Simultaneous recognition of HIV-1 TAR RNA bulge and loop sequences by cyclic peptide mimics of Tat protein

Abstract

The interaction of the HIV-1 transactivator protein Tat with its transactivation response (TAR) RNA is an essential step in viral replication and therefore an attractive target for developing antivirals with new mechanisms of action. Numerous compounds that bind to the 3-nt bulge responsible for binding Tat have been identified in the past, but none of these molecules had sufficient potency to warrant pharmaceutical development. We have discovered conformationally-constrained cyclic peptide mimetics of Tat that are specific nM inhibitors of the Tat-TAR interaction by using a structure-based approach. The lead peptides are nearly as active as the antiviral drug nevirapine against a variety of clinical isolates in human lymphocytes. The NMR structure of a peptide–RNA complex reveals that these molecules interfere with the recruitment to TAR of both Tat and the essential cellular cofactor transcription elongation factor-b (P-TEFb) by binding simultaneously at the RNA bulge and apical loop, forming an unusually deep pocket. This structure illustrates additional principles in RNA recognition: RNA-binding molecules can achieve specificity by interacting simultaneously with multiple secondary structure elements and by inducing the formation of deep binding pockets in their targets. It also provides insight into the P-TEFb binding site and a rational basis for optimizing the promising antiviral activity observed for these cyclic peptides.

Transcription of the HIV-1 RNA in infected cells is strongly activated by the complex between the viral Tat protein and its cognate transactivation response (TAR) RNA, a 59-nt RNA found at the 5′ end of all nascent viral transcripts (Fig. 1A). Tat and its cellular cofactor, the transcription elongation factor-b (P-TEFb), are recruited to the elongating RNA polymerase II (RNAP II) through interactions with TAR and are required for reactivation of the integrated proviral genome in latently infected cells (1). The cooperative binding of Tat and P-TEFb to TAR activates the CDK9 kinase of P-TEFb that phosphorylates RNAP II and the repressive NELF factors, leading to greatly enhanced RNAP II processivity (1–3).

Fig. 1.

RNA and peptide constructs used in NMR studies. (A) Sequence and secondary structure of the HIV TAR construct, numbered according to the start of transcription of the viral RNA. (B) Sequence of the L-22 cyclic peptide.

The Tat-TAR complex is an attractive target for developing new antivirals because the interaction between Tat and TAR is unique to the virus, whereas P-TEFb is used widely for transcription of most host genes. Furthermore, TAR is extremely conserved among viral isolates and P-TEFb plays a key role in promoting infectivity through the Tat-TAR complex and in the emergence from latency. These considerations have led to the synthesis and evaluation of numerous small-molecule and peptidic inhibitors of the Tat-TAR interaction during the last 15 years (4–7). However, none of these molecules had sufficient potency or selectivity to progress into preclinical studies and, indeed, inhibiting the Tat-TAR interaction has proven challenging. First, there is little precedent for the pharmacological disruption of protein-RNA complexes. Although the bacterial ribosome is a very well-validated drug target, most ribosome-binding antibiotics are large natural products with fairly unspecific RNA-binding activity. Mimicry of such compounds, often the aminoglycoside class of antibiotics, has so far not been fruitful. Second, RNA–protein interfaces are large; as with protein–protein interfaces, it remains challenging to inhibit them with small drug-like molecules, despite recent progress (8). Third, the structure of the Tat-TAR complex has so far been inaccessible, although structures of TAR bound to a variety of small-molecule ligands and Tat-derived peptides have been reported (9, 10). These studies have convincingly demonstrated that Tat binding to TAR induces a characteristic conformational change that can be mimicked even by the binding of just argininamide (9, 10), but various critical aspects of the structure remain unclear, most significantly the conformation of the apical loop where P-TEFb binds.

In seeking new approaches to inhibiting viral transactivation, we considered mimicking Tat protein with peptidic molecules. Although proteolytically stable peptide derivatives have been used previously to target TAR (5, 11), these linear peptide analogues were highly flexible and therefore bound various RNAs nonspecifically, preventing sufficient activity in cells and the development of clear structure–activity relationships or structural information. Thus, we focused our attention on the development of conformationally constrained mimics of HIV-1 Tat. We based the design of these compounds on the structure of the bovine immunodeficiency virus (BIV) complex whose high-resolution structure is available (12–14). BIV Tat protein and TAR RNA are highly homologous to HIV-1 Tat and TAR, and we recently demonstrated that it is possible to mimic BIV Tat protein by using cyclic peptides that bind to BIV TAR with nM affinity (15, 16).

Here, we report the discovery of a family of structurally constrained β-hairpin cyclic peptide mimics of HIV-1 Tat protein that bind to TAR RNA with nM affinity and have greatly improved specificity compared with previous small-molecule or peptidic ligands for TAR. The complex of one of the most potent peptides with TAR demonstrates that the peptides interact with the apical loop where P-TEFb binds and with the bulge where Tat binds. The cyclic peptides use a critically important bis-arginine motif, observed previously in small molecule–TAR complexes, to induce conformational changes in the TAR bulge region. Unexpectedly, the interaction with the loop nucleotides buries the peptide in a much deeper groove than previously observed in other structures of TAR, the formation of which relies on a conserved looped-out nucleotide to stabilize the peptide–RNA complex. By defining the molecular basis for the high affinity and specificity of this class of inhibitors of the Tat-TAR complex, the present structure provides a template for the optimization of this family of antiviral leads and introduces structural principles for RNA recognition.

Results

Identification of nM Inhibitors of the HIV-1 Tat-TAR Interaction.

We used structure-based methods to discover low nM inhibitors of the Tat-TAR interaction in BIV (15, 16). This was achieved by mimicking (14) the β-hairpin formed by the RNA-binding domain of BIV Tat protein upon binding its cognate TAR (12, 13) with cyclic peptides stabilized by the heterochiral d-Pro-l-Pro template (17). Given the similarities in sequence and secondary structures between BIV and HIV Tat and TAR, we reasoned that the same family of constrained cyclic peptides may also be able to act as HIV Tat inhibitors. We varied the amino acids on the peptide surface predicted to face the TAR RNA to optimize binding in HIV and then combined multiple individual amino acid changes.

Peptide binding to TAR was measured by EMSAs. Fig. 2A presents EMSA data for one of the peptides, L-22 (Fig. 1B,K_d = 30 nM; Table 1) that also efficiently inhibits formation of the ADP-1–TAR complex (Fig. 2B); ADP-1 is a 37-mer linear peptide containing the entire RNA-binding activity of Tat protein (10). EMSA experiments also demonstrated that L-22 inhibits formation of the complex between TAR RNA and a chimeric Tat/Cyclin T₁ protein at submicromolar concentrations. In remarkable contrast to linear peptides of similar length, binding to TAR could not be disrupted by a 10,000-fold excess of tRNA present in the binding buffer (Fig. 2A).

Fig. 2.

Binding of cyclic peptides to HIV-1 TAR and inhibition of the Tat-TAR complex by EMSA. (A) Binding of L-22 peptide (all concentrations are in μM) to HIV TAR (1 nM); the buffer contains a 10,000-fold excess of tRNA to ensure that the observed binding is specific. (B) Inhibition of the complex between the ADP-1 peptide (150 nM) and HIV TAR (1 nM) by L-22, in the presence of 10,000-fold excess of tRNA.

Table 1.

Sequences of the lead peptides L-22, L-50, and L-51 and summary of their activity for their interaction with HIV-1 and BIV TAR, as determined by EMSA

Mimetic	Position												Kd (HIV), nM	Kd (BIV), nM
	1	2	3	4	5	6	7	8	9	10	11	12
L-22	R	V	R	T	R	K	G	R	R	I	R	I	30	5
L-50	R	V	R	T	R	G	K	R	R	I	R	R	1	1
L-51	R	T	R	T	R	G	K	R	R	I	R	V	5	50

Among the ≈100 peptides investigated, 3 sequences stand out for their potency. L-22, L-50, and L-51 all have activity in the low nM range (Table 1); the K_ds for L-50 and L-51 are well below 10 nM. The interaction we observe is also specific: in addition to withstanding competition from a 10,000-fold excess of tRNA, some of these peptides discriminate even between HIV-1 and BIV TAR. For example, L-22 and L-51 have binding affinities of 30 and 5 nM, respectively, for HIV-1 TAR and 5 and 50 nM, respectively, for BIV TAR. The selectivity of L-51 toward HIV-1 TAR is remarkable given the similarities between the 2 RNAs and the origin of these peptides as structural mimics of BIV Tat. Furthermore, the results indicate that small changes in peptide sequence lead to significant changes in activity even for RNAs as closely related as the TAR RNAs from BIV and HIV-1. This finding is again in marked contrast to previous linear peptidic and peptidomimetic inhibitors of HIV-1 Tat-TAR, where changes in sequence generally led to only modest changes in activity and specificity against unrelated RNAs (5, 11). Satisfactorily, when nucleotide U23 was removed altogether or mutated to cytidine, no binding was observed, as established by EMSA (Fig. S1). Thus, these peptides discriminate at least 1,000-fold against RNA lacking the critical U23 nucleotide.

To understand the molecular basis for the activity of this class of Tat mimics, and to further improve their characteristics, we determined the structure of the L-22–TAR complex. This peptide was selected for structure determination because of the very large number of intermolecular NOEs observed in NOESY spectra in the complex with TAR.

Structure of the TAR–L-22 Complex.

The structure of the complex relied on 143 intermolecular NOEs to accurately define the relative position of the RNA (Fig. 3 and Table 2) and the peptide (Fig. 3A), although the RNA apical loop remains somewhat disordered (Fig. 3B) as also observed by relaxation studies (18). Remarkably, most arginine guanidinium NH₂ protons were assigned from the combined analysis of filtered and conventional NOESY spectra; presumably, strong interactions with TAR slowed down bond rotation and exchange with solvent sufficiently to allow their observation. As a consequence, at least 1 intermolecular NOE was visible from the guanidinium group for every arginine residue, facilitating their precise positioning in the structure.

Fig. 3.

Structure of the HIV-1 TAR/L-22 complex. (A) The orientation of the peptide in the RNA major groove is identical to what was observed in the complex between BIV TAR and a related peptide (14). (B) Superposition of 10 lowest-energy structures: guaninosines in purple, cytosines in black, adenosines in red, and uracils in gray. (C) The peptide does not simply stick into the major groove, but organizes the bulge and loop residues to partially close around it, creating a much deeper binding pocket than seen in other structures of TAR.

Table 2.

Experimental restraints and structural statistics for the L22–TAR complex

Total number of restraints	1,107
Average no. of restraints per hydrogen atom	2.3
NOE-derived restraints	809
Intermolecular NOEs	143
RNA (intramolecular)	464
Peptide (intramolecular)	204
Dihedral restraints	193
Dipolar coupling restraints	25
Hydrogen-bonding restraints	66
Planarity restraints	12
Average rmsd values from experimental restraints
Distance, Å	0.02
Dihedral,°	0.89
RDC, Hz	0.49
Average rmsd values from ideal geometries
Bonds, Å	0.003
Angles, °	0.64
Improper, °	0.50
Heavy atom rmsd from mean structure, Å
L-22 (all backbone atoms)	0.33
L-22 (all heavy atoms)	0.98
TAR RNA (all heavy atoms)	1.9
TAR RNA core (G18–U23, G26–C29, G36–C44)	0.64
L-22 and TAR core	0.86
Entire structure	1.9

The L-22 peptide binds TAR in the major groove of the upper RNA helix (nucleotides 26–29 and 36–39) with the d-Pro-l-Pro template facing down toward the lower helix (Fig. 3C). This orientation is analogous to what has been observed in the structure of the precursor peptide BIV-2 bound to BIV TAR RNA (14); this observation was important to the design of the HIV-directed library of peptides and it was reassuring to observe it in the final structure.

The Cyclic Peptide Makes Intimate Contacts with the Apical Loop of TAR, Burying It from Solvent into an Unusually Deep Binding Groove.

A very surprising feature is the position of loop residue A35. It is flipped out from the RNA as in other structures of TAR (9, 10), but it is also drawn down in an unprecedented way toward the UCU bulge through an interaction with the guanidinium group of Arg-11 (Fig. 4A), which resides near the d-Pro-l-Pro end of the peptide. In doing so, A35 draws other loop residues to partially close around the peptide so as to create a much deeper binding pocket than observed in any other structure of TAR (Fig. 3B). A35 draws with it the RNA backbone between residues G33 and G34 to partially close over the strand of the peptide between Arg-8 and Ile-12. As a consequence of the interaction between A35 and Arg-11, even Val-2, which was postulated to remain solvent exposed when the peptides were designed, is partially protected from solvent by interacting with the minor groove face of A35. These are key elements of the potency and selectivity of this class of peptides: replacement of Arg-11 with even Lys in a peptide closely related to L-22 led to the loss of any detectable binding. Substitution of A35 with a cytidine (A35C), however, results in a much smaller 5- to 10-fold loss of affinity for the peptide. NMR data for the A35C RNA–peptide complex reveals that the RNA adopts the same conformation as the wild-type RNA, but most intermolecular NOEs are no longer visible, suggesting that the interaction is less intimate than observed with the wild-type RNA.

Fig. 4.

Hydrophobic interactions of peptide side chains into the RNA groove stabilize formation of the base triple. (A) The base of nucleotide A35, which is normally flipped into solution, is oriented toward the d-Pro-l-Pro end of the peptide and interacts with Arg-11. The RNA is represented as a gray surface with nucleotides of interest in purple; the peptide is in orange. (B) Comparison of the analogous hydrophobic interaction observed in the HIV-1 TAR-L22 complex (Left) and the BIV TAR-BIV-2 complex (Right) between Ile-10 and the U10 and U24 bases. The RNA and peptide are represented as in Fig. 3; the Van der Waals surface of the Ile-10 side chain is represented as dots.

Other interactions with the apical loop appear to be of less importance, and indeed the remainder of the loop remains locally unstructured (Fig. 3B). The aliphatic portion of the Arg-8 side chain interacts with the major groove face of G34, positioning the guanidinium group to stack on top of the last base pair in the upper helix. The aliphatic portion of the Arg-9 side chain lies next to the riboses of G33 and G34, and the corresponding guanidinium group interacts with the base of G33 as well.

Hydrophobic Interactions Analogous to the BIV Tat-TAR Complex Stabilize a Base Triple and Dock the Peptide in the RNA Groove.

The present structure contains a base triple between U23 and the A27/U38 base pair that had been proposed for the complex between TAR and argininamide (9), but never directly observed with 2 hydrogen bonds in HIV-1 TAR. It is clearly established in this complex because the U23 NH resonance is protected from exchange and downfield-shifted upon formation of a hydrogen bond with A27. The NOE data indicate a canonical pairing for U23 and A27 in the base triple, and both A27 amino hydrogen atoms are protected from solvent exchange and observable in NOESY experiments.

The presence of the base triple enables a continuously stacked pseudohelical conformation between the upper and lower stems that progresses through the bulge, despite the extrusion of nucleotides C24 and U25 in the bulge away from the rest of the RNA. It is facilitated by a tight turn in the RNA backbone similar to what had been observed with argininamide and linear mimics of Tat protein (9, 10). Although U25 is flipped well out into solution, NOE interactions between C24 and the peptide residues Thr-4, Arg-5, and Lys-6 reveal that the base of C24 interacts with the Lys-6 side chain, creating a small pocket for this residue by covering its solvent-exposed surface opposite the major groove.

Two of the 3 hydrophobic residues, Ile-10 and Ile-12, are buried in the RNA major groove. Burial of the Ile-10 methyls against the U23–A27–U38 base triple closely imitates Ile-79 in the BIV Tat-TAR complex (12, 13) and Ile-10 in our BIV-2/BIV TAR structure (14), complexes upon which the L22 peptide was modeled. The methyl groups of Ile-10 are sequestered away in a pocket (Fig. 4B) that was also occupied by a benzoylic group in a small-molecule complex with TAR (19). Thus, this hydrophobic interaction likely drives base triple formation between U23 and the A27–U38 Watson–Crick base pair as do the analogous interactions of Ile-10 in the complex of the precursor peptide BIV-2 with BIV TAR (Fig. 4B) and Ile-79 from the BIV Tat protein. No comparable hydrophobic residue exists within the basic region of HIV Tat (residues 48–57), but a similar structural role could be played by residues from the hydrophobic stretch immediately to its N terminus (Tyr-47, Ile-45, Leu-43, or Phe-38). Burial of the Ile-10 and Ile-12 hydrophobic groups, along with the aforementioned protection of Val-2 by A35, leave very little of the hydrophobic surface area of the peptide exposed to solvent. Even the methyl group of Thr-4 is partially protected by hydrophobic interactions with Gly-7 in the β-turn at the tip of the peptide.

Polar Interactions Between Arg Side Chains and the Base Triple and Bulge Region of TAR.

Multiple polar interactions involving Arg side chains and the RNA bases and phosphodiester backbone are observed at or near the bulge region of TAR. The guanidinium group of Arg-1 contacts the RNA backbone between G21 and A22 and is within hydrogen-bonding distance of the A22 O2P atom (Fig. 5A). The guanidinium group of Arg-3 establishes an electrostatic interaction with the U23 phosphate and lies against the major groove edge of the A22 base. The aliphatic portion of the Arg-5 side chain lies next to the ribose of U23, whereas the Arg-5 guanidinium group stacks on top of the U23 base to provide a cation-π interaction (Fig. 5B). The interaction of Arg-5 with U23 is closely reminiscent of the upper half of a bis-arginine sandwich previously described in the complex of small molecule rbt203 with HIV-1 TAR RNA (19). The guanidinium group of Arg-5 is adjacent to the base of G28 and forms a hydrogen bond with the N7 of G28; a second hydrogen bond to the O6 of G28 (Fig. 5C) is less well constrained in the structure because of a lack of NOE data surrounding the O6. We do not observe any interaction with the G26 major groove face, as had been postulated (9, 20).

Fig. 5.

Polar interactions between Arg and Lys side chains and the RNA bases and backbone at or near the TAR bulge; the RNA and peptide are represented as in Fig. 3. (A) The guanidinium group of Arg-1 contacts the RNA backbone between G21 and A22 and is within hydrogen-bonding distance of the A22 phosphate (dotted line). (B) The Arg-5 guanidinium group stacks on top of the U23 base to provide a cation–π interaction that stabilizes the base triple (hydrogen bonds are shown by dotted lines), while the aliphatic portion of the Lys-6 side chain lies over Arg-5, packing its guanidinium group against U23. (C) Arg-5 is also adjacent to the base of G28, forming hydrogen bonds with the O6 and N7 of G28.

A Lys Residue in the Turn Provides Optimal Electrostatic Interactions with the Phosphate Backbone at the Interhelical Junction.

In the design of our library of inhibitors, both Lys-6–Gly-7 and Gly-6–Lys-7 were used in the turn, yet Lys-6–Gly-7 was considerably more potent than Gly-6–Lys-7. The structure reveals that the Lys-6–Gly-7 arrangement contributes favorably to the binding energy. The charged terminus of the Lys-6 side chain is positioned in a pocket of phosphate groups formed by the backbone atoms of C24, A27, and G28. The aliphatic portion of the Lys-6 side chain lies over Arg-5, packing the guanidinium group against U23, while being covered by the base of C24.

Discussion

The Tat-TAR complex has long been considered an attractive target for the discovery of novel antivirals because of its central role in promoting infection by up-regulating HIV transcription. However, numerous previous attempts to discover compounds with sufficient potency and selectivity to warrant pharmaceutical development have been unsuccessful (4, 6, 7, 19). Because finding small molecules that are specific and potent inhibitors of RNA–protein interactions is inherently difficult, we reasoned that constrained peptides would provide more effective scaffolds to inhibit these large interfaces. They would also provide more specific ligands, compared with peptide mimics studied in the past, because of their increased rigidity. Here, we report that proteolytically stable β-hairpin cyclic peptides bind specifically to HIV-1 TAR RNA with low nM activity, and we describe the structural basis for their activity and selectivity.

The structural rigidity imposed by the d-Pro-l-Pro template locks the pharmacophores onto a β-hairpin structure with side chains emerging from each face. The most potent ligands discriminate very strongly against tRNA (>10,000-fold), ≈10-fold between RNAs that are as closely related as the TAR elements from BIV and HIV (Table 1), and binding is abolished by substitution or deletion of a single critical nucleotide, U23. This specificity allows them to compete with Tat protein and displace it from preformed complexes with TAR (Fig. 2B), while the structural rigidity and high affinity allowed us to determine a structure with TAR of unprecedented quality.

The most surprising result was the observation that the peptide simultaneously organizes the bulge and loop residues to partially close around it, creating a much deeper binding pocket than seen in all previous structures of TAR (Fig. 3B). The simultaneous formation of contacts with multiple secondary structure elements resembles a previously observed complex between HIV-2 TAR and a cyclic aminoglycoside (21). Here, the RNA conformation buries ≈51% of the peptide surface, including most of its hydrophobic side chains. The formation of a deep binding pocket is reminiscent of what has recently been observed with small-molecule inhibitors of protein–protein interfaces (8), which achieve high inhibitory potency by inducing the formation of deeper pockets within the contact surface than seen at the natural protein–protein interface targeted for disruption. This structural feature in TAR is generated by the interaction between Arg-11 in the peptide and A35, a highly conserved nucleotide that is bulged out in all structures of the TAR loop (22); however, the remainder of the loop makes few contacts with the peptide and remains poorly structured (Fig. 3B) and conformationally flexible (18). This observation suggests that further improved potency may be obtained by optimizing the peptide sequence to interact with the apical loop that is the binding site for the essential cellular cofactor P-TEFb.

Binding of the peptide induces formation of a U23–A27–U38 base triple that was proposed (9) but never directly observed in HIV-1 TAR; the UAU base triple has a canonical structure without any disruption of the A27–U38 base pair (23). The base triple is conclusively identified here by the observation of a signal from the U23 NH. Although the hydrogen bond to the N7 was demonstrated for HIV-2 and HIV-1 TAR (24, 25) and was consistent with previous structures of TAR bound to Tat-derived peptides (10), the second hydrogen bond was controversial because it is inconsistent with chemical interference data suggesting that chemical modification of the A27 amino group does not lead to loss of ADP-1 binding to TAR (26). Examination of the chemical-shift data in various complexes reveals much larger shifts for L-22, relative to the free TAR, than in complexes with argininamide or ADP-1 (Table S1), suggesting that the conformational change has only been “completed” in the present complex. It is possible that the base triple observed here is not present in the complex of HIV-1 Tat with TAR; however, it is also possible that Tat protein, in the absence of P-TEFb, only forms a highly dynamic complex with TAR that can tolerate modifications of its interaction with TAR as substantial as the disruption of the second hydrogen bond stabilizing the base triple. Further stabilization was proposed to occur through hydrogen bonds between argininamide and G26 and with a so-called “arginine fork” involving the same guanidinium group and the phosphates of A22 and U23 (9, 20). We do not see any evidence for these interactions in our complex; the conformational change is instead driven by the burial of the hydrophobic side chain of Ile-10 in the major groove of the RNA. This interaction is closely reminiscent of what has been observed with the BIV Tat-TAR complex (12, 13) and recapitulated in our structure of BIV TAR with a related peptide mimic BIV-2 (14), and in the interaction of hydrophobic groups from small molecules with TAR (6, 19). In addition, Arg-5 forms hydrogen bonds with G28 above the base triple and a cation-π interaction with U23, reinforced by electrostatic interactions of Lys-6 with A27 (above the base triple) and Arg-3 below.

These cyclic peptides are potent inhibitors of viral replication with activity only 10-fold lower than the non-nucleoside reverse transcriptase inhibitor neviparine in primary T lymphocytes. They also display many of the properties of a promising drug lead: they are broad-spectrum HIV inhibitors and are active against a wide range of viral isolates from each of the HIV-1 clades without any cytotoxicity to at least 1 mM; their constrained cyclic structure makes them stable to proteolysis. Finally, mechanistic studies show that the peptides do not inhibit HIV interaction with its receptor or viral entry, off-target activity that stunted the development of several linear peptide and small-molecule inhibitors of the Tat-TAR interaction (11). Instead, the compounds inhibit early steps in reverse transcription and Tat-dependent HIV transcription. We are optimistic that cyclic peptides in this class will provide new antiviral drug leads active against a wide range of viral strains, including strains that are resistant to the current range of drugs. The structure presented here provides rational insight toward the optimization of their antiviral activity.

Methods

Peptide and RNA Synthesis.

The synthesis of the cyclic peptides was conducted as described (15, 16). The ADP-1 peptide (10) was prepared on MBHA-Rink amide resin by using Fmoc chemistry on an Applied Biosystems 433A peptide synthesizer. HIV-1 TAR RNA was prepared in vitro from commercially synthesized DNA templates (IDT) with in-house purified T7 RNA polymerase and either unlabeled nucleotides (Sigma) or 98% ¹⁵N, ¹³C-labeled nucleotides (Isotec) or 98% deuterated nucleotides (Isotec) and purified as described (27).

Binding Assays.

HIV-1 TAR RNA preparation, 3′-end labeling of TAR, and EMSA assays were done as described (15, 16). Inhibition assays were performed in the same way but by preforming the complex of HIV TAR with the ADP-1 peptide (150 nM) before peptide addition and fractionation by electrophoresis.

NMR Spectroscopy and Spectral Assignments.

NMR spectra were collected on Bruker 500- and 600-MHz spectrometers and a Varian Inova 800-MHz spectrometer equipped with cryo-cooled probes. Data were processed with NMRPipe (28) and analyzed in Sparky (29). NOESY experiments were recorded at multiple mixing times (100–300 ms) at either 25 °C (D₂O) or 4 °C (H₂O). The 2D ¹³C-filtered NOESY and total correlation spectroscopy (TOCSY) spectra of the “F1fF2f” type (30) were collected to facilitate assignments of peptide resonances; spectra using partially perdeuterated RNA (with only H6/8, H2, H1′, and H2′ protons present) had sharper line widths that facilitated NOE assignment. Resonance assignments started from those already available for free HIV-1 TAR RNA and TAR bound to argininamide and the ADP-1 peptide (10) and were extended to this complex by using HCCH-COSY, HCCH-TOCSY, and HSQC-NOESY experiments. Initial proton assignments for the peptide were made from F1fF2f NOESY spectra and based on the observation of the typical Hα(i) HN(i + 1) pattern of NOEs common in β-sheets, then completed with the partially perdeuterated RNA that removed much of the overlap between peptide and sugar resonances.

Structure Determination.

Structures of the HIV-1 TAR RNA/L-22 complex were calculated with Xplor-NIH (31) using the experimental restraints of Table 2. The imino proton of U23 was clearly hydrogen-bonded, because a sharp resonance was observed in the water NOESY spectra; however, the interaction between U23 and A27/U38 was not constrained by any hydrogen-bond restraint until structure calculations showed that U23 was poised to make 2 hydrogen bonds between U23 O4 and A27 HN″ and U23 H3 and A27 N7. Dihedral angle restraints were experimentally established with reported methods (27), and backbone dihedral angle restraints for the peptide were derived with TALOS (32). Structures were originally calculated without residual dipolar couplings (RDCs); then these restraints, obtained from a sample partially aligned using Pf1 phage (Asla), were applied with a harmonic potential well. The powder pattern-like distribution of RDCs was used to calculate starting values for the orientation parameters that were then optimized with a grid search (33). The final 10 converged structures of the complex were selected from the population of structures with no NOE restraint violations >0.5 Å and the lowest violations of all experimental restraints.