Induction of Hepatitis B Core Protein Aggregation Targeting an Unconventional Binding Site

The hepatitis B virus (HBV) infection is a major global health problem, with chronic infection leading to liver complications and high death toll. Current treatments, such as nucleos(t)ide analogs and interferon-α, effectively suppress viral replication but rarely cure the infection. To address this, new antivirals targeting different components of the HBV molecular machinery are being developed. Here we investigated the hepatitis B core protein (HBc) that forms the viral capsids and plays a vital role in the HBV life cycle. We explored two distinct binding pockets on the HBV capsid: the central hydrophobic pocket of HBc-dimers and the pocket at the tips of capsid spikes. We synthesized a geranyl dimer that binds to the central pocket with micromolar affinity, and dimeric peptides that bind the spike-tip pocket with nanomolar affinity. Cryo-electron microscopy further confirmed the binding of peptide dimers to the capsid spike tips and their capsid-aggregating properties. Finally, we show that the peptide dimers induce HBc aggregation in vitro and in living cells. Our findings highlight two tractable sites within the HBV capsid and provide an alternative strategy to affect HBV capsids.


Introduction
The hepatitis B virus (HBV) infects the liver and can cause acute and chronic hepatitis.In childhood and infancy, the virus is particularly dangerous, as the recovery rate among children is approximately 50%, while among infants infected through perinatal transmission, only 10% will naturally recover, the remainder will develop chronic infection. [1,2]On the global scale the most effective approach to address hepatitis B is through preventive treatment with vaccinations.However, the goals of achieving sufficient vaccination coverage and timely immunization have yet to be met. [1,3]Furthermore, vaccinations are ineffective for individuals who are already infected. [4]With about 300 million chronic carriers and over 800,000 hepatitisrelated yearly deaths, chronic hepatitis B is a global health problem [2,5] that requires a solution.
Currently, there are two approved classes of medications for the treatment of chronic hepatitis B: nucleos(t)ide analogs (NAs) and interferon-α and its derivatives (IFN-α). [6,7]NAs compete for binding with the natural nucleotide substrates, inhibiting the viral protein P in charge of the reverse transcription of the viral pre-genomic RNA (pgRNA) into HBV DNA. [8]IFN-α serves as both an immunomodulator and immunostimulant, activating genes with diverse antiviral functions to target various steps of viral replication.Additionally, it indirectly suppresses HBV infection by modifying cell-mediated immunity. [9]sent treatments effectively suppress HBV replication, reduce liver inflammation, fibrosis, and the risk of cirrhosis and hepatocellular carcinoma (HCC), but IFN-α treatment is associated with significant adverse effects and NAs typically require long-term oral administration, often lifelong, as treatment discontinuation frequently leads to relapse.Moreover, although current therapy allows effective management of the disease, clinical cure is rarely achieved, and the risk of HCC, although reduced, still persists. [7]Consequently, various classes of direct-acting antivirals and immunomodulatory therapies are currently under development, aiming to achieve a functional cure following a finite treatment duration. [10] HBV antivirals capitalize on the enhanced understanding of the viral life cycle and can be categorized into several classes (Table 1): Entry inhibitors that disrupt HBV entry into hepatocytes by blocking the binding to the sodium/taurocholate co-transporting polypeptide (NTCP) receptor. [11]HBsAg inhibitors based on nucleic acid polymers that interfere with the production of HBV surface antigens, and viral gene repressors based on nucleases.
Translation inhibitors based on small interfering RNAs or antisense oligonucleotides that silence HBV RNA, thereby decreasing the viral antigen production.Finally, the capsid assembly modulators (CAMs) target the hepatitis B core protein (HBc) that participates in multiple essential steps of the HBV life cycle. [7]able 1.Direct-acting HBV antivirals

HBsAg inhibitor
Inhibition of the host HSP40 chaperone DNAJB12, that mediates spherical HBV assembly.Reduces the HBsAg in the circulation and lowers intracellular HBsAg. [15]P 2139 [16] Phase II Nucleic acid polymer

Translation inhibitors
Antisense oligonucleotide (ASO) or small interfering RNAs (siRNA) [17] that target HBV messenger RNAs and act to decrease levels of viral proteins.

Capsid assembly modulators
Target the hydrophobic pocket located at the dimer-dimer interface near the C termini of HBc subunits and induce misassembly of the core protein, thereby impeding the formation of infectious progeny virions. [20,22,23]nocapavir [23] Phase II Small molecule EDP-514 [24] Phase I Small molecule Among the direct acting antivirals that are in preclinical or clinical studies a third are CAMs. [6]sids are attractive targets due to the absence of human homologues for HBc and their involvement in crucial stages of the HBV life cycle, including nuclear entry, encapsulation of the pgRNA and polymerase, optional nuclear recycling to replenish the covalently closed circular DNA (cccDNA) pool, and eventual coating and secretion from infected cells. [7,22] capsid is composed of 120 units of HBc dimers, assembling into a T = 4 icosahedron.
Within this structure, 60 asymmetric units are formed by 4 HBc monomers each, designated as A, B, C, and D, or AB dimers and CD dimers. [25]The ultrastructure formed by the HBc dimer reveals several binding pockets that can be exploited as potential targets for modulating protein activity (Figure 1).
CAMs target the hydrophobic pocket at the HBc dimer-dimer interface, upon binding they strengthen the association energy between HBc-dimer subunits, thereby promoting capsid assembly, rather than inhibiting it. [26,27]As a result, abnormal or empty capsids may form, sometimes accompanied by the aggregation of core proteins, consequently inhibiting HBV DNA replication.Additionally, CAMs can disrupt the disassembly of incoming virions and the intracellular recycling of capsids, thereby impeding the establishment and replenishment of cccDNA. [7,22,28,29]ently, a new potentially druggable site was discovered in the HBc dimera hydrophobic pocket formed at the base of the spike.This site was targeted by the detergent Triton X-100 (TX100), ultimately causing conformational alterations in the capsid structure. [30,31]In addition to the spike-base hydrophobic pocket, there is another less-explored interacting domain located on the spike tip of the HBc dimer.Previous studies have shown that peptides targeting the cleft on the spike tip reduced viral replication in a cell model, likely by interfering with viral assembly through modulation of the HBc interaction with the surface antigen. [32]These two effector sites could serve as a foundation for the development of new types of HBc modulators and provide alternatives ways for controlling HBV infections.
In this study, we characterize and explore these alternative HBc binding pockets at the innerdimer interface in the center and at the tips of capsid spikes [30,31,33] , and unveil the HBc aggregating properties of a spike-binding dimeric peptide.

Results
HBV capsid assembly modulation via the binding pockets on the HBc multimer ultrastructure represents a promising pharmacological strategy but until now only one site located on the HBc dimer-dimer interface was explored (Figure 1A). [22,23]We designed and synthesized bivalent binders that target the hydrophobic pocket in the center of HBc-dimers or the tips of the spikes (Figure 1) with avidity enhanced binding affinities and evaluated their impact on HBV capsids in vitro and in living cells.

Geranyl dimer targets the central hydrophobic pocket of HBc-dimers with micromolar affinity
Structural and thermodynamic studies identified a distinct hydrophobic pocket in the center of HBc-dimers, with TX100, a nonionic surfactant with a polyethylene oxide chain, as a ligand. [30,31,34,35]Several of the pocket forming amino acids, such as K96 and 129-PPAY-132, [36] and the natural occurring point mutations HBcP5T, L60V, F97L and P130T [35,[37][38][39][40] are involved in secretion of enveloped virions from the cell.These findings lead to the infectious HBV particles signal hypothesis where this hydrophobic pocket is involved in the regulation of the envelopment of nucleocapsids and thus could be an alternative druggable pocket to block virus envelopment. [34]rophobic post-translational modifications, such as myristylation of the Large Hepatitis B Virus Surface Protein (L-HBs), are essential for HBV infectivity and play a role in mediating viral assembly. [41]Additionally, farnesylation, another hydrophobic post-translational modification, is involved in the envelopment of hepatitis D virus [42] , which relies on the presence of HBV and its protein machinery for propagation.We therefore reasoned that the binding partners of the hydrophobic pocket of HBc could likely be myristylated (4), geranylgeranylated (2), or farnesylated (1) (Figure 2A).Importantly, these naturally occurring modifications share features with the previously identified ligand TX100, namely hydrocarbon chains, that could mimic the binding interactions.However, farnesyl pyrophosphate, geranylgeranyl pyrophosphate and myristic acid are poorly soluble.Therefore, we explored geraniol as a water-soluble mimetic of farnesyl and geranylgeranyl as well as n-Decyl-beta-Dmaltopyranoside (DM) (5) as soluble mimetic of myristic acid (Figure 2A).The isothermal calorimetric titration (ITC) of HBc capsids with DM (5) resolved micromolar affinity (KD = 133 +/-38 µM) to all four hydrophobic pockets of HBc capsids' asymmetric unit (N=1.05+/-0.1) (Figure 2B and Supplementary Figure 7C).
ITC of geraniol with HBc showed a slightly enhanced micromolar affinity (KD=94 +/-8 µM) (Figure 2A,B, supplementary tables 1, 2 and Supplementary Figure 7A,B) and a stoichiometry of N=1.01 +/-0.04,implying that all four hydrophobic pockets of the asymmetric unit are occupied simultaneously.To confirm geraniol's binding to the capsids and to resolve the molecular details of this interaction we conducted cryo-EM of a mixture of HBc with excess of geraniol followed by single particle analysis.This experiment resolved an additional density for geraniol in all four hydrophobic pockets within the asymmetric unit of HBc capsids (Figure 2D, Supplementary Figure 8), confirming the thermodynamic binding data and further defining the underlying molecular interactions of the involved HBV residues P5, L60, K96 and F97.Encouraged by the enhanced geraniol affinity to the central hydrophobic pocket we designed and synthesized a dimeric version of geraniol capable of simultaneous binding to the HBc dimer.We connected the two geranyl moieties with a polyethylene glycol (PEG) linker that could bridge the distance of 38 Å between the two opposing hydrophobic pockets (Supplementary Figure 1).After synthesis, purification and mass spectrometric validation (Appendix 1) we determined the HBc capsid binding parameters of the geranyl dimer via ITC.
The analysis suggested that the dimer engages with both HBc binding sites simultaneously resulting, however, only in a moderately enhanced micromolar affinity of 63 +/-8 µM (Figure 2B,C).

Targeting the pocket of capsid spike tips with nanomolar affinity peptide dimers
Although geraniol and geranyl dimer displayed improved affinity to HBc and allowed structural insights on a binding pocket located at the center of HBc dimers, micromolar affinity is suboptimal for a functional compound.Therefore, we proceeded to explore another binding site located on the capsid spike tips formed by HBc dimers (Figure 1). [32]lier studies have shown that phage display-derived peptides were binding to the spike tips of recombinant HBc capsids.These peptides were also observed to disrupt the interaction between HBc and HBV's surface protein, L-HBs. [43]Recently we have shown that these peptides MHRSLLGRMKGA (P1), GSLLGRMKGA (P2) and the core binding motif SLLGRM bind to wild-type (wt) and mutant HBc variants (P5T, L60V and F97L) with intermediate micromolar KDs of 26, 68 and 130 µM, respectively. [33]e, we designed dimeric peptides with a PEG linker capable of bridging the distance of 50 Å between the capsid spikes, thus tailoring our binders for simultaneous binding of two HBc dimers (Figure 1B, Supplementary Figure 1).The three distinct dimeric peptides, the minimal SLLGRM dimer (7), the P2 dimer (8), and the P1 dimer, were synthesized, purified, and validated using mass spectrometry (Appendix 1).Subsequently, their binding to the HBV capsid was evaluated through ITC (Figure 3A, Supplementary Table 2, Supplementary Figures 3,7).With a KD value of 4.9 +/-0.7 µM, the SLLGRM-dimer (7) has the lowest affinity to HBc, followed by the P2-dimer (8) (KD =1.9 +/-0.4 µM).Finally, the P1-dimer (9) displayed a nanomolar affinity of 312 nM.Thus, P1-, P2-and SLLGRM-dimers show 83-, 36-and 27fold increased affinities compared to their monomeric counterparts.
The significant increase in affinity of the P1-dimer over the monomer, by almost two orders of magnitude, may not be solely attributed to binding to two sites simultaneously.Once the P1dimer binds it can interact with up to four binding partners in its vicinity (Supplementary Figure 5). [44]This may enable detachment and immediate reattachment to a nearby binding partner, further enhancing the local concentration and the overall binding strength of the P1 dimer.
Finally, our Cryo-EM experiments have confirmed that the peptide dimers occupy the binding site at the spike tip of HBc dimers.Among these, the dimerized P1 exhibited a higher occupation of the binding site, as illustrated in Supplementary Figure 9. Sequence requirements of the HBc Spike binding site.Full positional scan of the P1 peptide sequence in µSPOT format, in which each residue was varied to each other proteogenic amino acid.Note that a drop in binding intensity upon variation of the core motif SLLGRM (highlighted in bold) substantiates its critical involvement in HBc binding.
Refer to supplementary table 3 for the corresponding absolute greyscale values.
Notably, while performing the ITC titrations we have noticed fast fluctuations of the heat signature baseline across the tested ligands (Figure 3; Supplementary Figure 3), at least for the P1 dimer this phenomenon may be attributed to aggregation.To further substantiate and quantify a possible dimer-induced HBc aggregation we next performed a turbidity assay. [45]found that P1 dimer induces turbidity of a HBc solution already at 1:10 equivalents of HBc, whereas the P2 dimer was slightly less potent and the SLLGRM dimer did not induce turbidity at the same conditions and further required significantly higher concentrations ratio relative to HBc (Supplementary Figure 2).To shed light on the seemingly sequence-specific aggregation properties of the different dimers we analyzed the binding of 240-point mutated P1 peptide variants in array format (Figure 3D, Supplementary Table 3).The analysis recapitulated our earlier resolved sequence requirement for HBc binding and substantiated that the minimal sequence SLLGRM is the major mediator of HBc binding.It further indicates that the additional N-terminal residues in P1 sequence are neither conserved nor critically required for binding despite their importance in inducing HBc aggregation.

P1dC aggregates HBc in living HEK293 cells
The strong nanomolar affinity of the P1-dimer, along with its ability to induce capsid aggregation in vitro, prompted us to evaluate its effect on HBV core protein in living cells.To adapt the peptide for the intracellular delivery, we synthesized a C-terminally cysteinated version of P1-dimer, P1dC (10) (Figure 3A), and its scrambled counterpart scrP1dC (scr10) as well as a thiol-reactive polyarginine-based cell penetrating peptide (CPP), containing a cysteine, with a 5-thio-2-nitrobenzoic acid (TNB) modified thiol (Figure 4A, Appendix 1).At the core of this intracellular delivery method is the in-situ conjugation of the cargo molecule to an excess of a CPP via a disulfide bond, and the application of this reaction mix on living cells.
The excess of the CPP reacts with the cell membrane to facilitate the penetration of the cargo-CPP conjugate.In turn, the disulfide bond between CPP and the cargo would be reduced in the cytosol, separating the cargo from the CPP, allowing unhindered activity of the cargo molecule within the cell (Figure 4B). [46,47]verify that the P1dC performs similarly to P1-dimer we performed another ITC assay to determine the affinity of the compound to HBc.The ITC confirmed that P1dC has an affinity of 420 +/-38 nM, comparable to P1-dimer, while the scrambled peptide did not display binding to HBc (Figure 3, Supplementary Figure 3, Supplementary Table 2).Thereafter we transfected mammalian cells (HEK293) with a plasmid coding for HBc.The cells expressed the protein for two days and were then treated for 1 hour with the thiol-activated cell penetrating peptide and P1dC or the respective negative control peptide scrP1dC (Figure 4B).Afterwards the cells were immediately washed and fixed and HBc was visualized with anti HBc antibody and a secondary DyLight650 conjugated antibody.Transfected but otherwise untreated cells showed a homogeneous distribution of recombinant HBc molecules in the nucleus and to a lesser extent in the cytoplasm (Figure 4C).Yet, upon administration of 10 µM of P1dC, we observed aggregates of HBc (in form of large bright spots) within the cells (Figure 4C, Supplementary Figure 4).At a concentration of 10 µM, the scrambled dimer scrP1dC did not induce aggregation and the distribution of HBc remained largely homogenous.A) A polyarginine cell-penetrating peptide containing a cysteine with a TNB-activated thiol (gray highlight, ( 11)).B) The live cell experiment flow.First, mammalian cells are transfected with HBc coding plasmid.Then, after the cells express the protein, a mix of ( 11) and ( 10) is applied.The excess CPP facilitates membrane permeation, allowing Our live cell experiments have corroborated our in vitro findings, providing us a visual proof of P1dC-mediated HBc aggregation in a living cell.Thus, the peptide dimer causes an aggregation that resembles the HAP induced aggregation of the core protein and, like CAMs, can be expected to have the potential to disrupt the HBV life cycle.

Cryo-EM confirms peptide-induced HBc aggregation
To affirm the capsid-aggregation property of our peptide dimers, we incubated solubilized purified capsid-like particles (CLPs, spherical capsid-like HBc multimers purified from E. coli) with an excess of SLLGRM-dimer or P1dC, applied them on carbon grids, and imaged them using cryo-EM.The effect of peptide dimers on CLPs was already seen on the microscale Cryo-EM images (Figure 5), with P1dC inducing large protein aggregates with multimicron diameter.The less potent SLLGRM-dimer also induced visible aggregation, although with smaller aggregate size, while geraniol-treated samples showed minimal aggregation and the smallest observed aggregate sizes.In the nanoscale we observed clumped CLPs (Figure 6A,B) and resolved the binding of both peptide dimers to the spike tips (Figure 6C,D, Supplementary Figure 9).The densities corresponding to bound peptide-dimers in both EM-  The asymmetric unit of HBc capsids (T=4) is a tetramer consisting of an A/B-and C/D-dimer which have slightly different 3D-structures.Interestingly, the SLLGRM-dimer binds the A/Bdimer as well as the C/D-dimer, in contrast to the monomeric SLLGRM which binds only to the tip of the C/D-dimer [33] , in line with the multivalent binding and the higher affinity we measured with ITC.
The live cell experiments showed the formation of HBc aggregates upon incubation with P1dC.
Yet, in live cells HBc may exist as a monomer, a dimer, a multimer or a whole capsid, therefore, the observed aggregates were not necessarily formed by whole capsids.The Cryo-EM experiment, however, provides confirmation that the peptide dimers have the capability to interact with complete CLPs, an important feature, that implies that the dimers have the potential to affect intact capsids upon cell infection.

Discussion
In this study, we focused on the HBV core protein, a protein essential for HBV proliferation and virulence.We explored the druggability of two alternative, non-HAP, binding pockets on the HBc ultrastructure and developed synthetic dimers that target these pockets with nanomolar affinity resulting in the aggregation of HBc.
We have hypothesized that compounds sharing structural similarities to farnesyl phosphate or myristic acid could interact with the hydrophobic binding pocket in the center of HBc-dimers, as farnesylation and myristylation mediate viral envelopment and secretion. [41,42]Our findings revealed that geraniol and a geranyl dimer we synthesized indeed bind to this pocket with micromolar affinity, however this interaction strength is sub-optimal for an HBc effector.A recent study described Triton X-100 derived effectors of the central hydrophobic pocket of HBc-dimers.Some of these novel effectors had low-micromolar affinities to HBc.By combining the multimerization approach and rational linker design, these effectors may be evolved to even more potent binders of the hydrophobic pocket, that may have a pharmacological effect on HBV. [48]) (7)  Cryo EM resolved a second effector pocket situated at the tips of capsid spikes at the inner dimer interface of the HBc-dimers.Using our structural knowledge of the capsid, particularly the distances between the spikes, we designed peptide dimers with the ability to simultaneously bind to neighboring spikes on the same capsid or attach to two distinct capsids (Supplementary Figure 5).Our in vitro assays demonstrated that these peptide dimers display a robust affinity ranging from low micromolar to nanomolar levels (Figure 3 and Supplementary Table 2).Specifically, the peptide dimer (P1dC) with nanomolar affinity (KD=420 +/-40 nM) is a promising candidate for a lead molecule for new a new class of CAMs.The peptide dimer, but not its scrambled dimeric counterpart, induced HBc aggregation in live mammalian cells expressing HBc.An effect resembling the aggregation observed after a treatment with the classical CAM HAP.[49] While these results are highly encouraging, application in complex organisms may require an alternative means for delivery, an investigation of HBV proliferation in HBV infection models and the study of immunogenicity and stability.Nevertheless, the insights given in this study on the yet untapped pharmacological potential of the two binding pockets on the capsid surface, and the compounds targeting them, can pave way to the development of new compounds capable of affecting the viral capsids.In contrast to classical CAMs, the peptide dimers have a different mechanism of action and might act synergistically with CAMs or other antivirals.Further biological investigations will illuminate the applicability and potency of compounds targeting the non-HAP binding pockets.

Materials and Methods
Unless otherwise noted, all resins and reagents were purchased from IRIS biotechnologies or Carl Roth and used without further purification.All solvents were HPLC grade.All watersensitive reactions were performed in anhydrous solvents under positive pressure of argon.
Then, regardless of the resin type, Fmoc was removed using 20% piperidine in DMF solution and the resin was washed with DMF and dichloromethane (DCM).After washes the peptide chain was elongated by adding the desired amino acid (AA, 3 eq.)with ethylcyanohydroxyiminoacetate (Oxyma, 3 eq.)and N,N'-diisopropylcarbodiimide (DIC, 3 eq.).
Capping was done with N,N-diisopropylethylamine (DIEA, 50 eq.)and acetic anhydride (50 eq.) in N-methyl-2-pyrrolidone for 30 min.Coupling efficiency was monitored by measuring the absorption of the dibenzofulvene-piperidine adduct after deprotection.The peptide chain was elongated with further identical deprotection-conjugation cycles and after the completion the peptides were cleaved from the resin using a cocktail of 94% trifluoracetic acid (TFA), 3% H2O, 3% Triisopropylsilane (TIPS), for 4 hours at RT.The peptides were precipitated in icecold ether and then purified with HPLC and analyzed by LC-MS, as described below.

Purification and characterization of peptide-based probes
The compounds were purified from the crude reaction mix by reverse phase HPLC using a water acetonitrile gradient with 0.1% formic acid (FA).Semi-preparative HPLC was performed on Shimadzu Prominence equipped with a diode-array detector (DAD) system using a C18 reverse-phase column (Phenomenex Onyx Monolithic HD-C18 100×4.6 mm or Onyx and the corresponding isotherms were fitted using a one site model.The peptide and geraniol dimers are bivalent and have 120 or 240 potential binding sites on CLPs, respectively.The two binding sites of peptide and geraniol dimers are not identical but very similar.This also true, for the binding sites on capsids, so the binding energetics of the dimers are very similar and are best represented by a one site model.All obtained thermodynamic parameters refer to concentrations of monomeric HBc.All ITC experiments were complemented with control experiments where solutions of peptide dimers were titrated into the dialysis buffer.

Turbidity assay
All peptides were dissolved in buffer A (40 mM HEPES, 200 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, pH 7.5) and the capsid solutions were filtered once (centrifugal filter unit Ultrafree MC, pore size of 100 nm, Merck KGaA, Darmstadt, Germany).The concentrations of the P1, P2 and SLLGRM dimers were varied between 0.1-100 µM and the concentration of HBc was kept constant at 10 µM for the P1 and P2 dimer and at 50 µM for the SLLGRM dimer.All experiments were performed using a standard photometer (GENESYS™ UV/VIS spectral photometer, Thermo Fisher Scientific, Hillsboro, OR, USA) at RT and at a wavelength of 350 nm using disposable UV transparent cuvettes (SARSTEDT AG & Co. KG, Sarstedtstraße 1, 51588 Nümbrecht/Germany).

Cryogenic grid preparation of capsids in complex with peptid-dimers
In a plasma cleaner (model PDC-002.Harrick Plasma, Ithaca, NY, USA) holey carbon grids (R1.2/1.3,300 mesh Cu grids, Quantifoil Micro Tools, Jena, Germany) were made hydrophilic by plasma cleaning.This was done at a pressure of 29 Pa for 2 minutes using ambient air as plasma medium at "medium power" of the instrument.Solutions of purified HBc (200 µM) in complex with the P1dC-and the SLLGRM-dimer (each 400 µM) were prepared in buffer A.
After the end of ITC experiment with geraniol and HBc (Figure 2), a sample from the cell of the ITC instrument was retrieved and used for freezing grids.3.5 µl aliquots of each sample were applied onto the grids.For plunge freezing of grids ethane was used as medium (liquefied by liquid nitrogen) with the help of a Vitrobot (mark IV, FEI Company, Hillsboro, OR, USA) using filter papers of Whatman (type 541).The Vitrobot had the following settings: no wait and drain times, 6 s of blot time, blot force of 25 and a nominal humidity of 100 %.The frozen grids were stored in liquid nitrogen for at least one night before being used for image acquisition.

Cryo-EM and image processing
Cryo-EM was done as previously described. [1]Shortly, movies were acquired with the software EPU on a Krios G3 electron microscope equipped with a Falcon III camera (Thermo Fisher Scientific, Hillsboro, OR, USA) in integrating mode at a magnification of 75,000 with an accelerating voltage of 300 kV.The total exposure was 40 e − /Å² and was fractionated over 20 fractions.For HBc CLPs with bound P1dC, 3 movies were acquired per hole and one hole was acquired per stage position.For HBc CLPs with bound SLLGRM dimers or bound geraniol, at each stage position three movies were acquired per hole from the central hole and from the four closest neighboring holes.The different movie positions at the same stage position were centered with image shift.Movies were motion corrected, exposure weighted and averaged with MotionCorr2.Figure S6a shows representative corrected movie averages, which were imported to Relion for further processing.Each image shift position was treated as a different optics group in the subsequent image processing.Image processing was done with Relion 3.1 or Relion 4. As previously described [1] imposing icosahedral symmetry.At the end of the image processing with Relion (for CLPs with bound P1dC or SLLGRM-dimers), particle images were imported into CryoSparc 4.02 and were further refined with none uniform refinement [2] , including global and local CTF refinement and Ewald's sphere correction.Final maps were filtered with deepemhancer, or B-factor sharpened (CryoSPARC "Sharpen" or "relion_postprocess").The resolution of the final maps was estimated by Fourier-Shell-Correlation (FSC=0.143;after gold standard refinement) with "relion_postprocess" (Figure S6).Parameters of the image acquisition and the processing are summarized in table S3.

Modelling of Cryo-EM maps, refinement of PDB files and their validation
For modelling of the EM densities of HBc in complex with the peptide dimers the PDB file 7od6 [3] was used as starting model.This model represents the asymmetric unit of the HBc capsids with T=4 packing.After slight modifications, the PDB model was fitted into the EMmap as a rigid body and refined iteratively by the software packages Coot [4] and Phenix [5] and validated by MolProbidity. [6]The resolution of the density at the tips of the capsids which we attributed to the binding segments of the peptide dimers was low.Therefore, these densities could only be modelled as poly-alanine chains.All figures showing EM-densities with or without the corresponding PDB models were prepared with Chimera. [7]ning Full length wild type (fl wt) HBc (genotype D; strain ayw; GenBank: V01460.1,MQLFHLCLIISCSCPTVQASKLCLGWLWGMDIDPYKEFGATVELSFLPSDFFPSVRDLLDTA SALYREALESPEHCSPHHTALRQAILCWGELMTLATWVGVNLEDPASRDLVVSYVNTNMG LKFRQLLWFHISCLTFGRETVIEYLVSFGVWIRTPPAYRPPNAPILSTLPETTVVRRRGRSPR RRTPSPRRRRSQSPRRRRSQSRESQC) [8] was cloned into the pEGFP-C2 vector (Clontech) using Gibson assembly [9] by replacing the gene sequence coding for eGFP.The vector and insert were amplified by PCR and purified by gel extraction (FastGene Gel/PCR Extraction Kit,

Wide field fluorescence microscopy
The coverslips with the cell samples were inserted in an imaging chamber (Ludin Chamber Type 1, Life Imaging Services) and imaged in PBS.The measurements were taken from distinct samples with a sample size ≥ 2, for each group.A series of images, used to generate the datapoints, were acquired from different regions of the sample, each region having a distinct group of cells.
The samples were imaged on an inverted Leica DMI6000B microscope with a 100x oilimmersion objective (NA 1.49) using a Leica DFC9000 GTC VSC-05760 sCMOS camera (16bit, image pixel size: 130 nm).The 628/40 excitation and 692/40 emission filter was used for DyLight650, 10 images were acquired at a frame rate (exposure time) of 100 ms and constant illumination intensity to ensure comparability.n≥10.

Peptide Microarray-Binding Assay
The microarray slides were blocked for 60 min in 5% (w/v) skimmed milk powder (Carl Roth) phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4).After blocking, the slides were incubated for 15 min with 55 nM (monomer equivalent) of HBc in the blocking buffer, then washed 3× with PBS.HBc was detected with a primary 1:2500 diluted mAb16988 (anti-hepatitis B virus antibody, core antigen, clone C1-5, aa 74-89, MilliporeSigma, Darmstadt, Germany) and a secondary 1:5000 diluted HRP-coupled Anti-mouse antibody (31,430, Invitrogen).The antibodies were applied in blocking buffer for 15 min, with three PBS washes between the antibodies and after applying the secondary antibody.The chemiluminescent readout was obtained using SuperSignal West Femto maximum sensitive substrate (Thermo Scientific GmbH, Schwerte, Germany) with a c400 Azure imaging system (lowest sensitivity, 90 s exposure time).
Binding intensities were quantified with FIJI [11] using the "microarray profile" plugin (OptiNav Inc, Bellevue, WA, USA).The raw grayscale intensities for each position were obtained for the left and right sides of the internal duplicate on each microarray slide, n = 3 arrays in total.

Figure 1 .
Figure 1.HBc binding pockets and mode of action of dimeric binders.A) Left: Close-up view of the three addressable effector sites within HBc-dimers (shown as cartoon model with transparent surface in grey) together with representative ligands shown as stick models: SLLGRM peptide (marine blue, PDB: 7PZN); Geraniol, resolved here (Cyan); heteroaryldihydropyrimidine (HAP, green, PDB: 5WRE).HAP is a representative example of a canonical CAM, that targets a hydrophobic pocket mediating HBc-dimer multimerization, an essential step in capsid assembly.Right: The general architecture of an HBc-dimer is depicted as a cartoon with transparent surface model in grey and the three ligands that target distinct binding pockets are in color.The binding sites of two HBc dimers can be linked by dimeric ligands, here exemplified with the peptide ligand.B) Hypothetical mode of action of HBc aggregation triggered by cross linking the spikes of individual HBc dimers, HBc multimers or the whole capsid.

Figure 2 .
Figure 2. The central hydrophobic pocket of HBc-dimer is targeted by hydrophobic molecules containing isoprene units.A) Structures of different substances used for the ITC and cryo-EM experiments.N-Decyl-beta-D-maltopyranoside (DM, (5)), is a soluble mimetic of myristic acid (4), their hydrocarbon chains contain 10 and 13 methyl units, respectively (cyan).Farnesyl pyrophosphate and geranylgeranyl pyrophosphate have 3 and 4 isoprene units, respectively (blue), but are poorly soluble in aqueous buffers.Their soluble, isoprene-containing mimetic is geraniol(3).Using geranic acid we synthesized Geranyl dimer (6), a dimeric binder forked by a lysine and having a linker of six dioxaoctanoic units.B) Representative ITC heat signatures of geranyl dimer(6), geraniol (3) and DM(5)   with HBc capsids.Heat release is detected upon titration of the ligands to the HBc solution, indicating a binding interaction.4 mM geraniol (3) was titrated into a solution of 210 µM HBc.A solution of 2 mM geranyl dimer (6) was titrated into a solution 200 µM HBc.1.6-2 mM solutions of DM(5) were titrated into solutions with 90, 100 and 150 µM HBc, respectively.The control experiments where geraniol, geranyl dimer and DM were titrated into buffer are depicted in Supplementary Figure7.C) Integrated heat signatures in kcal⋅mol -1 plotted against the molar ratio of titrants to HBc.Binding isotherms (solid lines) were determined using a curve fitting procedure based on a one-site model.Among the ligands, the geranyl dimer has the strongest affinity to HBc, expectedly surpassing the monovalent geraniol by 2-fold.D) Close-up view of the geraniol (cyan) binding site within HBc.EM-density of a hydrophobic pocket of the A/B-dimer (grey) was rendered transparent to visualize the model of the geraniol-HBc complex in ribbon representation.Geraniol and residues (P5, L60, K96 and F97) involved in HBV's envelopment (green) with natural phenotypes are depicted in stick representation.The EM density of geraniol is shown in the zoom-out by a magenta mesh.

Figure 3 .
Figure 3. Dimeric peptide spike binders display strong low micromolar and nanomolar affinity.A) Chemical structures of the dimeric peptides, all contain the core binding sequence -SLLGRM and share the same PEG linker and a lysine as the branching element of the dimer.B) Exemplary ITC thermograms showing the titration heat signature of HBc with dimers.A solution of 1500 µM (7) was titrated into a solution 150 µM HBc.A solution of 125 µM (8) was titrated into a solution 25 µM HBc.A solution of 200 µM (9) was titrated into a solution 25 µM HBc.A solution of 100 µM (10) was titrated into a solution 25 µM HBc.C) The peptide dimers display low micromolar to nanomolar affinity to HBc, the affinity increases with the elongation of the binding sequence.D)

( 10 )
to enter the cell after a brief incubation.Once inside, (10) is separated from the CPP and can interact with the capsids.C) After 1 hour incubation with (10) or scr10 the cells were immediately washed, fixed and labelled with anti HBc mAb16988 and a secondary DyLight650 conjugated antibody.The cells were visualized on wide-field fluorescent microscope with identical conditions and are presented with the same grayscale range.Transfected and untreated cells display diffuse HBc distribution, with clear fluorescence at the nucleus.Transfected cells treated with (10) display bright aggregates, whereas transfected cells treated with (scr10) have similar diffuse labelling as the untreated cells.Non-transfected cells are non-fluorescent.Scale bar 20 µm.
reconstructions have volumes which can accommodate a peptide chain of approximately 6 amino-acid residues (Figure6C,D).

Figure 5 .
Figure 5. Cryo-EM confirms strong capsid aggregation with peptide dimers.Low magnification cryo-EM images of CLPs + P1dC (10) (A), (B) CLPs + SLLGRM-dimer (7) and (C) CLPs + geraniol (3).The micrographs are part of the grid-atlas of the respective data acquisition.Each image shows 4 meshes of the respective grid atlas at a similar ice thickness.For representation, the images were aligned to show a similar orientation of the meshes.CLP aggregates are seen as dark speckles (yellow arrow).The size of the aggregates is largest in P1dCtreated samples, while aggregates are frequent and smaller in samples treated with SLLGRM-dimer.Geraniol treated samples have very few aggregates which are generally smaller than 1 µm.

Figure 6 .
Figure 6.Nanoscale resolution of the dimer binding sites by Cryo-EM.HBc capsids with bound SLLGRMdimer (7) or P1dC (10) imaged by electron cryo-microscopy.(A) and (B) show selected areas of micrographs of CLPs treated with (7) or with (10).One exemplary aggregate of multiple HBc capsids is indicated by an arrow in each micrograph.(C) and (D) show close-ups of the asymmetric unit of HBc capsids with bound SLLGRM dimers or with bound P1dC.Models of a single asymmetric unit consisting of two HBc dimers is fitted into the asymmetric unit.Both maps show a density at the tips of the spikes (arrow) that accounts for approximately six amino acids of the peptide-dimer.The flexible linker between the peptides was not resolved.The position of the symmetry axes of the icosahedral capsid is labelled with numbers in (C).
positional scan data.Obtained raw greyscale values from P1 full positional scan in µSPOT format.Each box corresponds to a single point variation of the P1 peptide sequence (horizontal) as indicated in the first column.The raw intensity values presented here were used for calculating the fold intensity change of each point variation against the wildtype sequence.Data are presented as mean of n=3 microarray slides with standard deviation.