Retroviral RNase H: structure, mechanism, and inhibition

Abstract

All retroviruses encode the enzyme, reverse transcriptase (RT), which is involved in the conversion of the single-stranded viral RNA genome into double-stranded DNA. RT is a multifunctional enzyme and exhibits DNA polymerase and ribonuclease H (RNH) activities, both of which are essential to the reverse-transcription process. Despite the successful development of polymerase-targeting antiviral drugs over the last three decades, no bona fide inhibitor against the RNH activity of HIV-1 RT has progressed to clinical evaluation. In this review article, we describe the retroviral RNH function and inhibition, with primary consideration of the structural aspects of inhibition.

1. Introduction

A unique and key enzyme encoded by retroviruses is reverse transcriptase (RT). RT has DNA polymerase and ribonuclease H (RNH) activities, both of which are critical during the viral life cycle [1,2]. The polymerase and RNH activities of RT are well-conserved among retroviruses [3-6], but the most studied, especially with regard to drug development, is HIV-1 RT, considering the impact that this retrovirus has had on human health. In the late 1980s, the AIDS pandemic triggered massive efforts toward anti-HIV-1 drug development, targeting the key enzymes of the virus: protease (PR), RT, and integrase (IN) [7-9], as well as against other key steps in HIV-1 replication. Currently, several classes of HIV-1 antivirals are in clinical use [10-12]. Despite these developments, no FDA-approved inhibitor against the RNH activity of RT is available. In fact, among the HIV-1 encoded enzyme activities, i.e., protease, polymerase, RNH and integrase, the RNH is the only one for which inhibitors have failed to reach the clinic. As such, an important scientific question, that lends itself to understanding retroviral RNH generally, is why potent, selective HIV-1 RNH inhibitors with minimal toxicity have not been identified or developed to date.

In this review article, we describe the retroviral RNH domain function and inhibition, mainly from a structural biology point-of-view. In retroviruses, RNH has been defined by its structure, as a domain at the C-terminus of RT (Figure 1a) [13-15] or as a distinguishing activity of RT, different from that of the polymerase [6,16-18]. To avoid confusion, we consider RNH as a domain within the RT molecule since the isolated RNH domain is not functional, as described below. For additional RNH resources, we recommend these review articles: genetics of RNH [19-22]; RNH in the context of the retroviral integrase superfamily [19,23,24]; RNH activities [3,25-28]; and RNH inhibition [29-35].

Figure 1.

RNH domains in retroviruses and retrotransposon: (a) HIV-1 RT in the Pol region of GagPol, (b) mature RT, comprising p66 and p51 subunits, (c) M-MLV RT that functions as a monomer, (d) Ty3/Gypsy RT in retrotransposon. Domains are indicated with the following colors: PR, pink; polymerase, blue; connection (Conn.), yellow; RNH, orange; IN, dark green; RNH-fold, cyan. (See text for the references)

2. RNH function in retrovirus

2.1. Structure

The unique retroviral enzyme RT possesses two activities: DNA polymerase and endonuclease, particularly ribonuclease H. Endonucleases cleave the RNA strand of RNA/DNA hybrids and are widely found in both prokaryotes and eukaryotes, as well as in retroviruses [19,28,36,37]. RNH activity is also encoded within retrotransposons, transportable elements that are related to retroviruses, and found in eukaryotes [19-22]. There are three RNH lineages: rnhA, rnhB and rnhC (by gene symbols, or RNase HI, HII, HIII by enzyme definition). rnhA is found in many Eubacteria and all Eukarya, rnhB is found in Archaea, Eubacteria and Eukarya; rnhC appears only in some Eubacteria [19,20]. In retroviruses, rnhA RNH is encoded at the C-terminal end of RT [16,38].

Structurally, RNH belongs to the retroviral integrase superfamily [19,23,24], which has also been called the RNH-like superfamily [39]. Retroviral RT comprises several domains, including a connection domain that links the N-terminal polymerase domains to the RNH domain. Interestingly, the connection, RNH, and IN domains that are encoded in retroviral Pol (Figure 1a) share a similar RNH-like fold, a five stranded β–sheet with α–helices [19,23,24] (Figure 2a-c). These three RNH-like folds are encoded relatively close to each other in the genome (Figure 1a) and, among them, the connection domain lacks any catalytic site and has α–helices on only one side of the β–sheet (Figure 2a), compared to the RNH and IN domains (Figure 2b and 2c, respectively) [14]. The mature, functional form of HIV-1 RT is a heterodimer, comprising 66 kDa and 51 kDa subunits (Figure 1b), and the connection domain directly interacts with nucleic acids via positively charged side chains to help heterodimer formation and possibly position the substrate in the RNH active site [14,15]. In retroviruses with a functional monomeric RT, such as Moloney murine leukemia virus (M-MLV)(Figure 1b), it is still unknown how the connection and RNH domains position relative to one another in the RT-substrate complex, because only a substrate-free structure is available [40].

Figure 2.

Comparison of the RNH folds in HIV-1 Pol: (a) Connection domain (residues 329 to 426 in PDB 3LP2 [78]), (b) RNH domain (residues 427 to 557 in PDB 3LP2 [78]), and (c) IN central domain (residues 56 to 190 in PDB 1EX4 [123]). In each ribbon presentation, α-helices are shown by yellow. The active site residues D443, E478, D498 and D549 in (b) and D64 D116, E152 in (c) are shown with sticks.

Importantly, the isolated RNH domain from HIV-1 is not functional by itself, even though the active site conformation is similar to that of the domain in RT. The absence of activity in the isolated domain can be explained by a short substrate interaction site (dashed circle in Figure 3a). The concept that a longer substrate interaction site is needed for HIV-1 RNH domain activity was confirmed by restoration of RNH activity in a chimeric RNH that extends the substrate interaction site by approximately 20 amino acids with a sequence from E. Coli RNH [41].

Figure 3.

Metal coordination in HIV-1 RNH domain: (a) the entire RNH domain with 2 Mg²⁺ bound (inhibitor bound form), (b-d) close-ups of the RNH domain active site in: (b) 2 Mg²⁺ bound form (inhibitor bound form), (c) one Mg²⁺ bound form (with dsDNA), and (d) one Mn²⁺ bound form, with gapped RNA/DNA and NNRTI. In (a) and (b), structural presentations were generated using PDB 5J1E [84]. In (c) and (d), structural presentations were generated using PDB 3KK2 [124] and 4Q0B [125], respectively. The active site residues D443, E478, D498 and D549, are shown with sticks. In (a), a short substrate binding region, compared to that of E. Coli is marked with a dashed red circle (see text).

RNH domain-containing RT has been found widely in both retroviruses and long terminal repeat (LTR) retrotransposons [6]. In LTR retrotransposons, the RNH domain is located immediately after the polymerase domain without a connection domain between the two (Figure 1d). A crystal structure of the LTR retrotransposon RT of S. cerevisiae Ty3, which belongs to the Gypsy family of retrotransposons, demonstrates that the RNH domain locates at a position corresponding to that of the connection domain and interacts with RNA/DNA hybrid [42].

2.2. Function

For the catalytic activity of rnhA RNH proteins, site-directed mutagenesis, thermal stability and structural determination studies have indicated that an active-site DEDD motif and conserved active-site histidine are required [43-50]; for example, in E. Coli RNase HI, these are Asp 10, Glu 48, Asp 70 and Asp 134 (DEDD) and His 124. A homologous DEDD motif is also found in retroviruses [16,19,20,38]. In HIV-1 RNH (Figure 3a), the catalytic acidic residues are D443, E478, D498 and D549 with the essential H539 near the active site pocket (Figure 3b) [51]. Interestingly, the His residue is not conserved in the LTR retrotransposons [19].

In the retroviral life cycle, RT synthesizes double-stranded DNA from viral single-stranded RNA. The RNH activity of RT is responsible for hydrolysis of RNA of RNA/DNA hybrids during synthesis and, thus, is critical for virus replication [3,25-28]. Retroviral RTs, including HIV-1 RT and avian myeloblastosis virus (AMV) RT, which function as heterodimers, as well as M-MLV RT, which functions as a monomer, have three modes of RNA hydrolysis by RNH: 3'-end DNA directed cleavage, 5'-end RNA directed cleavage, and non-directed or internal cleavage [26-28,31,52-54]. These modes were named according to the interaction of RT with its substrate, i.e. RNA/DNA hybrid. During DNA polymerization from the RNA template, the 3'-end of the elongating DNA is positioned in the polymerase active site and orients the RNA strand in the RNH active site, which allows RNH to cleave the RNA at a position 15-20 nucleotides away from the 3’-DNA terminus [27,28,55,56]. This mode is polymerase-dependent and occurs during pauses in DNA synthesis, leaving nicks in the RNA as RT slides along it [27,28]. To complete RNA degradation, polymerase-independent modes are required. The 5'-end RNA-directed cleavage is determined by the position of the 5’-recessed RNA (from nicks) in the DNA polymerase active site and removes longer RNA fragments that remain base-paired to the minus DNA [26-28,52-54]. Non-directed or internal cleavages take place within larger segments of RNA/DNA duplex and might depend, in part, on the sequence of the RNA [28,52,55].

2.3. Metal interaction

Endonuclease reactions, such as that of RNH, require divalent metal ions, either Mg²⁺ or Mn²⁺, as co-factors [47]. In HIV-1 RNH, Mg²⁺ is the physiologically relevant metal ion for enzyme activity [57], while M-MLV RNH, which has 22 % sequence identity with HIV-1 RNH, exhibits higher catalytic activity in the presence of Mn²⁺ compared to Mg²⁺ [58]. Since metal coordination significantly impacts the design of active-site inhibitors of RNH, an understanding of the metal coordination is critical for drug development.

In HIV-1 RNH, two different models for metal coordination in the endonuclease reaction have been proposed, one that invokes binding by a single ion [47,59] (Figure 3c and 3d) and one that invokes binding by two metal ions [50,60-62] (Figure 3b). The one-metal ion model was proposed based on the dissociation constant and stoichiometry of metal binding to RNH active sites [47,59]. The two-metal ion model was proposed based on structure and mutagenesis studies of RNH [50,60-62]. In support of the latter, two Mg²⁺ ions were observed in RNH bound to an RNA/DNA hybrid [49]. Based on these structural and functional experiments, the two metal-binding mechanism has been widely accepted [27,28,30,31,63,64].

A noteworthy discrepancy in the studies exploring Mg²⁺ stoichiometry at the HIV-1 RNH active site arises upon comparison of thermodynamic studies that suggest one Mg²⁺ at the active site [59] and a subset of titration experiments, monitored by NMR, that supported two Mg²⁺ at the active site (K_D, ~ 3 mM and 35 mM) [65,66]. NMR titration is a well-established sensitive method to elucidate binding affinity of ligand and conformational changes upon ligand binding [67-72]. To understand the discrepancy between these thermodynamic and NMR experiments, titration experiments with Mg²⁺ at different salt conditions were conducted [73].

In the absence of NaCl, chemical shift changes at low MgCl₂ concentrations (0 ~ 10 mM, Figure 4a) were small, compared to those observed in the presence of 50 mM NaCl or in a constant Cl⁻ concentration condition (Figures 4b and 4c). In addition, the former titration curve did not fit a simple 1:1 stoichiometry model. A possible explanation is that, in the absence of NaCl, the titrated MgCl₂ is consumed non-specifically at the initial phase of the titration, as a salt (Figure 4d), while, in the presence of NaCl, Mg²⁺ is less non-specifically consumed (Figure 4e). When Cl⁻ concentration was kept constant (Figures 4c), the initial chemical shift changes were similar to those in the presence of 50 mM NaCl (Figure 4b), but severe signal broadening was not detected (Figures 4c), suggesting that Na⁺ increases the kinetic rate of specific Mg²⁺ interaction at the active site of RNH in the constant Cl⁻ concentration condition (Figure 4f). Thus, non-specific charge interactions can influence interpretation of Mg²⁺ titration curves by NMR. Once this non-specific effect is discarded, binding of only one Mg²⁺ to the RNH domain is observed, consistent with the calorimetry data [59]. These solution observations in the absence of substrate or inhibitor do not necessarily help to explain the catalytic mechanism of the RNH but do provide important information to consider for active-site inhibitor design (see discussion in section 3.3 below).

Figure 4.

RNH titration with Mg²⁺ monitored by NMR. (a-c) ¹H-¹⁵N HSQC spectra, i.e. signals of backbone amides, showing an example of the chemical shift changes of a single residue, T477, at different Mg²⁺ concentrations and (d-f) cartoons indicating possible roles of cations in each experiment. The NMR experiments were performed in (a) 20 mM bisTris buffer without any additional salt, (b) 20 mM bisTris buffer containing 50 mM NaCl, and (c) 20 mM bisTris buffer at constant Cl⁻ concentration (160 mM) [73]. In (a)-(c), dash lines indicate 0, 10 mM and 80 mM Mg²⁺ concentrations. Cartoons envision the following: (d) at low MgCl₂ concentration (< 20 mM), titrated MgCl₂ binds charged protein surfaces, as a salt, as well as the active site, resulting in the relatively slow migration of NMR peaks that is observed in (a), compared with (b) or (c); (e) in the presence of NaCl, as in (b), Na⁺ interacts with charged protein surfaces, leaving more Mg²⁺ available to bind the active site; (f) in the constant Cl⁻ condition of (c), the number of cations is larger than in (a) or (b), and Na⁺ ions can temporarily occupy the active site, even in the presence of Mg²⁺, which weakly binds the site (K_D > mM). An abundance of cations, not limited to Mg²⁺, may enhance the on/off exchange rate at the active site, resulting in sharp NMR signals, compared to (b).

3. RNH inhibition

3.1. Specific features in the HIV-1 RNH inhibitor assay

Ten RT polymerase inhibitors are currently approved for clinical use in anti-HIV therapy, and like the RT polymerase activity, RT RNH activity is critical for the HIV-1 life cycle, yet the RNH domain remains the only HIV enzyme not targeted by approved drugs. Many review articles have described efforts to develop RNH inhibitors [29-35]. Here, we briefly discuss specific points of HIV-1 RNH inhibitor screening and characterization. As mentioned above, because the isolated RNH is not enzymatically active, a chimeric RNH domain, p15-EC, that contains a short helix from the E. coli RNH has been used to screen for inhibitors and to characterize the bound structures [74-76], although less used these days, presumably because structures of RNH inhibitor-bound HIV-1 RNH or RT can confirm the inhibitor binding site(s) [77-80]. In addition, a high-throughput (HTP) system for detection of cleavage products of different lengths, using fluorescent-labeled RNA, was developed [81]. To identify whether positive hits of compound screens inhibit only the RNH activity or also inhibit polymerase or IN strand transfer (INST) activity, these two enzymes are also assessed for inhibition by the screened compounds [31]. The IN assay is important because both the RNH domain and the IN have similar RNH-like folds (Figure 2). Through these efforts, two broad categories of RNH inhibitors have been identified: active-site inhibitors and allosteric inhibitors. During the past several years, significant progress has been made in the development of active-site RNH inhibitors [32-35]. Here, we provide an update on RNH inhibitor development, from the structural biology point-of-view. For additional detail on RNH inhibitor development, we refer the reader to other review articles [29-35].

3.2. Active-site inhibitors

Several pharmacophores for RNH active site inhibitors are: diketo acid, N-hydroxyimide, hydroxytropolone, pyrimidinol carboxylic acid, N-hydroxy naphthyridinone, pyrido-pyrimidinone, nitrofuran-2-carboxylic acid and thiocarbamates [31,32]. Most of these are metal chelating fragments. Diketo acid, N-hydroxyimide and hydroxytropolone exhibit INST inhibition as well as RNH inhibition, as explained in a recent review article [34]. Derivatives of diketo acid have been synthesized and tested, with a goal to increase binding specificity or to enhance dual enzyme inhibition [82,83]. Among these derivatives, pyrrolyl diketo acids have shown good potency against IN and moderate inhibition of the RNH, with antiviral EC₅₀ ~μM, demonstrating that identification of dual IN-RNH inhibitors is an attractive strategy for new drug development [83].

The mechanisms of inhibition by small molecules have been structurally characterized. The crystal structure of RT with a hydroxytropolone inhibitor, β-thujaplicinol, showed the inhibitor in the active-site with chelation of two Mn²⁺ (Figure 5a) and prompted postulation on how RNA/DNA substrate interaction with the inhibitors may occur, providing insight into the mechanism of inhibition [77]. Crystal structures of RT with N-hydroxy naphthyridinone inhibitors also experimentally demonstrated direct RNH active-site interaction by the inhibitors and showed that the position of H539 sterically prohibits the naphthyridinones from flipping to an opposite orientation [78] (Figure 5b). Structures of p15-EC or RT with pyrimidinol carboxylic acid and N-hydroxy quinazolinedione compounds also showed binding to the active site [80]. In the latter paper, the authors discussed a lack of specific interactions between the phenyl moiety and the protein [80].

Figure 5.

Active-site inhibitor interaction with the RNH: (a) β-thujaplicinol and (b) naphthyridinone-based inhibitor. Active site residues D443, E478, D498 and D549 are shown by stick presentation. The active site Mn²⁺ ions are shown by pink spheres. Protein surfaces are shown for residues within 6 Å from the inhibitor. The structural presentation was generated with (a) PDB 3IG1 [77] and (b) PDB 3LP0 [78], using VMD software [126].

More recently, newer chemotypes for RNH active-site inhibitors have been developed [35], including hydroxypyridonecarboxylic acid (HPCA) [84], 3-hydroxypyrimidine-2,4-dione (HPD) [85,86], N-hydroxy thienopyrimidine-2,4-diones (HTPD) [87], and double-winged HPD [88]. HTPD exhibited an RNH IC₅₀ below μM, antiviral EC₅₀ in the μM range and CC₅₀ above 100 μM [87]. Double-winged inhibitors showed high selectivity against INST inhibition, with RNH IC₅₀ below μM and antiviral EC₅₀ in the μM range [88]. A molecular docking study of HPD showed that substrate interaction with the protein hampers a favored HPD interaction with H539 in the RNH domain [88]. Based on this observation, a double-winged inhibitor was designed to allow interactions with both H539 with or without substrate [88]. The designed inhibitors exhibited highly potent and selective inhibition against RNH and inhibited HIV-1 in cell culture [88].

A comprehensive study for RNH-inhibiting compounds, based on the INST inhibitor scaffold consisting of a 1,8-naphthyridine has been performed [89]. Specifically, the mechanisms of inhibition and specificities have been assessed along with a crystal structure of RT bound with a representative 1,8-naphthyridine ring compound, XZ462 [89]. Comparison of the XZ462-bound form of RT with the structures of RT bound to the N-hydroxy naphthyridinone inhibitors, MK1 and MK2 [78], elucidated structural differences in the contacts for the naphthyridine pharmacophore [89]. Importantly, XZ462, but not MK1 or MK2, has a hydrogen-bonding interaction with the conserved H539 residue [89].

Another newly reported pharmacophore is pyrrolyl pyrazole; these inhibitors, as non-diketo acid inhibitors, have inhibitory activities in the low μM – sub-μM range and have selectivity against INST inhibition [90]. Molecular docking studies indicate that the inhibitors interact at the RNH active site [90]. These newer scaffolds might help to further optimize interactions between inhibitor and the RNH domain.

3.3. Allosteric site inhibitors

As allosteric inhibitors, acylhydrazones, 1,2,4-triazoles, thiocarbamates and vinylogous urea have been described [31]. Acylhydrazones inhibitors were initially proposed to bind to the RNH active site but also have allosteric inhibition [91,92]. These acylhydrazones contain phenol and/or hydroxy-naphthyl or a biphenyl ring that is expected to recognize hydrophobic surfaces of the RNH or RT [92-94]. Further, a crystal structure study demonstrated acylhydrazone binding to a polymerase domain in HIV-1 RT but at a site distinct from the non-nucleoside reverse transcriptase inhibitor (NNRTI) site [95]. A recent study of acylhydrazones described chelating motifs at the active site of RNH [96]. These results may suggest that the RT interaction site is determined by the chemo groups attached to the acylhydrazone.

Derivatives of 1,2,4-triazole were initially identified as NNRTIs [97], while 1,2,4-triazolo[1,5-a]pyrimidines were recently proposed to work as allosteric inhibitors of the RNH [98]. A 2-amino-6-(trifluoromethyl) nicotinic acid scaffold was proposed to act as an allosteric dual-site compound, inhibiting both HIV-1 RT enzymatic functions [99]. Also, neo-clerodane diterpenes showed moderate RNH inhibition [100]. However, as far as we know, no RT crystal structure has shed light on the mechanism of allosteric inhibition of the RNH, except for those that bind at or near the NNRTI binding site. Indeed, NNRTIs were found to modulate RNH activity upon recognition of the NNRTI binding site in the polymerase domain [101-103]. Further, mutagenesis studies and hydrogen-exchange experiments indicate a long-range structural change at the RNH domain upon NNRTI interaction in the polymerase domain [102,104].

RT dimerization has also been proposed as a target for inhibition [105-113]. Since heterodimer formation is essential for the RNH activity, as well as the polymerase activity of RT, dimer dissociation is a reasonable approach to inhibit both activities in RT [105,114]. However, how this high-affinity (K_D of the heterodimer is 1 nM - 300 nM [105,115]) and large interface (~3000 Ų [116,117]) can be overcome by a small molecular inhibitor is not clear. Nevertheless, this approach is supported by the observation that a single substitution mutation in RT significantly lowers the dimerization affinity, suggesting that it is possible to reduce dimerization by targeting key contacts in the interface [107,118-121]. The proposed target sites for dimerization inhibitor binding have been summarized by Tachedjian and colleagues [122], and among these, recent studies have targeted two: the β7-β8 loop and a Trp-rich region [111-113]. Importantly, nanoparticle-based delivery of a peptide-type dimerization inhibitor to cells resulted in an antiviral submicromolar EC₅₀ [112]. Again, structural verification of this mechanism of inhibition is not well developed, compared to the active-site inhibitors. Further structural characterizations are anticipated.

4. Summary

In the last two decades, a large number of RNH and RNH-like molecules in retrovirus and retrotransposons have been discovered or proposed. However, only some have been structurally and/or functionally characterized. Given that retroviral RNHs are validated but challenging drug targets, such structure and functional studies are likely to aid drug development. Indeed, HIV-1 RNH function has been studied in detail as described above and the development of active-site HIV-1 RNH inhibitors have significantly improved in the last 10 years. In particular, interaction of the active-site inhibitors at the RNH site has now been structurally verified and used for improvement of the inhibitors.