Analysis of polypurine tract-associated DNA plus-strand priming in vivo utilizing a plant pararetroviral vector carrying redundant ectopic priming elements.

Initiation of DNA plus-strand synthesis in most reverse-transcribing elements requires primer generation by reverse transcriptase-associated RNase H at one or more template polypurine tracts (PPTs). We have exploited infectious clones of the plant pararetrovirus cauliflower mosaic virus carrying redundant ectopic plus-strand priming elements to study priming in vivo. Ectopic priming generated an additional discontinuity in progeny virion DNA during infection of plants. We found that altering the length of the 13-base pair PPT by +/-25% significantly reduced priming efficiency. A short pyrimidine tract 5' to the PPT, highly conserved among diverse reverse-transcribing elements, was shown to play an important role in PPT recognition in vivo. The predominant DNA plus-strand 5' end remained 3 nucleotides from the PPT 3' end in mutant primers that were longer or shorter than the wild-type primer. Use of an ectopic redundant primer to study replication-dependent priming was validated by demonstrating that it could rescue infectivity following destruction of the wild-type priming elements. We propose a model for plant pararetroviral plus-strand priming in which pyrimidines enhance PPT recognition during polymerase-dependent RNase H cleavages, and suggest that fidelity of primer maturation during polymerase-independent cleavages involves PPT length measurement and 3' end recognition by RNase H.

Retro-elements (REs), 1 the propagation of which involves copying genomic RNA into DNA mediated by reverse transcriptase (RT), are found in organisms representing the major groups of eukaryotes. REs include animal retroviruses, animal and plant pararetroviruses, and retrotransposable elements (1). Although REs are diverse, most share several common features. Important among these is the pol gene product specifying multiple enzyme activities associated with reverse transcriptase (RT) involved in RE DNA synthesis and processing (2,3).
REs have a different origin of replication for synthesis of each strand of the double-stranded (ds)DNA phase. First, reverse transcription of the genomic RNA into minus-strand DNA is initiated in most REs by a host tRNA (4). During minus-strand synthesis, RT-associated RNase H activity degrades the RNA template (5,6). This degradation appears to occur in at least two stages. Processive DNA minus-strand synthesis is accompanied by a polymerase-dependent, non-processive RNase H degradation of the template leaving RNA fragments. These fragments are probably further degraded by subsequent polymerase-independent RNase H cleavages (7)(8)(9). The RNA primer for DNA plus-strand synthesis is generated during these steps by RNase H cleavage at a specific site called the polypurine tract (PPT), 13-18 nucleotides long. For REs producing an integrating linear double-stranded DNA with long terminal repeats (LTRs), precision of plus-strand initiation is essential for RE viability as one LTR border is defined by this step. Therefore, understanding the role of the PPT is important, particularly in retroviruses where plus-strand initiation is a possible target for therapeutic control.
Several studies of PPT function have involved analysis of products made by RT added to RNA primer/DNA template combinations in vitro (10 -13) but it is not clear how such reactions relate to plus-strand priming in vivo. Mutagenesis of the LTR-associated PPT could render an RE non-functional in vivo limiting usefulness of this approach. Members of the lentivirus and spumavirus groups of retroviruses, and the yeast retrotransposon Ty1, have a second central PPT colinear with the pol gene (14 -16), which could be more amenable to analysis in vivo. Studies have shown that mutations in the pol-associated PPT, leaving the colinear pol amino acid sequence unchanged, impaired priming and retroviral replication in transfected cells (17,18) but did not affect Ty1 transposition in yeast (16). Charneau et al. (18) also inserted a 29-bp linker containing the PPT sequence into a new site in the HIV-1 genome, where it primed DNA plus-strand synthesis. With this approach, the primer could be moved potentially to a new genome site for analysis without interference from colinear functions.
Plant pararetroviruses, such as cauliflower mosaic virus (CaMV), offer great potential to study plus-strand priming because clones of their circular double-stranded DNA genome can be infectious to plants, and, some members of the group have multiple PPT-associated priming sites (19). Different plant pararetroviruses have 1-3 PPTs close to virion DNA discontinuities, which are structurally similar to the HIV-1 pol-associated DNA discontinuity (Fig. 1). CaMV has two PPTs; one is colinear with a non-essential gene and can be deleted without loss of viral infectivity (20,21). Priming at the CaMV PPTs in vivo is less that 100% efficient (22,23), but additional priming seems to occur at other PPT-like sequences (24) and by hairpin fold-back (22,25), although it is not known if these extra events contribute to viral replication. CaMV PPTs have functional polarity (26), and we have recently shown that plus-strand priming can be initiated in vivo by insertion of a PPT at an ectopic genome site (27). These features have enabled us to investigate the pararetroviral plusstrand priming mechanism in vivo.

EXPERIMENTAL PROCEDURES
Virus, Plants, and Kits-CaMV isolates used were Cabb B-JI and Aust (21), and CM4 -184 (20) maintained as infectious clones. Infection studies were with turnip plants (Brassica rapa-rapifera, cv. "Just Right"). Celite abrasive was supplied by Celite Corp. pGEM cloning vectors were from Promega. Mutagenesis was performed using the QuikChange procedure (Stratagene). Primer extension was performed with the T7 sequencing kit supplied by Amersham Pharmacia Biotech.
Infection of Plants-Turnip seedlings were grown in a greenhouse (minimum day length, 16 h) at 20°C. Plants were mechanically inoculated using Celite abrasive with sap from infected plants, or cloned viral DNA constructs (1-4 g/plant) digested with a restriction enzyme to liberate viral DNA from the cloning vector. Symptoms were scored when consistent on a minimum of 4 plants.
Construction of Mutant Viral Genomes-Molecular biological manipulations were based upon those described by Sambrook et al. (28). Artificial plus-strand primers, (ϩ)ps elements, were made by annealing pairs of complementary, 5Ј-phosphorylated oligonucleotides (Table I, oligonucleotides 1-28) generating terminal NheI sites. The annealed oligonucleotides were cloned into the unique NheI site introduced into infectious CaMV 35 S promoter deletion mutants M4 and M10 (29) as described by Noad et al. (27). Insert orientation was checked by PCR.
Authentic CaMV (ϩ)ps elements were inactivated by mutagenesis as follows. For the (ϩ)ps1 element colinear with gene V, the clone MP12 (27) was digested with PstI and the fragment (3430 -5386) containing the cloning vector (pGEM5) inserted at the unique SalI site (4836), and the site to be mutated (ϳ4210), was self-ligated and cloned. Mutagenesis was performed using the QuikChange procedure with mutagenic oligonucleotides (Table I, oligonucleotides 29 and 30). The mutation was checked by sequencing, and an NaeI fragment (pGEM5 site to CaMV site 3910) containing the change was isolated and used to replace the wild-type fragment in the full-length CaMV clone pBJI5Sal, generating clone PB1. The (ϩ)ps2 priming element colinear with gene II was mutated by exploiting the adjacent unique XhoI site to design an oligonucleotide containing this site and the mutations (Table I, oligonucleotide 31). A fragment was amplified between this and an upstream site (740 -761), and an NsiI-XhoI fragment (1042-1644) was isolated from the PCR product. This fragment was used to replace a similar fragment in a BstEII (126)-XhoI subclone of the CaMV genome (producing clone BB5) and this, in turn, replaced the wild-type fragment generating the full-length clone BB8. The (ϩ)ps2 sequence was also deleted using a strategy similar to that for BB8 but employing an oligonucleotide with a 33-bp in-frame deletion (Table I, oligonucleotide 32) relative to the wild-type CaMV sequence, generating clone BB12. Clone BB9, containing both mutant (ϩ)ps elements, was made by replacing the BstEII/XhoI-cut fragment from BB8 with that from BB5 producing clone BB9. A mutant genome containing the modified (ϩ)ps1 and the deleted (ϩ)ps2 was created by replacing the SpeI(110)-XhoI fragment of PB1 with that from BB12 producing clone BB21. Construct PM1 comprised the deleted gene II region of CaMV isolate CM4 -184, the mutated (ϩ)ps1 element, and a linker in the 35 S promoter. PM1 was made by ligating the 3.0-kb StuI fragment from clone MP12 (27) to the complementary 7.6-kb StuI fragment of clone PB1. Construct GG5 was a modification of PM1 in which an artificial (ϩ)ps element (Table I, oligonucleotides 1-2) was introduced in forward orientation into the NheI linker in the 35 S promoter. All mutations introduced into the viral genome were checked by sequencing before infection and after recovery from plants if constructs were infectious.
Analysis of Virion DNA Discontinuities-Priming of DNA plusstrand synthesis was determined by detecting discontinuities in CaMV virion DNA purified as described by Covey et al. (30). Denaturing gel electrophoresis of CaMV virion DNA and detection by blot-hybridization was as described by Noad et al. (27). Efficiency of plus-strand priming at the ectopic site (ϩ)ps3 was determined semiquantitatively by scanning autoradiograms following denaturing Southern blot analysis. Scanned lanes were analyzed using ImageQuant software to integrate peak areas. Precise determination of DNA plus-strand 5Ј ends was by primer extension using appropriate primers upstream of the (ϩ)ps2 and (ϩ)ps3 elements, respectively. Primer extension was performed on 1-2 g of virion DNA using the method described by Medberry et al. (31).

RESULTS
The Polypurine Tract and 5Ј-Pyrimidines Are Both Required for Plus-strand Priming in Vivo-We have adopted a new descriptor for plant pararetroviral plus-strand priming sequences, (ϩ)ps, since the term PPT is an incomplete definition and previous abbreviations in plant pararetroviruses referred to virion DNA gaps rather than priming elements themselves. The two authentic CaMV (ϩ)ps elements are co-linear with viral genes (Fig. 1). The CaMV reverse transcriptase gene (gene V) contains (ϩ)ps1, and the aphid transmission protein gene (gene II) contains (ϩ)ps2. Although gene II is not essential for viral replication and can be deleted, our studies 2 of (ϩ)ps2 have shown that mutagenesis can interfere with co-linear functions affecting CaMV pathogenicity. To circumvent this problem, we exploited our discovery that a redundant 31-bp copy of (ϩ)ps2 inserted into the CaMV 35 S promoter (Figs. 1A and 2A) efficiently primes DNA plus-strands during infection of plants with no detected effect on viral pathogenicity (27). Our assay for functionality of the inserted primer is to measure generation of a new discontinuity in progeny CaMV DNA. Following denaturing gel electrophoresis, virion DNA of the M4 vector, lacking the inserted priming element ( Fig. 2A), resolves as three major single-stranded fragments (Fig. 2B) comprising the genome-length product of minus-strand priming (␣-strand, 8 kb), and two fragments resulting from plus-strand priming at (ϩ)ps1 (␤-strand, 5.4 kb) and (ϩ)ps2 (␥-strand, 2.6 kb). A 3.8-kb fragment (␤Ј-strand) is a product of failed second strand transfer during reverse transcription (Fig. 2, B and C). The presence of an additional primer in the 35 S promoter, in vector M4Gf, produces a new discontinuity in the ␤-strand (Fig. 3). The efficiency of priming at the new site was measured in denaturing gels by assessing the reduction in amount of the 5.4-kb single-stranded DNA fragment and replacement by two additional fragments of 3.1 and 2.3 kb. The amount of residual ␤-strand DNA in M4Gf shows that priming at the new site is less than 100% effective. By scanning gels, integrating the peak areas (see Fig. 2B), and comparing amounts of ␤-strand, we 2 N. Al-Kaff, D. Turner, and S. Covey, unpublished observations.  (Table I) is recognized in 85-90% of virion DNA molecules.
The relative activity of mutant (ϩ)ps elements was then tested at the ectopic site (Fig. 3). This was quantified by comparing the intensity of the 3.1-kb band (␦-strand) with that of the 2.6-kb band (␥-strand) as described under "Experimental Procedures." Both of the authentic CaMV (ϩ)ps elements have a similar organization, although the sequences differ slightly. In each case, the PPT is preceded immediately by a pyrimidinerich sequence then an A-rich region 5Ј of this (see Fig. 7A). We first tested the involvement of these conserved upstream elements in priming at the ectopic site. Removal of the 7-nucleotide A-rich sequence in construct (ϩ)ps3:p had no significant effect on priming compared with (ϩ)ps3:o. However, deletion of the 5 5Ј pyrimidines, conserved among REs, in construct (ϩ)ps3:q, appeared to abolish priming (Fig. 3B). Deletion of 6 nucleotides, mostly purines, immediately upstream of the Gbox in construct (ϩ)ps3:s, reduced the frequency of priming to about 13% of the level of the full-length ectopic sequence, while deletion of the G-block, construct (ϩ)ps3:r, eliminated detectable priming. The seemingly essential nature of the pyrimidines was unexpected since there is evidence that they are not required for retrovirus plus-strand priming (12). However, deletion of the pyrimidines and the upstream purines, leaving the PPT intact (construct (ϩ)ps3:u), allowed a level of priming approximately one-fifth that of the full-length primer. Further deletion of the 6 5Ј nucleotides, predominantly purines, within the PPT in construct (ϩ)ps3:v, also abolished priming. Substitution of 2 G-box G nucleotides for A nucleotides in construct (ϩ)ps3:t, reduced priming efficiency to about 68%. A more significant reduction was seen in construct (ϩ)ps3:w and, to a lesser extent in (ϩ)ps3:x in which the PPT was shortened by 1 and 2 nucleotides, respectively (Fig. 2). These experiments show that maximal priming activity requires both a PPT sequence of optimal length together with upstream pyrimidines.
Further mutants were constructed to investigate the effects of small deletions and insertions on primer efficiency and primer fidelity (described below). We confirmed involvement of the pyrimidines by removing 2 of the 5 T nucleotides in a construct lacking upstream purines that interfere with priming, (Fig. 4, construct (ϩ)ps3:b) and observed a reduction in the level of priming efficiency to 75% that of the full-length element. In an attempt to increase priming efficiency, we added 3 extra T nucleotides (construct (ϩ)ps3:a), but this produced priming with an efficiency indistinguishable from the wild-type sequence (Fig. 4). Removal of 3 of the 7 G nucleotides (construct (ϩ)ps3:d) significantly reduced priming efficiency to 11%. A slight reduction in efficiency was also found with construct (ϩ)ps3:c containing 2 extra G nucleotides (Fig. 3). These experiments define the limits of the cis-acting elements required for CaMV plus-strand priming in vivo. Thus, maximal priming activity requires 5 5Ј pyrimidines and about 13 nucleotides in the PPT. Loss of more than 3 purines virtually abolishes priming, as does loss of all of the pyrimidines. Increasing the length of the PPT also reduced priming efficiency.
Factors Affecting Plus-strand Priming Fidelity in Vivo-The 5Ј ends of the authentic CaMV plus-strands have been mapped previously to fixed positions (32), as is the case with other REs. Therefore, we wanted to determine (i) whether priming fidelity was maintained at the ectopic site and (ii) the effects of mutations on priming precision in vivo. We first checked the plus- strand 5Ј end at the authentic (ϩ)ps2 site in gene II, by primer extension. The 5Ј end was found to be at a G residue 3 nucleotides upstream of the 3Ј end of the PPT portion of the (ϩ)ps2 site (Fig. 5A). This is 1 nucleotide upstream of that previously reported by Richards et al. (32). In addition to the predominant 5Ј end, lesser secondary ends were observed 1 or 2 nucleotides to the 5Ј side of the major end. Interestingly, a gene II variant (d2) we have obtained following mutagenesis, 2 with 2 of the PPT G nucleotides substituted for T nucleotides, showed essentially the same priming fidelity as the wild-type sequence (Fig. 5, A and D).
We then determined priming fidelity at the (ϩ)ps3 ectopic site and compared it with that at the authentic site. Priming by (ϩ)ps3:o gave the same predominant 5Ј end as was found at the wild-type (ϩ)ps2 priming site (Fig. 5, compare B with A). In different primer extension experiments, the degree of heterogeneity around the major 5Ј end sometimes varied. The distribution of ends shown in Fig. 5D has been assessed from two or more primer extension experiments. None of the mutations  studied altered the position of the major 5Ј end relative to the 3Ј end of the PPT. However, the relative abundance and positions of the secondary 5Ј ends varied in some mutants. For instance, (ϩ)ps3:t showed a slight bias of minor ends to the 5Ј side and (ϩ)ps3:x to the 3Ј side of the major 5Ј end (Fig. 5, B and  D). Mutations affecting the 5Ј pyrimidines had no significant effect on the major 5Ј end. The most significant effect on plusstrand priming fidelity was observed in construct (ϩ)ps3:c in which 2 additional G residues had been inserted into the PPT. This construct produced two major 5Ј ends with the predominant end located 3 nucleotides from the PPT 3Ј end, as before. However, the second major end was displaced 2 nucleotides to the 5Ј side, which places this end at the same relative position from the 5Ј end of the PPT as in the wild-type sequence (Fig. 5D). These experiments show that DNA plus-strand 5Ј end positioning by the ectopic priming element is essentially the same as at the authentic site. Additionally, the start site appears to be controlled by the length of the PPT portion of the (ϩ)ps element with a primary fixing point at the PPT 3Ј end.
Priming from Tandemly Arranged (ϩ)ps3 Elements-During assembly of one (ϩ)ps3 construct into CaMV vector M10, we isolated a clone with three tandem (ϩ)ps3:o elements, each in forward orientation (Fig. 6A). This construct was infectious and relatively stable, although a recombinant with a single (ϩ)ps3:o element was cloned from PCR reactions on progeny virion DNA. Primer extension of the (ϩ)ps3:3o virions showed a correctly positioned DNA 5Ј end at each of the three priming sites (Fig. 6B). However, the discontinuities resulting from priming at the upstream elements would only be detectable by primer extension if priming at the respective downstream element(s) had not occurred (Fig. 6C, i-iii). Priming at all three tandem sites in any one virion DNA molecule (Fig. 6C, iv) would not be detectable by primer extension, so we cannot exclude more than one priming event per molecule. An additional 5Ј end was mapped to a non-authentic purine-rich sequence in the most distal element (Fig. 6B) that we had not seen in other primer extension experiments. This experiment indicates that the presence of adjacent primers does not affect primer fidelity, although there may have been some competition for selection and generation of additional ends due to the complexity of the sequence environment.
Pararetrovirus Infectivity Rescued by an Ectopic Plus-strand Primer-The experiments above demonstrated the relative efficiency of variant (ϩ)ps elements when they were redundant to viral replication. Although primer redundancy is a normal feature of CaMV, we wanted to be sure that priming at the ectopic site could be fully functional in replication. To achieve this, we resolved to destroy authentic CaMV plus-strand priming and to rescue infectivity with a single ectopic priming element. The CaMV (ϩ)ps1 element is colinear with the RT gene; therefore, destruction of this primer must not disrupt RT. We introduced single nucleotide changes converting purines to pyrimidines and, vice versa, without changing the RT amino acid sequence, producing construct PB1, which was infectious (Fig. 7, A and  B). Progeny PB1 virion DNA isolated from plants had only one discontinuity in each strand producing predominantly genomelength single-stranded DNA fragments (8 kb) on denaturation together with a predicted fragment of 6.4 kb arising from defective second strand transfer (Fig. 7, C and D). Next, the (ϩ)ps2 element was disrupted with third nucleotide changes, in construct BB8. This was also infectious with characteristics suggesting priming from only one (ϩ)ps site (Fig. 7, C and D). To test the effectiveness of both sets of mutations, we produced construct BB9, which should have been completely disabled in plus-strand priming. In the first infectivity test, we inoculated 15 turnip seedlings with BB9 DNA and none showed symptoms. In a second round of BB9 inoculations, 1 plant from 15 showed symptoms. Analysis of the virion DNA from this plant indicated that priming had occurred only at the mutated (ϩ)ps2 site but that the mutated sequence had been retained. 3 Because of the residual infectivity exhibited by BB9, we made two further constructs in which the (ϩ)ps2 element was deleted. BB21 had a 33-bp in-frame deletion removing (ϩ)ps2, which, together with the (ϩ)ps1 mutations, completely abolished infectivity. We also constructed a hybrid CaMV, PM1, comprising the gene II region from CaMV isolate CM4 -184 with its 421-bp deletion removing (ϩ)ps2, the inactivated (ϩ)ps1 element, and a linker in the 35 S promoter capable of receiving a new (ϩ)ps element. PM1 was not infectious but infectivity was rescued by insertion of a (ϩ)ps:o element (Table  I) into the linker generating construct GG5 (Fig. 7). This clearly demonstrates that CaMV infectivity can be made dependent upon an artificial plus-strand primer in an ectopic genomic location and validates the functional assay of mutants at the redundant ectopic site.

DISCUSSION
Priming of DNA plus-strand synthesis is a critical step in RE propagation. We report here the first detailed study of the cis-acting sequence requirements for this process during RE viral infection of a whole organism. This was facilitated by the ability to manipulate a redundant ectopic (ϩ)ps element in infectious clones of the plant pararetrovirus CaMV. In particular, we were able to study priming, independent of any requirement in replication or pathogenicity. By measuring the percentage of virion DNA molecules containing a new discontinuity at the ectopic (ϩ)ps3 location, we estimated the approximate efficiency of priming at this site as about 85-90%. We have previously shown that priming at the authentic CaMV plus-strand primers is less than 100% efficient, generating a proportion of progeny DNA molecules lacking a discontinuity at each of the two sites, respectively (23). Therefore, we would not expect priming at the new site to be more efficient than at the authentic primers so 85-90% probably approximates the efficiency of the authentic primers as well.
The shortest artificial primer that gave maximal activity (construct p) was 19 nucleotides long. We conclude that the complete element probably encompasses 18 nucleotides assuming that the 5Ј-most A residue in this construct is not a part of the primer. Further mutational analysis should confirm this. Deleted variants of the artificial (ϩ)ps3 element which reduced the size of the 13 nucleotide PPT to 10 or fewer purines significantly reduced the efficiency of priming. Priming was still detectable when only the 7 G-box residues of the 13 nucleotide PPT were retained, but deletion of the 7 G nucleotides leaving 5 purines abolished priming. Priming efficiency was also diminished when the length of the PPT was increased by 3 or more nucleotides. This suggests a simple relationship between PPT length and priming efficiency, but other factors must also be important because deletion of the first PPT purine caused a greater loss of priming efficiency than loss of the first 2 purines (Fig. 3).
Unexpectedly, deletion of the 5 pyrimidines (T nucleotides) located immediately 5Ј of the PPT (construct q) abolished detectable priming (Fig. 3). Loss of the T nucleotides caused juxtaposition of an upstream stretch of A nucleotides next to the PPT increasing the PPT length from 13 to 20 nucleotides. This change was most likely inhibitory to priming, supporting our conclusion that the PPT has both a minimal and maximal length for optimum efficiency. However, other mutants showed a clear involvement of the 5Ј pyrimidines in primer recognition. Although pyrimidine-rich sequences 5Ј to the PPT are highly conserved among REs, in vitro experiments with retroviral PPTs have suggested that the 5Ј pyrimidines are not necessary for priming (see ,e.g., Refs. 11 and 12). The differences between CaMV and retroviruses could reflect a fundamental divergence in the priming mechanism. However, due to the conservation of the 5Ј pyrimidines among REs, we favor the explanation that some priming reactions in vitro do not accurately reflect all of the priming steps in vivo, as discussed by Klarmann et al. (13). One explanation for this difference might reside in the design of primer/template combinations used in in vitro studies. For example, use of a continuous RNA primer annealed to a DNA template (12) allows assay of only the polymerase-independent RNase H cleavages and excludes the initial round of polymerasedependent RNase H cleavages generating RNA fragments during reverse transcription (6). Thus, we could rationalize our observation that the upstream pyrimidines play an enhancing role in primer recognition by proposing that they function in vivo during the polymerase-dependent phase.
From our observations, we can propose a hypothetical model for plus-strand priming in CaMV (Fig. 8). We suggest that, 3  during synthesis of the DNA (Ϫ)strand, polymerase-dependent RNase H cleavages are directed toward the 5Ј pyrimidines. It is possible that, at this stage, the PPT is completely separated from adjacent sequences (Fig. 8A, iii), leading to enhanced recognition on the subsequent round of polymerase-independent RNase H cleavages. The PPT is then processed to leave a precise 3Ј end from which the DNA plus-strand originates. The primer is recognized by RT, which extends the DNA plusstrand on the DNA minus-strand template. Completion of the plus-strand occurs after one circuit of the circular template, where it slightly displaces the DNA plus-strand 5Ј end. Further progression by strand displacement is prevented by inefficient strand-displacing activity of the CaMV RT. The RNA primer on the DNA 5Ј end "ages" leaving one or a few oligoribonucleotides (34). Aspects of this model could be tested by comparing in vivo functions with in vitro reactions using purified CaMV RT.
Fidelity of priming at the artificial (ϩ)ps3 site was the same as at the authentic (ϩ)ps2 site, although this was 1 nucleotide different from that previously reported by Richards et al. (32). Since the 5Ј pyrimidines had no apparent effect on priming accuracy, processing of the remnants of the pregenome template RNA to give the precise cleavage for priming must only involve interaction between the ribo-PPT and the active site of RNase H on a DNA template (Fig. 8A, iv-vii). Our results are consistent with a model (Fig. 8B) in which there is an apparent preference for interaction between the 3Ј portion of the PPT and RNase H since various deletions of the PPT did not alter the position of the primary cleavage site 3 nucleotides upstream of the PPT 3Ј end (Fig. 4D). Increasing the length of the PPT by 2 nucleotides again produced the major priming site 3 nucleotides upstream from the PPT 3Ј end, suggesting a primary measuring recognition between RNase H and the PPT 3Ј end. However, the presence of a major secondary site 2 nucleotides upstream of this suggests alternative PPT length measurement from the PPT 5Ј end (Fig. 8B, ii and iii). This might also explain heterogeneity in priming. An alternative model for primer recognition would involve changes in shape of the RNase H active site to fit around primers of different lengths rather than variations in the way the PPT fits into a fixedshape active site (Fig. 8B, iv).
It is likely that other structural features are important in defining the 5Ј end, as suggested for HIV-1 by Powell and Levin (12). Various studies have shown that purine for pyrimidine substitutions in the PPT affect priming fidelity (11,16,18,33). This remains to be determined for CaMV (ϩ)ps elements, but, in two of our primer mutants in which 2 adjacent PPT G residues were substituted by 2 A and 2 T nucleotides, respectively, no change in the predominant DNA 5Ј end was observed. The presence of immediately adjacent, tandem (ϩ)ps elements in one of our constructs also showed no effect of primer proximity on priming fidelity, although there appeared to be competition for primer selection and possibly a novel 5Ј end generated in a purine-rich region in the last tandem primer. The precision of the plus-strand 5Ј ends in pararetroviruses is intriguing since they do not integrate an LTR-containing provirus with precise boundaries like retroviruses and retrotransposons. This conserved precision suggests that the RTassociated enzyme activities responsible for generating plusstrand primers are inextricably linked to production of a precise start site rather than a frayed end, suggesting that pararetroviral RTs evolved from retroviral or retrotransposon RTs rather than the converse.
We also confirmed that the artificial (ϩ)ps element functioned accurately and efficiently independent of its sequence environment. It also seems to function equally well in at least three other randomly chosen insertion sites (data not shown). Full functionality was confirmed by use of the ectopic primer to rescue infectivity following mutational inactivation of the authentic (ϩ)ps sites. Suppression or removal of the two authentic CaMV (ϩ)ps elements resulted in complete abolition of viral infectivity, suggesting that any minor additional priming sites, previously reported by Maule and Thomas (24), were not sufficient to establish infectivity. These observations raise the question as to the requirement for redundant plus-strand primers in CaMV and other REs. It has been suggested that an additional (ϩ)ps site increases the rate of replication of HIV-1, although it may also be involved in a novel replicative strategy (35). There is no evidence that multiple (ϩ)ps sites in plant pararetroviruses confer a replicative advantage. Pathogenically, CaMV variants with one (ϩ)ps are only marginally less severe than those with two, although this difference could be due to an effect on colinear elements. One explanation for A, i, part of the RNA template for reverse transcription adjacent to a plus-strand priming site shows the PPT and the 5Ј-pyrimidines (pyr) with the advancing DNA minus-strand; ii and iii, polymerase-dependent RNase H introduces initial cleavages into the RNA template during DNA synthesis (ii) and, into the pyrimidines (iii) to enhance recognition of the PPT; iv, polymerase-independent RNase H cleavages generate the precise primer 3Ј end; v and vi, selection and initiation of DNA plus-strand synthesis (v) and DNA extension and displacement of RNA fragments (vi); vii, limited displacement of the DNA plus-strand 5Ј end and loss of primer ribonucleotides produces the characteristic "gap" structure. B, hypothetical processing of the ribo-PPT to produce the correct plus-strand DNA 5Ј end as in A (iv), above. i, affinity of the ribo-PPT 3Ј end for the RNase H active site positions the cleavage 3 nucleotides from the 3Ј end. ii and iii, an overlong PPT can fit in alternative ways generating different cleavage sites. iv, alternative cleavage sites or different structured ribo-PPTs could be accommodated by changes in the RNase H active site. plus-strand primer redundancy in plant pararetroviruses might lie in their inherent inefficiency. Two or more primers could confer a slight adaptive advantage over a single primer in marginal environments. Even so, the ability to manipulate pararetroviral plus-strand priming mechanisms in vivo opens up new possibilities to understand the relationship between in vitro and in vivo studies of RT and to engineer new types of RT and RE tools.