Examining Ty3 Polypurine Tract Structure and Function by Nucleoside Analog Interference*

We have combined nucleoside analog interference with chemical footprinting, thermal denaturation, NMR spectroscopy, and biochemical studies to understand recognition of the polypurine tract (PPT) primer of the Saccharomyces cerevisiae long terminal repeat-containing retrotransposon Ty3 by its cognate reverse transcriptase. Locked nucleic acid analogs, which constrain sugar ring geometry, were introduced pairwise throughout the PPT (-)-DNA template, whereas abasic tetrahydrofuran linkages, which lack the nucleobase but preserve the sugar phosphate backbone, were introduced throughout the (-)-strand DNA template and (+)-strand RNA primer. Collectively, our data suggest that both the 5′- and 3′-portions of the PPT-containing RNA/DNA hybrid are sensitive to nucleoside analog substitution, whereas the intervening region can be modified without altering cleavage specificity. These two regions most likely correspond to portions of the PPT that make close contact with the Ty3 reverse transcriptase thumb subdomain and RNase H catalytic center, respectively. Achieving a similar phenotype with nucleoside analogs that have different effects on duplex geometry reveals structural features that are important mediators of Ty3 PPT recognition. Finally, the results from introducing tetrahydrofuran lesions around the scissile PPT/unique 3′-sequence junction indicate that template nucleobase -1 is dispensable for catalysis, whereas a primer nucleobase on either side of the junction is necessary.

) (4,5). A detailed biochemical study of these processes has been hampered by the lack of purified LTR-containing retrotransposon reverse transcriptases (RTs). However, the availability of active recombinant Ty1 and Ty3 enzymes (6,7) now allows interactions with their cis-acting signals to be compared and contrasted with those of the retroviral counterparts.
A focus of our recent studies (8 -13) has been understanding the mechanism that positions RT on the PPT-containing RNA/DNA hybrid for accurate cleavage at the PPT/unique 3Ј-sequence (U3) junction. Such specificity is required to create the (ϩ)-strand primer and later to remove it from nascent (ϩ)-DNA, thereby creating an integration-competent, double-stranded DNA provirus. Surprisingly, although a compilation of PPT sequences (14,15) suggests a common "modular" make-up of homopolymeric rA:dT and rG:dC blocks in retroviruses, this organization is absent in LTR-containing retrotransposon PPTs. Despite this, the Ty1 and Ty3 PPTs are accurately cleaved by their cognate RTs (7,16). In the absence of any obvious sequence homology, we have considered the possibility that structural features of the Ty3 PPT upstream of the PPT/U3 junction facilitate enzyme positioning in an orientation placing the RNase H catalytic site over the scissile bond. At the same time, subdomains of the cognate RT would be expected to make specific contacts with both nucleobases and the sugar phosphate backbone. Crystallographic studies with human immunodeficiency virus type 1 (HIV-1) RT (17)(18)(19) showed that ϳ14 bp of duplex nucleic acid are accommodated between its RNase H catalytic center and the p66 thumb subdomain. Targeted nucleobase removal from the PPT (Ϫ)-DNA template showed that modifying positions Ϫ15, Ϫ14, and Ϫ13 (defining position Ϫ1 as the first base pair 5Ј to the PPT/U3 cleavage junction) altered cleavage specificity, suggesting that contacts mediated by the p66 thumb subdomain had been disrupted (13). Such a correlation indirectly suggests that protein/nucleic acid interactions involving either extremity of the PPT and defined by the spatial separation of the RNase H active site and thumb subdomain could be important modulators of cleavage specificity. The goal of this study was to test this hypothesis with a non-retroviral RT and "non-canonical" PPT.
One strategy to investigate nucleic acid geometry is to introduce nucleoside analogs into the duplex in a manner that does not disrupt sequence context, but that imparts increased or decreased flexibility. Using a scanning strategy with locked nucleic acid (LNA) analogs (20,21) and abasic tetrahydrofuran (H 4 folate) linkages (22,23), which introduce rigidity and flexibility, respectively, into nucleic acids, we have defined two regions of the Ty3 PPT important for accurate cleavage of the scissile bond at the PPT/U3 junction. As might be predicted, one of these is the PPT/U3 junction, where LNA insertions and nucleobase removal as far as position Ϫ3 are capable of redirecting cleavage from the junction. The second region is the 5Ј-extremity of the PPT, where nucleoside analog insertion between positions Ϫ11 and Ϫ8 also induces altered cleavage specificity. However, the intervening region from posi-tions Ϫ7 to Ϫ4 is more tolerant to analog insertion. Despite differing from HIV-1 in both PPT sequence and quaternary structure of its cognate polymerase, the results presented here with Ty3 RT suggest a common feature inasmuch as protein/nucleic acid interactions involving both extremities of the RNA/DNA hybrid cooperate to position the RNase H active site over the Ϫ1/ϩ1 scissile bond.

EXPERIMENTAL PROCEDURES
Nucleic Acid Substrates-rSpacer (introducing abasic H 4 folate RNA lesions) and dSpacer (introducing abasic H 4 folate DNA lesions) were purchased from Glen Research Corp. (Sterling, VA). Hereafter, these are referred to as RAb and DAb, respectively. LNA phosphoramidites were purchased from Proligo Reagents. Substituted 39-nucleotide DNA and 29-nucleotide RNA oligonucleotides were synthesized at a 1-mol scale on a PE Biosystems Expedite 8909 nucleic acid synthesizer by standard phosphoramidite chemistry, with the exception that coupling times for LNA-cytidine were extended to 15 min. The fluorescent analog Cy3 was chemically substituted at the 5Ј-end of the oligonucleotides by a standard chemical synthesis method. Stepwise coupling yields for incorporation of the analogs were Ͼ98% as determined by trityl cation monitoring. Nucleic acid sequences and sites of substitution are illustrated in Fig. 2A and are described below. Deprotection and cleavage of the DNA oligonucleotides from the Controlled-Pore Glass support were carried out by incubation in 30% NH 4 OH for 36 h at 25°C, whereas that of the RNA oligonucleotides were carried out following the instructions of Proligo Reagents. Oligonucleotides were purified by preparative PAGE and quantified spectrophotometrically (260 nm), assuming a molar extinction coefficient equal to the sum of the constituent deoxyor ribonucleotides. An unmodified 29-nucleotide RNA containing the Ty3 PPT and flanking sequences at its 5Ј and 3Ј termini was also synthesized. Hybrids were prepared by heating a 1.5:1 RNA/DNA mixture to 95°C in 10 mM Tris-HCl (pH 7.8) and 25 mM NaCl for 5 min, followed by slow cooling to 4°C. Hybridized duplexes were stored at Ϫ20°C.
Melting Temperature (T m ) Measurements-For T m measurements, equimolar concentrations of (ϩ)-strand RNA primer were hybridized to (Ϫ)-strand DNA templates by heating to 90°C and slow cooling in a degassed solution of 10 mM Na 2 HPO 4 /NaH 2 PO 4 (pH 7.0) and 80 mM NaCl. Annealed substrates were analyzed on a 10% nondenaturing polyacrylamide gel and visualized by ethidium bromide binding to determine that complete hybridization had been achieved. 10 g/ml solutions of RNA/DNA hybrids were analyzed in a Beckman DU 640 spectrophotometer. E 260 was measured at 0.2°C intervals from 30 to 80°C. The T m of each hybrid was calculated by the "first derivative" method described by the manufacturer.
Double quantum filtered (DQF) COSY experiments were performed using samples of unmodified (1.2 mM) and LNA-substituted (ϳ0.5 mM) 20-bp PPT-containing RNA/DNA hybrids in 10 mM phosphate (pH 7.0) and 80 mM NaCl dissolved in 90% H 2 O and 10% D 2 O at 30°C. The spectrum of the unmodified Ty3 RNA/DNA hybrid was acquired using a four-channel Varian 800-MHz spectrometer equipped with an actively shielded z-axis gradient triple-resonance probe, and those of the LNA-substituted hybrids were acquired using a three-channel Bruker BioSpin Avance 600-MHz spectrometer equipped with an actively shielded z-axis gradient triple-resonance TXI CryoProbe. All spectra were processed using NMRPipe on a Linux PC workstation.
KMnO 4 Footprinting-RNA/DNA hybrids containing analogs in the (Ϫ)-DNA template were incubated at room temperature for 5 min in 20

Ty3 Polypurine Tract Structure
mM Tris-HCl (pH 8.0), 100 mM NaCl, and 100 mM MgCl 2 . The total reaction volume was 20 l. Modification reactions were initiated by adding 2 l of freshly prepared 25 mM KMnO 4 solution and terminated after 30 s with 2 l of 14 M ␤-mercaptoethanol. After ethanol precipitation and desiccation, samples were treated with 100 l of 1 M piperidine for 30 min at 90°C. Piperidine was removed by vacuum desiccation. Nucleic acids were washed three times with 50 l of water and vacuum-dried after each resuspension. Samples were finally resuspended in 89 mM Tris borate (pH 8.3), 2 mM EDTA, and 95% formamide containing 0.1% (w/v) bromphenol blue and xylene cyanol and fractionated by electrophoresis on 15% polyacrylamide gels containing 7 M urea.
Fluorescence Anisotropy Measurements-Anisotropy measurements were carried out using a HORIBA Jobin Yvon FluoroMax 3 fluorometer equipped with a 75-watt arc lamp and both emission and excitation monochromators. Samples containing 10 nM RNA/DNA hybrids in 50 mM Tris-HCl (pH 7.8) and 80 mM NaCl were pipetted into a 100-l ultramicro cell, and fluorescence emission was recorded from 550 to 570 nm using an excitation wavelength of 542 nm at 25°C. Slits widths on both channels were set to 5 nm. All reported values represent the average of measurements obtained from three different samples. The affinity of Ty3 RT for PPT variants was measured by titrating increasing concentrations of enzyme against 10 nM Cy3-labeled PPT substrate. Anisotropy of the substrate was recorded after each enzyme addition. The results of each titration were plotted as a relative change of anisotropy versus enzyme concentration and were fitted to an equation using nonlinear regression as described by Heyduk and Lee (24).
RNase H-mediated PPT Selection-Ty3 PPT selection was evaluated as described previously (10). 5Ј-End labeling of the PPT-containing RNA was performed with T4 polynucleotide kinase and [␥-32 P]ATP. Substrates were generated by annealing radiolabeled RNA to wild-type or substituted (Ϫ)-strand DNA. Hydrolysis was initiated by adding Ty3 RT to RNA/DNA hybrids in 10 mM Tris-HCl (pH 7.8), 80 mM NaCl, 5 mM dithiothreitol, and 9 mM MgCl 2 at 30°C, with the enzyme and RNA/DNA hybrid present at final concentrations of 15 and 50 nM, respectively. Reactions were terminated after 10 min by adding an equal volume of formamide-based gel loading buffer (95% (v/v) formamide containing 0.1% (w/v) bromphenol blue and xylene cyanol), and the hydrolysis products were fractionated by high voltage electrophoresis on 15% (w/v) polyacrylamide gels containing 7 M urea. Products were visualized by autoradiography and/or phosphorimaging.
DNA-dependent DNA Polymerase Activity-DNA synthesis was measured on 39-nucleotide DNA templates annealed to a 5Ј-end-labeled 21-nucleotide DNA primer containing the PPT sequence. Polymerization was initiated by adding 10 nM enzyme to a mixture containing 50 nM template-primer, 200 M dNTPs, 25 mM Tris-HCl (pH 7.8), 80 mM NaCl, 6 mM dithiothreitol, and 9 mM MgCl 2 at 25°C and terminated after 10 min by adding an equal volume of gel loading buffer. DNA synthesis products were fractionated by high voltage electrophoresis on 15% (w/v) polyacrylamide gels containing 7 M urea in Tris borate/EDTA buffer. After drying, gels were subjected to autoradiography and/or phosphorimaging using a Personal Molecular Imager FX system (Bio-Rad).

RESULTS
Experimental Strategy-The strategy for our nucleoside analog interference studies with the Ty3 PPT is presented in Fig. 2. Position Ϫ1 is defined as the first base pair 5Ј to the PPT/U3 junction, and duplex regions are defined as upstream or downstream of this junction ( Fig.  2A). Overlapping LNA insertions were introduced pairwise throughout the (Ϫ)-DNA template between positions Ϫ16 and ϩ6, whereas abasic H 4 folate linkages were introduced singly into the (Ϫ)-DNA template and (ϩ)-RNA primer between positions Ϫ12 and ϩ2. Fig. 2B shows the structure of an LNA monomer, which induces a C-3Ј-endo configuration of the C-4Ј/C-2Ј methylene-linked sugar (20). In contrast, H 4 folate linkages in the DNA or RNA strand (Fig. 2C) remove the nucleobase while preserving the sugar phosphate backbone (25). Prior to evaluating the effect of analog insertion on cleavage specificity, the duplex stability and affinity of Ty3 RT for modified RNA/DNA hybrids were determined to eliminate the possibility that the altered enzyme activity was a consequence of gross conformational changes of the substrate.

Ty3 Polypurine Tract Structure
Thermal Denaturation Profiles-Under our experimental conditions, a T m of 70°C was determined for the unsubstituted RNA/DNA hybrid. Nucleobase removal from the RNA primer was more destabilizing, reducing the T m by between 8°C (Ϫ1RAb) and 10.2°C (Ϫ8RAb). This result likely reflects loss of both intrastrand stacking and interstrand hydrogen bonding. The destabilizing effect was slightly less pronounced for substitutions in the (Ϫ)-DNA template, where the T m was reduced by between 6.0°C (Ϫ12DAb) and 9.1°C (Ϫ6DAb). However, no individual position within the Ty3 PPT appeared particularly sensitive following nucleobase removal from either strand (data not shown).
In contrast, an unexpected pattern was obtained for RNA/DNA hybrids containing LNA insertions in the (Ϫ)-DNA template (Table 1). Consistent with a previous report (26), insertions between positions Ϫ16 and Ϫ12 and between positions Ϫ4 and ϩ6 increased the T m by between 1.0°C (Ϫ14/Ϫ13LNA) and 6°C (Ϫ4/Ϫ3LNA). In contrast, LNA insertions between positions Ϫ12 and Ϫ4 uniformly decreased the T m by between 5°C (Ϫ11/Ϫ10LNA) and 7°C (Ϫ5/Ϫ4LNA). Resynthesis of representative (Ϫ)-DNA templates containing the Ϫ7/Ϫ6LNA and Ϫ6/Ϫ5LNA substitutions resulted in a similar reduction in the T m of the hybrid, excluding the simple possibility of experimental error during oligonucleotide synthesis or annealing (data not shown). The data of Table 1 thus suggest that constraining sugar ring geometry upstream of the Ty3 PPT/U3 junction has a destabilizing effect, possibly through disrupting stacking interactions at a particularly sensitive region of the RNA/DNA hybrid. Interestingly, the complementary sequence between positions Ϫ12 and Ϫ4 contains three adjacent and overlapping 5Ј-rArGrA-3Ј steps. Kopka et al. (27) noted that the ringover-ring stacking of purines of such a sequence is not observed in an RNA/DNA hybrid modeled after the HIV-1 PPT, suggesting that it may contribute to deformability of this sequence. The presence of multiple steps of 5Ј-rArGrA-3Ј in the Ty3 PPT may therefore render the duplex highly malleable and, by inference, highly susceptible to LNA modification.
NMR Analysis of LNA-modified PPT Hybrids-To determine whether structural perturbations in the Ty3 PPT hybrid might explain the decreased thermal stability that results from LNA doublet substitution, a series of homonuclear NMR experiments were performed with the Ϫ2/Ϫ1LNA-, Ϫ6/Ϫ5LNA-, and Ϫ11/Ϫ10LNA-substituted PPTs, allowing an initial qualitative comparison with the unmodified hybrid. DQF-COSY spectra were acquired for the Ϫ2/Ϫ1LNA-, Ϫ6/Ϫ5LNA-, and Ϫ11/Ϫ10LNA-substituted hybrids and compared with that of the unmodified hybrid. Differences in the chemical shift patterns and line widths observed for H-5/H-6 correlations in the DQF-COSY spectra of the Ϫ2/Ϫ1LNA-substituted (Fig. 3B), Ϫ6/Ϫ5LNA-substituted (Fig.  3C), and Ϫ11/Ϫ10LNA-substituted (Fig. 3D) hybrids provide a direct indicator of changes in base stacking geometry and dynamics in the  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 hybrid. Strikingly, chemical shift perturbations were observed for almost all H-5/H-6 correlations in each of the LNA-substituted hybrids relative to the unmodified PPT, indicating that the LNA doublet insertion results in an altered structure over the entire PPT, with long-range effects observed for residues up to 16 bp distal to the site of the substitution (Fig. 3A), and not simply a local structural perturbation as has been observed in previous studies of LNA-substituted RNA/DNA hybrids (21,26). The observed chemical shift perturbations for other proton resonances in the LNA-substituted PPTs also indicate that these insertions result in structural alterations over the entire PPT (data not shown). In addition, the structural perturbations appear to propagate more strongly upstream of the position of LNA substitution, with particularly large effects immediately adjacent to the PPT between positions Ϫ16 and Ϫ13. Significant line broadening was also observed for a number of H-5/H-6 resonances in each spectrum, with the most significant resonance line broadening resulting from the Ϫ10/Ϫ11LNA doublet substitution. Taken together, the NMR data demonstrate that the insertion of LNA doublets produces a structural perturbation that appears to significantly alter the hybrid conformation both within and immediately upstream of the Ty3 PPT. In this regard, the PPT-containing RNA/DNA hybrid appears unique in that a fixed altered local backbone geometry resulting from the LNA substitutions is propagated as long-range structural perturbations throughout the entire element. It is noteworthy to point out that LNA substitutions in RNA/DNA hybrids are known to alter the secondary structure of the nucleic acid locally, mostly restricted to the flanking 3Ј-and 5Ј-bases (28). However, there has been no report on the long-range global structural perturbations associated with local LNA substitutions in general RNA/DNA context as observed with Ty3 PPTs. The global nature of the structural effects could possibly explain the unusual observation of decreased T m values for hybrids with LNA doublet substitutions between positions Ϫ12 and Ϫ4. However, more detailed structural analysis of the LNA-substituted hybrids will be required before specific structural changes can be correlated with altered thermal stability of the Ty3 PPT.

Ty3 Polypurine Tract Structure
Affinity of Ty3 RT for Modified RNA/DNA Hybrids-To assess whether analog insertion affected the affinity for Ty3 RT for the modified PPT, equilibrium dissociation constants were determined for several substituted hybrids via fluorescence anisotropy (29,30). For sub-strates containing RAb substitutions, the fluorophore Cy3 was introduced into the (Ϫ)-DNA template, whereas for substrates containing DAb and LNA substitutions, the fluorophore was introduced into the (ϩ)-RNA primer. As shown in Fig. 4A, K d values of 11.13 and 13.24 nM were determined for the unmodified hybrids containing Cy3 on the DNA and RNA strands, respectively. Although the affinity of Ty3 RT for the PPT-containing RNA/DNA hybrid was slightly reduced, these values are consistent with the K d of ϳ3.6 nM previously determined with duplex DNA (31). The dissociation constants for Ty3 RT with PPT variants containing LNA doublet substitutions at positions Ϫ2/Ϫ1, Ϫ6/Ϫ5, Ϫ11/Ϫ10, and Ϫ13/Ϫ12 were in general agreement with those for the unmodified RNA/DNA hybrids (Fig. 4B), ranging from 13.85 nM (Ϫ13/Ϫ12LNA) to 19.01 nM (Ϫ11/Ϫ10LNA). This result indicates that the modifications did not compromise recognition by Ty3 RT. Surprisingly, the affinity of Ty3 RT for substrates lacking template or primer nucleobases Ϫ1, Ϫ6, Ϫ9, and Ϫ12 increased ϳ5-fold for duplexes substituted in either the RNA and DNA strand. This result might indicate that increased duplex flexibility, resulting from local alterations to base stacking and hydrogen bonding, stabilizes the nucleoprotein complex.
Cleavage of LNA-substituted PPT Variants by Ty3 RT-As shown in Fig. 2A, LNA substitutions were introduced pairwise into the (Ϫ)-DNA template throughout the PPT. In addition, the adjacent upstream sequences between positions Ϫ16 and Ϫ13 and downstream sequences between positions ϩ2 and ϩ6 were substituted to determine whether the LNA effect was restricted to the PPT-containing duplex or simply reflected a more general alteration of nucleic acid architecture. The results of this analysis are illustrated in Fig. 5.
For the wild-type RNA/DNA hybrid, the major hydrolysis sites were the PPT/U3 junction (defined as position Ϫ1) and positions ϩ4 and ϩ6 (Fig. 5, lane W). LNA substitutions between positions Ϫ16 and Ϫ13 had little effect on cleavage at the PPT/U3 junction (Fig. 5, lanes a-d), although the Ϫ14/Ϫ13LNA and Ϫ13/Ϫ12LNA substitutions eliminated cleavage at position ϩ4 (Fig. 5, lanes c and d, respectively). In contrast, reduced cleavage at the PPT/U3 junction and enhanced cleavage immediately ahead of this were evident when LNA substitutions were relocated between positions Ϫ12 and Ϫ8 (Fig. 5, lanes e-h). The adjacent Ϫ8/Ϫ7LNA and Ϫ7/Ϫ6LNA substitutions appeared to selectively inhibit PPT/U3 cleavage (Fig. 5, lanes i and j, respectively),

Ty3 Polypurine Tract Structure
whereas low level PPT/U3 cleavage was observed with hybrids substituted at positions Ϫ6/Ϫ5 and Ϫ5/Ϫ4 (Fig. 5, lanes k and l, respectively). LNA doublets between positions Ϫ4 and ϩ1 had the most pronounced effect, severely reducing cleavage both at the PPT/U3 junction and ahead of this (Fig. 5, lanes m-p). For substitutions flanking the PPT/U3 junction (Fig. 5, lanes p-s), elimination or reduction of cleavage suggests that altering duplex geometry most likely influences its trajectory within the RNase H active site because the affinity of Ty3 RT for this hybrid was not seriously affected (Fig. 4B). PPT/U3 cleavage was gradually restored when the position of LNA substitution was relocated downstream of the scissile bond, although the ϩ4 and ϩ6 hydrolysis products were absent (Fig. 5, lanes t-u). Fig. 5 thus indicates that duplex geometry at either end of the PPT is sensitive to substitution, reflecting regions where either (a) altered geometry is incompatible with correct positioning of Ty3 RT or (b) enzyme contacts have been compromised, possibly through local alterations to sugar puckering (20,21,26). These possibilities will be discussed below.
Removal of PPT Template Nucleobases- Fig. 6 illustrates the effect of introducing H 4 folate lesions between positions Ϫ12 and ϩ2 of the PPT (Ϫ)-strand DNA template. Removing nucleobase Ϫ12 did not affect specificity, but rather enhanced cleavage at the PPT/U3 junction (Fig. 6,  lane a). In contrast removing nucleobases between positions Ϫ11 and Ϫ8 reduced cleavage at the PPT/U3 junction while at the same time increasing cleavage in the U3 region downstream of this (Fig. 6, lanes  b-e). Although subtle, a novel cleavage site, "tracking" ϳ12 bp from the position of analog insertion, was evident. As an example, substrate Ϫ11DAb exhibited enhanced cleavage at position ϩ1, whereas substrate Ϫ9DAb rendered position ϩ3 hypersensitive (Fig. 6, lanes b and  d, respectively). Removing template nucleobases Ϫ7, Ϫ6, and Ϫ5 (Fig. 6,  lanes f-h) was less deleterious in that, although cleavage ϳ12 bp from the site of analog insertion persisted (e.g. at position ϩ5 on substrate Ϫ7DAb) (Fig. 6, lane f), the PPT/U3 junction was recognized. In contrast, Ty3 RT failed to recognize the PPT/U3 junction when template nucleobases Ϫ4, Ϫ3, and Ϫ2 were removed, but rather was redirected to cleave within the U3 region between positions ϩ8 and ϩ10 (Fig. 6,  lanes i-k).
Template nucleobases Ϫ1 and ϩ1 were removed to determine the importance of hydrogen bonding and/or base stacking opposite the scissile PPT/U3 junction or, in fact, the necessity for a template nucleobase because a previous study with HIV-1RT (9) demonstrated that this junction is surprisingly tolerant to substitution. For Ty3 PPT, our analysis showed that the template nucleobase 3Ј to the PPT/U3 junction (position Ϫ1) could be removed without any major consequences (Fig.  6, lane l), whereas cleavage was eliminated when the nucleobase on the 5Ј-side of the scissile bond was removed (Fig. 6, lane m). Finally, a wild-type processing profile was observed when template nucleobase ϩ2 was eliminated (Fig. 6, lane n).
Susceptibility of Template Thymines to KMnO 4 Oxidation-Thymine sensitivity to KMnO 4 oxidation can be exploited to determine whether such bases are unpaired (32,33) or are structurally distorted but retain weak hydrogen bonding (34). Because thymines are evenly dispersed throughout the Ty3 PPT (Ϫ)-DNA template, chemical footprinting of the RNA/DNA hybrid was performed to gain insight into the structural consequences of targeted nucleobase removal (Fig. 7). Regardless of the position of nucleobase removal from the (Ϫ)-DNA template, only a single thymine in its immediate vicinity was KMnO 4sensitive. This result suggests that the reduced T m values accompanying nucleobase removal reflected local modification of duplex geometry. Also, as might be expected, removing any thymine eliminated that particular hydrolysis product. Eliminating nucleobase Ϫ10C (despite being 3Ј-and 5Ј-flanked by thymines) induced the reactivity of exclusively its 3Ј-neighbor, Ϫ11T (Fig. 7, lane c), suggesting that the 5Ј-thymine at position Ϫ9 retains a stably stacked and base-paired configuration. Selective reactivity of the 3Ј-thymine was also noted with substrates Ϫ8DAb and Ϫ6DAb, in which the modified nucleobase was likewise 3Јand 5Ј-flanked by thymine (Fig. 7, lanes e and g). Although supporting data with other RNA/DNA hybrids are unavailable, differential thymine reactivity has been noted for duplex DNA by McCarthy et al. (35). Although Ϫ4C is only 3Ј-flanked by a thymine, this base (Ϫ5T) still displayed KMnO 4 reactivity (Fig. 7, lane i). However, as the site of nucleobase removal approached the PPT/U3 junction, a different pattern emerged. Removing Ϫ3C enhanced the reactivity (although  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 weakly) of the adjacent 5Ј-thymine, Ϫ2T (Fig. 7, lane j), and likewise, removing Ϫ2T rendered the 5Ј-thymine (Ϫ1T), KMnO 4 -sensitive (Fig.  7, lane k). At the same time, Ϫ2T was minimally KMnO 4 -sensitive following removal of nucleobase Ϫ1T. These data imply that, in the absence of protein factors, the RNA/DNA hybrid likely adopts an unusual geometry near the PPT/U3 junction. This postulate is in keeping with our recent NMR studies showing an inversion of sugar pucker near the scissile bond (36). Finally, removing nucleobase ϩ1C restored the original pattern in that the 3Ј-flanking thymine at position Ϫ1 was rendered KMnO 4 -reactive.

Ty3 Polypurine Tract Structure
Cleavage Specificity following Removal of PPT Primer Nucleobases-We next examined the targeted removal of PPT (ϩ)-RNA primer nucleobases (Fig. 8). Because the RNA primer was 5Ј-end-labeled, hydrolysis products will lack the nucleobase and therefore migrate slightly faster than the unsubstituted RNA or its RNase H products in lane W. Moreover, we observed that hydrolysis products lacking guanine migrated slightly faster than those lacking adenine, producing the "staggered" appearance of the PPT/U3 hydrolysis product. Overall, removal of (ϩ)-primer nucleobases had similar positional effects on PPT/U3 cleavage specificity as that of their (Ϫ)-DNA template counterparts. Substitutions between positions Ϫ11 and Ϫ8 (Fig. 8, lanes b-e) severely reduced PPT/U3 cleavage, whereas cleavage was less pronounced when nucleobases were removed from the intervening region between positions Ϫ7 and Ϫ4 (lanes f-j). Removing primer nucleobases Ϫ2, Ϫ1, and ϩ1 completely eliminated PPT/U3 cleavage (Fig. 8, lanes  k-m), suggesting that they mediate important contacts at the RNase H catalytic site. This result contrasts with results from substitutions in the same region of the (Ϫ)-DNA template, where nucleobase Ϫ1 could be eliminated without inhibiting PPT/U3 cleavage (Fig. 6, lane l). Relocation of the RNase H catalytic center a constant distance (ϳ12 bp) relative to the site of analog insertion was most pronounced with the Ϫ8 and Ϫ4 substitutions, which promoted cleavage at positions ϩ4 and ϩ8, respectively (Fig. 8, lanes e and i). Also, in the vicinity of the PPT/U3 junction, the Ϫ2 and Ϫ1 substitutions induced hypersensitivity at positions ϩ10 and ϩ11, respectively (Fig. 8, lanes k and l).
DNA Polymerase Activity of Ty3 PPT Variants-Finally, we examined, F9 DNA-dependent DNA synthesis on duplexes containing LNA insertions (Fig. 9B) or H 4 folate lesions (Fig. 9C) in the (Ϫ)-DNA template. For these experiments, a 5Ј-end-labeled DNA primer was employed to avoid cleavage at the PPT/U3 junction and removal of nascent DNA or the use of mutant enzyme lacking RNase H activity. A prominent feature of LNA insertions in the (Ϫ)-DNA template was stalling of the replication complex shortly after initiation of DNA synthesis on duplexes with substitutions between positions Ϫ4 and Ϫ1 (Fig.  9B, lanes m-p). As an example, a Ϫ1/ϩ1 insertion led to stalling at positions Pϩ4 and Pϩ5 of the DNA template (Fig. 9B, lane p). This result suggests that, when this insertion occupies positions Ϫ5 and Ϫ6 of the duplex following translocation of the polymerizing machinery, the structural alteration induced by constrained sugar ring geometry impedes contact with important protein motifs of the enzyme. In a similar fashion, the Ϫ2/Ϫ1LNA substitution induced stalling at position Pϩ4 of the template (Fig. 9B, lane o), i.e. when the LNA doublet would again be predicted to occupy positions Ϫ5 and Ϫ6 of the polymerizing complex. This effect continued as far as the Ϫ5/Ϫ4LNA sub-  (37)(38)(39)(40)(41). Another feature, although small, was pausing at multiple positions throughout the template for substrates containing LNA insertions between positions Ϫ11 and Ϫ4 (Fig. 9B, lanes e-l), corresponding   FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 to the region of the PPT-containing RNA/DNA hybrid destabilized by LNA insertion (Table 1).

Ty3 Polypurine Tract Structure
DNA synthesis on PPT derivatives containing H 4 folate template lesions was unaffected in general, with the exception of the region between positions Ϫ4 and ϩ1 (Fig. 9C, lanes i-l). Removing template nucleobase Ϫ4 induced pausing at positions Pϩ1 and Pϩ2 (Fig. 9C, lane i), suggesting again that, when this substitution (which alters base stacking) occupies positions Ϫ5 and Ϫ6 of the translocation complex, contacts with the Ty3 RT thumb subdomain are compromised. Likewise, removing template nucleobase Ϫ3 induced pausing at positions Pϩ2 and Pϩ3 of the translocation complex (Fig. 9C, lane j). In contrast, removing template nucleobase Ϫ2 did not induce aberrant stalling, indicating that, following addition of 2-3 nucleotides to the primer terminus, the polymerization machinery has acquired sufficient energy (possibly generated through dNTP hydrolysis) to bypass the lesion. As might be expected, removing template nucleobase Ϫ1 was most deleterious for DNA polymerase function because this would most likely affect positioning of the primer 3Ј-OH in the active site. These data again support our recent mutagenesis studies on the Ty3 thumb subdomain showing that alteration of the counterpart of ␣-helix H severely influences the stability of the replication complex (42). Thus, when Ty3 RT adopts the orientation with its RNase H catalytic center positioned for cleavage at the PPT/U3 junction, it is not unreasonable to assume that abnormal nucleic acid geometry upstream of this junction is also sensed by its thumb domain.

DISCUSSION
Although retroviral and LTR-containing retrotransposon PPTs differ significantly in both their length and sequence (5,14), they are accurately removed from (ϩ)-RNA and (ϩ)-DNA by their cognate RTs, suggesting that common structural features of these RNA/DNA hybrids contribute to the sequestration and positioning of the viral polymerase such that its RNase H catalytic center is correctly located over the scissile bond at the PPT/U3 junction. For the HIV-1 PPT, x-ray crystallography (19), chemical footprinting (8), and nucleoside analog interference (9,11,13) studies make a strong case for an interaction between the p66 thumb subdomain of RT and a destabilized ("unzipped") portion of the duplex 12-15 bp upstream of this junction (19). This region of the PPT has been implicated in an induced fit with HIV-1 RT as a potential mechanism to correctly locate its RNase H active site (8,15). Whether a similar mechanism applies to PPT recognition by LTR-containing retrotransposon enzymes has been addressed here by scanning mutagenesis of both the DNA and RNA strands using analogs that increase or decrease the stability of the Ty3 PPT-containing duplex.
Although they might be expected to have different consequences for duplex architecture, LNA analogs and H 4 folate linkages at the 5Ј-end of the Ty3 PPT produce the same positional effect, implying that they influence duplex geometry rather than affect specific protein/nucleic acid contacts. Another consequence is the appearance of novel cleavage sites that appear to track a constant distance from the site of analog insertion. This is evident for duplexes with LNA substitutions between positions Ϫ11 and Ϫ8; incubation with Ty3 RT led to enhanced cleavage in the U3 region ϳ12 bp from the site of insertion (Fig. 5, lanes f-i).
In contrast, immediately adjacent LNA substitutions between positions Ϫ16 and Ϫ13 in duplexes, i.e. outside the PPT, had minimal effects (Fig.  5, lanes a-d), eliminating a general property of LNA insertion into RNA/DNA hybrids. Scenarios that might explain these novel cleavage sites include (a) a long-range effect of LNA substitution on nucleic acid geometry at and ahead of the PPT/U3 junction and (b) local alterations in duplex architecture promoting novel binding sites for Ty3 RT. LNA analogs in the (Ϫ)-DNA template will constrain the sugar ring in a C-3Ј-endo or N-type configuration, which is contrary to its normal orientation (20), whereas NMR studies have shown that LNA insertions can influence the sugar pucker of adjacent 5Ј-and 3Ј-residues (21,28). Thus, the LNA doublet insertions evaluated here could locally alter sugar puckering over as many as 4 bases of the (Ϫ)-DNA template. Experimentally, we observed that PPT substitution between positions Ϫ11 and Ϫ8 induced novel cleavage at several sites ahead of the PPT/U3 junction. In keeping with our HIV-1 data, we therefore suggest that a structural element of Ty3 RT located 11-12 bp from its RNase H catalytic site "senses" an altered sugar pucker as a stable binding site. A logical protein motif would be the thumb of its DNA polymerase domain, which studies with HIV-1 RT (37-41) have shown is particularly sensitive to alterations in nucleic acid geometry. The DNA synthesis data of Fig. 9, combined with our recent mutagenesis studies on the Ty3 RT thumb subdomain (42), show that it, too, is responsive to subtle alterations to duplex architecture. Thus, the model we envisage proposes that each LNA-induced alteration to the sugar pucker between positions Ϫ11 and Ϫ8 can provide a transient interaction site for the Ty3 RT thumb subdomain, which would relocate the RNase H catalytic center over sites ahead of the PPT/U3 junction. Our recent NMR studies on the Ty3 PPT-containing RNA/DNA hybrid further support a contribution from sugar puckering to PPT selection (36).
The observation that H 4 folate linkages in the template and primer between positions Ϫ11 and Ϫ7 (Figs. 6 and 8, respectively) have a similar effect on PPT/U3 cleavage suggests that their effect might also involve alterations in sugar puckering. In this respect, molecular dynamics simulations of model duplexes have indicated that the abasic sugar adopts a variety of puckers centered around the C-3Ј-endo pucker (43). Thus, for the wild-type PPT-containing duplex, our data suggest that malleability between positions Ϫ11 and Ϫ8 is an important mediator of interactions with the thumb subdomain of its cognate RT. Because most nucleic acid-polymerizing enzymes are known to bend the duplex in the vicinity of the catalytic center (44), flexibility inherent to this region of the Ty3 PPT could be envisaged as an important modulator of an induced fit with RT.
Ty3 hybrids with substitutions between positions Ϫ7 and Ϫ5 are cleaved at the PPT/U3 junction, albeit with reduced efficiency. The crystal structures of HIV-1 RT bound to duplex DNA (17,18) and an RNA/DNA hybrid (19) indicate that considerably fewer contacts are made with nucleic acid in this region, which would occupy its connection subdomain. Thus, provided the appropriate contacts are made with these N-and C-terminal domains of Ty3 RT, manipulating the intervening nucleic acid structure appears to be less deleterious. A second and abrupt alteration in PPT cleavage specificity follows substitution in the adjacent region between positions Ϫ4 and Ϫ2, which most likely perturbs contacts with critical protein motifs close to the RNase H catalytic center. The RNase H primer grip motif of HIV-1 RT (19) has been shown to contact nucleic acid several bases upstream of the RNase H catalytic center, possibly as a means of guiding the trajectory of the RNA strand into the active site for cleavage. A recent crystal structure of Bacillus halodurans RNase H complexed with an RNA/DNA hybrid (45) has defined a similar region as the "phosphate-binding pocket," indicating that it mediates contact with the DNA strand 2-4 bp upstream of the scissile bond. Altering sugar pucker in the equivalent region of the Ty3 PPT (Ϫ)-DNA template might therefore be predicted to compromise interactions with its phosphate-binding pocket, thereby inhibiting RNase H activity.
Structural studies with RNases H from Escherichia coli (46 -49) and B. halodurans (45) have implicated several active-site residues in con-tacting the RNA strand in the catalytic center (46 -49). For the E. coli enzyme, it has been proposed that Gln 72 is hydrogen-bonded to the 2Ј-OH of the nucleoside 2 bases 5Ј to the scissile bond, whereas Cys 13 interacts with the 2Ј-OH of the ribonucleotide ahead of this. In addition, an outer sphere complex between the catalytic Mg 2ϩ ion and the 2Ј-OH group of the nucleoside 5Ј to the scissile bond has been predicted. Because a detailed structure for Ty3 RT is unavailable, assigning this function to equivalent residues becomes speculative. However, the recent structure of a co-crystal of B. halodurans RNase H with an RNA/ DNA hybrid indicates that two of its catalytically critical carboxylate residues (Glu 109 and Asp 132 ) contact the ribose moiety of the RNA strand on either side of the scissile bond. For Ty3 RT, the equivalent residues are Glu 401 and Asp 426 , respectively. Thus, constraining nucleic acid geometry or altering sugar pucker around the Ty3 PPT/U3 junction would be predicted to interfere with these critical contacts. Finally, the data of Fig. 6 show that template nucleobase Ϫ1 is dispensable for hydrolysis, possibly indicating separation of the DNA and RNA strands as the hybrid enters the catalytic center. This observation concurs with our recent studies of HIV-1 RT demonstrating that template nucleobases can also be removed without affecting cleavage specificity (13).
In the absence of crystallographic data, our studies with Ty3 RT have by necessity required comparison with and extrapolation from structures of the HIV-1, E. coli, and B. halodurans enzymes. In particular, we have proposed that the thumb subdomain of Ty3 RT interacts with a particularly malleable region of the RNA/DNA hybrid between positions Ϫ11 and Ϫ7 of the PPT and that nucleoside analogs alter duplex architecture in a manner that subtly changes these contacts. The thermal melting data of Table 1 support this notion and indicate that LNA insertions between positions Ϫ12 and Ϫ4 destabilize the hybrid, in contrast to published data. The same region exhibits unusual KMnO4 reactivity in that thymines 5Ј to H 4 folate lesions remain well stacked, whereas those 3Ј to the lesion are oxidized (Fig. 7). The unusual KMnO4 reactivity of template thymines immediately adjacent to the PPT/U3 junction (Fig. 7) suggests that the duplex geometry here also deviates from standard Watson-Crick base pairing. Taken together, these observations provide a strong argument that the Ty3 PPT has a pre-organized structure that facilitates RT binding in a productive hydrolytic mode. As an initial step in confirming this proposal, we have initiated NMR studies on both the wild-type and substituted duplexes, and preliminary data indeed indicate that a single substitution in the (Ϫ)-DNA template alters the sugar pucker of both its RNA complement and several neighboring RNA bases, indicating considerable malleability at the 5Ј-end of the Ty3 PPT (36).