Inhibition of HIV Replication by Cyclic and Hairpin PNAs Targeting the HIV-1 TAR RNA Loop

Human immunodeficiency virus-1 (HIV-1) replication and gene expression entails specific interaction of the viral protein Tat with its transactivation responsive element (TAR), to form a highly stable stem-bulge-loop structure. Previously, we described triphenylphosphonium (TPP) cation-based vectors that efficiently deliver nucleotide analogs (PNAs) into the cytoplasm of cells. In particular, we showed that the TPP conjugate of a linear 16-mer PNA targeting the apical stem-loop region of TAR impedes Tat-mediated transactivation of the HIV-1 LTR in vitro and also in cell culture systems. In this communication, we conjugated TPP to cyclic and hairpin PNAs targeting the loop region of HIV-1 TAR and evaluated their antiviral efficacy in a cell culture system. We found that TPP-cyclic PNAs containing only 8 residues, showed higher antiviral potency compared to hairpin PNAs of 12 or 16 residues. We further noted that the TPP-conjugates of the 8-mer cyclic PNA as well as the 16-mer linear PNA displayed similar antiviral efficacy. However, cyclic PNAs were shown to be highly specific to their target sequences. This communication emphasizes on the importance of small constrained cyclic PNAs over both linear and hairpin structures for targeting biologically relevant RNA hairpins.


Introduction
The transcriptional transactivation of the HIV-1 genome requires a specific interaction between the highly conserved TAR RNA hairpin fragment with the viral Tat protein and cellular factors (PTEFb-cyclin T1-CDK9 kinase complex). Both the six-nucleotide loop and the three-nucleotide bulge of TAR RNA (Figure 1(a)) are involved in the formation of this complex [1][2][3]. Therefore, molecules that can bind to the bulge or the loop of TAR are of great therapeutic interest, since disruption of the ternary complex formation leads to abortive mRNA synthesis and, consequently, to inhibition of viral replication.
During the last decade, a wide number of TAR ligands have been described [4,5]. Among them, one can cite R06 aptamers (such as R06 24 or R06 18 , Figure 1(b)), which were identified initially by in vitro selection [6]. These aptamers are folded RNA stem-loop structures which recognize the mini-TAR fragment (Figure 1(a)) not only on the basis of sequence complementarity, as classical antisense oligomers, but also on the basis of the tertiary structure of their target. This leads to highly stable and specific loop-loop complexes, also called "kissing complexes." The key features for the establishment of such complexes are the hairpin structure of R06 aptamers as well as the octameric loop constituted by the 5 -UCCCAG-3 sequence complementary to the TAR hexaloop, flanked by a G and a A residues. Although these two G/A residues are not directly involved in the loop-loop interaction, they were shown to be crucial for the formation of a stable kissing complex [7][8][9][10].
In a cellular compartment, RNA aptamers are rapidly degraded by nucleases, limiting their potential as therapeutic agents. Thus, several chemically-modified R06 derivatives were prepared with the view of improving both the pharmacological properties and TAR affinity. N3-> P5 phosphoramidate [11,12], 2-O-methyl RNA [13,14], and some Journal of Nucleic Acids hexitol nucleic acids (HNA)/RNA mixmers [15] were shown to display an improved nuclease resistance while maintaining a similar TAR-binding constant. TAR-binding properties of R06 analogs containing LNA residues were also studied [10,[16][17][18][19]. While the fully modified LNA version of R06 proved to be a poor TAR ligand, some chimeric LNA/DNA, and LNA/2 -OMe RNA aptamers displayed binding properties of interest. However, the identification of such chimeric aptamers is laborious, because it requires a systematic screening of all possible combinations, as no rule dictates the number and positions at which LNA nucleotides have to be incorporated to allow a strong loop-loop interaction.
Concerning the biological activity of these aptamer analogs, although some of them were shown to inhibit specifically Tat-mediated transcription in cell-free assays [12,13,15,20,21] or in cell assays when transfected with cationic lipids [17], none of them was evaluated as anti-HIV agents. However, it was shown that, when expressed endogenously in HeLa cells, the RNA aptamer R06 was able to inhibit HIV replication [22], highlighting the antiviral potential of nuclease resistant molecules that recognize the TAR loop through both their primary sequence and their tertiary structure.
Based on these results, we previously devised small synthetic constrained structures derived from the R06 aptamer derivatives, and reported that they were able to interact with the TAR loop through "kissing-like" complexes of high affinity [23]. These structures are constituted by an octameric PNA ( Figure 2) 5 -GTCCCAGA-3 sequence identical to the one found in R06 aptamers, head-to-tail cyclized via polyamide linkers of different length (1a-c, Figure 2(b)). We chose to introduce PNAs as RNA mimics since they are highly stable in biological media and they hybridize strongly with complementary RNA sequences [24]. A limitation to their in vivo applications is their poor ability to cross cell membranes. Thus, a lysine residue was incorporated in cyclic PNAs to allow their subsequent conjugation with cell penetrating moieties [25][26][27].
Here, we report the synthesis of cyclic PNAs (1-3, Figures 2(b) and 3) and hairpin PNAs (4-6, Figure 3) conjugated to a triphenylphosphonium (TPP)-based cell penetrating vector (Figure 4(a)). This vector is constituted by a TPP lipophilic cation capable of transporting PNAs across the lipid bilayer [28,29], bound via a disulfide bridge, to a mercaptoethoxycarbonyl moiety connected to the PNA. Intracellularly, the reduction of the disulfide bond leads to a spontaneous decomposition that releases the PNAs (Figure 4(b)) [30]. The cyclic PNAs exhibited potent anti HIV-1 activity in comparison to other derivatives, confirming the therapeutic potency of these conjugates.

General Methods.
Reagents and solvents were obtained from commercial sources and used without further purification unless indicated. Analytical HPLC chromatograms were obtained using an HP1100 UV detector set at 260 nm and a Beckman Ultrasphere RP-C18 (5 μm) column at 55 • C.
Purifications using semipreparative HPLC were done on the same instrument using a Phenomenex Jupiter column RP-C18 (5 μm). Elution solvent A was water (0.1% TFA); elution solvent B was acetonitrile (0.1% TFA). ESI mass spectra were recorded with a Bruker Esquire 3000 plus. Concentrations of cyclic PNAs, hairpin PNAs, and TPP conjugates were determined by UV spectroscopy, using the usual extinction coefficients [31]. The mini-TAR RNA fragment used for thermal denaturation studies was purchased at Dharmacon Inc. (Lafayette, USA). Thermal denaturations of mini-TAR/PNAs complexes were carried out on a Varian Cary 300 Scan spectrophotometer.

Synthesis of Cyclic PNA 3 and Hairpin PNAs 4-6.
These PNAs were synthesized in Merrifield vessels on MBHA resin (100-200 mesh, 0.63 mmol/g, Merck Schuchardt OHG, Hohenbrunn, Germany), on a 100-μmol scale. Elongation was carried out starting from Boc/Z protected PNA monomers, using HBTU as the coupling reagent, and NMP as solvent. Compound 3 was synthesized as previously described for cyclic PNAs 1a-c and 2 [23]. The lysine residue at the 5 -end of hairpin PNAs 4-6 was introduced after elongation, by means of Boc-Lys(2-Cl-Z)-OH and HBTU as the activator. Acetylation of the lysine residue was performed after Boc deprotection (TFA/TIS 10%, 4 mL for 15 min), using an Ac 2 O/pyridine/NMP 1/1/8 v/v/v solution (2 × 4 mL for 15 min). Compounds 4-6 were deprotected and cleaved from the resin using a TFMSA/TFA/TIS solution (1 : 8 : 1) for 4 h, then precipitated in cold anhydrous diethyl ether. Crude products were isolated by centrifugation (3,000 min −1 , −4 • C), washed twice with diethyl ether (10 mL), and purified by semipreparative HPLC using the following method: 55 • C, A/B 100/0 for 7 min, then from 100/0 to 50/50 for 45 min, with a flow rate of 2 mL/min.    Journal of Nucleic Acids Figure 1: Sequence and secondary structure of (a) HIV-1 mini-TAR RNA, (b) R06 24 and R06 18 aptamers reported in this study. Bold bases indicate complementarity between aptamer and TAR loops. The crucial G and A residues flanking the R06 aptamers loop are in italics.

Thermal Denaturation Studies.
One nmol of mini-TAR was solubilized in 250 μL (4 μM concentration) of R buffer solution at pH 7.3, that buffer containing cacodylate (20 mM), NaCl (20 mM), KCl (140 mM), and MgCl 2 (0.3 mM). The solution was heated at 90 • C for 2 min, immediately cooled at 4 • C, and maintained at this temperature for 10 min, then kept at 20 • C. For preparing hairpin PNAs 4-6, a solution of each compound in R buffer (4 μM) was heated for 3 min at 95 • C, then cooled to 20 • C with a rate of 0.5 • C/min [31]. Individual compounds 3, 4-6 and mini-TAR in R buffer (2 μM final concentration of each) were mixed, then incubated at 5 • C for 1 h. Thermal denaturation was generated by increasing the temperature from 5 • C, to 90 • C at 0.4 • C/min, then followed by UV absorption (260 nm). Melting temperatures (T m ) were determined as the maximum of the first derivative of the melting curves.

Transfection and Production of HIV-1 Virions.
For production of highly infectious pseudotyped HIV-1 virions, 293T cells grown in complete Dulbecco's modified Eagle's medium (DMEM) were cotransfected with pHIV-1JR-CSFlucenv(−) and pVSV-G, using a calcium phosphate transfection system (Invitrogen Carlsbad, CA, USA) [32,33]. The culture supernatant was saved at 24, 48, and 72 h after transfection, then pooled and analyzed for p24 antigen using the ELISA p24 antigen kit (ZeptomMetrix, Buffalo, NY, USA). The pseudotyped HIV-1 virions were then isolated from the culture supernatant by filtration through a 0.45 μm pore size PVDF membrane (Millipore Bedford, MA, USA) and then by ultracentrifugation at 70,000 g for 45 min. The viral pellet was resuspended in complete Dulbecco's medium and stored at −80 • C.

Anti-HIV-1 Activity in CEM Cells
. CEM CD4+ lymphocytes 12D7 were grown in RPMI-1640 medium supplemented with 10% fetal calf serum and 4 mM L-glutamine at 37 • C in 5% CO 2 containing humidified air [34]. Earlyto mid-log-phase cells were harvested and washed with an equal volume of PBS without Ca 2+ and Mg 2+ . Approximately

Results and Discussion
Previously, we have shown that cyclic PNAs 1a-c tightly bind to TAR (Figure 2(c)), with a higher affinity than that of a R06 aptamer (R06 18 , Figure 1) and that they were highly specific for TAR despite the limited number of bases constituting them, since the introduction of a single mismatch in the PNA sequence was strongly deleterious for TAR binding. Indeed, compound 2 (Figure 2(b)), in which the C4 residue was replaced by an A4 residue, showed no affinity for TAR [23]. The first goal of the present study was to assess whether these PNA structures, which are cyclized in a covalent way, are more advantageous for targeting the TAR loop than hairpin structures in which the loop is not covalently closed. Although PNAs are among the best nucleic acid mimics, no PNA analogue of R06 aptamers has been reported so far. Thus, we have prepared hairpin PNAs (compounds 4 and 5, Figure 3) containing the same octameric PNA sequence than in cyclic PNAs 1a-c, closed by two and four G-C pairs, respectively, and measured their affinity for TAR. The second goal of this study was to determine whether, as for R06 aptamer derivatives, the G and A PNA residues flanking the loop sequence are necessary for the establishment of stable loop-loop complexes. For this purpose, we synthesized the G-and A-deleted cyclic PNA 3 and hairpin PNA 6 ( Figure 3), and studied their interaction with TAR. Finally, in order to evaluate the ability of both cyclic (1a-c and 2-3) and hairpin (4-6) PNAs to inhibit HIV replication in infected cells, we conjugated them to a cell-penetrating vector, via their lysine residue (Figure 4(a)). The vector chosen in this study is an intracytoplasmic biodegradable triphenylphosphonium (TPP)-based moiety, which was shown to allow the uptake and release of a "naked" PNA into cytoplasm (i.e., without any residual TPP moiety attached to PNAs, Figure 4(b)) [30]. For antiviral activity studies, a previously described TPP conjugate of a 16-mer PNA TAR targeting the apical stemloop of TAR was taken as a reference compound (Figure 4(c)) [30].

Chemistry.
The synthesis of compounds 1a-c and 2 was previously reported [23]. Cyclic PNA 3 was prepared following a solid-phase strategy via on-resin cyclization, using a glutamic acid-anchored MBHA resin, as for cyclic PNAs 1a-c and 2. Hairpin PNAs 4-6 were synthesized on a MBHA resin, using standard procedures. Briefly, the elongation was performed using Boc/Z-protected PNA monomers and HBTU as coupling reagent. The lysine residue at the 5 -end was introduced after elongation by means of Boc-Lys(2-Cl-Z)-OH and HBTU. After Boc cleavage under acidic conditions (TFA/TIS 10%), the α-amino group of the lysine residue was acetylated using an Ac 2 O/pyridine/NMP solution. Hairpin PNAs 4-6 were obtained after deprotection and cleavage from the resin using a TFMSA/TFA/TIS solution The TPP-conjugates of cyclic and hairpin PNAs were obtained in almost quantitative yields from their corresponding cyclic 1a-c, 2-3 and hairpin 4-6 precursors, in one step, using an excess of the key para-nitrophenyl carbonate reagent 7 in the presence of sodium azide, as previously described [28]. The TPP-conjugates were purified by semipreparative HPLC, with an RP-C18 column and water (0.1% TFA) and acetonitrile (0.1% TFA) as the elution solvents. Their structures were confirmed by ESI-MS experiments in which the corresponding spectra displayed characteristic (M+nH) (n+1)+ /(n+1) peaks (n = 1 to 5) (see experimental protocols in Supplementry Material).

Thermal Denaturation Study.
The affinity of compounds 3-6 for the mini-TAR RNA fragment was evaluated by thermal denaturation monitored by UV absorption spectroscopy (λ max = 260 nm) in R buffer, as previously reported for cyclic PNAs 1a-c and 2 and R06 18 aptamer [23]. Melting temperatures (T m ) of the TAR/cyclic PNA 3 and TAR/hairpin PNAs(4-6) complexes are summarized in Table 1, together with the melting temperatures of hairpin PNAs 4-6 alone (i.e., without TAR).
Thermal denaturation studies of hairpin PNAs 4 and 5 alone exhibited a single transition at, respectively, 73.0 • C and 84.0 • C ( Table 1), independently of PNA concentration, indicating that they fold to form highly stable hairpins [31]. The difference between their T m values (ΔT m = 11 • C) reflects the higher stability of the double strand in PNA 5 than in PNA 4, due to the presence of two additional canonical CG pairs in PNA 5 relative to PNA 4. The melting profiles obtained with mixtures of mini-TAR and hairpin PNAs 4 or 5 displayed two transitions (e.g., see Figure 5). The highest one forms a broad peak, resulting from the 6 Journal of Nucleic Acids   Concerning the 5 -G and 3 -A deleted cyclic PNA 3, the thermal denaturation study clearly demonstrates that it does not bind to TAR, highlighting the importance of the G and A flanking residues for the formation of a stable complex between TAR and cyclic PNAs. Conversely, the 5 -G and 3 -A deleted hairpin PNA 6 is able to smoothly interact with TAR. However, the melting temperature of the corresponding complex (38.4 • C) is higher than the melting temperature of the hairpin PNA 6 itself (29.8 • C). Thus, it is likely that the formation of the complex with TAR occurs, at least in part, on the unfolded form of 6.
It has been earlier shown for TAR/TAR RNA aptamer complexes that the presence of G and A loop closing residues is a key structural determinant conferring a high stability both to the RNA aptamer alone and the TAR/RNA aptamer complex. Substituting the GA pair by the AU one (or GC, CA. . .) decreases the T m of both the RNA aptamer and the TAR/RNA aptamer complex by 17 • C and 14 • , respectively [9]. Comparing PNA 5 and 6 shows that the presence of the GA pair also leads to a drastic increase in the stability of both the PNA hairpin alone and the TAR/PNA complex (ΔT m of 54 • and 14 • C, resp.). NMR [8] and molecular dynamics studies [9] have shown that the presence of these two residues increases the stability of both the aptamer and the kissing complex by increasing the stacking at the stem-loop junctions, via the stabilization of two hydrogenbond base pairs located at these stem-loop junctions of the kissing complex: the intramolecular G-A pair itself, via hydrogen bonding of N1-N1 carbonyl-amino type, and the intermolecular A-U pair, via the classical Watson-Cricks network. By contrast, when the loop of the aptamer is closed by the classical AU pair, the very high tension in the loop causes the opening of both this AU intramolecular pair and of the intermolecular AU one, leading to a less stable kissing complex. It is possible that such events also occur in the case of PNA 4-6 hairpins and corresponding TAR complexes but for the moment, no proof supports this hypothesis.
It is also possible that larger loop size of PNA 4/5 with a nonhydrogen bonding GA pair may offer greater stability to stem region as compared to small loop size of PNA 6 with a hydrogen-bonding GT pair closing the loop. The smaller loop size may cause strain on the stability of the stem region.

HIV-1 Inhibition in Cell Culture by Anti-TAR PNA.
In order to evaluate the in vitro antiviral efficacy, we incubated CEM CD4+ lymphocytes, infected with highly infectious VSV-G pseudotyped HIV-1 virions expressing the firefly luciferase reporter gene [30], with varying concentrations of individual TPP-[PNA 1a-c, 2-6]. Similar experiments were carried out with unconjugated PNAs 1a-c and 4-6. The TPP-conjugate of a 16-mer antisense [PNA] TAR targeting the apical stem-loop of TAR was taken as a reference compound [30] (Figure 4(c)). We have previously showed that this compound inhibited HIV replication in infected cells at a micromolar concentration. To measure the effect of TPP-PNA conjugates on HIV-1 production in CEM cells, we monitored the expression of the firefly luciferase reporter gene, cloned instead of the nef gene in the HIV-1 virions. After 48 h incubation followed by cells lysis, the extracts were normalized for total protein content and analyzed for quantitative levels of luciferase expression. An arbitrary value of 100 was assigned to the luciferase activity obtained in infected cells in the absence of compounds; values relative to this control value were given to the other samples. Median dose effects (IC 50 ) for individual compoundwere then determined using CalcuSyn software ( Figure 6). As expected, no antiviral activity was detected for the unconjugated PNAs 1a-c and 4-6, probably as a consequence of their poor cellular permeation (data not shown). Median dose effects (IC 50 ) obtained for individual TPP-[PNA 1a-c, 2-6] are summarized in Table 2.
In all cases, except TPP-conjugates of PNAs 2 and 3, a substantial decrease in HIV-1 replication was observed as the concentration of TPP-conjugate was increased, the IC 50 values ranging from 1.24 to 3.70 μM ( Table 2). The micromolar inhibitory effects measured for TPP conjugates of cyclic PNAs 1a-c are very encouraging, because they are similar to those obtained with the heavier TPP-conjugate of  (Figure 4(c)). As previously noticed for the complexes stability (T m from to 41 • C to 43.4 • C, Table 1), the length of the linker closing cyclic PNAs 1a-c (n = 3, 4, 5, Figure 2(b)) has little influence on the antiviral activity of their TPP conjugates (IC 50 from 1.24 to 1.94 μM). In addition, it appears that the antiviral activity of TPP-[PNA 1a-c] conjugates is specific. Indeed, TPP-[PNA 2] and TPP-[PNA 3], which, respectively, derive from the mismatched cyclic PNA 2 and from the GA-deleted cyclic PNA 3, have no effect on viral replication. These results, together with the fact that no interaction between mini-TAR and these two cyclic PNAs was detected, tend to further demonstrate that Altogether, these results emphasize the advantage of cyclic PNA structures over hairpin ones for inhibiting HIV-1, through the targeting of the TAR RNA loop.

Conclusion
We demonstrated that the small cyclic PNAs targeting the HIV-1 TAR RNA loop inhibit viral replication when conjugated to a cell penetrating vector, as efficiently as do higher molecular weight compounds, such as hairpin PNAs or an anti-TAR 16-mer PNA antisense targeting both the stem and loop of TAR. Furthermore, despite their short PNA sequence, they are highly specific for their RNA target, since the introduction of a single mismatch in the PNA sequence is detrimental both for TAR binding and HIV 8 Journal of Nucleic Acids inhibition. In addition, these results, combined with the high PNA stability in biological media, indicate that such cyclic compounds hold potential as new anti-HIV agents. On the other hand, these results emphasize the advantage of using small constrained cyclic structures over both linear antisense oligonucleotides and hairpin ones for targeting biologically relevant RNA hairpins.