A highly selective TCR-mimic antibody reveals unexpected mechanisms of HBV peptide-MHC recognition and previously unknown target biology

ABSTRACT

Curative therapies for chronic hepatitis B virus infection (CHB) are needed, and T-cell redirection is a promising approach, with peptide-MHC complexes (pMHC) being attractive targets. HBV core_18–27 peptide (C18, 10-mer) presented by HLA-A*02:01 (C18-MHC) has two major variants (C18-V or C18-I, differing in the C-terminal residue), both of which are known to be targeted by CD8⁺ T cells in HBV-infected individuals. Through an extensive screening campaign, we identified a highly selective anti-C18-MHC antibody clone MUR35. A MUR35-based T-cell engager (TCE) potently killed HBV-infected hepatocytes but had no activity on uninfected hepatocytes, on other HBV-negative cell types or on host peptides with sequence similarity to C18. Crystal structures of MUR35 bound to both C18-I- and C18-V-MHC revealed a unique binding mode with contacts mediated exclusively by the light chain complementarity-determining regions (CDRs), suggesting that high specificity is achievable without a typical T-cell receptor-like binding mode involving both heavy and light chain CDRs. Although MUR35 exhibits similar binding affinity and structural contacts with C18-V and C18-I, TCE killing was only detected on hepatocytes producing C18-V. To better understand the cause of this discrepancy, we conducted a quantitative proteomics study in an HBV-infected humanized mouse model and found that C18-V was expressed at approximately 300 copies/cell, while C18-I expression was below the limit of detection. Unexpectedly, the proteomics studies revealed that previously unreported 9-mers missing the N-terminal phenylalanine of C18-I and -V were expressed at an average of 508 and 142 copies/cell, respectively, and therefore could be alternative targets for HBV pan genotypic coverage. Our data suggest unexpectedly large differences in antigen presentation efficiency between highly conservative amino acid substitutions in C18 peptide and reveal potentially novel HBV targets for future studies.

Introduction

Worldwide, an estimated 254 million people are infected with chronic hepatitis B (CHB), resulting in approximately 1.1 million deaths every year.¹ Antiviral nucleos(t)ide analogue (NUC) therapy suppresses hepatitis B virus (HBV) replication, but rarely leads to a functional cure due to persistence of viral covalently closed circular DNA (cccDNA) and integrated HBV DNA.² Treatment with pegylated interferon-alpha (pegIFN-α) has substantial adverse events and variable efficacy, with only a limited number of individuals experiencing HBV surface antigen (HBsAg) loss.³ Therefore, more effective therapies are needed to achieve a functional cure of HBV.⁴

HBV-specific CD8⁺ T cells mediate immune clearance of HBV+ hepatocytes,^5–7 but CHB is characterized by low T cell abundance coupled with tolerance and/or exhaustion of the remaining HBV-specific T cells.^8–10 Recent clinical studies have explored the potential of restoring endogenous HBV-specific T cell function in CHB patients using multiple approaches. Two of the major approaches include suppression of HBV antigen production^11–13 and use of immunomodulators such as toll-like receptor agonists,^14,15 a RIG-I/NOD agonist,¹⁶ PD-1/L1 blockade,¹⁷ and therapeutic vaccines.¹⁸ Curative efficacy with these agents has been limited, emphasizing the difficulty of reinvigorating dysfunctional HBV-specific T cells.

Alternatively, HBV cure could be achieved by re-directing non-exhausted, highly functional T cells to eliminate HBV+ hepatocytes. T cell receptor-engineered T cells (TCR-T),¹⁹ chimeric antigen receptor T cells (CAR-T),²⁰ HBsAg-targeted bispecific T cell engagers (TCEs)^21–24 and engineered T cell receptors²⁵ have all shown antiviral efficacy in preclinical studies. TCR-T cells showed substantial antiviral effects in a Phase 1 trial in HBV-associated HCC patients,^19,26 confirming that redirecting T cells to HBV pMHCs can be efficacious. TCEs targeted to HBV pMHCs have a similar mechanism of action to TCR T cells, but are simpler to manufacture, more convenient to administer, less costly, and easier to distribute to a large patient population.

Target selection is critical for a TCE approach to HBV cure. For example, targeting plasma membrane HBsAg would enable broad patient coverage, but could be subject to antigen-mediated antibody sink effects due to the large amount of circulating HBsAg.²⁷ In contrast, targeting viral pMHCs would not be subject to a sink effect because HBV pMHCs have much lower cell surface abundance and are specific to HBV+ cells.²⁸ Discovery of truly specific pMHC targeting antibodies is challenging, however, due to the wide variety of host peptides that can load on the same HLA, some of which may have sequence similarity with the target peptide.^24,29 Use of nonselective pMHC binders for T cell redirection could have fatal consequences in the clinic, as observed with MAGE A3 pMHC directed TCR-T cells in cancer patients.³⁰ This highlights the importance of thorough selectivity assessment of potential TCE leads.

HBcAg_18–27 (C18) is a well-characterized immunodominant 10-mer epitope of HBV.³¹ Two major variants of this peptide having C-terminal valine (C18-V) and isoleucine (C18-I) are predominant in HBV genotypes B/C and genotypes A/D/E,²⁴ respectively, which cumulatively cause > 96% of global HBV infections.³² CD8⁺ T cell responses to both variants have been reported in CHB patients.^33–36 TCR transduced T cells exhibit activity on cells loaded with either variant peptide,³⁷ suggesting both variants are processed and presented in HBV-infected hepatocytes.

Here we report the discovery of a novel HBV core_18–27 peptide-HLA-A*02:01-binding antibody (anti-C18-MHC) which we used to characterize structural mechanisms of selectivity and variations in target peptide biology.

Results

Identification of highly selective anti-C18-MHC antibodies

Highly selective TCR-like antibodies are challenging to discover since only the 9 or 10 amino acids of the peptide differentiate the pMHC of interest from all other off-target pMHCs. To enrich for peptide-selective hits, Trianni mice were immunized with a purified C18 peptide/HLA-A *02:01 complex (C18-MHC). Serum titers against C18-MHC or an off-target HBV peptide E183/HLA-A *02:01 complex (E183-MHC) were measured by ELISA, and mice were selected based on higher C18-MHC specific titer. Bone marrow, lymph nodes and spleen tissues were harvested, and plasma cells were screened using single B-cell microfluidics platform at AbCellera (Vancouver, BC, Canada). B cells that bound specifically to C18-MHC but not the irrelevant pMHC were identified and sequenced. From 600,000 B-cells, 170 unique C18-MHC antibody sequences were grouped into 90 complementarity-determining region (CDR) H3 families based on 75% identity and representative member(s) of each family were expressed in antigen-binding fragment (Fab) format.

Binding affinities of the 90 Fabs to recombinant C18-MHC were measured by Octet RED384 and had K_D values ranging < 1 nM to 100 nM (Figure 1A). On- and off-target cell binding analysis was carried out using T2 cells pulsed with the target C18 peptide or an irrelevant HBV envelope peptide E183,³⁸ as T2 cells provide a noncompetitive environment for presentation of exogenously added peptides due to defective HLA-I presentation of endogenous antigens.³⁹ Of the 90 Fabs tested, 25 bound C18-pulsed T2 cells with EC₅₀ values of 0.3– 69.0 nM (Figure 1B) and did not bind T2 cells pulsed with the irrelevant peptide (Figure 1C). Seventeen unique CDR H3 sequence families from the 25 hits were identified for further characterization.

Figure 1.

Identification of C18-MHC specific antibodies. (A) binding potencies of Fabs determined by Octet biosensors using purified C18-HLA-A*0201 (C18-MHC) complex or (B) flowcytometry of C18 peptide pulsed (50 μg/ml) T2 cells. (C) representative Fab binding profiles on T2 cells that were pulsed with the target C18 peptide or an irrelevant HBsAg peptide E183 (FLLTRILTI). The number of Fabs with similar binding profile noted at the top of each graph. (D) schematic of alanine-scanning. Fab binding to T2 cells that were pulsed with alanine scanning C18 variants (50 μg/ml) was assessed using flowcytometry. The T cell and antibody schematic was created in BioRender https://www.biorender.com/loxldiz. (E) binding heat-map of all Fabs with alanine-scanning variants shown. The heatmap is scaled by area, which is expressed as %binding relative to the canonical C18-pulsed T2 binding. Binding values lower than 33% are highlighted by black boxes. Data represent mean of 2 (B, E) or 3 (A) independent experiments; (C) one of two independent experiments shown.

To determine the binding footprint of the 17 selective Fabs, alanine scan analysis was carried out by pulsing T2 cells with variants of C18-V with single alanine substitution at each non-anchor position (P1 and P3–9), as well as C18-I (Figure 1D). A selective Fab should make multiple peptide contacts, particularly focused on the center of the peptide (P4–7). Among the 17 screened antibodies, 4 were sensitive to Ala substitution at positions 5–9 and another 4 at positions 5–8, suggesting antibody contact with these residues. The remaining 9 Fabs showed sensitivity in only 2–3 residues of C18-V (Figure 1E and Supplementary Figure S1). Notably, 10 of 17 Fabs had comparable binding to cells pulsed with C18-I or C18-V (Figure 1E).

Target specificity of C18-MHC TCEs

Although alanine-scanning reveals antibody binding footprint, reliable detection of off-target pMHC may require a more sensitive functional assay measuring T cell activation. To this end, the 14 lead binders with at least 3 points of contact in alanine scan were reformatted as TCEs to evaluate their ability to selectively induce potent on-target T cell activation. The TCEs were screened for nonspecific activity in killing assays using co-cultures of CD8⁺ T cells and the HBV-negative, HLA-A *02:01+ glioblastoma-derived cell line U-87 MG at an effector to target cell ratio of 1:1. Eleven of the 14 TCEs showed T cell activation on U-87 MG despite the absence of C18 peptide (Figure 2A), suggesting recognition of host pMHC complexes. Three of the TCEs (MUR 35, MUR 14 and HICON06) did not induce T cell activation in unpulsed U-87 MG cells, suggesting selectivity for C18-MHC. To confirm that these three TCEs were capable of on-target killing, U-87 MG cells were pulsed with C18-V and T cell activation was detected by measuring IFN-γ production. MUR35 and MUR14 TCEs had comparable sub-nanomolar potencies, while HICON06 TCE was much less potent (Figure 2B). The observed T cell activation selectivity of MUR14 and MUR35 is consistent with their alanine scan profile, which demonstrates sensitivity at five non-anchor C18 residues (Figure 1F). Interestingly, HICON06 was selective in the U-87 MG killing assay (Figure 2B) despite having only 3 residues (P4–6) selective in alanine scan (Figure 1F) and had a distinct pattern of contact residues from MUR14 and MUR35.

Figure 2.

On- and off-target activities of C18-MHC TCEs. IFN-ɣ levels induced by C18-MHC TCEs at 48 hours after coculture of partial HLA-matched CD8⁺ T cells with: (A) HLA-A*02:01+ U-87 MG cell line without the target C18 peptide, (B) C18-pulsed (5 μg/mL) or unpulsed U-87 MG cell line, or (C) indicated HLA-A*0201+ cell lines that originated from different tissues. Data represent mean ± SEM of 3 replicates. Cells were cocultured at a ratio of 1:1 effector to target cells.

Since host pMHC repertoire may vary by tissue-type,⁴⁰ the specificities of the three antibodies were further evaluated as diabody-Fc TCEs on a panel of 10 HLA-A*02:01+ cell lines from different human tissues. A TCE with known nonspecific activity was used as a positive control, and an anti-respiratory syncytial virus (anti-RSV, palivizumab)⁴¹ CD3 diabody-Fc (anti-RSV x CD3) as a negative control. No off-target T cell activation was observed, suggesting all three TCEs exhibited strong C18-MHC binding selectivity (Figure 2C).

Detailed characterization of antibody specificity for target peptide

The peptide specificity of the pMHC binding domain in each of the three TCEs was characterized using X-scan assays, which systematically evaluate the tolerance for amino acid substitutions at each position within the C18 peptide.⁴² Binding was quantified on T2 cells pulsed with all 19 single-amino acid substitution variants of the C18 peptide. A binding score for each variant was expressed as a percentage relative to the mean fluorescence intensity of wild-type C18-V peptide (Figure 3A). MUR35 and MUR14 had similar specificity profiles overall and appeared highly selective for C18 positions 5–9. HICON06 had a distinct specificity profile characterized by high selectivity at positions 4–6, but relatively permissive binding elsewhere (Figure 3B and Supplementary Figure S2).

Figure 3.

MUR35 TCE has no activity on human peptides partially matching with C18 peptide sequence. (A) schematic of X-scan binding assay. T2 cells were pulsed with X-scan variants of C18 peptide (50 μg/mL) and binding intensities of indicated pMHC binders in diabody-Fc TCE format was detected using flowcytometry at EC₉₀ of TCE concentration. (B) sequence logos and heatmaps showing relative binding to all single amino acid variants of C18. Data expressed as %binding relative to canonical C18 peptide pulsed T2 cell binding. Number of in silico predicted host peptide binders were determined based on X-scan binding scores of each pMHC binder and noted at the bottom of the heatmaps. Black boxes denote binding to cells pulsed with the C18-V peptide. (C) Off-target T cell activation by MUR35 TCE on coculture of CD8⁺ T cells and primary human hepatocytes at 1:1 ratio. Hepatocytes were pulsed (5 μg/mL) with each of the ten in silico predicted human peptides before incubation with diabody-Fc TCE and CD8⁺ T cells. Data represent average of 2 independent experiments (B) or average ± SEM of 2 replicates from one experiment.

The X-scan results were then used to predict potential antibody binding to host derived peptides, further assessing specificity against peptides that may not have been present in previously screened cell lines. All 10-mer peptides in the human proteome were assigned an in-silico score based on the X-scan binding profile of each pMHC binder (see Methods), with peptides having a predicted binding score > 10% of the canonical C18-V peptide classified as potential host off-targets. Ten off-target peptides were predicted for MUR35 TCE, compared to 202 for MUR14 and 408 for HICON06 (Figure 3B). Since MUR35 had the lowest number of predicted off-target binders and the highest on-target potency (Figure 2B), we selected this as our lead. Additional killing assays showed that MUR35 TCE had no activity with the 10 predicted host off-target peptides on T2 peptide pulsed cells, even at high concentrations of both TCE (200 nM) and peptide (5.0 μg/mL), indicating exquisite selectivity for the C18-pMHC viral target (Figure 3C).

MUR35 TCE selectively kills HBV+ hepatocytes

HBV-infected primary human hepatocytes (PHH) represent the most physiologically relevant cells for assessing the potential of TCEs to kill HBV+ hepatocytes in vivo. Accordingly, the killing activity of the TCEs was next evaluated in a CD8⁺ T-cell coculture assay with HLA-A*02:01 PHHs infected with HBV genotype (GT) D (expressing C18-V) at an effector to target ratio of 1:1 (Figure 4A). None of the TCEs mediated killing of uninfected PHHs, as measured by caspase‐cleaved keratin 18 (CK18) release (Figure 4B), or demonstrated T cell activation as measured by IFN-γ, TNF-a, perforin and granzyme B (Figure 4C). In contrast, all three TCEs induced killing of HBV-infected PHHs as well as CD8⁺ T cell activation (Figure 4B,C). To further compare the activity of the three TCEs, killing kinetics was assessed by Incucyte® measurement of caspase 3/7 activity with a fixed concentration of TCE (EC₉₀ in CK18 release assay). MUR14 and MUR35 TCEs induced caspase activation with a similar magnitude and kinetics, whereas activation was slower with HICON06 TCE (Figure 4D). Caspase activation kinetics were consistent with the higher potency of the MUR14 and MUR35 TCEs (EC₅₀ ~1.0 nM) compared with HICON06 TCE (EC₅₀ ~5.0 nM) in the CK18 release assay (Figure 4B). Direct killing of HBV-infected PHH by the MUR35 TCE was also assessed by measuring intracellular HBV core protein by immunocytochemistry. The potency of MUR35 TCE as measured by core staining (EC₅₀ = 0.4 nM) was comparable to that measured by CK18 (EC₅₀ ~1.0 nM), and all detectable HBV core positive cells were eliminated at ≤10 nM (Figure 4E).

Figure 4.

C18-MHC TCEs selectively kill HBV-infected hepatocytes. (A) schematic of killing assay using coculture of bispecific antibody, CD8⁺ T cells and HBV-infected primary human hepatocytes. Hepatocytes were infected with HBV genotype D propagated from HepAD38 cells. The schematic was created in BioRender https://BioRender.com/rxe8w1l. Comparison of 3 lead diabody-Fc TCEs in the hepatocyte killing assay showing release of (B) CK18 levels – a measure of hepatocyte death and (C) cytolytic and non-cytolytic cytokine levels produced by activated T cells. (D) Incucyte® caspase 3/7 induction kinetics on HBV+ hepatocytes with 7.0 nM concentration of indicated diabodies. (E) representative images of HBV core immunostaining at day 5 of hepatocyte killing assay at 7.0 nM TCE concentration. Effect of antibody titration on reduction of HBV core positive cells shown in the plot. An anti-RSV diabody-Fc TCE was used as a negative control. *p < 0.05; ****p < 0.0001; 2-way ANOVA. Data represent mean ± SEM of 3 replicates using a primary human hepatocyte donor.

Structural basis of peptide selectivity

To understand the structural basis of pMHC selectivity, we characterized Fab-C18-HLA-A*02:01 complexes of MUR35 and MUR14 by X-ray crystallography (Figure 5A, Supplementary Table S1, PDB accession code: 9PIX and 9PKC). MUR35 and MUR14 showed closely related Ala-scan/X-scan binding profiles (Figure 1F, Figure 3B) despite differing substantially in their CDRH3 sequences. For both antibodies, the C18 contacts are mediated by light chain CDRs (Figure 5B,C), which have closely related sequences (Figure 5D), whereas MHC contacts are distinct, and mediated largely by the heavy chain CDRs (Figures 5E,F). This light chain-mediated pMHC binding mode differs from canonical antibody-antigen binding mode where antigen specificity is largely driven by antibody CRH3 binding with the cognate antigen.⁴³ In the Fab/C18-MHC complexes, peptide positions 5D-9S make direct contacts with the light chain CDRs (Figure 5B,C) and are relatively intolerant to mutations as determined by X-scan analysis (Figure 3B). In contrast, position 4S is solvent-exposed (Figure 5B,C) and tolerates a broad range of amino acid substitutions.

Figure 5.

MUR35 and MUR14 contact C18 peptide via light chain CDR residues. (A) overview of Fab-C18-HLA-A*02:01 complexes, shown in same orientation. For both Fabs, heavy chain and light chain are shown in red and blue, respectively. C18 peptide residues are shown in yellow, while HLA in gray. (B) MUR14 and (C) MUR35 Fab-C18-HLA interfaces with key interactions (direct or occurring via single ordered water) highlighted. Multiple C18 peptide residues are contacted by MUR14 and MUR35 light chain CDRs, and the two leads form similar sets of interactions. Coloring scheme is same as in A. (D) light and heavy chain CDR sequences of MUR14 and MUR35. Sequence variations between the two pMHC binders are highlighted in gray. CDRH3s of (E) MUR14 and (F) MUR35 form contacts only with HLA. (G) C18 residues P3 and F6 form a buried intramolecular packing contact.

Despite their structural similarities, comparison of MUR14 and MUR35 X-scan profiles reveals that MUR35 is less permissive to C18 residue substitutions, particularly at position S9, correlating with a lower number of host peptides predicted to bind MUR35 than MUR14 (Figure 3B). While the structural basis for these differences is not clear, subtle alterations in contacts at the Ab:pMHC interface between the two leads may contribute to their divergent selectivity profiles. For example, in the MUR14 structure, S9 contacts CDR residues Y92 and N93, while in the MUR35 structure this C18 residue forms an additional interaction with HLA residue K146. MUR35 binding also induces an intramolecular interaction between C18 residues S9 and V10; this interaction is absent in the MUR14 complex structure (Supplementary Figure S3). It is possible that these interactions with/near S9 are more of a constraint for MUR35 than for MUR14, leading to enhanced MUR35 selectivity at this position.

Peptide residue 3P does not make any direct contacts with MUR14/MUR35 CDRs (Figure 5B,C), yet most substitutions at 3P substantially reduced binding in X-scan; only amino acids with small side chains (Ala/Gly/Ser) were well tolerated. Proline at position 3 does not appear to be necessary for HLA loading or stability, as this position is more tolerant to substitution in the context of HICON06 binding (Figure 3B). Closer structural examination revealed that the 3P residue packs closely against residue 6F (Figure 5G), which interacts directly with the MUR14/MUR35 CDRs. It is possible that selectivity of MUR14/MUR35 toward peptide position 3 arises from this intramolecular interaction of 3P and 6F, such that substitution of 3P with a larger amino acid could shift 6F and disrupt MUR14/MUR35 binding.

MUR35 interacts similarly with two predominant C18 peptide variants

Based on the predicted superior selectivity profile and higher on-target potency, we characterized our lead MUR35 TCE for binding and activity against the two major C18 variants. Sequence analysis of 1145 HBV patient serum samples confirmed the predominant presence of C18-I in patients with HBV genotypes B and C, and C18-V in genotypes A and D (Figure 6A). Multiple assays were used to confirm near-identical engagement of MUR35 to both variants, as lack of activity against either variant would considerably reduce the treatable CHB population. MUR35 in the TCE format had identical binding EC₅₀ values on T2 cells pulsed with C18-I or C18-V (Figure 6B), and similar on-rate, off-rate, and K_D on purified pMHC with either variant (Figure 6C).

Figure 6.

MUR35 interacts identically with C18-I-HLA and C18-V-HLA. (A) abundance of C18 variants determined from serum HBV DNA sequences of patients (n = 1145) infected with indicated HBV genotypes. (B) MUR35 binding affinity to T2 cells pulsed with 50 μg/ml of indicated peptide, measured by flowcytometry. (C) Octet binding kinetics of MUR35 in TCE format with indicated pHLAs. Binding affinity, K_on and K_off values are noted in the table. (D) Biochemical stability of purified pHLA complexes. Proportion of intact pHLA complexes was measured using size exclusion chromatography. (E) Alignment of X-ray structures of MUR35/C18-V-HLA (2.5 Å resolution) and MUR35/C18-I-HLA (2.6 Å resolution) complexes show close alignment of peptides and Ab CDRs.

C18-I has ~ 9-fold lower affinity for HLA-A*02:01 than C18-V,^31,44 although pMHC stability has been reported as a better predictor of T-cell immunogenicity than peptide affinity for MHC.⁴⁵ Stability assessment at 37°C for 170 hours showed that purified C18-I-MHC and C18-V-MHC molecules had a similar stability profile with a half-life of about 100 hours (Figure 6D), and comparable melting temperatures (67°C vs 72°C). Structural analysis of MUR35 Fab-pMHC with both C18 variants also showed virtually identical peptide conformations and interactions with light chain CDRs (Figure 6E).

PHH killing activity of MUR35 TCE is restricted to C18-V peptide

To confirm MUR35 TCE killing against C18-I-MHC, HLA-A*02:01+ PHHs were infected with HBV strains encoding C18-V (GT-D AD38) or C18-I (GT-C consensus strain) and CD8⁺ T cell co-culture assays were performed. Unexpectedly, the TCE had no detectable killing on PHHs infected with the C18-I-encoding virus (Figure 7A). IFN-γ release was detectable but reduced by > 100-fold (Figure 7b), suggesting minimal T cell activation. Similar results were obtained on PHHs infected with a series of additional GT-B and GT-C HBV strains encoding C18-I (Figure 7C), despite having higher infection efficiency particularly with GTC viruses relative to GTD (Figure 7D).

Figure 7.

Activity of MUR35 TCE is restricted by C18 peptide variant. (A) CK18 levels and (B) T-cell produced IFN-ɣ levels induced by TCE on primary human hepatocytes after infection with HBV genotype D from HepAD38 cells (GTD) or a patient isolate of genotype C (GTC). (C) MUR35 TCE induced CK18 levels on hepatocytes infected with indicated strains of HBV genotypes. CK18 was measured after 48 hours of CD8⁺ T cell and hepatocyte coculture at 1:1 effector to target ratio. (D) HBeAg levels of culture supernatants from panel C were measured with an MSD kit to verify infection levels with different HBV genotypes after 6 days of PHH infection. (E) CK18 levels induced by a C18-specific cytotoxic T cell clone (CTL) on PHHs infected with indicated HBV genotypes. (F) schematic of mass spectrometry detection of HLA associated HBV peptides in humanized mouse livers. Created in BioRender https://BioRender.com/ghb8n19. (G) copies of indicated HBV peptides per hepatocyte. Each dot represents one individual mouse. Peptide numbers were normalized assuming 6.67 × 10⁴ hepatocytes per mg of liver tissue. Data represent mean ± SEM of 3 replicates using one primary human hepatocyte donor (A-E), or 3 mice per HBV genotype (G). ***p < 0.001; ****p < 0.0001 one-way ANOVA.

To confirm that C18-I-MHC is indeed produced in our infected PHH system, PHHs infected with C18-V or I-encoding viruses were co-cultured with a C18-specific cytotoxic T-lymphocyte (CTL) clone previously demonstrated to have reduced but detectable killing on C18-I versus C18-V peptide pulsed cells.³¹ Consistent with earlier observations, the CTL clone killed PHHs infected with HBV strains encoding both C18 variants, albeit with reduced killing on C18-I (Figure 7E). These results suggested that PHHs likely present both C18-V- and C18-I on HLA-A*02:01, but potentially with different efficiencies.

A human hepatocyte mouse model of HBV infection was used to assess cell surface density of C18-V and C18-I. The uPA-SCID mouse has constitutive expression of urokinase-type plasminogen activator in mouse hepatocytes, causing hepatic injury and permitting the expansion of transplanted human hepatocytes achieving up to 95% reconstitution of the liver⁴⁶ and enabling nearly complete HBV infection.⁴⁷ These mice were reconstituted with HLA-A*02:01+ PHHs (Supplementary Figure S4a) and infected with HBV GT-C (C18-I) or GT-D (C18-V). Liver samples were collected at day 76 when infection reached near steady state (Supplementary Figure S4b,c) and C18-MHC was detected by HLA class I pull-down followed by quantitative mass spectrometry (Figure 7F). Two HBsAg peptides, S208 (ILSPFLPLL)⁴⁸ and E183 (FLLTRILTI), were also quantified as controls for potential differences in infection or replication efficiency between genotypes. Mass spectrometry detected similar levels of these HBsAg peptides in GT-D and GT-C infected livers. In contrast, much larger differences were observed between C18-V and C18-I. Using a targeted proteomic approach, an average of 315 C18-V peptides per human hepatocyte were detected in GT-D infected livers, whereas C18-I was below the limit of detection in livers infected with GT-C virus (Figure 7G), suggesting HLA presentation of C18-I is inefficient.

Given the reported CTL and patient T-cell responses to C18-I,^33–37 it was unexpected that the C18-I peptide was below the limit of detection in GT-C infected livers. Therefore, we performed a non-quantitative global proteomic search in the same liver samples to confirm the results. This analysis also failed to detect C18-I, confirming the quantitative data. However, the non-quantitative analysis detected 9-mer core peptides LPSDFFPSI and LPSDFFPSV that overlap C18 but lack the initial F residue. Subsequent quantitative analysis detected over 100 copies per cell of the 9-mer peptides in both GT-C- and GT-D-infected livers (Supplementary Figure S5), indicating that a highly conserved and well-presented core peptide may exist that could serve as an alternative target for future TCE development efforts.

Discussion

Directed elimination of HBV positive cells is a promising approach for curative CHB therapies. Here we describe the discovery, engineering, and proof-of-concept results of a novel T-cell engager antibody targeting HBV core pMHC. MUR35 TCE showed high target specificity despite a non-canonical binding mode, and structural evaluation revealed insights into the molecular basis of peptide selectivity. It bound similarly to pMHCs with both of the two major target peptide variants, C18-V and C18-I, but could only direct selective killing of HBV-infected hepatocytes expressing C18-V. Quantitative proteomic analysis of HBV-infected humanized mouse livers suggested that this may be due to unexpectedly large differences in pMHC density between the variants.

TCR transduced autologous T cells (TCR-T) to HBV E183 (HBsAg-derived peptide, also called S20) pMHC have shown encouraging antiviral efficacy in HBV-positive hepatocellular carcinoma patients.^19,49 TCE antibodies have a similar mechanism of action to TCR-T, but avoid the costly and complicated manufacturing processes of cell therapy^50,51 and typically have lower levels of adverse events.^52,53 A single dose of the TCR-T SCG101 induced ≥ Grade 3 adverse events in a considerable proportion of treated patients,²⁶ which was most likely due to rapid killing kinetics of HBV positive hepatocytes and tumor cells. Compared to TCR-T cells for which killing is uncontrolled, TCEs could be more amenable to dose escalation, modulating the on-target killing rate and minimizing adverse events. A bispecific fusion protein (ImmTAV) consisting of an affinity-matured HBsAg-specific TCR and an anti-CD3 single-chain variable fragment²⁵ is also being evaluated in CHB patients. Initial results from three CHB patients who received a single low dose of ImmTAV (IMC-I190V) showed acceptable safety,⁵⁴ but higher doses remain to be tested. Furthermore, TCRs have evolved as membrane-associated proteins and present challenges in the drug development process as soluble proteins.⁵⁵ In contrast, TCR-mimic (TCRm) antibodies are an alternative approach to circumvent the complexities associated with soluble TCR development. Antibodies are generally thermostable, have higher native affinities, and can be manufactured using traditional processes.

The discovery and development of TCRm antibodies has historically been challenging because most pMHC antibodies emerging from discovery campaigns lack target specificity⁵⁶ and so are unsuitable for clinical use. The limited surface area of the peptide-MHC interface and possible sequence similarities with host peptides make cross-reactivity a common issue.^24,29 Additionally, both TCRs and TCRm antibodies require sufficient affinity to mediate T cell activation at the low to ultra-low target density observed for pMHC.^57,58 Identifying leads meeting required affinity and selectivity criteria is therefore a considerable challenge, and as such there have been very limited discovery of HBV pMHC-targeted TCEs²⁴ and none have been evaluated in patients. We used a combination of primary screening on a large number of B cells and subsequent assays of increasing stringency to discover a potent, highly selective anti-C18-MHC antibody MUR35. When reformatted as a TCE, MUR35 had sub-nanomolar on-target killing potency on HBV-infected hepatocytes and no detectable off target activity on predicted host pMHCs, a broad range of A* 02:01+ target negative cell lines, or uninfected primary human hepatocytes.

The high selectivity of MUR35 derives from a combination of non-canonical structural features. Direct interactions between the target peptide and antibody⁵⁹ or TCR⁶⁰ drive specificity and typically involve interaction with CDRH3, which is the site of greatest antibody CDR diversity.^43,61 In the case of MUR14/MUR35, all C18 peptide contacts are light chain mediated. Given this finding, we carried out a retrospective analysis of CDR L3 families. Applying the same 75% identity threshold used to group CDR H3 families, we identified 27 CDR L3 families. Twenty six of the 90 Fabs screened shared a CDR L3 family with MUR35 or MUR14. Most of the 26 Fabs showed off-target binding in T2 peptide pulse assays or a poor alanine scan profile. Notably, only 2 Fabs that shared both a CDR L3 and CDR H3 family with our leads demonstrated C18-MHC-specific T cell activation. This likely reflects the cooperative role of both light chain and heavy chain CDRs in positioning the antibody on the C18-MHC complex for optimal peptide specificity. Additionally, X-scan analyses of both MUR35 and MUR14 revealed little tolerance for mutations at 3P despite the crystal structures of both leads showing no direct contact between this residue and antibody CDRs. The apparent paradox is heightened by the fact that there was almost no selectivity at the adjacent 4S position, despite this residue being more proximal to the antibody CDRs (Figure 5B,C). Further structural analysis suggested an intramolecular interaction between 3P and 6F (Figure 5G) that appears dispensable for peptide loading to the HLA, but important for MUR35 and MUR14 binding. The residue 6F lies near the center of the C18 contact region for both antibodies (Figure 5B,C) and binding is highly sensitive to its substitution with most amino acids (Figure 3B). Changes to 3P that result in structural re-positioning of 6F could therefore be expected to impair antibody-HLA interaction, yielding a unique structural mechanism for achieving high peptide specificity at a residue not directly contacted by any CDR.

Another interesting facet of MUR14 and MUR35 is the pairing of similar light chain CDR interactions with very different CDRH3 interactions with the HLA, resulting in qualitatively similar, yet quantitatively different, X-scan profiles. The pattern of X-scan selectivity was similar overall, exhibiting selectivity at P3 and P5–9. But when X-scan results were used for computational identification of host peptides that may potentially bind to either antibody, the number of predicted host peptides was over 20-fold higher for MUR14 than MUR35. Upon closer inspection, the X-scan for MUR35 showed quantitatively higher selectivity than MUR14 to some substituted peptides, with C18 positions 8 and particularly 9 showing the greatest divergence between the two antibodies (Supplementary Figure S2). This divergent selectivity prediction could also be related to slightly differing antibody-peptide-MHC contact interfaces between MUR35 versus MUR14. For example, in the complex with MUR14, HLA residue K146 interacts with C18 residue V10, while for MUR35 this residue is shifted to interact with C18 residues P8 (through water) and S9 (directly). Additionally, MUR35 induces an intramolecular contact between C18 residues S9 and V10 and alters a nearby Ab:HLA contact (Supplementary Figure S3). Collectively, these differences emphasize the degree to which very subtle structural differences may have important implications for selectivity of pMHC-targeted antibodies.

Despite having desirable target specificity and potency, MUR35 TCE was unable to kill cells infected with HBV viruses that encode C18-I. This finding was surprising given that T-cell responses to both C18-I and C18-V were previously detected in CHB patients,^31,33–35 and that MUR35 has nearly identical binding to C18-I and C18-V on both recombinant pMHC- and peptide-pulsed T2 cells. This discrepancy may result from artificially high pMHC density on pulsed T2 cells, which could mask differences in the efficiency of peptide loading that become evident in HBV-infected hepatocytes. Consistent with this notion, mass spectrometry analyses revealed that C18-I-MHC was below the limit of detection in HBV infected humanized mouse livers. While this data suggests proteasomal processing and/or MHC presentation of C18-I may be inefficient in hepatocytes, the cause of this possible inefficiency remains elusive. Preferential flanking sites for cleavage by the 20S proteasome has been shown to affect selection of peptides for MHC-I presentation.⁶² However, sequence analysis of HBV genotype D and C core protein showed no variation in the 5 amino acids flanking the N and C terminus of C18 peptide. However, the 6^th amino acid flanking the N terminus is a Serine for C18-I and Threonine for C18-V, although it is unclear if this variation could affect MHC presentation of C18-I. The very low C18-I density was nevertheless capable of mediating low-level killing with C18-specific CD8⁺ T cell clone, suggesting more efficient targeting of ultra-low copy pMHC by the CTL clone used here than by the MUR35 TCE.

Taken together, this study demonstrates a successful experimental approach for identification of potent and selective HBV pMHC-targeted TCEs and illuminates structural determinants of specificity that are likely to be applicable to other pMHC targeting strategies. At the same time, our results reveal surprisingly large impacts of conservative amino acid substitutions on peptide processing and/or presentation efficiency, emphasizing the challenges in achieving broad coverage of viral strains and genotypes. The serendipitous identification of a previously undescribed but efficiently presented 9-mer core peptide, HBcAg_19–27, may be of help in future efforts to meet that challenge.

Materials and methods

pMHC production

MHC class I heavy chain and hβ2 m preparation, purification, and folding were performed based on published protocols.⁶³ Proteins were produced in E. coli BL21(DE3) cells. Cells were sedimented and lysed by microfluidizer in lysis buffer (50 mM Tris·Cl pH 8.0, 1 mM EDTA, and 1 mM TCEP). Inclusion bodies were harvested by centrifugation, washed with detergent buffer (50 mM Tris·Cl pH 8.0, 1% Triton X 100, 5 mM EDTA, and 1 mM TCEP), followed by lysis buffer and then dissolved in 8 M urea and 20 mM tris pH 8 for 16 h at 4°C. Following high-speed centrifugation at 40,000 × g for 20 min, the soluble denatured protein in the supernatant (concentrations usually about 15 mg/ml for hβ2 m, 7 mg/ml HLA) was stored at − 80°C until use.

To assemble the heavy chain-hβ2 m-peptide complexes, 3 μM of heavy chain and 6 uM of β2 m and 10 μM of peptide as indicated were diluted into 250 ml of refolding buffer (100 mM Tris·Cl pH 8, 0.5 M arginine, 2 mM EDTA, 0.5 mM oxidized glutathione, 5 mM reduced glutathione), and the reaction was incubated at 4°C for 3 days, then centrifuged at 40,000 × g for 30 min to sediment aggregates. Concentration of soluble protein was then determined using the Bradford assay. The refolding mixture was concentrated with a Centricon-20 (Amicon, Beverly, MA) and loaded onto a Superdex S200 column (GE Healthcare) in 20 mM Tris pH 8 100 mM NaCl. The peak corresponding to the MHC complex was analyzed by SDS-Page before being pooled. The presence of all three pMHC complex components was confirmed by mass-spectrometry.

Antibody discovery, production, and purification

Antibody discovery was carried out at AbCellera. Trianni humanized mice (Trianni Mouse^TM) were immunized with recombinant C18-HLA*02:01 protein (C18-MHC). Lymph nodes, spleen, and bone marrow were harvested, and plasma cells were isolated for B-cell microfluidic screening. B-cells producing antibodies that selectively bound fluorescently labeled C18-MHC, but not an unrelated peptide-MHC were sequenced. From 600,000 B-cells screened, 170 unique sequences were identified. The sequences were grouped by CDR H3 at 75% identity to identify clonal families. A total of 90 representative member(s) covering each clonal family were selected for production and screening in downstream assays. During sequencing analysis, a breeding error was discovered where two of the Trianni mice had wild-type mouse HCs and human LCs.

For Fab production, CHO codon optimized fragments encoding the paired variable light (VL) and variable heavy (VH) sequences of anti-C18 pMHC clones were synthesized by GeneArt and cloned in frame into a CMV promoter driven expression vector containing CL and CH1 with a C-terminal HisFLAG tag coding sequence. Proteins were expressed in Expi293™ cells and clarified by centrifugation/filtration. His-containing samples were purified using IMAC-Phytips (Biotage) on HamiltonStar liquid handling platform and buffer-exchanged into 1x phosphate-buffered saline (PBS) (pH7.4).

Effector-silenced diabodies were expressed in Expi293™ (ThermoFisher) cells. Expression supernatants were loaded on a Protein A column (Cytiva MabSelect SuRe) and eluted with 100 mM acetate pH 3.7, which was immediately neutralized with 6% (by volume) of 1 M Tris pH 8.0. The pool was clarified using centrifugation and 0.2 µm filtration before being purified using a CEX column (Cytiva SP Sepharose HP). Some diabodies required further polishing with a Cytiva Superdex 200 column to achieve > 95% purity. All antibody solutions were buffer exchanged to 20 mM histidine pH 5.8 with 9% sucrose and 0.05% Tween-80 for long-term storage.

T2 cell binding assays

T2 cells (174xCEM.T2; CRL-1992) from American Type Culture Collection (ATCC, USA) were pulsed with 5 μg/ml of recombinant hβ₂m (BD Biosciences, USA) and 50 μg/ml of a synthetic C18 peptide variant or an irrelevant peptide E183 (FLLTRILTI) for 4 hours at 37°C. Cells with C18 variant or E183 peptide were separately membrane stained with PKH67 or PKH26 fluorescent dye (Sigma Aldrich, USA) and mixed at a 1:1 ratio. After staining with a viability dye, test antibodies were added and incubated for 1 hour at 4°C followed by incubation with a secondary detection antibody for 30 minutes at 4°C. Stained cells were fixed with 4% paraformaldehyde and acquired on the BD Fortessa flow cytometer. Data were analyzed using FlowJo.

For alanine scan, T2 cells were similarly pulsed with a panel of C18 peptide variants with single alanine substitutions in each variant. X-scan used a panel of 200 C18 variants encompassing every single amino acid substitution at each position. Antibody binding (mean fluorescence intensity) to variant peptide-pulsed cells was expressed as a percentage or number score relative to binding to wild-type C18 peptide pulsed cells at EC₉₀ antibody concentration. The antibody binding potential with all human decapeptides was determined after query of a local copy of the UniProtKB/Swiss-Prot database with splice variants. Each human peptide’s anti-C18-MHC binding score was calculated by multiplying the numeric binding scores of all amino acids in the decamer. For this calculation, the X-scan binding score of the C18 variant matching the amino acid of the human peptide at a given position was utilized. For example, if position 1 of a human peptide contained proline, the X-scan binding score of the C18 variant with proline at this position was assigned for this amino acid in human peptide. The naïve probability of a human peptide binding to C18-MHC was defined as positive if the combined binding score of all amino acids in that peptide was > 0.1, corresponding to > 10% binding relative to wild-type C18.

BioLayer interferometry immunosorbent assay

Samples or buffer were dispensed into polypropylene 384-well black flat-bottom plates (Greiner, Germany) and all measurements were performed at 30°C with agitation at 1000 rpm. Octet® Streptavidin (SA) Biosensors (Pall ForteBio, CA) were used to capture biotinylated C18-MHC complex. For kinetic measurements biosensors were first pre-wetted with assay buffer to remove protective sucrose coating followed by the recommended regeneration conditions for standardization of biosensor surface. Biotinylated C18-MHC protein was immobilized on the SA sensor for 900 seconds. The pMHC loaded biosensors were then blocked with 20 μg/ml biocytin (Sigma) for 200 seconds and then dipped into assay buffer containing wells for 200 seconds to remove any nonspecific protein or unbound pMHC protein. C18-MHC loaded and biocytin blocked biosensors were then transferred into fresh assay buffer for 200 seconds to collect a baseline read. Kinetic measurements for anti-pMHC C18 Fab and bispecific antibody binding were performed by dipping the pMHC-coated biosensors into wells containing multiple concentrations of Fab or antibody (0–300 nM) for 100 seconds followed by a 300-second dissociation time by transferring the biosensors into assay buffer-containing wells. The pMHC loaded sensors were then regenerated for the next kinetic measurement. Regeneration conditions consisted of 3 cycles of 5 seconds of dipping the biosensors in Regeneration Buffer (10 mM Glycine HCl pH 1.5) followed by Assay Buffer (10 mM sodium phosphate pH 7.4, 140 mM sodium chloride, 0.005% Tween20 with 0.2% bovine serum albumin). All sensorgrams were referenced for buffer effects and fitted to a one-site binding model using Octet Data Analysis Software V.11 (Pall ForteBio, CA) generating affinity values for association (k_on), dissociation (k_off) rate constants, and the equilibrium dissociation constant (K_D).

Killing assays, Incucyte, and immunocytochemistry

Peptide pulse assays. HLA-A*02:01-positive primary human hepatocytes (PHH) or U-87 MG cells were plated on 96-well collagen coated plates a day prior to pulsing with 5 ug/ml of recombinant beta 2 microglobulin (BD Biosciences, USA) and 5 μg/ml of on-C18 peptide (FLPSDFFPSV) or host peptides (Genscript, USA) and incubated at 37°C for 24 hours. Cells were subsequently washed of excess peptide and co-cultured with primary human CD8⁺ T cells at a 1:1 effector to target (E:T) ratio and a titration of test antibody at 37°C for 48 hours. Supernatants were collected and activity of cytotoxic CD8⁺ T cells was assessed by the measurement of IFN-γ using an MSD kit (Meso Scale Discovery, USA) and, for PHH, cytokeratin 18 cleavage measured by M30 Apoptosense® CK18 Kit (#P10011, DiaPharma).

Human cell line assays. A panel of 11 HLA-A*02:01+ cell lines that originated from different human tissues (U87-MG, AGS, MCF-7, A375, HEK293, HepG2, U2OS, HCC1599, DMS-79, RT4, and HUVEC) were procured from ATCC. Cells were cultured for 48 hours in 96-well plates (Life Technologies, USA), then co-cultured with CD8⁺ T cells in presence of TCEs and IFN-γ levels measured 48 hours later as above.

Infected hepatocyte assays. PHH were seeded in 96-well cell culture plates (Life Technologies, USA) and infected with consensus or patient-isolated HBV strains of genotypes B-D for 8 days at 500 viral genomes per cell. CD8⁺ T cell co-culture, IFN-γ MSD, and CK18 ELISA were then performed as above. For Incucyte kinetic experiments, CellEvent Caspase 3/7 green dye (#C10423, Invitrogen, USA) was added at the time of TCE and CD8⁺ T cell addition to infected PHHs. The plates were then imaged on the Live Cell Imaging Incucyte® S×5(Sartorius) for 5 days. Data was analyzed with Incucyte software and plotted using GraphPad Prism. Results shown in Figure 7E were derived from experiments where core-specific CTLs were used in place of CD8⁺ T cells and test antibody. HBV core-positive cells were measured by 4% paraformaldehyde fixation of infected PHH, blocking with fetal bovine serum, and staining with HBV core-specific rabbit IgG 366–2⁶⁴ (Gilead Sciences Inc.) and secondary goat anti-rabbit IgG (Alexa Fluor 647) (# A32733, Invitrogen) and 4′,6-diamidino-2-phenylindole (#62248, Thermo Fisher Scientific). Imaging and quantification of HBV core-positive cells were performed using Cellnsight C×7High Content Analysis Platform (Thermo Fisher Scientific).

X-ray crystallography

For Fab/pMHC complex formation, pMHC complexes were mixed with individual Fabs in a 1:1.2 ratio and injected into a Superdex-200 16 60 column equilibrated with a buffer containing 20 mM Tris pH 8, 100 mM NaCl. The UV profiles of the mixtures were superimposed with those of pMHC and Fab to monitor complex formation by comparison of elution volumes. Following chromatographic purification, complexes were concentrated to 10–12 mg/mL total protein concentration for use in crystallization experiments. Crystals of the MUR14 Fab/C18V pMHC complex were grown via sitting-drop vapor diffusion in 96-well plate format, by equilibrating a drop of 0.1 μL protein complex +0.1 μL precipitant (0.2 M NH4F + 20% PEG3350; JCSG Core Suite I condition E5) versus a reservoir of 50 μL precipitant. Crystal size was improved via microseeding of larger drops (1 μL +1 μL versus 500 μL reservoir) in 15-well plate format. Crystals were cryoprotected by briefly transferring to mother liquor supplemented with 20% ethylene glycol prior to plunging in liquid nitrogen. Diffraction data were collected at the Advanced Light Source (sector 5.0.2).

Crystals of MUR35 Fab with C18-V or C18-I pMHCs were grown in a 15-well plate using the same precipitant condition as above, by seeding with microcrystals of the MUR14/C18-V pMHC complex. Crystals of MUR35/C18-V pMHC and MUR35/C18-I pMHC complexes were cryoprotected by briefly transferring to a solution containing 0.15 M NH4F, 15% PEG3350, and 25% ethylene glycol prior to plunging in liquid nitrogen. Diffraction data for crystals of both MUR35 complexes were collected at IMCA-CAT at the Advanced Photon Source.

Diffraction data for all structures were processed using the XDS package,⁶⁵ symmetry assigned using POINTLESS,⁶⁶ and final merging statistics generated in AIMLESS.⁶⁷ For all datasets, five percent of reflections were flagged for calculation of Rfree.⁶⁸ Structures were solved via molecular replacement in PHASER⁶⁹; for the structure of the MUR14/c18V pMHC complex, Fab and pMHC components from an unpublished X-ray structure were used as initial search model. Structures of MUR35 with c18V and c18I pMHCs were solved using the structure of the MUR14/c18V pMHC complex as an initial search model. All structures were refined to convergence in phenix.refine.⁷⁰ The structure of MUR14/c18V pMHC was refined to 2.35 Å resolution, while the structures of MUR35 Fab with c18V and c18I were refined to 2.5 Å and 2.6 Å resolution, respectively. Full data processing and model refinement statistics are included in Supplementary Table S1.

HBV sequencing

Using baseline samples from two clinical trials, GS-US-320–0108 (HBeAg-) and GS-US-320–0110 (HBeAg+) (n = 1145), testing tenofovir alafenamide (TAF) in CHB participants, whole viral genome sequencing was performed using Illumina DNA-sequencing (DNA-seq) with 50 base pair (bp) paired-end reads. Each pair of reads was merged using PEAR v0.9.6⁷¹ to create continuous fragments and then aligned to the hepatitis B virus (HBV) genome using bwa mem v0.7.⁷² Next, samtools v.1.10⁷³ was used to extract aligned reads overlapping the coordinates of the 10-mer C18 peptide. The extracted reads were then trimmed to obtain just the nucleotide sequence of the C18 before the nucleotide sequence was translated into its respective amino acid sequence. The frequencies of C18 variants with c-terminus valine or isoleucine were determined per individual participant (inter-patient analyses) and were stratified by HBV genotype.

pMHC stability testing

The pMHC complexes were dispensed into thin-walled PCR tubes and subjected to thermal stability testing at 37°C for 36 hours using a thermal cycler. Samples were collected at 0.5, 1, 2-, 4-, 12-, and 36-hour time points. After centrifugation at 87,000 × g for 20 min, purity of the pMHC complex in the supernatant was determined using a Superdex 200 Increase 5/150 GL Analytical SEC column, pre-equilibrated with PBS.

Mass spectrometry of HBV-infected humanized mouse liver

HBV-infected human liver-chimeric PXB-mouse® was generated by PhoenixBio Co., Ltd., as previously described.^46,74 Prior to infection with consensus strains of HBV genotypes C or D, mouse livers were reconstituted with human hepatocytes from an HLA-A*0201-positive donor (BioIVT, Westbury, NY). Establishment of stable infection was confirmed by measurement of HBsAg, HBeAg, and HBV DNA in serum samples. At Day 70 post-infection, livers were collected and snap-frozen until further analyses.

HLA-I isolation was performed at Cayman Chemical (Ann Arbor, MI) and mass spectrometry detection at MS Bioworks (Ann Arbor, MI) according to a previously described method.⁷⁵ Briefly, frozen tissue blocks were homogenized in lysis buffer (1.0 mL buffer per 65 mg tissue) containing 0.76% CHAPS (Panreac AppliChem, Darmstadt, Germany) and protease inhibitor cocktail (Complete Mini EDTA-free, Roche). Clarified lysates were subjected to immunoaffinity purification of pMHC using resin conjugated with the human pan HLA-I antibody W6/32.⁷⁶ HLA-I molecules were eluted with 0.1 M acetic acid and 0.1% trifluoroacetic acid. Peptides were concentrated and desalted via Waters μHLB (Waters) solid phase extraction before being analyzed by mass spectrometry at MS Bioworks.

Stable isotope labeled C18-V, C18-I, S208, and E183 HBV peptides were synthesized by Vivitide (Gardner, MA). A mixture of labeled peptides containing 200 fmol of each peptide was added to each liver lysate sample. Peptides (100%) were analyzed by nano LC-MS/MS using a Waters NanoAcquity system interfaced to a ThermoFisher Fusion Lumos mass spectrometer. Peptides were loaded on a trapping column and eluted over a 75 μm analytical column at 350 nL/min; both columns were packed with Luna C18 resin (Phenomenex). A 2 hr gradient was employed. The mass spectrometer was operated using a custom data-dependent method, with MS performed in the Orbitrap at 60,000 FWHM resolution and sequential MS/MS performed using high-resolution CID and EThcD in the Orbitrap at 15,000 FWHM resolution. All MS data were acquired from m/z 300–800. A 3s cycle time was used for all steps. Data were searched using a local copy of PEAKS (Bioinformatics Solutions) and PEAKS output was further processed using Microsoft Excel.

Statistical analyses

GraphPad Prism 10 software was used for data analysis. Statistical significance among different TCEs for induction of cytokines or caspase 3/7 signals were determined by 2-way ANOVA followed by Tukey’s multiple comparison test. Significance among CK18 levels induced by different HBV genotypes was determined by one-way ANOVA with Dunnett’s multiple comparison test. All experimental data were included in statistical analyses.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/19420862.2025.2574103)

Funding Statement

All research was funded by Gilead Sciences.

Disclosure statement

S. Khan, J. Lum, H. Stephenson, P. Bir Kohli, D. Mortenson, D. Ramakrishnan, M. Hung, S. Ding, E. Seto, S. Lu, R. Yen, D. Jin, B. Lee, S. Clancy, N. Schirle Oakdale, N. Novikov, D. Kang, R. Li, D. Pan, R. Dave, E. Lansdon, S. P. Fletcher, N. Thomsen and S. Balsitis are employees of Gilead Sciences, Inc. A. Garg is a former employee of Gilead Sciences, Inc., currently employed by BeOne Medicines USA, Inc. at the time of this submission.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2025.2562998.