Carbonyl J Acid Derivatives Block Protein Priming of Hepadnaviral P Protein and DNA-Dependent DNA Synthesis Activity of Hepadnaviral Nucleocapsids

Abstract

Current treatments for chronic hepatitis B are effective in only a fraction of patients. All approved directly antiviral agents are nucleos(t)ide analogs (NAs) that target the DNA polymerase activity of the hepatitis B virus (HBV) P protein; resistance and cross-resistance may limit their long-term applicability. P protein is an unusual reverse transcriptase that initiates reverse transcription by protein priming, by which a Tyr residue in the unique terminal protein domain acts as an acceptor of the first DNA nucleotide. Priming requires P protein binding to the ε stem-loop on the pregenomic RNA (pgRNA) template. This interaction also mediates pgRNA encapsidation and thus provides a particularly attractive target for intervention. Exploiting in vitro priming systems available for duck HBV (DHBV) but not HBV, we demonstrate that naphthylureas of the carbonyl J acid family, in particular KM-1, potently suppress protein priming by targeting P protein and interfering with the formation of P-DHBV ε initiation complexes. Quantitative evaluation revealed a significant increase in complex stability during maturation, yet even primed complexes remained sensitive to KM-1 concentrations below 10 μM. Furthermore, KM-1 inhibited the DNA-dependent DNA polymerase activity of both DHBV and HBV nucleocapsids, including from a lamivudine-resistant variant, directly demonstrating the sensitivity of human HBV to the compound. Activity against viral replication in cells was low, likely due to low intracellular availability. KM-1 is thus not yet a drug candidate, but its distinct mechanism of action suggests that it is a highly useful lead for developing improved, therapeutically applicable derivatives.

INTRODUCTION

More than 350 million people are chronically infected with hepatitis B virus (HBV) and at a greatly increased risk of developing severe liver disease, including cirrhosis and hepatocellular carcinoma (57). Current treatments are limited to immunomodulation by type I interferons and five FDA-approved nucleos(t)ide analogs (NAs) that act as reverse transcriptase (RT) inhibitors, namely, lamivudine (3TC), adefovir, entecavir, telbivudine, and tenofovir. Either type of therapy achieves sustained responses in, at most, 30 to 40% of patients (22). Interferons can cause severe adverse effects; the major drawbacks of NAs are the requirement for possibly life-long treatment, which may cause nephrotoxicity and myopathy (20), and the emergence of resistant virus variants (58). Though most pronounced with the early drugs, even the low resistance to entecavir in treatment-naive patients rises dramatically in 3TC-experienced patients (1.2% versus 57% after 6 years) (58), pointing to cross-resistance as an ever-more-important problem (31).

The relative success of highly active antiretroviral therapy against human immunodeficiency virus type 1 (HIV-1) infection (37) relies on a combination of drugs with different mechanisms of action, e.g., NAs plus nonnucleosidic RT inhibitors (NNRTIs) plus protease or integrase inhibitors. Similarly, the recent development of differently acting anti-hepatitis C virus drugs may soon render chronic hepatitis C a curable disease (39). In contrast to these two viruses, HBV, with its small 3-kb genome, offers its RT, P protein, as the only enzymatic target. Some drugs targeting nonenzymatic processes, especially nucleocapsid assembly, have been reported (26), including the heteroaryldihydropyrimidine BAY41-4109 (14, 17) and the phenylpropenamide AT-130 (18, 25). However, even P protein has functions beyond the RNA-dependent and DNA-dependent DNA polymerase and RNase H activities it shares with other RTs that provide attractive alternative drug targets. In particular, its interaction with an RNA stem-loop structure on the pgRNA, ε, is indispensable for pregenomic RNA (pgRNA) encapsidation and initiation of reverse transcription.

HBV and the related hepadnaviruses from, e.g., woodchucks (woodchuck hepatitis virus) or ducks (duck HBV [DHBV]), all replicate by protein-primed reverse transcription (8, 30). In an infected cell, the nuclear covalently closed circular HBV DNA serves as the template for cellular RNA polymerase II to generate several transcripts, including pgRNA, which acts both as mRNA for the viral core protein and P protein and as a precursor for new DNA genomes. P protein binding to the 5′-proximal ε RNA stem-loop triggers pgRNA encapsidation into newly forming nucleocapsids and initiates reverse transcription (priming). Inside the nucleocapsid, reverse transcription proceeds to yield the typical P protein-linked relaxed circular DNA (RC-DNA) genome found in infectious virions. Crucial for the peculiar protein priming mechanism is the unique terminal protein (TP) domain at the N terminus of hepadnaviral P proteins (Fig. 1A). It provides a specific Tyr residue (Y63 for HBV, Y96 for DHBV) that substitutes for the 3′ end of a conventional nucleic acid primer and serves as an acceptor for the first nucleotide of negative-strand DNA (8, 30). Beyond mere ε RNA binding, the formation of a priming-active initiation complex requires structural rearrangements in both P protein (50) and RNA (7). Only then can a specific nucleotide in a bulged region of the ε template incorporate the first plus two or three more nucleotides into a TP-linked oligonucleotide that, after translocation to a 3′-proximal acceptor site on the pgRNA, is extended into full-length negative-strand DNA. Concomitant degradation of the pgRNA by P protein's RNase H (RH) activity spares the 5′-terminal nucleotides which serve, upon a second template switch, as the primer for positive-strand DNA, and a third template switch mediates circularization into RC-DNA (8, 30).

Fig 1.

(A) Domain architecture of P proteins and suitability of recombinant miniDP-MBPH6 protein for in vitro protein priming. P proteins contain an RT (DNA polymerase) and an RH domain; boxes A to G are motifs conserved among all RTs. The unique TP domain provides a Tyr residue (Y96 in DHBV, Y63 in HBV) for protein priming. Numbers are amino acid positions for the DHBV P protein. In recombinant miniDP-MBPH6, regions dispensable for in vitro priming activity were deleted as indicated. (B) Recombinant miniDP-MBPH6 preparations used. Purified fusion protein from the soluble fraction of the bacterial lysates (native) or from inclusion bodies (refolded) was analyzed by SDS-PAGE and Coomassie blue staining. Sizes of the marker proteins in kDa are indicated on the left. Small arrowheads denote the major impurities in the two preparations. (C) Specific labeling by in vitro priming of the P protein part in miniDP-MBPH6. Incubation of MiniDP-MBPH6 with Dε RNA, followed by addition of Mn²⁺ and [α-³²P]dGTP, led to covalent labeling of the 87-kDa fusion protein, as detected by SDS-PAGE and autoradiography. Incubation with 1 μg (+) or 2 μg (++) TEV protease released a 44-kDa fragment, as expected for the miniDP protein part, at the expense of the full-length fusion protein. (D) MiniDP-MBPH6 has authentic RNA template requirements. MiniDP-MBPH6 was offered WT Dε RNA (1.2 μM final concentration), two concentrations (+, 1.2 μM; ++, 12 μM) of Dε variants with reduced template activity (DεV5, DεL4), or tRNA. Signals were quantified by phosphorimaging; the signal produced from WT Dε was set at 100%. rel., relative.

Hence, the unique and highly specific protein-priming reaction is evidently an attractive antiviral target. Some NAs, e.g., the dGTP analog lobucavir (42), appear to block priming. However, resistance mutations in P protein that protect the virus from the action of such NAs during chain elongation would likewise undermine their antipriming activities.

Most insights into the protein-priming mechanism have come from in vitro systems comprising DHBV P protein and its cognate P-DHBV ε (Dε) RNA; HBV P protein has thus far proven refractory to displaying enzymatic activity in vitro (19, 23). The first system (52) consisted of DHBV P protein translated in rabbit reticulocyte lysate (RRL); in the presence of Dε RNA and deoxynucleoside triphosphates (dNTPs), this leads to synthesis of the TP-linked oligonucleotide. The suspected role of RRL factors was confirmed by the successful reconstitution of Dε-dependent priming with bacterially expressed DHBV P protein fused to solubility-enhancing partners such as glutathione S-transferase (24), NusA, GrpE (6, 9), or maltose binding protein (MBP) plus purified chaperones (minimally, Hsc70/Hsp40 plus ATP) (46). Chaperone dependence is overcome (6, 54) by using truncated P proteins (miniPs) that lack parts of the TP domain, the spacer, the RH domain, and the C-terminal part of the RT domain (Fig. 1A). According to homology-based modeling for HBV (3, 16) and DHBV (5, 55), the latter part forms a thumb subdomain similar to that in HIV-1 RT. Structural similarity between the two hepadnaviral P proteins is further supported by their similar sensitivities against various NAs in the context of viral infection or transfection.

In vitro priming with DHBV P protein should thus provide a useful surrogate model to search for specific priming inhibitors (31). Any disturbance of the interaction between P protein and ε RNA may be inhibitory; hence, potential inhibitors could act on the P protein (and/or its interactions with chaperones) or on the RNA. Indeed, hemin (iron protoporphyrin IX), present in RRL, and some related porphyrins have recently been identified as in vitro priming inhibitors (28); they appear to target the TP domain and impair its function in ε RNA binding.

Here, on the basis of the likely structural similarity between the RT domains of P proteins and HIV-1 RT, we tested whether naphthylurea derivatives of the carbonyl J acid [7,7′-carbonylbis(azanediyl)bis(4-hydroxynaphthalene-2-sulfonic acid); CAS no. 134-47-4] family, known to inhibit primer/template binding by the HIV-1 enzyme (44, 53), would affect hepadnaviral protein priming. Indeed, we found that carbonyl J acid and in particular its derivative KM-1 potently blocked in vitro priming. According to all data, the compound competes with Dε RNA for overlapping binding sites on the P protein. Exploiting this property, we found that the stability of the P-Dε RNA complex increases during the different stages of protein priming. KM-1 also inhibited the DNA-dependent DNA polymerase activity of DHBV and, importantly, of HBV nucleocapsids. Equal inhibition of a highly NA-resistant HBV variant further confirmed the non-NA-like mechanism of KM-1. Anti-HBV activity of KM-1 in cells was low, likely because of poor intracellular availability. Hence, while clinical applications will require substantial pharmacological improvements, our data establish carbonyl J acids as valuable leads for the development of new anti-HBV antivirals that target multiple crucial steps in the HBV replication cycle.

MATERIALS AND METHODS

Compounds and reagents.

Carbonyl J acid, calcomine orange, hemin, 3TC, and monoclonal antibody (clone MBP-17) against MBP were purchased from Sigma-Aldrich. 3TC-TP was provided by Yung-Chi Cheng as a 5 mM stock solution, and KM-1 was provided as a dry substance by D. H. Eargle, Jr. Working solutions of hemin were prepared as previously reported (28), carbonyl J acid was dissolved in water, and calcomine orange and KM-1 were first dissolved in dimethyl sulfoxide and then diluted 500-fold in water.

Plasmid constructs.

The Escherichia coli expression vector pET28-miniDP-MBPH6 encodes a miniDP comprising amino acids (aa) 75 to 575 of full-length Dpol (from DHBV16; GenBank accession no. K01834) with the spacer (aa 221 to 348) replaced with a Ser-Pro-Gly linker (5), followed by a TEV protease cleavage site and C-terminally His₆-tagged MBP. Constructs for the expression of separate TP (pET28-MBP-TP) and RT domains (pET28-MBP-RT) encoded TP (aa 75 to 220) and RT (aa 349 to 575) fused to the C terminus of MBP; in variant pET-MBP-TP-Y96D, TP priming residue Y96 was replaced with D, and in variant pET-MBP-RT-YMHD, the active-site YMDD motif was changed to YMHD (36). For in vitro translation of full-length DHBV P protein in RRL, pT7AMVDpol16H6 was used (4). Eukaryotic expression vectors for the DHBV core (DHBc) protein carried genes for full-length DHBc (aa 1 to 262) or DHBc lacking the C-terminal domain (CTD) (aa 1 to 195) (33) in pAAV-MCS (Stratagene). For in vitro transcription of Dε RNA (positions 2557 to 2624), plasmid pDε1 (10) was used. pCD16 (34) and pCH-9/3091 (32) harbor about 1.1-fold genomes of DHBV16 and HBV2 (genotype D, subtype ayw; GenBank accession no. J02203) (21) under the control of the CMV immediate-early enhancer/promoter. In the 3TC-resistant variant pCH-9/3091-rtL180MM204V, the designated residues in the RT domain, numbered in the unified RT system (47), were mutated as indicated. pHBV-1.3×97 contains the 1.3-fold genome of HBV strain 97 (GenBank accession no. AF411411) under the control of the natural HBV core promoter (56). All constructs were verified by DNA sequencing.

Cell culture and transfection.

Chicken LMH and human Huh7 hepatoma cells were cultured in Iscove's modified Dulbecco's medium (Invitrogen) and Dulbecco's modified Eagle's medium (DMEM; PAA), respectively, supplemented with 10% fetal bovine serum, 100 U/ml penicillin G, and 100 μg/ml streptomycin, and 0.2 mM l-glutamine, at 37°C in a humidified atmosphere of 5% CO₂. Cells were transfected with TransIT-LT1 reagent (Mirus) as recommended by the manufacturer.

Protein expression and purification.

Bacterial expression and purification of miniDP, MBP-TP, and MBP-RT fusion proteins were performed in a manner analogous to that previously reported (5). In brief, lysates from induced cultures of appropriately transformed E. coli BL21(DE3) cells were cleared by centrifugation. Soluble proteins from the supernatants were purified by immobilized metal ion affinity chromatography (IMAC) with an Ni-nitrilotriacetic acid Superflow column (Qiagen), followed by size exclusion chromatography using a Superdex 75 HiLoad 16/60 column (GE Healthcare). Insoluble proteins from the pellets were dissolved in 7 M guanidinium hydrochloride containing buffer and refolded by rapid dilution into refolding buffer (5).

In vitro transcription.

Unlabeled Dε RNA was obtained by in vitro transcription from ClaI-linearized pDε1 plasmid using the Ampliscribe T7 high-yield transcription kit (Epicentre). ³²P-labeled Dε RNA was obtained analogously with Riboprobe system T7 (Promega).

Electrophoretic mobility shift assay (EMSA).

Two hundred nanograms of refolded miniDP-MBPH6 protein was mixed with 1 μl ³²P-labeled Dε RNA (8 × 10⁵ cpm/μl) in an 8-μl reaction volume containing 40 mM Tris-HCl (pH 8.0), 50 mM NaCl, 2 mM dithiothreitol (DTT), and 8 U of RNasin (Promega). Inhibitor and/or other additives were added as specified in Results. The binding mixture was incubated at 37°C for 1 h and then separated on a 5% native polyacrylamide gel at 15 V/cm in 0.5× Tris-borate-EDTA (TBE) buffer. Signals were recorded by phosphorimaging of the dried gels. Densitometry was performed using MultiGauge V2.2 software (Fujifilm).

In vitro protein priming.

Two hundred nanograms of recombinant miniDP was incubated with 0.6 μM Dε RNA (unless indicated otherwise) in a 10-μl volume containing 40 mM Tris-HCl (pH 8.0), 50 mM NaCl, and 8 U of RNasin at 37°C for 1 h to allow the formation of P-Dε complexes. Priming was initiated by the addition of MnCl₂ (2 mM final concentration) and 2 μCi [α-³²P]dGTP (3,000 Ci/mmol) and allowed to proceed for 1 h at 37°C. Reactions were stopped by adding an equal volume of 2× SDS loading buffer (100 mM Tris-HCl [pH 6.8], 4% [wt/vol] SDS, 20% [vol/vol] glycerol, 2% [wt/vol] β-mercaptoethanol, 0.2% [wt/vol] bromophenol blue), and the products were separated by SDS-PAGE on 10% polyacrylamide gels. Products were visualized by autoradiography or phosphorimaging of the dried gels. For protein priming in RRL, the respective proteins were in vitro translated with a coupled transcription-translation system (TNT Quick system T7; Promega). Ten microliters of translated products was mixed with 0.6 μM Dε RNA and 10 μCi of [³²P]dGTP (3,000 Ci/mmol; Perkin-Elmer) in a 20-μl volume containing 10 mM Tris-HCl (pH 7.5), 2 mM MnCl₂, and 6 mM MgCl₂. The mixture was incubated at 30°C for 1 h and then analyzed as mentioned above.

Native agarose gel electrophoresis of capsids and Western blotting.

Electrophoresis was conducted with 1% agarose gels (12, 56). After capillary transfer to a nitrocellulose membrane using TNE buffer (10 mM Tris-HCl [pH 7.5], 50 mM NaCl, 1 mM EDTA), DHBV capsid proteins were detected using monoclonal antibody (MAb) 2B9-4F8 (51) and HBV capsid proteins were detected using MAb 35/312 (38), followed by peroxidase-conjugated anti-mouse secondary antibody and ECL chemiluminescence reagent (GE Healthcare).

EPR.

Nucleocapsids from cytoplasmic lysates of appropriately transfected cells were pelleted by centrifugation through a 20% sucrose cushion (TLS55 rotor, 55,000 rpm, 90 min) and resuspended in capsid buffer (20 mM Tris-HCl [pH 7.5], 50 mM NaCl, 1 mM EDTA, 0.01% Triton X-100, 0.1% NP-40, 0.05% β-mercaptoethanol). For the endogenous polymerase reaction (EPR), the nucleocapsids were incubated in EPR buffer (50 mM Tris-HCl [pH 7.5]; 75 mM NH₄Cl; 1 mM EDTA; 10 mM EGTA; 20 mM MgCl₂; 0.1% [vol/vol] β-mercaptoethanol; 0.5% [vol/vol] NP-40; 0.4 mM [each] dATP, dGTP, and dTTP; 10 μCi [α-³²P]dCTP [3,000 Ci/mmol]) and then 30 min later in 10 μM unlabeled dCTP at 37°C overnight. Viral DNA was released by the addition of 20 mM EDTA, 0.5% SDS, and 0.5 μg/μl proteinase K and incubation at 50°C for 2 h. After extraction by phenol-chloroform and ethanol precipitation, DNAs were separated on 1% agarose gels in 0.5× TBE buffer. Products were visualized by phosphorimaging of the dried gels.

Limited V8 proteolysis.

Three hundred nanograms of natively purified MBP-miniDPH6 protein was incubated with 6 μM KM-1 or 1 μM Dε RNA for 45 min at 37°C in a 10-μl volume containing 40 mM Tris-HCl (pH 8.0), 50 mM NaCl, and 2 mM DTT and then incubated for 15 min at room temperature with 0.1 μg V8 protease (endoproteinase Glu-C; New England BioLabs). Products were separated by Tricine-SDS-PAGE (40) using 12% acrylamide gels. P-protein-containing fragments were detected using anti-DHBV P protein MAbs 5, 6, and 25 (15).

Calculation of 50% inhibitory concentrations (IC₅₀s) and statistical analysis.

For IC₅₀ determinations, priming or P-Dε complex (EMSA) signal intensities (from at least three independent experiments each) obtained after various treatments relative to the untreated control (set at 100%) were plotted against log [inhibitor]. IC₅₀s were calculated from the dose-response curves by sigmoidal curve fitting as implemented in the Origin 8.0 software (OriginLab). One-way analysis of variance (ANOVA) for statistical analyses was also carried out by using Origin 8.0.

KM-1 cytotoxicity assay.

Huh-7 cells were seeded at a density of 1 × 10⁴/well into a 96-well plate and treated with increasing concentrations of KM-1. Four days later, cells were washed twice with PBS and refed with 100 μl fresh DMEM per well. Cell viability was measured using the CellTiter 96 aqueous nonradioactive cell proliferation assay (Promega). The absorbance at 490 nm of the group without KM-1 was set as 100%. The 50% toxic concentration (TC₅₀) was derived analogously to the IC₅₀.

RESULTS

Recombinant miniDP-MBPH6 protein displays authentic Dε-dependent protein-priming activity.

To take advantage of the simplicity of the chaperone-independent miniP, we used a truncated DHBV P protein, miniDP-MBPH6, that harbors all of the parts of the TP (aa 75 to 210) and RT (aa 349 to 575) domains required for in vitro priming activity (5); a C-terminally fused MBP, connected through a TEV protease cleavage site-containing linker, served to enhance solubility (Fig. 1A). The protein was well expressed by E. coli, with substantial amounts present both in the soluble form and in inclusion bodies. Purification of the soluble fraction under native conditions via IMAC yielded a product with the expected molecular mass of 87.8 kDa (Fig. 1B, left side) plus two minor degradation products. Size exclusion chromatography suggested that the protein forms soluble oligomers (data not shown). Purification from the inclusion bodies under denaturing conditions, followed by refolding as recently described (5), yielded an even purer preparation (Fig. 1B, right side). To corroborate specific priming activity, the natively purified protein (Fig. 1C) and the refolded protein (not shown) were incubated with Dε RNA and [α-³²P]dGTP, which constitutes the first nucleotide of the authentic DNA primer. Generation of a strong ³²P-labeled band at the expected 87-kDa position indicated successful priming by the fusion protein. Treatment with TEV protease reduced the intensity of the 87-kDa band and generated a new ∼44-kDa product, as expected for clipped-off miniDP. Hence, miniDP-MBPH6 was specifically labeled in the miniDP part. To confirm template specificity (Fig. 1D), we replaced Dε RNA, at the same concentration (lanes labeled +) or a 10-fold higher concentration (lanes labeled ++), with tRNA or two previously described Dε variants with reduced template activity in the RRL system, V5 and L4 (10); L4 carries just a single U → C mutation in the apical loop of Dε. tRNA did not generate any signal, whereas both Dε variants produced much weaker signals (5% and 20%) than wild-type (WT) Dε. Thus, miniDP-MBPH6 had the same RNA template requirements for protein priming as the full-length DHBV P protein in RRL. Identical results were obtained with the refolded protein, making both preparations suitable for inhibitor testing.

Carbonyl J acid derivatives inhibit in vitro protein priming by DHBV P protein.

Carbonyl J acid is a nonnucleosidic HIV-1 RT inhibitor that does not bind in the common NNRTI pocket (44). Azo derivatives similar to calcomine orange 2RS had shown enhanced inhibitory potency, in particular KM-1 (Fig. 2A). To test whether these compounds could interfere with hepadnaviral protein priming, we incubated miniDP-MBPH6 with Dε RNA (0.6 μM) in the simultaneous presence of increasing concentrations of the compounds, followed by the addition of Mn²⁺ and [α-³²P]dGTP to initiate priming. Hemin (28) was included as a control. All of the compounds used reduced the priming signals in a dose-dependent fashion (Fig. 2B). Around 20 μM carbonyl J acid and around 6 μM hemin were required for 50% inhibition, whereas both azo derivatives were about 10- to 20-fold more potent. Quantitative determination (n = ≥3) of KM-1-mediated inhibition yielded an IC₅₀ of 1.13 ± 0.49 μM for natively purified miniDP-MBPH6 and 1.40 ± 0.26 μM for refolded miniDP-MBPH6 (Fig. 1D). Compared to HIV-1 RT DNA polymerase inhibition, carbonyl J acid and KM-1 were 4- to 10-fold less effective against hepadnaviral protein priming (IC₅₀s of 20 μM versus 5 μM and 1.1 μM versus 0.09 μM, respectively) but the trend toward stronger inhibition by the bulkier compounds was preserved. The reported protein-priming IC₅₀ for hemin of 1.1 ± 0.3 μM (28) is somewhat lower than in our experiments, likely because of the nonidentical reaction conditions. As all of our experiments were performed in parallel, we conclude that, at least under our experimental conditions, KM-1 is a five to six times more potent priming inhibitor than hemin.

Fig 2.

Inhibition of miniDP protein priming by carbonyl J acid derivatives. (A) Chemical structures. (B) Dose-dependent priming inhibition by carbonyl J acid derivatives and hemin. MiniDP-MBPH6 protein (200 ng) was simultaneously incubated with 0.6 μM Dε RNA and the indicated final concentrations of the compounds. Priming was then initiated by the addition of Mn²⁺ and [α-³²P]dGTP, and the products were analyzed by SDS-PAGE and autoradiography. Signal intensities were determined by phosphorimaging and are given relative (rel.) to that of the signal without inhibitor (100%). (C) The order of KM-1 versus Dε addition affects sensitivity to priming inhibition. MiniDP-MBPH6 was incubated with KM-1 before Dε RNA (top) or after Dε RNA (bottom); priming was then initiated. (D) Dose-response curves for KM-1-mediated priming inhibition in different experimental settings. For IC₅₀ determination, the mean relative priming efficiencies from three independent experiments were each plotted against the compound concentration (in log[KM-1]), followed by sigmoidal curve fitting as implemented in the Origin 8.0 software. All other IC₅₀s in this study were obtained analogously. Note the shift in IC₅₀ when KM-1 was added before, together with, and after Dε RNA. nat., natively purified; refold., refolded.

The order of KM-1 versus Dε RNA addition significantly affected the apparent IC₅₀. Compared to coincubation, adding KM-1 before the RNA (Fig. 2C and D) reduced the IC₅₀ to 0.65 ± 0.05 μM, whereas adding the compound after the RNA increased the IC₅₀ by 4- to 5-fold to 3.65 ± 0.86 μM; this increase was significant (see Fig. 4C). A possible explanation was that KM-1 and Dε RNA compete for P protein binding and that priming conditions stabilize the P-Dε complex against the drug, as confirmed below.

Fig 4.

KM-1 inhibits protein priming by full-length DHBV P protein. (A) In vitro priming in RRL. Full-length DHBV P protein (top) and miniDP-MBPH6 (bottom) were translated in vitro in RRL and subsequently incubated with Dε RNA, [α-³²P]dGTP, and the indicated concentrations of KM-1. Priming products were analyzed as described in the legend to Fig. 2. rel., relative. (B) Chaperone-dependent in vitro priming with recombinant nearly full-length DHBV P protein. Grp-DP, a GrpE fusion protein lacking just the spacer and the C-terminal 25 aa of DHBV P protein, was activated by using Hsc70/Hsp40 and ATP as described previously (46). Prior to priming, KM-1 at the indicated final concentrations was added either together with (top) or after (bottom) Dε RNA. Detection of the priming products and IC₅₀ determinations were conducted as described in the legend to Fig. 2. (C) Addition of Dε RNA prior to KM-1 significantly increased the IC₅₀ for chaperone-independent (miniDP) and chaperone-dependent (GrpDP) in vitro priming. The graphs compare the mean IC₅₀s (n = ≥3; error bars represent standard deviations) obtained by varying the order of addition. I, KM-1 before Dε RNA; II, KM-1 with Dε RNA; III, KM-1 after Dε RNA. Data for miniDP were obtained from the inhibition curves shown in Fig. 2D, and data for GrpDP are from analogous curves derived from the experiment shown in Fig. 4B plus further repeats. Significance was assessed by one-way ANOVA as implemented in the Origin 8.0 software, and Tukey's posttest was used for pairwise comparisons. P values of ≤0.05 were considered significant.

MiniP proteins can support Dε RNA-dependent in vitro priming at alternative (“cryptic”) residues to Y96, including at Y561 in the RT domain (5, 13). Exploiting the ability of separate TP and RT domains to trans complement each other, we generated MBP fusions carrying either the WT TP domain or a mutant form with a defective natural acceptor site (TP-Y96D), or a WT RT domain or an enzymatically disabled mutant form (YMDD → YMHA). Both authentic and cryptic priming depended on an active RT domain, as reported, and KM-1 inhibited both reactions, with IC₅₀s between 0.79 and 1.69 μM (Fig. 3). Hence, inhibition of Dε-dependent priming was not restricted to a specific acceptor site.

Fig 3.

KM-1 inhibits authentic priming at Y96 and also cryptic priming in the RT and TP domains. (A) Recombinant TP and RT fusion proteins used. TP and TP-Y96D, lacking the authentic protein-priming residue, and RT and RT-YMHD, a polymerase active-site mutant, were expressed in E. coli as His-tagged MBP fusions, purified from inclusion bodies, refolded, and analyzed by SDS-PAGE and Coomassie blue staining. The sizes of marker proteins in kDa are indicated on the left. (B) trans complementation analysis. The indicated proteins were subjected to priming in the presence of Dε RNA and [α-³²P]dGTP, and products were analyzed by SDS-PAGE and subsequent autoradiography. Generation of priming signals required TP and active RT (no signal with either domain alone or upon trans complementation of TP with RT-YMHD). Labeling was strongest in the WT TP domain (arrowhead marked MBP-TP) but consistent with the cryptic priming sites recently reported (5, 13) also occurred in the RT domain (MBP-RT) and at alternative sites in the TP domain when the authentic Y96 acceptor was mutated. (C) Similarly efficient inhibition of authentic and cryptic priming by KM-1. Equimolar mixtures of MBP-TP and MBP-RT or of MBP-TP-Y96D and MBP-RT were simultaneously incubated with Dε RNA and KM-1 at the indicated concentration, and then priming (prim.) was initiated by the addition of Mn²⁺ and [α-³²P]dGTP. Products were visualized as in panel B and quantified by phosphorimaging. Mean IC₅₀s ± standard deviations are derived from three experiments.

Because miniDP lacks substantial parts of the P protein, we next investigated whether KM-1 could inhibit the priming of the full-length DHBV P protein translated in RRL. As shown in Fig. 4A (top), priming was indeed inhibited, though with an about 10-fold higher IC₅₀ (∼14 μM). One explanation was that potential KM-1 binding sites were shielded in the full-length protein. However, a comparably increased IC₅₀ was seen with RRL-translated miniDP-MBPH6 (Fig. 4A, bottom). A reduced inhibitory activity in RRL has also been reported for hemin (28). Possibly, the high concentration of proteins and other factors in RRL reduced the effective concentration of the compounds. Conversely, KM-1 efficiently blocked in vitro priming by a nearly full-length P protein (lacking only the spacer and the C-terminal 25 aa of the RH domain) fused to GrpE, termed GrpDP, which was activated for Dε RNA binding by purified Hsc70, Hsp40, and ATP (6, 45). KM-1 present during the initial RNA binding step showed an IC₅₀ of 0.46 ± 0.06 μM (Fig. 4B, top), which increased about 7-fold (to 3.44 ± 1.91 μM) when KM-1 was added after Dε RNA (Fig. 4B, bottom). As for the chaperone-independent miniDP protein, this difference was significant. Hence, both data sets are in line with an enhanced stability of the P-Dε RNA complex under priming conditions and/or upon successful priming.

KM-1 targets P protein, not Dε RNA.

Priming inhibition by KM-1 and its relatives could be due to interactions with P protein, with Dε RNA, or with another factor in the reaction that is required for priming. In the miniP system, the only such factors are bivalent metal ions, which are crucial for P activity (5, 29). However, these were present at millimolar concentrations whereas the IC₅₀ for KM-1 inhibition was around 1 μM, ruling out metal ion sequestration. To distinguish between P protein and Dε RNA binding, we performed competition experiments in which the priming reaction mixtures were supplemented with additional RNAs or proteins that might trap the compound and reduce its effective concentration.

In the first set of experiments, we used KM-1 at a 10 μM concentration, which had inhibited in vitro priming by >95% (Fig. 2B). Dε RNA was used at the standard concentration of 0.6 μM. As shown in Fig. 5A, adding tRNA in 5.5-fold or 28-fold excess over Dε RNA did not relieve inhibition, excluding nonspecific RNA binding by KM-1. Adding a 20-fold excess of Dε RNA variants V5 and L4 even enhanced the inhibition by KM-1 (not shown), most likely by competition with the WT RNA. However, increasing the concentration of WT Dε RNA to 16.7 μM, thus exceeding the 10 μM KM-1 concentration (Fig. 5A, middle), led to a modest but well-detectable signal increase. In the absence of a compound, higher concentrations of Dε RNA did not enhance the priming signal over that at 0.6 μM RNA, in accord with previous data (6, 46). This excluded the possibility that KM-1 sequesters the Dε RNA into a stable equimolar complex and suggested that the drug instead competes with the RNA for P protein binding.

Fig 5.

KM-1 targets P protein, not Dε RNA. (A) Partial rescue from KM-1-mediated priming inhibition by excess Dε RNA and high concentrations of BSA but not tRNA. Two hundred nanograms of miniDP-MBPH6 was coincubated in a total volume of 10 μl with 0.6 μM Dε RNA and the indicated concentrations or amounts of KM-1 and tRNA (top), Dε RNA (middle), or BSA (bottom) and then subjected to priming. Product analyses were performed as described in the legend to Fig. 2. In accord with previous data (6, 46), excess Dε RNA did not increase the priming signals in the absence of inhibitor but increased the priming signal at 10 μM KM-1 whereas tRNA did not. The slight increase in the priming signal caused by 5 μg BSA likely reflects the general affinity of BSA for hydrophobic compounds. rel., relative. (B) Specific rescue from KM-1-mediated priming inhibition by separate TP and RT domains. MiniDP-MBPH6 (200 ng) was coincubated with Dε RNA and KM-1 at the indicated concentrations plus 500 ng of MBP, BSA, recombinant MBP-RT (RT), or MBP-TP (TP); MBP-TP was also used at 2.5 μg (5x TP). TP and RT but not MBP or BSA reduced sensitivity to KM-1. Data collection and IC₅₀ determinations were performed as described in the legend to Fig. 2.

To test for KM-1 acting via P protein binding, we first added 2 μg and 5 μg bovine serum albumin (BSA; 10- and 25-fold mass excesses, respectively, over miniDP-MBPH6; total BSA concentrations, 30 and 77 μM) to priming reaction mixtures in the presence of 10 μM KM-1 (Fig. 5A, bottom) and found a slight signal increase at the high BSA concentration. Given the ability of BSA to bind hydrophobic compounds, this would be consistent with the trapping of some KM-1 and hence a somewhat reduced effective concentration. To look for specific P protein binding, we supplemented priming reaction mixtures with 0.5 μg (2.5-fold mass excess over miniDP-MBPH6) of BSA or MBP or of the separately expressed MBP TP and RT domains (Fig. 5B). Their activity in trans complementation (Fig. 3) indicated that they were not grossly misfolded. In contrast to BSA and MBP, both MBP-TP and MBP-RT reduced KM-1-mediated priming inhibition (Fig. 5B), and this reduction became more pronounced when the amount of MBP-TP was doubled (Fig. 5B, 5x TP), with a corresponding increase in the apparent IC₅₀. Together, these data suggested specific binding of KM-1 to P protein and the existence of binding sites (or a composite site) in both the TP and RT domains.

Limited proteolysis confirms specific binding of KM-1 to P protein and suggests overlapping Dε RNA binding sites.

To validate this interpretation, we used limited proteolysis to detect potential differences in the protease accessibility of P protein in the presence versus the absence of KM-1 and, for comparison, also of Dε RNA. This approach had previously been used by us (45) and others (29, 49, 50) to reveal structural alterations imposed on the P protein during chaperone activation and priming. DHBV P protein contains numerous E residues that, if accessible, are substrates for V8 (Glu-C) protease. Specific fragments can be assigned by their sizes and differential immunodetection by anti-DHBV P protein MAbs that recognize linear epitopes within the TP domain or an epitope in the RT domain (15). Here we used MAbs 5 and 6 (which recognize aa 143 to 147 and aa 191 to 197 in the TP domain) and MAb 25 (which recognizes aa 561 to 565 in the RT domain). To simplify analysis, we employed the construct MBP-TEV-miniDP575 (5), in which a P protein part identical to that in miniDP-MBP-H6 is N terminally fused to MBP (Fig. 6A). Hence, the MAb 25 epitope is at the very end of the protein and the C termini of all MAb 25-reactive products are defined in relation to this physical end or one of the E residues downstream of aa 565 (E571, E572, and an artificial E residue preceding the His₆ tag). The maximal uncertainty in the molecular mass at the C-terminal border of a fragment is thus only 1.7 kDa (V8 cleavage at E571 versus an unprocessed C terminus).

Fig 6.

KM-1 reduces V8 (GluC) protease accessibility in the TP and RT domains similarly to Dε RNA. (A) Schematic view of the MBP-miniDP protein used. MBP-miniDPH6 (5) contains the same miniDP part as miniDP-MBPH6 but fused to the C terminus of MBP; this places the MAb 25 epitope (aa 561 to 565) close to the physical end of the protein (shown in the sequence above the scheme). Upward-pointing filled arrowheads denote all of the E residues in miniDP; downward-pointing arrowheads denote the closest upstream non-miniDP E residues. Open arrowheads indicate non-V8-dependent spontaneous processing sites likely targeted by E. coli proteases. (B) Potential and observed cleavage products detectable by MAb 25. The top line represents full-length (FL) MBP-miniDP; the epitopes of MAb 5 (m5), MAb 6 (m6) and MAb 25 (m25) are denoted by ovals; arrowheads are defined as in panel A; the values above each line refer to the calculated masses in kDa. T3 (aa 176 to 183) and RT1 (aa 385 to 415), proposed composite binding site for Dε RNA; sp, spontaneous non-V-8-dependent fragments; a to e, V8-dependent fragments detected by MAb 25 and differentially by MAb 5 and MAb 6; nd1 and nd2, V8-dependent fragments of sufficient size to be resolved by SDS-PAGE but not detected by MAb 25. On the left, the specific E residues are indicated where V8 cleavage would create fragments of sizes compatible with those shown in panel C. (C) Partial V8 proteolysis. MBP-miniDP was incubated in the absence or in the presence of 6 μM KM-1 or 1 μM Dε RNA and analyzed before (lanes without V8) or after V8 treatment by SDS-PAGE, followed by immunoblotting with the indicated MAbs. Note that the doublet appearance of the top bands (labeled FL) in the MAb 5 and MAb 6 blots reflects consumption of the luminescent substrate in the center of the band. The band labeled sp was present before V8 treatment and was recognized by all three MAbs; for MAb 25, it must result from cleavage within the N-terminal part of TP or further upstream. All MAb 25-reactive products larger than 44 kDa must derive from cleavage within MBP. Note that all of these products had similar intensities regardless of the presence or absence of KM-1 or Dε RNA, whereas the faster-migrating products (a to d) where suppressed by KM-1 and less strongly by Dε RNA. Ø, neither Dε nor KM-1 added.

Next, we subjected MBP-TEV-miniDP575 alone or in the presence of 6 μM KM-1 or 1 μM Dε RNA to V8 proteolysis; partial digestion was confirmed by the preservation of a substantial fraction of the full-length protein (Fig. 6C). To control for the presence of non-V8-dependent smaller products (from inadvertent cleavage in E. coli or during workup), one aliquot from each reaction mixture was set aside before V8 addition. Equal aliquots of the protease-treated reaction mixtures plus the nontreated control were then separated by Tricine-SDS-PAGE (40) and subjected to immunodetection with MAbs 25, 5, and 6. All nontreated samples showed the full-length protein plus a few additional, faster-migrating bands, most prominently one with an apparent molecular mass of ∼40 kDa (termed sp and marked by open arrowheads in Fig. 6C). A band at this position was detected by all three MAbs, indicating that it contained most of the P protein part and derived from inadvertent cleavage upstream of the MAb 5 epitope (aa 143 to 147). Cleavage by a bacterial protease within the TEV protease site between MBP and P protein would generate a product of 44.6 kDa. The slightly faster mobility of sp suggests cleavage within the N-terminal TP part. Candidate sites for a trypsin-like protease are R78 and K99, which would result in products of 42.6 and 40.3 kDa, respectively, either one of which is in line with the mobility observed. The sp band and the other, fainter, degradation products in the pretreatment samples provided a convenient reference to compare the three different blots.

A first telling result was that for all of the samples on one blot, the V8-induced band pattern above the 40-kDa sp band was nearly identical in terms of the number, position, and intensity of the bands. This excluded the possibility that KM-1 or Dε RNA inhibited V8 protease. In addition, the two most prominent large V8-dependent products detected by MAb 25 (collectively labeled e in Fig. 6C) ran just below and at the position of the 50-kDa marker; they must therefore result from cleavage within MBP, most likely a triple E cluster about 60 aa upstream of the MBP domain end (calculated mass, 53 kDa without C-terminal processing), and at an E residue in the TEV protease site linking MBP and TP (calculated mass, 45 kDa without C-terminal processing). Either product would still contain the MAb 5 and MAb 6 epitopes, and corresponding signals were indeed visible with these two MAbs. Cleavage within MBP must also apply to the MAb 25-reactive V8 product at about the 65-kDa position. The unaltered intensities of these large products demonstrated that neither KM-1 nor Dε RNA affected V8 accessibility in the MBP part of the fusion protein.

In strong contrast, the intensity of the smaller V8 products running ahead of the 40-kDa sp band was strongly reduced by KM-1 and also, though less pronouncedly, by Dε RNA. This is particularly evident from the MAb 25 blot, where four distinct V8 products with apparent masses of 20, 24, 30, and 37 kDa (labeled a to d in Fig. 6C) were detected. Knowing that all of these fragments must end after E571, or at most after the C-terminal His tag, allowed a rather accurate definition of the upstream cleavage sites (Fig. 6B), i.e., E401/E402 for fragment a, E369/E370/E371 for fragment b, E176 (less likely E164) for fragment c, and E124 (less likely E106) for fragment d. Consistently, MAb 6 detected only bands with a mobility corresponding to that of fragments c and d, and MAb 5 recognized only a band of fragment d-like mobility. No products arising from N-terminal cleavage at E499 (calculated mass, 8.8 or 10.3 kDa if C terminally unprocessed) or E199 (calculated mass, 28.2 or 29.8 kDa) were seen. While fragment c had approximately the proper size for E199 cleavage, a band of the same size was detected by MAb 6, whose epitope would be removed. Together, these data confirmed that KM-1 interacts specifically with P protein and that this interaction causes a reduction in V8 protease accessibility in both the TP and RT domains similar to that caused by the interaction with Dε RNA. Notably, two of the KM-1 shielded sites (E176, and E401/E402) reside within a proposed Dε RNA binding site composed of the T3 motif in TP (aa 176 to 183; E164 would be close) and the RT1 motif in the RT domain (aa 381 to 416).

KM-1 inhibits P-Dε RNA complex formation.

The above-described data suggested that KM-1 inhibits priming by preventing P protein from interacting with Dε RNA and thus blocks the formation of a priming-active complex. We therefore directly assessed P protein binding to ³²P-labeled Dε RNA in an EMSA (Fig. 7). In preliminary experiments, the natively purified miniDP-MBPH6 protein repeatedly produced a smear of slowly migrating material (data not shown), possibly because of its oligomeric nature (see above). In contrast, the refolded preparation gave a distinct band (Fig. 7A, lane 4) which was absent without miniDP-MBPH6 (lane 2); its formation was not inhibited by a 100-fold excess of tRNA (lane 1) but was strongly reduced by excess unlabeled Dε RNA (lane 5). Addition of MBP antibody generated a supershifted band high up in the gel (lane 7) that was not seen in the absence of miniDP-MBPH6 protein (lane 6).

Fig 7.

KM-1 inhibits P protein-Dε RNA complex formation. (A) Specific detection of P-Dε complex by EMSA. Refolded miniDP-MBPH6 was incubated with ³²P-labeled Dε RNA plus the indicated additives, and the products were detected after electrophoresis in native 5% polyacrylamide gels by autoradiography. The band labeled miniDP-Dε RNA complex was seen only in the presence of these two components, reduced not by tRNA but by excess Dε RNA, and supershifted by anti-MBP antibody (arrowhead). (B) KM-1 inhibits complex formation with an IC₅₀ similar to that with which it inhibits priming. KM-1 was added to miniDP-MBPH6 30 min before the RNA (top), together with the RNA (middle), or 30 min after the RNA (bottom), and the products were analyzed by EMSA. Signals from three experiments each were quantitated by phosphorimaging and used to calculate IC₅₀s as described in the legend to Fig. 2D. (C) Priming stabilizes the P-Dε RNA complex against KM-1. Preformed P-Dε complexes were incubated simultaneously with the indicated concentrations of KM-1 plus Mn²⁺ and dCTP, which does not allow priming with WT Dε RNA as the template (top), or plus Mn²⁺ and dGTP, which allows priming to occur (middle). Finally (bottom), KM-1 was added after priming with dGTP had been allowed to proceed for 45 min. Here, a detectable complex signal repeatedly remained even in the presence of 30 μM compound. (D, E) Significance of IC₅₀ differences in KM-1-mediated P-Dε binding inhibition. Data for IC₅₀ differences related to the order of KM-1 versus Dε RNA addition (D) and to the use of Mn²⁺ plus dCTP versus Mn²⁺ plus dGTP (E) were processed as described in the legend to Fig. 4C; as there, I, II, and III in panel D refer to KM-1 addition before, together with, and after the RNA, respectively.

Next, we performed analogous EMSAs in the presence of KM-1, which was added prior to the Dε RNA (Fig. 7B, top) together with the RNA (middle) or after the RNA (bottom). Quantitative evaluation (n = 3) revealed an increase in the IC₅₀ from 1.46 ± 0.27 μM over 1.82 ± 0.18 μM to 3.65 ± 0.86 μM. These values were close to those seen for priming inhibition (Fig. 2), consistent with the prevention of Dε RNA binding as the primary cause of priming inhibition. However, although significant (Fig. 7D), the apparent IC₅₀ increase when KM-1 was added after P-Dε complex formation was only 2-fold, compared to 4-fold (miniDP-MBPH6; Fig. 2A) or 7-fold (GrpDP; Fig. 4B) with priming as the readout. This indicated that the priming conditions, or priming per se, stabilized the P-Dε complex against KM-1.

For distinction, we repeated the binding experiments with three variations. First, we added KM-1, after P-Dε complex formation, together with 2 mM Mn²⁺, a cofactor that strongly boosts in vitro priming (5), but without dNTP. In this case, the IC₅₀ remained at about 3 μM (not shown). Adding KM-1 with dGTP but without Mn²⁺ also had no substantial impact (IC₅₀ = 4.06 ± 1.04 μM; not shown). However, when KM-1 was added together with both dGTP and Mn²⁺ to allow priming to occur, the apparent IC₅₀ was nearly doubled to 6.59 ± 1.42 μM (Fig. 7C, middle); the difference from the addition of only Mn²⁺ or only dGTP was significant (P = 0.012), and the difference from the simultaneous addition of Dε RNA and KM-1 under nonpriming conditions was highly significant (P < 0.001). Next, we replaced, under priming conditions, dGTP with dCTP, which is not detectably linked to P protein with WT Dε RNA as the template (5). In this case, the apparent IC₅₀ also remained at about 3.3 μM (Fig. 7C, top) and the difference between the addition of dCTP and that of dGTP was significant (Fig. 7E). Together, these data strongly suggested that priming increases the stability of the P-Dε complex against KM-1. In vitro priming under our conditions does not occur instantly but reaches a plateau within about 30 min. Adding KM-1 at the time of priming initiation (addition of Mn²⁺ and dGTP), as in the experiments above, would thus leave the compound time to displace some of the more weakly bound Dε RNA from not-yet-primed complexes. We therefore added KM-1 45 min after priming initiation (Fig. 7C, bottom), which gave an IC₅₀ of 7.71 ± 2.27 μM (n = 4). In this setting, some complexes repeatedly remained even at 30 μM KM-1. Together, these data indicated that KM-1 inhibits priming largely by preventing the P-Dε interaction and that priming reduces the sensitivity of the complex to KM-1.

KM-1 inhibits DNA-dependent DNA polymerase activity of DHBV nucleocapsids.

To assess whether KM-1 also inhibits DNA elongation by P protein, we used the EPR in which the encapsidated P protein extends immature DNAs when provided with dNTPs. To this end, DHBV nucleocapsids from transfected LMH cells were incubated for 16 h with dNTPs including [α-³²P]dCTP (for DNA analysis) or with unlabeled dNTPs (for capsid analysis) in the presence of increasing concentrations of KM-1. Isolated DNAs were separated from the ³²P-labeled EPR products by agarose gel electrophoresis and visualized by autoradiography (Fig. 8A). In the absence of a drug, two major bands at the positions of full-length double-stranded linear (DL) and single-stranded viral DNA were detectable. A low KM-1 concentration strongly reduced these signals (>50% reduction at 0.2 μM), and a 5 to 10 μM concentration of the compound led to the complete disappearance of all signals. Hence, KM-1 inhibited strand elongation with a potency similar to that with which it inhibited in vitro priming.

Fig 8.

KM-1 inhibits the DNA-dependent DNA polymerase activity of DHBV nucleocapsids. (A) EPR with DHBV nucleocapsids from transfected LMH cells. Nucleocapsids were subjected to EPR for 16 h in the presence of dNTPs, including [α-³²P]dCTP plus the indicated concentrations of KM-1. Viral DNAs were then separated by agarose gel electrophoresis and detected via autoradiography. The positions of DL-DNA (DL) and single-stranded DNA (ss) are indicated. M1, ³²P-end-labeled restriction fragment covering a unit length DHBV genome; M2, heat-denatured M1. (B) Likely interaction of KM-1 with the basic CTD of DHBc. Nucleocapsids from transfected LMH cells were incubated for 16 h with unlabeled dNTPs (+dNTPs) or in the same buffer without dNTPs (−dNTP), and then KM-1 was added at the indicated concentrations. Samples were subjected to NAGE, followed by immunodetection with a DHBc-specific MAb. Alternatively, full-length DHBc (FL-DHBc; aa 1 to 262) capsids generated in LMH cells (middle) or in E. coli (bottom left) or E. coli-derived capsids from truncated, CTD-deficient DHBc (aa 1 to 195) were subjected to the same procedure. KM-1 at 1 to 2 μM increased the mobility and compactness of the capsid band in all cases, except for the CTD-deficient variant.

To confirm the presence of equal amounts of nucleocapsids in the EPR products, the nonlabeled EPR products were subjected to native agarose gel electrophoresis (NAGE), where capsids run as a distinct band (12) and can be detected by immunoblotting (Fig. 8B). Unexpectedly, the presence of 0.2 to 2 μM KM-1 slightly enhanced the mobility of the capsids and led to a more focused appearance of the signal; no further change occurred at higher KM-1 concentrations.

The similar KM-1 concentration profiles for prevention of DNA elongation in the EPR and for the mobility change of dNTP-supplemented nucleocapsids (Fig. 8B, +dNTP) suggested that extension of the packaged DNA caused the effect. However, nucleocapsids not subjected to EPR showed the same mobility shift (Fig. 8B, −dNTP) and so did capsids from cells transfected with an expression vector solely encoding DHBc protein (Fig. 8B, middle). Hence, the shift was not dependent on encapsidated P protein and/or viral nucleic acid. As KM-1 carries multiple negative charges, we suspected that it might interact with the positively charged nucleic acid binding CTD of DHBc. We therefore tested KM-1's influence on E. coli-derived DHBV capsids (33) from full-length DHBc (FL-DHBc, aa 1 to 262) or a variant lacking the CTD (DHBc-CTD; aa 1 to 195). KM-1 again caused a forward mobility shift of the FL-DHBc capsids but had no effect on CTD-deficient capsids (Fig. 8B, bottoms). These data are consistent with an electrostatic interaction of KM-1 with the positively charged CTD of the core protein. Because the core protein CTD is involved in capsid-internal reverse transcription (27, 32), such an interaction could contribute to DNA synthesis inhibition by KM-1.

KM-1 inhibits DNA-dependent DNA polymerase activity of human HBV nucleocapsids.

For potential therapeutic applications, it would be important to directly demonstrate that KM-1 inhibits human HBV P protein. In the absence of an appropriate in vitro priming system, we used the EPR as described above as the readout. To assess the maturation status of the encapsidated DNA before the EPR, DNAs from cytoplasmic nucleocapsids, obtained from Huh7 cells transfected with the genotype D HBV expression vector pCH-9/3091 (32), were analyzed by Southern blotting after overnight incubation in EPR buffer containing or not containing dNTPs (Fig. 9A, left side). Without added dNTPs, DNA was detected largely as a smear between single-stranded DNA and DL-DNA plus a weak signal at the RC-DNA position. Incubation with dNTPs shifted these products toward mature forms, particularly RC-DNA. Essentially the same pattern was seen after a ³²P-labeled EPR in the absence of KM-1, whereas 1 to 2 μM KM-1 reduced in vitro DNA elongation by about 50%; higher concentrations blocked the formation of mature products (Fig. 9A, middle). The same results were obtained (Fig. 9A, right side) with nucleocapsids from the 3TC-resistant variant rtL180M/M204V (58). 3TC resistance was confirmed by EPR in the presence of 3TC-triphosphate (3TC-TP), the active form of the NA (Fig. 9B). DNA synthesis in WT nucleocapsids was inhibited by 50% at about 1 μM 3TC-TP, whereas inhibition of the mutant was seen only at a 50 μM drug concentration. KM-1 also inhibited the EPR activity of genotype C isolate FMC#97 (56), whose P protein sequence diverges by about 10% from that of the genotype D isolate used above (data not shown). These results independently confirmed that KM-1 acts by a mechanism that is distinct from that of NAs and is active against different HBV genotypes and NA-resistant variants.

Fig 9.

KM-1 inhibits DNA-dependent DNA synthesis in nucleocapsids from WT and 3TC-resistant human HBV. (A) Equally efficient inhibition of DNA elongation in nucleocapsids from WT and 3TC-resistant viruses. Nucleocapsids from Huh7 cells transfected with a vector for a genotype (gt) D WT HBV genome (pCH-9/3091) or its 3TC-resistant variant rtL180M/M204V were subjected to EPR conditions with (+dNTP) or without (−dNTP) unlabeled dNTPs, and the DNA products were assessed by Southern blotting (left side). In parallel, EPRs were performed in the presence of dNTPs, including [α-³²P]dCTP and the indicated concentrations of KM-1. Radiolabeled DNAs were separated and visualized by autoradiography as described in the legend to Fig. 8. Comparable inhibition was seen with the genotype C HBV isolate FMC#97. (B) Confirmation of the 3TC resistance phenotype of variant rtL180M/M204V. EPRs were performed as for panel A, except that KM-1 was replaced with 3TC-TP. (C) Likely interaction of KM-1 with the basic CTD of HBV core protein (HBc). NAGE analysis was performed as described in the legend to Fig. 8B, using nucleocapsids or capsids (from an HBc-only expression vector) from transfected Huh7 cells, E. coli (E. c.)-derived capsids from full-length (HBc aa 1 to 183) or CTD-deficient (HBc aa 1 to 149) core protein. KM-1 at concentrations of >1 μM retarded capsid mobility in all samples, except for the CTD-deficient variant.

As for DHBV, NAGE analysis (Fig. 9C) revealed an impact of KM-1 on HBV nucleocapsid mobility but in the opposite direction. Increasing concentrations of KM-1 retarded a fraction of the anticapsid antibody-reactive material, weakening the signal at the position of nontreated nucleocapsids (Fig. 9C, leftmost panel). This retardation also occurred with capsids from Huh7 cells transfected with a core protein-only expression vector (Fig. 9C, second panel from the left) and with E. coli-derived full-length core protein capsids (HBc183; Fig. 9C, third panel from the left) but not with capsids lacking the CTD (HBc149; Fig. 9C, rightmost panel) or HBc183 capsids in which the CTD was phosphorylated (35), which counterbalances the positive charges in the CTD (not shown). Hence, KM-1 appears to interact with the basic CTDs of both HBV and DHBV. Given the role of the CTDs in enabling reverse transcription inside the capsid, such interactions may also contribute to DNA synthesis inhibition.

Low in-cell anti-HBV activity of KM-1.

KM-1 inhibited HIV-1 replication in cells with an IC₅₀ of 2.5 μM, i.e., about 25-fold higher than the IC₅₀ for HIV-1 RT's DNA polymerase activity in vitro (44, 53), likely reflecting low intracellular availability. The TC₅₀ for the lymphocytic cell lines used was 112 μM (53). To test KM-1 for in-cell activity against HBV, we transfected Huh7 cells with the WT-HBV vector pCH-9/3091 or its 3TC-resistant rtL180M/M204V variant and cultured them in the presence of KM-1. Southern blotting for HBV DNAs revealed a slight reduction (about 25%) at 50 μM KM-1 for WT HBV (Fig. 10, left side), which was further enhanced in the 3TC-resistant variant by 80 μM KM-1 (Fig. 10, right side). Comparable results were obtained for the genotype C FMC#97 isolate (not shown). In an attempt to shed more light on the modest reduction of viral DNA levels at a high compound concentration, we measured the core protein (Fig. 10B) and assembled capsid (Fig. 10C) levels in cytoplasmic lysates. While no marked differences with increasing KM-1 concentrations were seen for core protein (normalized to actin content in the same sample), capsid levels decreased gradually and to an extent similar that of the DNA levels monitored by Southern blotting. The lack of visible capsid retardation likely reflects the much lower intracellular compound levels than in the direct incubation experiments (Fig. 9C). Together, these data are compatible with a direct effect on core protein assembly yet also with an indirect impact of P protein inhibition-mediated suppression of DNA synthesis on nucleocapsid stability. Because even further pleiotropic effects on the cell are currently not excluded, the mechanistic basis of the moderate reduction of intracellular HBV replication remains to be determined.

Fig 10.

Low intracellular anti-HBV activity of KM-1. Huh7 cells were transfected with vectors for the WT genotype (gt) D HBV or its 3TC-resistant variant and cultured in the presence of the indicated concentrations of KM-1. At 4 days posttransfection, DNAs from cytoplasmic nucleocapsids were analyzed by Southern blotting. DNA signals were quantified by phosphorimaging. Detectable inhibition required 50 or 80 μM KM-1. Higher concentrations were not tested because of beginning cytotoxicity. rel., relative. (B) KM-1 does not markedly affect HBV core protein levels. Core protein (HBc) in the same cytoplasmic lysates as used in panel A was monitored by SDS-PAGE, followed by immunoblotting and luminescence-based detection. Band intensities were then normalized to those of β-actin (actin) in the same sample, with the ratios obtained in the untreated samples set at 100%. (C) KM-1 reduces the levels of capsids and replicative DNA similarly. Aliquots of the same cytoplasmic lysates as used in panel A were analyzed for intact capsids as described in the legend to Fig. 9C. The lack of detectable retardation with increasing KM-1 concentrations probably reflects the low intracellular concentration of the compound.

To test for the feasibility of repeating the experiments with higher dosing, we determined the cytotoxicity of KM-1 for naive Huh7 cells (not shown). The measured TC₅₀ of 105 μM indicated a low therapeutic index for the in-cell anti-HBV activity of the compound. Thus, rather than an immediate drug candidate, KM-1 is a valuable lead for further development.

DISCUSSION

The limitations of current treatments for chronic hepatitis B urge the development of new antivirals with different mechanisms of action (26, 31). The highly specific interaction of P protein with the ε element is crucial for pgRNA encapsidation and priming of replication and thus an attractive target for intervention. Here we show that naphthylureas of the carbonyl J acid family, in particular KM-1, potently inhibit protein priming by DHBV P protein. KM-1 interferes with the formation of P-Dε initiation complexes by competing with the RNA for overlapping binding sites on the P protein. These data provide new insights into the hepadnaviral replication initiation mechanism, including a successive stabilization of the initiation complex during its maturation. In addition, KM-1 inhibited the DNA-dependent DNA polymerase activity of DHBV and, importantly, of HBV nucleocapsids, proving that the human virus, including a highly 3TC-resistant variant, is susceptible to the drug. Because of its low intracellular bioavailability, KM-1 is not an immediate drug candidate but it represents a useful lead for improved compounds that would inhibit multiple crucial steps in the HBV replication cycle by a non-NA-like mechanism.

Inhibitory mechanism of KM-1: replication initiation complex formation as the primary target.

In vitro priming with DHBV P protein provides a unique opportunity for mechanism-based identification of inhibitors that might not be found by cell-based assays. KM-1 (as well as carbonyl J acid and calcomine orange) inhibited in vitro priming by miniDP-MBPH6 protein (Fig. 2B) yet also chaperone-dependent in vitro priming by the nearly full-length GrpDP protein and by the full-length P protein in RRL (Fig. 4). The only major difference was a greater IC₅₀ in RRL than in the pure-component assays (14 μM versus around 1 to 7 μM). This clearly related to the RRL because the same was seen with in vitro-translated miniDP-MBPH6 (Fig. 4A). Hence, priming inhibition by KM-1 was not a peculiarity of the miniDP system.

Our RNA competition experiments (Fig. 5) excluded Dε RNA as a target of KM-1, while the modest rescue seen with excess WT Dε RNA hinted at competition between the drug and WT Dε RNA for an a binding site(s) on the P protein (see below). Conversely, adding an excess of the separate TP or RT domain increased priming in the presence of KM-1 (Fig. 5B), suggesting that the drug targets binding sites in both the TP and RT domains of the P protein, as proposed for Dε RNA (2, 45).

Evidence for competition between KM-1 and Dε RNA for overlapping binding sites on the P protein: implications for the mechanism of hepadnaviral protein priming.

Two lines of evidence further support the idea that KM-1 and Dε RNA compete for overlapping binding sites on the P protein, namely, the similar effects of the compound and Dε RNA on protease accessibility and the dose-dependent inhibition of Dε RNA binding seen in EMSAs.

The exact binding site(s) for KM-1 on HIV-1 RT is not known. For lack of direct structural information on hepadnaviral P proteins, identification of potential KM-1 binding sites is even more difficult. Clearly, however, KM-1 strongly reduced V8 protease cleavage at E residues in the P protein part of the MBP-miniDP protein that were quite accessible in its absence (Fig. 6C). Beyond the size of the V8 protease itself (∼30 kDa), the resolution of the partial-proteolysis approach is limited by the few epitopes for which specific MAbs are available (aa 143 to 147, MAb 5; aa 191 to 197, MAb 6; aa 561 to 565, MAb 25) (15). However, the C-terminal location of the MAb 25 epitope (Fig. 6B) allowed us to define the protease-shielded residues as E401/E402 and E369/E370/E371 in the RT domain and E176 (or E164) and E124 (or E106) in the TP domain, as was fully supported by the selective detection of some, but not all, of these products by MAb 6 and MAb 5 (Fig. 6C). In sum, KM-1 shielded several dispersed sites in the TP and RT domains from V8 attack and so did Dε RNA, albeit less efficiently; this may reflect a lower affinity of Dε.

Structural information on the P-Dε complex is limited to evidence that the T3 motif in the TP domain (aa 176 to 183) (15, 45) and the RT1 motif in the RT domain (aa 381 to 416) (2) form a composite binding site. Furthermore, the bulge of Dε specifying the sequence of the DNA primer and the protein-priming Y96 residue in the TP domain must occupy positions relative to the active site and dNTP pocket equivalent to those of the template and primer strands during DNA elongation (5, 55). A visual impression of the large surface area occluded by bound nucleic acid is given by the homology-based, mutagenesis-supported (11, 55) structural model of the DHBV P protein RT domain with bound primer/template DNA and dTTP (Fig. 11A). We note, however, that this model can give, at most, a coarse idea of the Dε-containing priming complex, not the least because TP could not be modeled for lack of homologs in the database. Nonetheless, because miniDP lacks the thumb subdomain, E571/E572 (cyan space fill models in the foreground) would be easily accessible for cleavage. In contrast, E401/E402 (cyan space fill models in the background), well accessible in free miniDP, would become shielded by nucleic acid, as observed. E499 (blue stick model on the left) is predicted to be part of a long α-helix comprising residues G483 to R503 (55), equivalent to the α-helix formed by residues G155 to K173 in HIV-1 RT. Within such a stable secondary structure, the C-terminal peptide bond of E499 would not be a protease substrate, consistent with the absence of corresponding cleavage products, regardless of the conditions. Less evident is how KM-1, roughly the size of a tetranucleotide (though less flexible), could cause similar shielding. One option are multiple KM-1 binding sites; another is the induction of a more global conformational change in the protein, similar to that induced by nucleic acid binding (49, 50). For HIV-1 RT, relative movements of the palm, finger, and thumb subdomains during substrate binding and polymerization cause the enzyme to switch between open and closed conformations (43). Hence, KM-1 may freeze miniDP in a closed-like conformation that restricts the V8 accessibility of various E residues which are solvent exposed in the open conformation. However, additional experiments are required to further verify this model.

Fig 11.

(A) Occlusion of the nucleic acid binding cleft in the DHBV P protein RT domain. The model (55) is based on homology to the HIV-1 RT-primer/template-dTTP complex (Protein Data Bank code 1RTD) and comprises residues 382 to 657 of the DHBV P protein (Fig. 1A); it makes no predictions about the TP and RH domains. The finger and palm subdomains are depicted with secondary structure elements, the C-terminal thumb subdomain, absent from miniDP, is depicted as a brown backbone trace, bound dTTP is depicted in red, and the YMDD loop is depicted as a stick model with dotted van der Waals radii in magenta. How Dε RNA is bound in the initiation complex is not known, but the nucleotide in the Dε bulge that templates the first DNA residue and the protein-priming Tyr residue in the TP domain must occupy positions equivalent to those of the 3′ end of the primer and the templating nucleotide in DNA-dependent DNA polymerization. Hence, at least some RT residues covered by DNA in the model might also be shielded by bound Dε RNA. Such shielding is evident for E401/E402 (cyan space fill models in the background) from the suppression of V8 fragment a (Fig. 6B and C), whose C terminus resides at E571 or E572 (cyan space fill models in the foreground) or, at most, a few amino acids downstream. Protection of the same residues by KM-1 supports the notion that the compound and Dε RNA have overlapping binding sites on the P protein. (B) Model of inhibition of hepadnaviral protein priming by KM-1. The cartoon integrates pertinent results from this study. KM-1 inhibited priming by full-length DHBV P protein and chaperone-independent priming by miniDP via dose-dependent interference with P-Dε RNA complex formation. Partial rescue from inhibition by separate TP or RT domains and a similar suppression of protease-accessible sites in the TP and RT domains by KM-1 and Dε RNA indicate the use of overlapping binding sites on the P protein. The most potent inhibition of Dε RNA binding and priming occurred when KM-1 was added before (1a) or together with (1b) the RNA. Hence, P–KM-1 complexes are favored over P-Dε complexes; partial rescue of priming by excess Dε RNA but not tRNA further implies that the formation of either complex is reversible. Adding KM-1 after the initial P-Dε complex formation (2), and particularly after allowing priming to occur (3), significantly increased the IC₅₀ for KM-1-mediated inhibition of RNA binding and priming. This suggests enhanced stability of the P-Dε complex after progression into the rearranged priming-active state (2) and further stabilization upon priming (3).

Direct evidence for binding of KM-1 to sites on the P protein that are important for productive interaction with Dε RNA comes from the KM-1-mediated inhibition of P-Dε RNA complex formation (Fig. 7). KM-1 dose dependently reduced complex formation with IC₅₀s in a low micromolar range similar to that seen for priming inhibition. Inhibition was most potent when the compound was added before or together with Dε RNA (IC₅₀, 1.4 to 1.8 μM), whereas preincubation with the RNA increased the IC₅₀ by around 2-fold. Allowing priming to occur (addition of Mn²⁺ plus dGTP) further increased the IC₅₀, especially when KM-1 was added 45 min after all of the other components (to ∼7.7 μM; Fig. 7C). Dependence on actual priming was demonstrated by the lack of effect when dGTP was replaced with dCTP (Fig. 7C).

These results fully support the proposed sequential mutual conformational adaptations in P protein (45, 50) and Dε RNA (7) upon complex formation and indicate an additional increase in affinity once the first nucleotide has been linked to TP. While significant, the change in complex stability was still modest, which makes biological sense, as primer synthesis must proceed further and eventually Dε is replaced as the template by a 3′-proximal region of the pgRNA. A model schematically summarizing these results is shown in Fig. 11B.

KM-1 also targets DNA-dependent DNA polymerase activity of both DHBV and HBV nucleocapsids.

Protein priming and subsequent DNA polymerization differ substantially in their template requirements. Hence, it was not trivial that a non-NA inhibitor that competes with the priming template, i.e., ε RNA, would block DNA elongation as well. Taking advantage of the EPR, which does not require that an inhibitor be cell permeating, our data clearly show that KM-1 inhibits positive-strand DNA elongation for DHBV and HBV (Fig. 8A and 9A). This could occur by the drug blocking polymerization chemistry, e.g., by misaligning the primer/template relative to the active center (43) or, analogously to priming inhibition, by displacement of the primer/template, as proposed for inhibition of HIV-1 RT (44, 53). Because the interaction of both P proteins with ε is crucial for pgRNA encapsidation, this step should as well be sensitive to KM-1. Finally, our data indicate that P protein molecules that have escaped inhibition of these early steps would still be subject to inhibition of subsequent DNA polymerization by a KM-1-based drug; in analogy to data obtained with HIV-1 RT (44), KM-1 may even block the RNase H activity of the P protein.

Potential interaction of KM-1 with DHBV and HBV capsids.

KM-1 affected the electrophoretic mobility of both DHBV and HBV capsids (Fig. 8B and 9B). The effect clearly depended on the highly positively charged CTDs of the two core proteins but was independent of encapsidated P protein and pgRNA or viral DNA. In low-percentage agarose gels such as those used here, capsid mobility is determined largely by the surface charge, with more negative charges increasing mobility toward the anode. Hence, for DHBV capsids, the negative charges of KM-1 (Fig. 2A) somehow contribute to an overall more negative surface charge; one interpretation is that part of the CTD is surface exposed, as earlier proposed (41), and neutralized by electrostatic interactions with KM-1. For HBV capsids, the broad distribution of slower-migrating material more likely reflects KM-1-induced aggregation. Further experiments are required to reveal the molecular basis of this phenomenon.

Of practical importance is whether an interaction of KM-1 with the capsid would interfere with its inhibitory activity against the packaged P protein. The holes in the capsid shell are large enough to allow dNTPs to enter, and packaged bacterial RNA in recombinant capsids is easily accessed by ethidium bromide (12); hence, KM-1 should be able to penetrate the capsid shell to interact with the P protein. A formal alternative is that EPR inhibition is caused by KM-1 binding to the capsid-internal CTDs, which are important for reverse transcription (27, 32). At present, our experiments do not distinguish whether and, if so, to what extent an interaction with the CTDs contributes to suppression of the EPR signals. Clearly, however, the data demonstrate potent inhibition of DNA synthesis inside authentic nucleocapsids. Notably, the IC₅₀ was in the same range as that for vitro priming where no capsid protein is present.

Aspects relevant to drug development.

Several of our inhibition data for KM-1 resemble those reported for selected porphine derivatives (28). Hence, the tetrapyrrole and naphthylurea derivatives may have similar inhibition mechanisms. Notably, like KM-1 hemin derivatives also block HIV-1 RT, supposedly by stacking interactions between the porphine ring and aromatic residues in the protein (1). Similar interactions may occur with the aromatic ring systems in the carbonyl J acid compounds. An additional important feature appears to be negative charges; only hemin and derivatives bearing carboxylates had activity against P protein (28). Similarly, carboxylates enhanced the anti-HIV-1 RT activity of carbonyl J acid derivatives (44). This suggests that different scaffolds fulfilling these two criteria would be useful for further exploration, e.g., by scaffold hopping (48), of improved priming inhibitors.

Like hemin, KM-1 itself is not yet a drug candidate although its selectivity for RTs, as opposed to other DNA polymerases, appears higher; for instance, KM-1 did not inhibit the Klenow fragment of E. coli DNA polymerase I (44) whereas hemin did (28). The most obvious shortcoming of KM-1, shared with hemin, was the much reduced intracellular activity (Fig. 10), which required concentrations (>50 μM) close to the TC₅₀ for Huh7 cells (105 μM). The intracellular anti-HIV-1 activity of KM-1 was also reduced by about 25-fold compared to in vitro inhibition (44) but because of the higher potency against HIV-1 RT was still substantially below the TC₅₀. Reduced intracellular activity is likely caused by a poor ability to permeate cells or by export from cells and/or metabolic modifications. Hence, there are several hurdles to be overcome, which do not, however, appear insurmountable given the advanced tools available in modern medicinal chemistry. Most importantly, KM-1-like drugs would add a new mechanism to the restricted repertoire of current anti-HBV drugs which, beyond inhibiting DNA-dependent DNA synthesis activity by a non-NA mechanism, targets protein priming, pgRNA encapsidation, and possibly RNase H activity as additional crucial steps. Hence, efforts to develop KM-1 into a useful drug appear highly warranted, as is further underlined by the potent inhibition of two different HBV genotypes and the highly 3TC-resistant variant selected during treatment with a currently approved drug.