A liver immune rheostat regulates CD8 T cell immunity in chronic HBV infection

Animals and viral infection models

Six-week-old C57BL6/J male mice were purchased from Janvier or Charles River. H-2K^bSIINFEKL-restricted OT-1 TCR-transgenic CD45.1⁺ mice and H-2K^bMGLKFRQL-restricted Cor93 TCR-transgenic CD45.1⁺ mice were purchased from Charles River and were bred under specific pathogen-free conditions at Animal Core Facility of the School of Medicine, TUM. Mice were housed with a 12 h light/12 h dark cycle. The temperature was set to 22 ± 2 °C, humidity to 55 ± 10% and checked daily. Guidelines of the Federation of Laboratory Animal Science Association were implemented for breeding and experiments. Experiments were approved by the District Government of Upper Bavaria (permission nos. ROB-55.2-2532.Vet_02-14-185, ROB-55.2-2532.Vet_02-16-55 and ROB-55.2-2532.Vet_02-18-100).

T cell transfer into mice

A total of 1 × 10² naive (CD44⁻CD62L⁺)H-2K^bSIINFEKL-restricted CD45.1⁺ CD8 T cells or 1 × 10⁴ naive (CD44⁻CD62L⁺)H-2K^bMGLKFRQL-restricted CD45.1⁺ CD8 T were isolated from TCR-transgenic OT-1 or Cor93 mice, respectively, by untouched immunomagnetic separation from spleens with more than 95% purity. Naive T cells were directly injected intravenously in PBS 1 day before infection with recombinant adenoviruses.

Generation of recombinant adenoviral vectors and transduction of hepatocytes in mice

Hepatotropic recombinant adenoviruses were generated as described previously^19,26. Two identical recombinant adenoviruses were generated expressing the cassette GOL, that is, genes for GFP, ovalbumin and luciferase, either under a minimal CMV promoter (resulting in acute resolved infection of hepatocytes) or under an hepatocyte-specific transthyretin promoter (TTR) (resulting in persistent hepatocyte infection) as described¹⁹. Recombinant adenoviruses for transduction of hepatocytes with replication-competent HBV were generated using a 1.3 overlength construct of HBV genome (genotype D) as previously reported⁵⁷. Recombinant adenoviruses were amplified in HEK 293 cells and infectious titres were determined by in vitro infection assays. High-titre adenoviral stocks were aliquoted and kept at −80 °C before use. For in vivo transduction of hepatocytes, recombinant adenoviruses were dissolved in saline immediately after thawing and injected intravenously through the tail vein.

In vivo bioluminescence imaging

In vivo bioluminescence from the expression of luciferase after transduction with recombinant adenoviruses coding for the GOL expression cassette was quantified with the in vivo imaging system IVIS Lumina LT-Series III (PerkinElmer). Mice were anaesthetized using 2.5% isoflurane and received 100 mg kg⁻¹ of body weight d-luciferin-K-salt (PJK) as substrate for luciferase. Regions of interest were defined in the upper right quadrant of mice and photons detected in this region were quantified. System calibration of the IVIS Lumina LT III performed before every experiment assured comparability of results.

Quantification of liver damage

Serum alanine transaminase was measured from peripheral blood of mice using the Reflotron Plus system (Roche Diagnostics).

Quantification of HBV replication in Ad–HBV and AAV–HBV-transduced hepatocytes

HBeAg titres were determined in peripheral blood using an Architect platform and the HBeAg reagent kit (6C32-27) with HBeAg quantitative calibrators (7P24-01, Abbott Laboratories).

Histology and immunohistochemistry

Histology and immunohistochemistry of liver tissue sections were performed as described previously¹⁹. In brief, tissues were fixed in 4% formalin and paraffin embedded. For haematoxylin/eosin staining or immunohistochemistry, 2 µm sections were made. Haematoxylin/eosin staining was performed according to standard protocols. Tissue sections were stained with anti-HBc and anti-GFP (1.5 µg ml⁻¹ polyclonal anti-HBcAg (Origene); 0.4 ng µl⁻¹ of polyclonal anti-GFP (Fitzgerald)) with a Leica Biosystems Bond MAX (Leica) and binding was visualized with DAB (Dako) as a brown precipitate. Slides were scanned with an Aperio System and analysed with Aperio Image Scope v.12.4.0 software (Leica) and QuPath v.0.2.3 (ref. ⁵⁸).

Isolation and culture of primary mouse cells

Splenocyte isolation

Spleens were passed through a 100 µm cell strainer, red blood cells were lysed with ammonium-chloride-potassium lysing buffer for 2 min and splenocytes were used for further experiments.

Isolation of liver-associated lymphocytes

Before excision, livers were perfused with PBS through the portal vein. Liver tissue was passed through 100 µm mesh cell strainers and digested with 125 µg ml⁻¹ of collagenase type II (Worthington) in GBSS (PAN Biotech) for 10 min at 37 °C. For enrichment of liver-associated lymphocytes, a density gradient centrifugation with 40%/80% Percoll (GE Healthcare) was performed at 1,440g for 20 min.

Isolation of primary mouse hepatocytes

Livers were perfused with 0.12 U ml⁻¹ of collagenase (SERVA) at 6 ml min⁻¹ for 8 min through the portal vein. Livers were then removed, mechanically disrupted and passed through a 300 µm cell strainer. Liver cell suspensions were filtered through a 100 µm mesh and pelletized at 50g for 2 min. Hepatocytes were purified by density gradient centrifugation with 50%/80% Percoll (GE Healthcare) at 600g for 20 min. For cytotoxicity assays, 10,000 hepatocytes per well were seeded on 96-well E-plates (ACEA Biosciences) coated with 0.02% collagenR (SERVA). Cell attachment was achieved in supplemented William’s E medium (PAN Biotech, 200 mM glutamine (Thermo Fisher Scientific), 1 M Hepes pH 7.4, 10⁴ U ml⁻¹ of penicillin/streptomycin, 50 mg ml⁻¹ of gentamycin (Merck), 0.005 n ml^–1 of insulin (INSUMAN rapid, Sanofi), 1.6% DMSO (Merck) and 10% FBS (PAN Biotech). Attached cells were cultivated in supplemented William’s E medium (as above) containing 1% FBS.

Isolation of primary mouse liver sinusoidal endothelial cells

Non-parenchymal liver cells were isolated from mouse livers after portal vein perfusion with collagenase in Gey’s balanced salt solution, followed by in vitro digestion with collagenase in a rotatory water bath at 37 °C and density gradient centrifugation. LSECs were then obtained by immunomagnetic separation using anti-CD146 coated microbeads (Miltenyi), reaching a purity of 95% or more, as previously described^59,60,61. To investigate the transfer of molecules from LSECs to T cells, LSECs were labelled with 10 µM CFSE (Invitrogen). LSECs were activated with 50 µg ml⁻¹ of IFNγ (Miltenyi) to increase adhesion before coculture experiments. To analyse cAMP signalling, LSECs were treated with 1 µM Celecoxib (Cayman Chemical) before coculture.

Ex vivo treatments of CD8 T cells

CD8 T cells were isolated from the liver or spleen and cultured in RPMI 1640 medium (GIBCO) supplemented with 10% FCS, 1% l-glutamine (200 mM), 1% penicillin/streptomycin (5,000 U ml⁻¹), 50 µM 2-mercaptoethanol. For ex vivo stimulation and intracellular cytokine staining, cells were stimulated with 10 nM recombinant SIINFEKL peptide (peptides&elephants), HB_core peptide MGLKFRQL (peptides&elephants) or 1× eBioscience cell stimulation cocktail (Thermo Fisher Scientific) with 3 µg ml⁻¹ of Brefeldin A (Invitrogen). To analyse cAMP signalling, T cells were incubated for 1 h with the adenylyl cyclase agonist Fsk (25 µM, Sigma-Aldrich), the PKA agonist Sp-8br-cAMPS (250 µM, Cayman Chemical), the EPAC agonist 8-pCPT-2′-O-Me-cAMP (30 µM, Tocris) or the adenosine A2A receptor agonist CGS21680 (100 nM, Tocris) solved in DMSO (Sigma-Aldrich).

CD8 T cells were cocultured at a 1:1 ratio with primary mouse LSECs or dendritic cells and were then separated from the antigen-presenting cells and directly analysed or restimulated with cognate peptide for 18 h. To analyse cAMP signalling in cocultures, T cells were pretreated with the EPAC inhibitor ESI-09 (10 µM, Tocris), PKA antagonist Rp-8-bromo-cAMPS (1 mM, Cayman Chemical), adenylyl cyclase antagonist MDL-12330A (100 µM, Tocris), A2AR antagonist SCH58261 (100 nM, Tocris) and PTPN22 inhibitor PTPN22-IN-1 (1.4 µM, MedChemExpress).

Isolation and culture of patient-derived cells and assessment of clinical parameters

Clinical diagnostics

Blood samples of participants with viral hepatitis were recruited at the Department of Medicine II of the University Hospital Freiburg, Germany, and at the Department of Gastroenterology, Hepatology and Endocrinology. Peripheral blood and liver fine-needle aspirations were collected from participants living with chronic hepatitis B at the Erasmus MC University Medical Center (Rotterdam, The Netherlands), the Toronto General Hospital (Toronto, Canada) and the Massachusetts General Hospital (Boston, United States). All participants provided written informed consent. This study was approved by institutional review boards at all three sites and was conducted in accordance with the declarations of Helsinki and Istanbul. Individuals were classified into different clinical phases of chronic or resolved HBV infection according to the European Association for the Study of the Liver guideline of 2017, which considers the presence of HBeAg, HBV DNA concentrations, transaminase concentrations (alanine transaminase and aspartate transaminase) and the presence or absence of liver inflammation⁶². HBeAg, serum HBV DNA and aspartate transaminase/alanine transaminase values were determined as part of the clinical diagnostics at the University Hospital Freiburg, Germany. Confirmation of HLA-A*02:01 was performed by HLA-typing by next-generation sequencing on a MiSeq system using commercially available primers (GenDx). Written informed consent was obtained from all participants before blood donation. The study was conducted according to federal guidelines, local ethics committee regulations of Albert-Ludwigs-Universität, Freiburg, Germany (no. 474/14) and the Declaration of Helsinki (1975).

Peripheral blood mononuclear cell isolation from patients

Venous blood samples were collected in EDTA-coated tubes. Peripheral blood mononuclear cells were isolated by density gradient centrifugation using lymphocyte separation medium (PAN Biotech). Isolated peripheral blood mononuclear cells were resuspended in RPMI 1640 medium supplemented with 10% FCS, 1% penicillin/streptomycin and 1.5% 1 M HEPES buffer (Thermo Fisher) and stored at −80 °C until used. Frozen peripheral blood mononuclear cells were thawed in complete medium (RPMI 1640 supplemented with 10% FCS, 1% penicillin/streptomycin and 1.5% 1 M HEPES buffer (ThermoFisher) containing 50 U ml⁻¹ of benzonase (Sigma).

Magnetic bead-based enrichment of \({{\bf{HBV}}}_{{{\bf{core}}}_{{\bf{18}}}}\) CD8 T cells from patients

A total of 1 × 10⁷ to 2 × 10⁷ peripheral blood mononuclear cells were incubated for 30 min with PE-coupled peptide-loaded HLA class I multimers. Enrichment was then performed with anti-PE beads using magnetic-activated cell sorting technology (Miltenyi) according to the manufacturer’s instructions. Enriched \({{\rm{HBV}}}_{{{\rm{core}}}_{18}}\)-specific CD8 T cells were subsequently used for transcriptome analysis.

Antibodies and multimers used for cell characterization by flow cytometry

Cell staining for flow cytometry was performed at 4 °C for 30 min. The following antibodies (clones, dilution, catalogue number) were used for staining of mouse cells: anti-CD8 (53-6.7, 1:250, 100752), anti-CD45.1 (A20, 1:200,110722, 110704 and 110748), anti-CXCR6 (SA051D1, 1:200, 151117, 151104, 151108, 151109 and 151115), anti-CX3CR1 (SA011F11, 1:200,149016, 149004 and 149006), anti-CD44 (IM7, 1:200, 103036), anti-CD69 (H1.2F3, 1:100, 104503), anti-TIM-3 (B8.2C12, 1:200, 134008), anti-TIGIT (1G9, 1:200, 142111), anti-IFNγ (XMG1.2, 1:200, 505808), anti-CD19 (1D3, 1:200, 152404), anti-CD335 (29A1.4, 1:200, 137606), anti-Lck pY394 (A18002D, 1:100, 933104), CD39 (Duha59, 1:200, 143805), anti-CD45.2 (104, 1:200, 109805), anti-CD3 (17A2, 1:200, 100217), anti-NK1.1 (PK136, 1:100, 108747), anti-CD4 (GK1.5, 1:200, 100449), anti-CD49a (HMa1, 1:200, 142606), all Biolegend, and anti-CD69 (H1.2F3, 1:100, 63-069-82), anti-PD-1 (J43, 1:200, 46-9985-82), anti-LAG-3 (eBioC9B7W, 1:200, 406-2239-42 and 12-2231-82), anti-TIM-3 (B8.2C12, 1:200, 12-2231-82) anti-TOX (TXRX10, 1:100, 12-6502-82), anti-granzyme B (GB11, 1:200, GRB04 and GRB05), anti-TNF (MP6-XT22, 1:200, 25-7321-82), anti-4-1BB (17B5, 1:100, 48-1371-82), anti-CD25 (PC61.5, 1:200, 48-0251-82), anti-Akt pS473 (SDRNR, 1:100, 25-9715-42), CD73 (TY/11.8, 1:200, 48-0731-82) (all Thermo Fisher Scientific) and anti-pPKA (47/PKA; BD Biosciences, 1:5,560205). MHC class I H-2K^bSIINFEKL-restricted or H-2K^bMGLKFRQL -restricted streptamers⁶³ were provided by D. Busch (Institute of Microbiology, TUM). For labelling of antigen-specific CD8 T cells, 0.4 µg of peptide-loaded streptamer per sample was incubated with 0.4 µl of Strep-Tactin-PE/APC (IBA Lifesciences) in PBS for 30 min on ice before incubation with cell suspensions. To exclude dead cells, fixable viability dye eFluor780 (Invitrogen) was included in the staining panels. For intracellular staining of cytokines, intracellular fixation buffer (Invitrogen) was used according to the manufacturer’s instructions. Staining of GzmB and TOX was performed in combination with Foxp3/transcription factor staining buffer set (Thermo Fisher Scientific) according to the manufacturer’s instructions. For staining of pPKA, cells were fixed in IC fixation buffer (Invitrogen) for 30 min and permeabilized with ice-cold methanol for 30 min before staining.

For staining of human cells, the following antibodies (clones, dilution, catalogue number, lot number) were used: anti-CD14 (61D3, 1:100, A15453, 2406638,), anti-CD19 (HIB19, 1:100, 17-0199-42, 2472560) (all eBioscience), anti-CD45RA (HI100, 1:200, 304178, 2327528), anti-CCR7 (G043H7, 1:20, 353244, B347205) (all Biolegend), anti-CD8 (RPA-T8, 1:200, 563795, 9346411) and anti-GZMB (GB11, 1:100, 563388, 3317967) (all BD Bioscience). Fixable viability dye eFluor 780 (65-086-14, eBioscience) was used for live/dead discrimination. HLA class I epitope-specific tetramers were generated through conjugation of biotinylated peptide/HLA class I monomers with PE-conjugated streptavidin (ProZyme) at a peptide/HLA I:streptavidin molar ratio of 5:1. Of note, targeted epitopes of HB_core-specific CD8 T cells were previously analysed for viral sequence mutations. T cell responses of patients harbouring viral sequence mutations in the targeted epitope were excluded. HLA-A*02:01/\({{\rm{HBV}}}_{{{\rm{core}}}_{18}}\), FLPSDFFPSV peptide was synthesized with standard Fmoc chemistry and a purity of more than 70% (Genaxxon).

Flow cytometry and cell sorting

Multicolour flow cytometry data were acquired on a Sony SP6800 spectral analyser (Sony Biotechnology) or a CytoFLEX S (Beckman Coulter). Cells were sorted with a Sony SH800 (Sony Biotechnology) or a MoFlo Astrios EQ (Beckman Coulter). Flow cytometry data were analysed with FlowJo software v.10.7.1 and v.10.8.0 (BD Biosciences), GraphPad Prism v.10.0.3 (Graphpad Software), R v.4.0.2 and R cytofkit GUI v.0.99.

Real-time impedance-based cytotoxicity assay

Ex vivo cytotoxicity assays were performed with timelapse xCELLigence-based cell impedance measurement. Primary murine hepatocytes were used as target cells and seeded on a collagenR-coated 96-well E-plate. Sorted CD8 T cells were added to peptide-pulsed or mock-treated primary mouse hepatocytes 24 h after isolation and cell impedance quantified as cell index was recorded with an xCELLigence RTCA MP instrument (ACEA Biosciences) as a measure of antigen-specific CD8 T cell cytotoxicity.

Confocal immunofluorescence imaging of liver tissue

Livers were perfused with 2.5 ml of Antigenfix solution (Diapath) through the portal vein, excised and fixed for 4 h in 1 ml of Antigenfix. Fixed liver lobes were embedded in Tissue-Tek O.C.T. (Sakura Finetek) and frozen at −80 °C, from which 50 µm cryosections were cut with a cryotome (Leica). Liver sections were permeabilized and blocked with 0.1 M Tris (AppliChem) containing 1% BSA, 0.3% Triton X-100 (Gebru Biotechnik), 1% normal mouse serum (Sigma) for 2 h or more. Sections were stained in blocking buffer with anti-CD3 (clone 17A2, 100240, 1:200, Biolegend), anti-CD45.1 (clone A20, 110732, 1:200, Biolegend), anti-CD146 (clone ME-9F1, 130-102-846, 1:100, Miltenyi) and Phalloidin DyLight 488 (21833, 1:100, Thermo Fisher Scientific) or anti-CD3 (clone 17A2, 100240, 1:200, Biolegend), anti-CD45.1 (clone A20, 110732, 1:200, Biolegend), anti-I-A/I-E (MHC class II) (clone M5/114.15.2, 107622, 1:200, Biolegend) and anti-CD103 (goat polyclonal, AF1990, 1:200, R&D Systems) followed by anti-goat IgG (donkey polyclonal, 705-625-147, 1:500, Jackson ImmunoResearch). Tissue sections were mounted with Mowiol and imaged using an inverted TCS SP8 confocal microscope (Leica). Images were analysed with Imaris 9.6 software (Bitplane).

Human liver immunohistochemistry

Human liver samples (formalin-fixed, paraffin embedded, n = 21; ethical approval: 518/19 S) were double-stained by RNAscope (CXCR6) and CD3 (MRQ39, 1:1,500). Briefly, after deparaffinization and standard pretreatment, slides were incubated with RNA probes for CXCR6 (468468, ACD, Bio-Techne), detected with a RNAscope 2.5 Leica Assay-brown (Leica Biosystems) followed by incubation with a primary antibody against CD3 (103R-95, CellMarque) and detection with a Bond Polymer Refine Red Detection Kit (Leica Biosystems) on a Bond Rxm system (Leica Biosystems). All slides were counterstained with haematoxylin, cover slipped and digitalized using an AT2 scanner (Leica Biosystems). The study was conducted according to federal guidelines, local ethics committee regulations of the Technical University of Munich, Germany (no. 518/19 S-SR)

RNA sequencing, bioinformatic and pathway analysis

Sample preparation for RNA-seq of OVA_257–264-specific CD45.1⁺ CD8 T cells

Liver-associated lymphocytes and splenocytes from mice with resolved Ad–CMV–GOL infection were sorted into CD45.1⁺CXCR6⁺CX₃CR1⁻ CD8 and CD45.1⁺CXCR6⁻CX₃CR1⁺ CD8 T cells. CD8 T cells derived from mice with persistent Ad–TTR–GOL infection were sorted into CXCR6⁺CX₃CR1⁻CD45.1⁺ CD8 and CXCR6⁺CX₃CR1⁺CD45.1⁺ CD8 populations. A total of 5,000 cells per sample were collected in 1× TCL lysis buffer (Qiagen) supplemented with 1% (v/v) 2-mercaptoethanol and immediately frozen on dry ice.

Library construction for bulk 3′-sequencing of poly(A)-RNA was performed as described previously⁶⁴. In brief, each sample was produced with a Maxima RT polymerase (Thermo Fisher) with barcoded complementary DNA. Unique molecular identifiers (UMIs) and template switch oligo (TSO) were used to elongate adaptor 5′ ends of the cDNAs. All samples were united and full-length cDNA was amplified with primers. The cDNA was complemented with the Nextera XT kit (Illumina) and 3′-end-fragments and supplemented with P5 and P7 Illumina overhangs. Library was sequenced using NextSeq 500 (Illumina). The UMI tables were spawned for samples and genes using Drop-seq pipeline (https://github.com/broadinstitute/Drop-seq). We annotated the reads using GRCm38 reference genome ENSEMBL annotation release 75. We used DESeq2 R package v.2.1.28.1 (ref. ⁶⁵) to extract the DEGs (log₂ fold-change 1 and P_adj ≤ 0.05). DEGs were visualized as volcano plot using ggplot2 R package v.3.3.2. Principal component analysis was executed using prcomp R function (in stats R package v.3.6.1) and pictured using ggplot 2 and ggrepel R v.0.9.4 packages. See Figs. 1 and 4 and Extended Data Fig. 2.

Sample preparation for RNA-seq of P14 LCMV-specific CD8 T cells

P14 cells were adoptively transferred into C57BL/6 mice and infected one day later with either LCMV clone 13 or LCMV Armstrong. Resident (CD69⁺CD101⁺CXCR6⁺CX3CR1⁻) and effector/effector-memory (CX₃CR1⁺) P14 cells from the liver were sorted at 27 d.p.i. Total RNA was isolated using the RNAdvance Cell v.2 kit (Beckmann-Coulter). Quality and quantity of isolated RNA was analysed with the Bioanalyzer RNA Pico Chip (Agilent). The cDNA synthesis was performed with the Smart-Seq v.4 Ultra Low Input RNA kit (Takara) following the manufacturer’s protocol with 12 cycles of PCR amplification. Input amount was 1 ng of each RNA sample. The cDNA was measured with Bioanalyzer DNA HS Chip (Agilent) and 300 pg of amplified cDNA were used for library preparation with the Nextera XP DNA Library Preparation Kit (Illumina). Libraries were analysed with a Bioanalyzer DNA HS Chip (Agilent) and quantified by quantitative PCR following guidelines from Illumina and using Kapa SYBR master mix (Kapa Biosystems). After the normalization of all libraries to 2 nM, 13 samples each were pooled and sequenced on two single-end runs (1× 100 base pairs, dual-index) on a HiSeq2500 (Illumina) using HiSeq Rapid v.2 chemistry (Illumina). See Extended Data Fig. 2.

Sample preparation for RNA-seq of core_93–100-specific CD45.1⁺ CD8 T cells

Liver-associated lymphocytes and splenocytes from mice with Ad–HBV infection were pregated on (CD19/Ly6G/TER119/CD335)⁻ CD8 T cells and sorted into liver CXCR6⁺CD45.1⁺, liver CD45.1⁺CX₃CR1⁺ CD8 T cells, spleen CD45.1⁺CX₃CR1⁺ CD8 T cells and liver CD45.1⁻ CD8 T cells from resolved infections and liver CD45.1⁺CXCR6⁺ and liver CD45.1⁺CXCR6⁺CX₃CR1⁺ CD8 T cells and liver CD45.1⁻ CD8 T cells from persistent infection. A total 100 CD8 T cells were directly sorted into 96-well plates prepared with 1× reaction buffer consisting of lysis buffer and RNase inhibitor for low input RNA-seq (Takara). Plates were spun down and immediately stored on dry ice or at −80 °C until further processing. Sample plates containing lysed T cells were subjected to cDNA library preparation using the Smart-Seq v.4 Ultra Low Input RNA Kit (Takara) followed by sequencing library preparation using the Nextera XT DNA Library Preparation Kit (Illumina) as per manufacturer’s instructions with minor modifications. Briefly, full-length cDNA was generated by reverse transcription, template-switching reaction and PCR pre-amplification of polyadenylated mRNA as previously described⁶⁶. The cDNA libraries were quantified using the Qubit dsDNA High Sensitivity Kit and quality was assessed on a bioanalyser using DNA high-sensitivity chips (Agilent). Double-stranded cDNA was subjected to fragmentation and PCR-based addition of Illumina barcoded sequencing adaptors at both fragment ends. Sequencing library quantity and quality was assessed as described above. The 50× cycles paired-end sequencing was performed on a NovaSeq 6000 instrument (Illumina) at a targeted read depth of 25 M per sample. See Fig. 2 and Extended Data Fig. 3.

Sample preparation for scRNA-seq of human HB_core-specific CD8 T cells

HBV_core18-specific CD8 T cells were enriched by magnetic bead-based sorting and surface staining was performed. In total, 1,152 live \({{\rm{HBV}}}_{{{\rm{core}}}_{18}}\)-specific CD8 T cells were sorted in 384-well plates (Bio-Rad) containing lysis buffer and mineral oil using FACS Melody Cell Sorter in single-cell sorting mode. Naive CD45RA⁺CCR7⁺ T cells were excluded. After the sorting, the plates were centrifuged for 1 min at 2,200g at 4 °C, snap-frozen in liquid nitrogen and stored at −80 °C until processed. The scRNA-seq was performed using the mCEL-Seq2 protocol, an automated and miniaturized version of CEL-Seq2 on a mosquito nanolitre-scale liquid-handling robot (TTP LabTech)^67,68. Twenty-two libraries with 96 cells each were sequenced per lane on an Illumina HiSeq 3000 sequencing system (pair-end multiplexing run) at a depth of about 130,000–200,000 reads per cell. Sequencing was performed at the sequencing facility of the Max Planck Institute of Immunobiology and Epigenetics (Freiburg, Germany). See Fig. 3 and Extended Data Fig. 4.

scRNA-seq of human HBV-specific CD8 T cells isolated from the liver by fine-needle aspiration

We analysed HBV-specific CD8 T cells from 23 cryopreserved fine-needle liver aspirates (three patients with HBV hepatitis, eight patients with HBe⁻HBV infection and ten patients with HBV functional cure). Cells were thawed and stained with lineage marker antibodies as well as HBV multimers for two distinct HBV-specificities. The live HBV-specific CD8 T cells were sorted in 96-well Armadillo plates (Thermo Fisher Scientific) containing RNA lysis buffer using a BD SORP FACS Aria in index single-cell sorting mode. After sorting, plates were centrifuged and snap-frozen on dry ice. The scRNA-seq was performed at the Broad Institute walk-up sequencing facility (Cambridge, MA, United States) using the Smart-Seq2 protocol and Illumina Nextseq500. After quality control, 977 HBV-specific cells from the 21 liver samples could be analysed using R v.4.1.2 with the Seurat package v.4.3.0. Raw counts were normalized and scaled using the Seurat v.4.3.0 NormalizeData and ScaleData functions, respectively, by dividing feature counts in each cell by the total counts of the cell, applying natural-log transformation to the result using log1p and scaling and centring expression levels for every gene. Subsequently, the HBV-specific cells in each sample were categorized on the basis of whether they did or did not exhibit CXCR6 expression. Gene expression levels were averaged per outcome group for each gene of the CREM signature according to CXCR6 status, followed by visualization in a heatmap. Liver fine-needle aspirations were collected from participants living with chronic hepatitis B at the Erasmus MC University Medical Center (Rotterdam, The Netherlands), the Toronto General Hospital (Toronto, Canada) and the Massachusetts General Hospital (Boston, United States). All participants provided written informed consent. This study was approved by institutional review boards at all three sites and was conducted in accordance with the declarations of Helsinki and Istanbul. See Fig. 3e.

Gene set enrichment and pathway analyses

We performed GSEA on gut, skin and lung tissue-resident memory T cell dataset⁶⁹ as follows: first, we downloaded raw microarray data pertaining from the GEO database (accession ID: GSE47045, tissue-resident memory T cells: gut, lung and skin versus tissue effector-memory cells (spleen)) and extracted DEGs from each comparison using Limma R package v.3.58.1 (ref. ⁷⁰). We used GSEA v.4.0.3 to perform enrichment analysis using DEGs which were ordered according to log₂-fold-changes delivered by DESeq2 v.2.1.28.1. We also performed core signature analysis using GSEA scores as follows. Initially, we extracted genes which contribute to core enrichment from the tissue-residency signature. The gene set associated with Hobit and Blimp was obtained from ref. ⁷¹ (GEO accession ID: GSE70813) and the raw dataset was processed using GREIN DB v.1 (ref. ⁷²). DEGs were determined using the DESeq2 v.2.1.28.1 R package⁶⁵. The gene set related to TCR signalling was obtained from the MsigDB BIOCARTA dataset (https://www.gsea-msigdb.org/gsea/msigdb/). We retrieved the calcium signalling pathway genes from the Molecular Genome Informatics database (http://www.informatics.jax.org/go/term/GO:0019722). Gene sets from Hobit-deficient, Blimp1-deficient cells were matched to DEGs from hepatic CXCR6⁺ CD8 T cells versus spleen CX₃CR1⁺ CD8 T cells. The gene sets dependent on CREM were obtained from ref. ⁷³. To create the gene set for cAMP signalling, gene symbols for all genes encoding adenylyl cylases, phosphodiesterases, PKA regulatory and catalytic subunits, kinase anchoring proteins, EZRIN, EPCA1, EPAC2 and small GTPases were downloaded from the human gene database GeneCards. The PreRanked tool from GSEA v.4.0.3 (ref. ⁷⁴), was used to evaluate the normalized enrichment score and FDR (q < 0.25) was used to measure the statistical significance of normalized enrichment score. See Figs. 1–4 and Extended Data Figs. 2 and 4.

Identification of transcription factors and network analysis

We performed transcription factor network analysis using DEGs CXCR6⁺ CD8 T cells from livers after Ad–CMV–GOL versus CXCR6⁺ CD8 T cells from livers during Ad–TTR–GOL infection. Transcription factors regulated in the transcriptomes were extracted using the transcription factor checkpoint database⁷⁵. Through this analysis we mined seven and two transcription factors from the transcriptome datasets. We evaluated transcription factor–transcription factor network: (1) promoter sequences (−1 kilobases (kb)) of significantly regulated DEGs were downloaded from Eukaryotic promoter database and UCSC (GRCm38/mm10) https://genome.ucsc.edu/cgi-bin/hgTrackUi?db=mm10&c=chrX&g=encode3RenEnhancerEpdNewPromoter and ref. ⁷⁶; (2) we extracted the transcription factor binding sites from the JASPAR core and HOCOMOCO databases^77,78; (3) finally, scanned promoter sequences (−1 kb promoters) of DEGs and transcription factors for binding sites using the custom Python v.3.12 script (https://zenodo.org/records/11040043). Transcription factor networks were generated and visualized in Cytoscape v.3.7.1 (ref. ⁷⁹). To evaluate the hierarchy of transcription factor networks, in (I) and out degrees (O) were computed for each transcription factor and their targets using the igraph R package v.2.0.2 (https://igraph.org/) and hierarchy height (H). \(H=(O-I)/(O+I)\) was calculated as explained previously⁸⁰. Hierarchy height score defined three and two levels of transcription factor–transcription factor network. See Fig. 1.

Analysis of RNA-seq data from P14 LCMV-specific CD8 T cells

Demultiplexing was done with the bcl2fastq software v.2.20.0.422. Reads were processed using snakemake pipelines⁸¹ as described at https://gitlab.lrz.de/ImmunoPhysio/bulkSeqPipe. Reads were filtered using Trimmomatic v.0.36 (ref. ⁸²). STAR v.2.5.3a (ref. ⁸³) was used for mapping to annotation release no. 91 and genome build no. 38 from Mus musculus (Ensembl GRCm38). Multimapped reads were discarded. Read counting was performed using htseq v.0.9.1 (ref. ⁸⁴) and DESeq2 v.1.24.0 (ref. ⁶⁵) was used for differential expression analysis. Genes showing total counts of less than 10 were discarded. Differences were considered significant when absolute log₂ fold-change greater than 1 and P_adj < 0.05. See Extended Data Fig. 2.

Analysis of scRNA-seq data from human HB_core-specific CD8 T cells

For data preprocessing, Fastq files were mapped to the human genome (v.GRCh38), annotated, demultiplexed and counted using the scPipe R package workflow v.1.12.0, R v.3.5.0. Cells with less than 150 UMI counts were filtered out. Cells were clustered using the Louvain method, UMAP projection and DEA were carried out using Seurat v.3.2.0. We scored the cells using the AddModuleScore function from Seurat v.3.2.0 with nbin = 5. For the human CD8 T cells blood signatures we used the signatures from ref. ⁸⁵. We removed signatures with less than ten genes and additionally clusters 11–13, which corresponds to marginal clusters in the Galletti study⁸⁵. Transcription factor activity levels were calculated using the pySCENIC pipeline (v.0.10.10). We selected 10 kb around the gene TSS for motif search. For analysing the CREM signature in circulating human HB_core-specific CD8 T cells, we performed unsupervised clustering of scRNA-seq data by calculating the principal components using the RunPCA function in the Seurat R package v.3.2.0. Next, we integrated four patient datasets using Harmony v.1.2.0. We identified clustering resolution (0.6) using the clustree R package v.0.4.0 (ref. ⁸⁶). Finally, we analysed the CREM signature using the UCELL R package v.1.2.4 (ref. ⁸⁷). See Fig. 3 and Extended Data Fig. 4.

Generation of conditional Icer-deficient mice

The genomic region encompassing the ICER-specific exon as well as the alternative promoters driving expression of ICER and smICER, respectively, was flanked by loxP sites using homologous recombination in mouse ES cells (Extended Data Fig. 8a). The neomycin-resistance cassette was flanked by FRT sites and removed by intercross with Flp-deleter mice, thereby generating the Icer^fl allele (B). An Icer^null allele is generated by Cre-mediated recombination (Extended Data Fig. 8b). Mice bearing the Icer^fl allele were backcrossed to the C57BL/6 background for more than ten generations. For specific deletion of ICER in T cells, Icer^fl/fl mice were intercrossed with Cd4^cre mice. Mice were maintained in a specific pathogen-free facility.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.