Cross-Linking of the Fingers Subdomain of Human Immunodeficiency Virus Type 1 Reverse Transcriptase to Template-Primer

By using BspMI cassette mutagenesis (4), an ApaI site was added to the p66 coding region of HIV-1 RT to make the construct RT (ApaI). This modification does not change the protein sequence of HIV-1 RT, but changes the codon usage at Pro52: -GGG CCT GAG→-GGG CCC GAG- GlyProGluGlyProGlu

RT (ApaI) was modified to produce RT (ΔScaI). In this plasmid, the HIV-1 RT coding region between codon 53 and codon 68 has been removed and replaced with aBamHI site. RT (ΔScaI) also has three new restriction endonuclease recognition sites created by site-directed mutagenesis: a SacI site at codons 79 and 80 (—GAA CTT—to—GAG CTC—), a MunI site at codons 91 and 92 (—CAA TTA—to—CAA TTG—), and an EcoRI site at codons 93 and 94 (—GGA ATA CCA—to—GGA ATT CCA—). None of these changes affects the protein sequence of HIV-1 RT.

To create the mutants K65C, D67C, K70C, and L74C, RT (ΔScaI) was digested with ApaI andSacI, and the large fragment was gel purified. Synthetic DNA fragments were generated by annealing complementary oligonucleotides. These synthetic DNA fragments span the region between codons 59 and 79 in the HIV-1 RT coding region and contain the codon changes to make the mutations K65C, D67C, K70C, and L74C. The synthetic DNA fragments have ends complementary to the overhangs created by ApaI andSacI. The synthetic DNA fragments were ligated to the gel-purified SacI-ApaI RT (ΔSacI) DNA. The resulting plasmids were first analyzed by digestion withBamHI (the desired clones do not have a BamHI site) and then analyzed for the ability to express the p66 protein. Selected plasmids were then sequenced to prove the desired mutations were present.

The mutant C38V was constructed by using the clone RT (ApaI). RT (ApaI) was digested withSmaI and ApaI, and the DNA was gel purified. Synthetic oligonucleotides were annealed to generate a DNA fragment, which regenerates the SmaI site, encodes a valine residue at codon 38, and has an overhang complementary to the ApaI site. This synthetic DNA fragment was ligated to theSmaI-ApaI-digested RT (ApaI). The resultant clones were screened for the expression of the p66 protein and then sequenced to ensure the mutation was present.

The C38V mutation was introduced into plasmids carrying the K65C, D67C, K70C, and L74C mutations by using the ApaI site. The C38V plasmid was digested with ApaI-HindIII, and the 1,500-bp band was gel purified. The larger fragment from C38V and the smaller fragment from the K65C, D67C, K70C, and L74C mutants were ligated to generate the double mutant. Each of the double mutant plasmids was further modified to produce HIV-1 RT proteins containing the C280S mutation and six histidines at the C terminus. This gives plasmids in which the p66 coding region contains only a single cysteine residue at the designated location. The six-histidine tag at the C terminus of p66 aids in protein purification.

The expression vector pUC12N/p51(−cys) is similar to the coexpression vectors previously described (3). The vector contains twolacZ promoters oriented in opposite directions. OnelacZ promoter transcribes a region encoding a p51 subunit with no cysteine residues (the specific mutations are C38V and C280S). The other lacZ promoter is oriented towards a polylinker, which contains a unique NcoI site and a uniqueHindIII site. The mutants described above, K65C(−cys)His, D67C(−cys)His, K70C(−cys)His, and L74C(−cys)His, were digested with NcoI-HindIII and cloned into pUC12N/p51(−cys). The resulting clones will coexpress the cysteine-less p51 subunit and a p66 subunit with a histidine tail and only one cysteine at the designated site.

A single colony ofEscherichia coli strain DH5α transformed with one of the plasmids mentioned above was inoculated into 750 ml of NZY medium and grown at 37°C for 12 to 14 h before harvesting by centrifugation. The expression system is based on pUC, so induction is not required for the production of HIV-1 RT. The bacterial pellet was washed once with Tris-buffered saline (TBS) (pH 7.5). The pellet was lysed and the RT was partially purified on a nickel-chelating affinity agarose column by using the six-His tag on the C terminus of p66. Pooled imidazole gradient fractions were dialyzed and then further purified on Q-Sepharose. Purity of the protein preparation was checked by denaturing polyacrylamide gel electrophoresis (PAGE).

HIV-1 genomic sequences were subcloned from the pNL4-3 clone (2) into the LITMUS 28 plasmid (New England Biolabs, Beverly, Mass.) and sequenced. The R-PBS template RNA was synthesized according to the instructions contained in the Ambion Megashortscript kit (Ambion, Austin, Tex.). In brief, an oligomer containing a T7 promoter, modified so that it contained the correct sequence for the 5′ end of the R region (5′-TAC GCCAAGCTACGTAATACGACTCACTATAGGTCTCTCTGGTTAGACC AGATCTGAGCCTGGGA-3′), and a second oligomer containing the primer binding site (PBS) sequence (5′-AGTCCCTGTTCGGGCGCCA-3′) were used to generate a PCR fragment from the pNL4-3 sequence cloned into LITMUS. The PCR fragment was used as the template for RNA synthesis. RNA was purified by electrophoresis on a 5% denaturing gel and then visualized under UV light, and the 200-base base band was excised. The gel slice was soaked overnight in a solution of 50 mM Tris (pH 8.0) containing 400 μg of proteinase K per ml (Promega, Madison, Wis.). The supernatant was recovered, extracted three times with an equal volume of phenol-chloroform, and ethanol precipitated. RNA was quantitated using a UV spectrophotometer. DNA oligomers were labeled with [γ-³²P-]ATP (Amersham Pharmacia, Piscataway, N.J.) and T4 polynucleotide kinase (New England Biolabs).

R-PBS template RNA was mixed with a fivefold molar excess of [³²P]PBS (5′-AGTCCCTGTTCGGGCGCCA-3′) and and incubated at 37°C for 30 min. A threefold molar excess of RT was added to the annealed primer-template and allowed to bind for 1 min at 37°C. Synthesis was initiated with the addition of RT start solution containing dNTPs (80 μM final concentration) and MgCl₂ (6 mM final concentration). Reactions were stopped after appropriate incubation time by the addition of an equal volume of a 90% formamide stop solution containing 1% sodium dodecyl sulfate (SDS), 4 μg of plasmid DNA per ml, bromophenol blue, and xylene cyanole. Reaction mixtures were heated to 95°C for 4 min, fractionated by electrophoresis on a denaturing acrylamide gel containing 0.05% SDS, dried under vacuum, and exposed to a PhosphorImager screen (Molecular Dynamics, Sunnyvale, Calif.). The total amount of DNA synthesis was used as an indicator of polymerase activity. The RNase H assay was performed as described earlier (20), the only difference being that there was a twofold excess of RT in the reaction mixtures, which contained 100 nM RNA-DNA template-primer.

Oligonucleotides were commercially synthesized by the phosphoroamidite method on a synthesizer with subsequent PAGE purification (BRL). Oligonucleotides were tagged by 5′ labeling with [γ³²P]ATP with T4 polynucleotide kinase (T4 PNK) obtained from Boehringer Mannheim and annealed at a 1:1 ratio for cross-linking experiments.

The oligonucleotide 30-mer 5′-GTG TGT GT[FdI] ATC GTG GCG CCC GAC AGG GAC-3′ was prepared on commercial dC-cpg (40 to 50 μmol/g) with a standard synthesizer protocol for generation of the 21-base sequence at the 3′ end. The 5′-DMT (dimethoxytrityl chloride)-protected phosphoramidite (15) derived fromO ⁶-(2-p-nitrophenylethyl)-2-fluoro-2′-deoxyinosine was coupled manually to the 5′ end of the oligonucleotide bound to the support (7). A typical 2-μmol synthesis utilized 20 mg (21 μmol) of the phosphoramidite and 150 μl of 0.5 M 4,5-dicyanoimidazole in acetonitrile for 16 h at room temperature; yield was estimated from the recovery of DMT cation after deprotection. End capping with acetic anhydride was omitted after the manual coupling step (7), and the support-bound oligonucleotide was directly oxidized (1 M tetrabutyl hydroperoxide in dichloromethane for 30 s) and returned to the synthesizer for addition of the remaining eight residues by the standard automated synthesis procedure. After removal of the 5′-DMT protecting group, the linker was coupled to the support-bound oligonucleotide by a modification of the procedure of Erlanson, Chen, and Verdine (15). The support-bound oligonucleotide was treated with a combination of 45 mg (178 μmol) of 3,3′-dithiobis(propylamine) dihydrochloride (17), which was prepared as described for the 4-carbon homolog (12), 60 μl (600 μmol) of triethylamine, and 100 μl of H₂O for 16 h at room temperature. In the course of preparingO ⁶-(2-p-nitrophenylethyl)-2-fluoro-2′-deoxyinosine, we observed cleavage of the nitrophenylethyl protecting group in the presence of wet tert-butyl ammonium fluoride. After reaction with dithiobis(propylamine), addition of concentrated NH₄OH (1.5 ml) containing 20 μmol of tert-butyl ammonium fluoride to the beads and solution, followed by heating at 60°C for 3 days, resulted in complete deblocking of the oligonucleotide; this procedure avoided the DBU (1,8-diazabicyclo[5.4.0]undec-7-ene)-formamide cleavage step and accompanying formylation (15) of the free amino group of the tether. After filtration, the oligonucleotide solution was dialyzed against 0.1 M triethylammonuim acetate buffer (pH 6.0) overnight to remove excess amines. The oligonucleotide was purified by high-performance liquid chromotography on a Hamilton PRP-1 column (7 μm, 10 by 250 mm) eluted at 3 ml/min with a linear gradient of acetonitrile in 0.1 M (NH₄)₂HCO₃buffer (pH 7.5) which increased the acetonitrile concentration from 0 to 17.5% over 20 min (retention time, 16.3 min). Mass spectrometry (electrospray) calculated, 9,477; found, 9,475.

Ideally, photoaffinity cross-linking is performed by using a reagent that is covalently attached to one of the members of a biomolecular complex (here, RT), and this reagent remains inert to all components of the reaction mixture until it is activated by irradiation with mild UV light. Photoactivation is initiated only after the relevant biomolecular complex has formed, and activation transforms the reagent into a highly active and (presumably) highly unselective chemical moiety that quickly reacts with its nearest neighbor whether it is a part of a biomolecular complex, a buffer component, or a water molecule. If there is a specific complex, the probability of cross-linking to the other component of the biochemical complex (in this case, DNA) is significantly higher than if there is no such complex. Even though nonspecific reactions with other components of the mixture (primarily water) reduce the efficiency of the cross-linking to DNA, the nonselectivity of the reaction means that the photocross-linking of two biological components (RT and DNA) can be interpreted as a direct interaction, and the shortest distance between the two cross-linked points in the natural complex can be considered to be no greater than the length of the cross-linker.

We used two types of photoactivatable thiol-specific reagents: a carbene-generating compound,N-bromoacetyl-N′-{2,3-dihydroxy-3- [3-(3-(trifluoromethyl)diazirin-3-yl)phenyl]propionyl}ethylenediamine (BATDHP), obtained from Biolinx LLC (Hagerstown, Md.) and a nitrene-generating compound, azidophenacylthiopyridine (APTP), obtained from Sigma (St. Louis, Mo.) (29). These were coupled to the SH- group of single Cys-containing RT mutants. Carbenes are among the most reactive moieties known. They are capable of reacting with any chemical bond present in a biomolecule, including aliphatic chains and aromatic rings. Reacting in nanoseconds, carbenes rapidly form covalent bonds with neighboring atoms. The significant electrophilicity of carbenes is “overpowered” by their high reactivity, and, in the absence of nucleophiles, carbenes will react even with C-H bonds (over 80% cross-linking to cyclohexane (35, 38). The high level of reactivity of carbenes with buffer components usually precludes high yields of cross-linked products.

Nitrenes, such as those generated from the azide-containing reagent APTP, are less reactive and tend to undergo intramolecular rearrangements that lead to less reactive products. They cross-link primarily to nucleophiles such as amino groups and, in a nonnucleophilic environment, can remain active for periods of up to several minutes. This makes them less reliable for the detection of close interactions, since selective cross-linking may occur to a relatively distant nucleophilic group that is only occasionally in the vicinity of the cross-linker. The efficiency of cross-linking with nitrenes formed from azides is thus higher, but there is a possibility of bias toward interactions with nucleophilic groups. Photocross-linking reagents were prepared as 10 to 20 mM stock solutions in dimethyl sulfoxide and stored in the dark at −20°C for not longer than 30 days.

All of the RT mutants were modified with photocross-linking reagents via a single Cys residue on p66. Fifty microliters of 1 to 10 μM solutions of RT were treated with 5 mM DTT on ice for 30 min to reduce the SH- group. DTT was than removed by gel filtration with Centrisep desalting columns (from Princeton Separations, Adelphia, N.J.) in buffer 1 (Tris-Cl [pH 8.0] 60 mM KCl, 10 mM MgCl₂, 1 mM CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate}). The reduced RT was allowed to react with 10- to 50-fold molar excess of a photoreagent in dark vials on ice for 4 to 12 h. Excess of the photocross-linking reagent was removed by gel filtration. All subsequent manipulations were carried in low light levels.

To estimate the extent of SH-modification, RT samples were reacted (before and after modification with thiol-specific cross-linking reagents) with thiol-specific biotin-maleimide (BMCC) from Pierce (Rockford, Ill.) at pH 5.2 according to the manufacturer's instructions. All of the reactions were performed in degassed buffers under argon gas. Samples were loaded on nonreducing PAGE in 1 M SDS and 3 M urea without boiling to avoid non-thiol-specific biotinylation at high pH and high temperatures. Reaction mixtures were analyzed by Western blotting with streptavidin conjugated to alkaline phosphatase from Sigma (St. Louis, Mo.).

Modified RT (1 μM) and template-primer (0.03 μM [5′ labeled with [γ³²P]ATP]) were incubated in buffer 1 for 5 min at 37°C and then UV irradiated with a handheld lamp (model UVM-57 from UVP, Upland, Calif.) for 15 min on ice with a glass plate as an additional filter (cutoff, 315 nm). Nonreducing denaturing PAGE was used for separation of the template-primer covalently cross-linked to RT. The cross-linked products were quantified with a PhosphorImager (Storm 860 from Molecular Dynamics, Inc., Sunnyvale, Calif.). The negative control samples were obtained by cleaving specific covalent bonds in the cross-links. APTP cross-links can be cleaved by reducing the disulfide bond formed with the SH group of modified Cys. BATDHP cross-links are cleavable in the presence of 10 mM NaIO₄ which oxidizes a cis-diol bond built into the reagent for this purpose.

The chemical cross-linking experiments were designed so that the reaction (S-S cross-linking) involved a particular chemical group on both biomolecules (RT and DNA). The reagents have only a minimal reactivity to other chemical moieties, including buffer components and water. Since there are no other targets in the reaction mixture, this type of chemical cross-linking is efficient. Unlike photoactivatable reagents, the chemical cross-linkers are reactive during the whole time of an experiment. This means cross-linking can occur whenever the modified elements are close enough for interaction, which does not necessarily require that these elements be appropriately complexed. However, in experiments in which an SH- group on DNA was cross-linked to the thumb of HIV-1 RT, cross-linking was specific and depended on appropriate alignment of the SH- groups on the protein and the nucleic acid (25).

The reaction mixture containing 0.03 μM template-primer with 1 to 2 μM RT in buffer 1 was incubated for various times up to 2 h at 37°C (Table 1). This assay was also performed with template oligonucleotides 5′- labeled with γ-³²P. Excess RT was present in the reaction mixture to ensure that all of the cross-linking occurs in RT bound to template-primer. The products were fractionated by nonreducing denaturing PAGE and quantified with a PhosphorImager. The disulfide bonds in the cross-linked products were cleaved to produce the negative control samples.

There was relatively efficient photoaffinity cross-linking of template-primer with a template that had an 11-nucleotide (nt) 5′ overhang and the modified RT proteins with Cys residues in positions 65, 67, 70, and 74 of the fingers domain of the p66 subunit. The relationship of the b3-b4 loop with template-primers and the scheme of the reactions are shown in Fig. 2. APTP-modified RTs produced 10 to 25% cross-linking, BAT-DHP-modified RTs produced lower yields of cross-linking (<8%). This is consistent with the chemical properties of the carbene-generating photocross-linker. Relative to APTP, BAT-DHP has a longer spacer (12 Å), and the reactive species has a shorter half-life (10⁻⁹ s). These results suggest that RT residues 65, 67, 70, and 74 are all located within 7 to 12 Å of the template-primer. Relative levels of cross-linking with the same template-primer are similar for modified cysteines at positions 65, 67, and 74 (up to 25% with APTP and 5 to 8% with BATDHP) and lower for position 70 (6 and 2%), respectively.

In order to determine which strand of the template-primer reacted with the photoaffinity reagent, we used template-primer duplexes in which either the 5′ end of the template or the primer was labeled. In these experiments, we used templates with an 11-nt 5′ overhang. To avoid the possibility of partial denaturation of dsDNA in a protein gel (which would overestimate the amount of cross-linking to the labeled strand), we used both a nonreducing protein gel and a Tris-borate-EDTA–urea 6% denaturing sequencing gel to determine the extent of cross-linking. All of the mutant RTs preferentially photocross-linked to the template strand of the dsDNA template-primer substrate. Formation of cross-links with the primer strand was about 5- to 10-fold less efficient for APTP and 3- to 8-fold lower for BATDHP reagent than cross-links to the template strand (Fig. 3). These experiments were also performed with different lengths of the template overhang, and similar results were obtained even if the template overhang was as short as +2 (data not shown).

Having established that the cross-linking primarily involved the template strand for the four different sites in the fingers subdomain, protein gels were used for subsequent analysis, and we assumed that cross-linking to the primer strand was only a minor portion of the total cross-linking.

Experiments comparing the efficiency of cross-linking with template overhangs of various lengths made it possible to estimate the preferred position of cross-linking to the template. The template overhang sequence was a (TG)n repeat that should have no propensity to form secondary structures and no heterogeneity in sequence. Assuming that the probability of cross-links to an extended template is a function of the spatial position of the individual amino acid residues relative to template overhang and not a function of the nucleotide sequence of DNA, these data can be used as a measure of distance between the defined mutant Cys and the position on the template. Since the photocross-linkers are on flexible arms and have no strong specificity for distinct positions along the templates we have used, cross-linking to a longer template will always be as efficient as, and will often be more efficient than, cross-linking to a shorter template. As might be expected, the efficiency of cross-linking plateaued as the template was extended. This allows us to estimate the range of template lengths for which the photocross-linkers make optimal contacts with the nucleic acid. By using two photocross-linkers with different chemistry, we should be able to obtain unambiguous results (Fig. 4). The data presented in Fig. 4are normalized for convenience of comparison and to enable averaging of multiple independent experiments that had similar patterns, but the absolute values obtained in the experiments varied. The actual yields of cross-linking at maximum (generally with TP+11) were as follows: K65C-APTP, 15% ± 3%; K65C-BATDHP, 4% ± 2% D67C-APTP, 20% ± 5%; D67C-BATDHP, 4% ± 2% K70C-APTP, 7% ± 3%; K70C-BATDHP, 2% ± 1% L74C-APTP, 25% ± 8%; L74C-BATDHP, 5% ± 2%.

The results obtained with the two cross-linkers are similar. For positions 65 and 70, the results obtained with BATDHP and APTP are offset by about 2 bases, which could reflect the difference in the length of the cross-linking arms. However, these differences are less obvious for positions 67 and 74. Modified RTs with photocross-linkers at positions 65 and 74 of p66 demonstrate significant cross-linking with template overhangs of +5 and longer. No additional increase in cross-linking occurs after +7 for K65C, and there is little increase beyond position +5 for L74C. The corresponding range for D67C is +7 to 11, and for K70C, it is +5 to 11. (Fig. 4). Comparing the efficiency of cross-linking suggests that the path taken by the DNA template extension beyond polymerase active site first passes by position 74, then passes 65 and 70, and finally goes past 67.

Crystallographic studies have shown that binding of an incoming dNTP to the DNA-RT complex alters the position of the p66 fingers subdomain (25). We wanted to study the effect of this type of structural change on photocross-linking to the extended template. To avoid incorporation of the incoming dNTP, experiments were performed with ddG at the 3′ end of the primer. Addition of a dNTP mix (100 μM each) to such RT-DNA complexes leads to the formation of a stable ternary complex with the incoming dNTP (19, 25, 36). Because the modified RTs retain full polymerase activity, we have assumed that they bind the incoming dNTP normally. The addition of 100 μM dNTPs to complexes formed with any of the modified RTs affects the efficiency of the photocross-linking reactions. To help understand how binding a dNTP affects the efficiency of photocross-linking, the results of experiments done with and without dNTP were compared. To be sure that there is no alteration of template-primer preference, experiments were done to determine the relative efficiency of the cross-linking to the template and the primer. In the presence of the incoming dNTP, the preference for cross-linking to the template is not affected for position 67, there is moderate decrease for position 70, and there a more significant decrease for positions 65 and 74 (see Discussion). In terms of efficiency of cross-linking, there was in general more cross-linking in the open configuration. However, when the template overhang was short, cross-linking was low, but there was more cross-linking with the closed configuration (Fig.5). These results are entirely consistent with what we know about the structure of the open and closed complexes. When the template extension is short, the closed conformation would bring the photocross-linker closer to the DNA, which would allow for some cross-linking to occur. However, when the template extension is long, the closed configuration could interfere with the ability of the fingers to interact with the extended template.

A chemically modified oligonucleotide was synthesized that contained an SH- tether linked at theN ² of guanine residue at a defined position of the template (Materials and Methods). The SH-modified template oligonucleotide was 5′- labeled and annealed with one of the primers that placed the modified guanine at a specific position on the template. By adding appropriate dNTPs and/or ddNTPs, these primers were extended by RT to change the position of the modified G∗ residue relative to the active site as desired and create either a binary complex (RT-DNA) or a ternary complex (RT-DNA-NTP). In the absence of a reducing agent, disulfide bonds can form if the free SH- groups of the protein and DNA are in close contact (25). The relative efficiency of such cross-linking was used as a measure of the relative positions of the two SH- groups (one on the DNA template overhang and the other on RT). To ensure that the cross-linking occurred only as a result of direct interaction, RT was used in molar excess relative to the modified template-primer. The cross-linking kinetics demonstrated that no measurable cross-linking occurred before 5 min, by which time all of the template-primer was bound and extended by RT (data not shown). The length of the tether connecting the sulfhydryl group to guanine was used as a measure of distance (≤4 Å) between two cross-linked points. This provided an independent measure of distance that could be compared with the photoaffinity cross-linking experiments. No RT-RT crosslinking was seen, which suggests that the cross-linking to the nucleic acid was specific (data not shown).

All of the mutant Cys residues (positions 65, 67, 70, and 74) reacted with the sulfhydryl group on the modified template. The cross-linking yields are as high as 80% for K65C, L74C, and D67C and 30 to 40% for K70C. These data suggest that the template-primer and the fingers of RT can interact in a fashion that puts the two sulfhydryls in close proximity and at appropriate angles.

To determine the position of the nearest nucleotide residue on the template overhang that can form a disulfide bond with the thiol group of the mutant RT, we performed a number of cross-linking experiments with the modified template annealed with DNA primers of various lengths. When binary complexes were formed with K65C and L74C, the optimal SH-cross-linking was observed at positions +4 to 5, while for D67C and K70C, the preferred positions were +5 to 6 relative to polymerase active site (Fig. 6). This is in reasonable agreement with the photoaffinity cross-linking experiments.

As has already been discussed, the position of the fingers changes when HIV-1 RT binds an incoming dNTP. To determine how differences in the position of the fingers affects cross-linking, ternary complexes (RT-DNA-NTP) were formed with primers of various lengths by addition of complementary ddNTP and then dNTP (100 μM each). The relative efficiency of SH cross-linking to the modified G∗ in binary complexes was slightly higher than in ternary complexes (Fig.7). There was no shift of the preferred position of cross-linking for K65C and K70C cross-linking; for D67C and L74C, the preferred position was shifted 1 base farther away (from +4 to +5).