer grip may have a role in recognition of the PPT. The crystal structure of HIV-1 RT in complex with the PPT RNA-DNA hybrid clearly documents direct interactions between the RNase H primer grip and the DNA of the RNA-DNA hybrid (30
). This provided invaluable information about the amino acids that make contacts and suggests they may participate in interactions with the PPT in vivo. However, it is not clear from the structure what function these interactions may have in vivo and whether these interactions are specific for PPT sequence.
More physiological support for a role of the RNase H primer grip in PPT recognition came from mutations in the C-helix of Moloney murine leukemia virus RT. This helix is adjacent to the RNase H primer grip and is absent in other retroviruses. Targeted alanine scanning mutations cause significant reductions in overall RNase H cleavage activity. However, some increases occur in the two-LTR circles with PPT sequence between the LTRs (18
). In the case of HIV-1, mutations I505A, Y501A, and N474A+Q475A were created in avector system that supports single round replication (8
). WhileI505A has little effect on virus titer, Y501A and N474A+Q475A reduced the titer significantly. These mutants were found to have 10-fold less initiation of DNA synthesis, a defect that is independent of the PPT. Analyses of two-LTR circles revealed that the majority of the cDNA produced by Y501A and N474A+Q475A had deletions in the U3 and U5 regions of the LTRs. Nevertheless, an increase in DNAs with PPT sequence between the LTRs was observed. The lack of specificity of the mutations in HIV-1 and Moloney murine leukemia virus compared to the mutations we isolated may be because the substitutions used were arbitrary and likely disrupted other important functions of RT.
The function of the RNase H primer grip of HIV-1 was further examined using recombinant RT with alanine mutations (28
). Although the mutations in the amino acids of the primer grip did not reduce DNA polymerase activity of RT, they did cause sharp reductions in polymerase independent RNase H cleavage of nonspecific RNA-DNA hybrids. However, the extent of these defects were more extensive when the mutant enzymes were challenged to cleave the PPT in the context of RNA or to remove the PPT from the 5′ end of a plus-strand RNA-DNA chimera. These data suggested the RNase H primer grip contributes specificity to the cleavage reactions. This specificity is not as pronounced as was observed with the RNase H of Tf1, perhaps again because the substitutions with alanine may have had secondary effects.
There is considerable evidence that the structure of the PPT plays a key role in resisting the RNase H mediated cleavages (27
). High resolution analyses of RNA-DNA hybrids with the sequence of the HIV-1 PPT identified structural anomalies (6
). The position of the anomalies corresponds to the cleavage sites, suggesting that it may be the PPT itself that positions the cleavages. However, the mutations we isolated indicate that Tf1 RT also plays a key role in directing the PPT cleavages. Taken together, our data are the first to indicate that RT contains amino acids that function solely to recognize the PPT. If the corresponding residues in HIV-1 RT provide the same specificity, then they become an important target for antiviral strategies. Inhibitors of PPT processing in combination with compounds that affect other activities of RT would significantly reduce the number of drug-resistant forms of HIV-1 RT.