Improved Functionality of Exhausted Intrahepatic CXCR5+ CD8+ T Cells Contributes to Chronic Antigen Clearance Upon Immunomodulation

Chronic hepatotropic viral infections are characterized by exhausted CD8+ T cells in the presence of cognate antigen in the liver. The impairment of T cell response limits the control of chronic hepatotropic viruses. Immune-modulatory strategies are attractive options to re-invigorate exhausted T cells. However, in hepatotropic viral infections, the knowledge about immune-modulatory effects on the in-situ regulation of exhausted intrahepatic CD8+ T cells is limited. In this study, we elucidated the functional heterogeneity in the pool of exhausted CD8+ T cells in the liver of mice expressing the model antigen Ova in a fraction of hepatocytes. We found a subpopulation of intrahepatic CXCR5+ Ova-specific CD8+ T cells, which are profoundly cytotoxic, exhibiting efficient metabolic functions as well as improved memory recall and self-maintenance. The intrahepatic Ova-specific CXCR5+ CD8+ T cells are possibly tissue resident cells, which may rely largely on OXPHOS and glycolysis to fuel their cellular processes. Importantly, host conditioning with CpG oligonucleotide reinvigorates and promotes exhausted T cell expansion, facilitating complete antigen eradication. The CpG oligonucleotide-mediated reinvigoration may support resident memory T cell formation and the maintenance of CXCR5+ Ova-specific CD8+ T cells in the liver. These findings suggest that CpG oligodinucleotide may preferentially target CXCR5+ CD8+ T cells for expansion to facilitate the revival of exhausted T cells. Thus, therapeutic strategies aiming to expand CXCR5+ CD8+ T cells might provide a novel approach against chronic liver infection.

INTRODUCTION CD8+ T cells are central to the control of viral infection by producing effector molecules such as granzyme B (GZMB), interferon gamma (IFN-g) and tumor necrosis factor (TNF-a) (1)(2)(3). During acute infection, CD8+ T cells kill antigen expressing cells, facilitating long-lived memory T cell formation (1)(2)(3). Contrarily, in chronic liver infections such as hepatitis B virus (HBV) and hepatitis C virus (HCV) infections, memory T cells fail to develop, due to T cell exhaustion (4,5). Exhausted T cells are marked by poor effector function and coexpression of multiple inhibitory receptors such as PD-1 and TIM-3, as well as deranged epigenetic, metabolic and transcriptional signatures (6)(7)(8). The persistence and duration of antigen exposure are important key features known to drive the development of exhausted T cells (8). Accordingly, immunomodulatory strategies that revive exhausted T cells via boosting the functionality and formation of tissue resident T cells are considered as an attractive therapeutic option to combat chronic liver infection. Indeed, T cell inhibitory receptor PD-1 blocking strategies have emerged that improves exhausted T cell function (8). Also, strategies triggering Toll-like receptor 9 (TLR9) signaling on myeloid cells have been proposed recently to improve exhausted T cell function (9). TLR9 ligand CpG ODN, which induces inflammation and promotes the maturation of myeloid cells has been suggested to boost T cell function in the presence of persistent liver antigen (9).
In models of chronic LCMV infection, exhausted T cells have been shown to be functionally and phenotypically heterogeneous (3). These heterogeneous lineages are made of progenitor CXCR5+ and terminal CXCR5-CD8+ T cell subsets (10)(11)(12)(13). The CXCR5+ CD8+ T cells are stem-like cells which generate terminally exhausted CXCR5-CD8+ T cells. In addition, the CXCR5+ CD8+ T cells are highly responsive to PD-1 blockade (10,11). In contrast, the CXCR5-CD8+ T cells are less responsive to PD-1 therapy and lack the ability to differentiate into CXCR5+ CD8+ T cells. On a molecular level, TCF-1 is critical for longterm maintenance of the CXCR5+ CD8+ T cells (10,14,15). Strikingly, different tissue localization has been described for CXCR5+ and CXCR5-CD8+ T cells. The terminal CXCR5-CD8+ T cells were profound in peripheral tissues and lymphoid organs while CXCR5+ CD8+ T cells are preferentially in lymphoid organsthe site of LCMV infection (10,11). The CXCR5 ligand, CXCL13, is predominantly expressed from cells in the lymphoid organ (16). While the initial insight into exhausted T cells was obtained from chronic LCMV infections, similar phenomena have been reported in patients with chronic HBV and HCV infections. However, most studies evaluating exhausted T cell responses in chronic HBV and HCV patients rely on peripheral T cells profiling, despite viral replication in the liver (17). Thus, these studies do not definitively demonstrate the phenotype of exhausted T cell at the site of antigen exposure. Moreover, the dynamic role CXCR5+ CD8+ T cells play in chronic liver infection is yet to be fully explored.
A deeper understanding of exhausted T cells in chronic liver infection is hindered by the lack of an appropriate infection model that recapitulates HBV and HCV infection. While overexpression of human specific entry factors (18) and elimination of mouse restriction factors (19) recently facilitated HCV infection in mice, it remains to be elucidated if these models can also reflect late infection states, i.e., the establishment of chronic infections and corresponding/accompanying diseases such as liver fibrosis, cirrhosis and eventually liver failure.
The liver has a unique microenvironment and can function as a lymphoid organ, which is distinct from classical lymphoid tissues. Among various immunomodulatory functions, the liver has the propensity to prime functional CD8+ T cell immunity (20)(21)(22). We previously introduced a transgenic mouse model in which intracellular Ovalbumin (Ova) expression is activated by Tamoxifen (Tam) inducible CreERT2 recombinase (23,24), which allows for elucidating T cell responses towards Ova antigen in defined tissues (25)(26)(27)(28). Using OvaXCre mice, in which CreERT2 is controlled by the albumin promoter, we documented that the frequency of Ova expressing hepatocytes can be adjusted by Tam titration (23,(25)(26)(27). Moreover, we demonstrated that adoptive transfer of Ova-specific CD8+ T cells (OT-1 cells) or Ova-specific CD8+ T cell induction by vaccination eliminates low frequencies of Ova expressing hepatocytes. In contrast, when Ova-specific CD8+ T cells are confronted with elevated frequencies of Ova expressing hepatocytes (high antigen conditions), CD8+ T cells lose their cytotoxic activity and rather exhibit an exhausted phenotype, characterized by the expression of multiple exhaustion markers such as Tim-3, PD-1 and Lag-3 (29,30). Thus, the high antigen condition in the OvaXCre model mimics the situation in early stage chronic hepatotropic viral infection, characterized by persistent viral antigen expression that drives exhausted T cell development, in the absence of end-stage liver diseases such as fibrosis, cirrhosis and finally cancer.
Here, using the OvaXCre mouse model, we show that CXCR5+ CD8+ T cells are also generated upon Ova antigen recognition in the liver. Importantly, the intrahepatic Ova-specific CXCR5+ CD8+ T cells preferentially express TCF-1 and Bcl6. The CXCR5+ CD8+ T cells are endowed with potent cytotoxic functions, an extensive proliferation rate as well as efficient mitochondria and nutrient uptake function. In addition, the CXCR5+ Ova-specific CD8+ T cells show profound in vivo and in vitro maintenance, survival, recall responses and self-renewal. Importantly, we show that CpG ODN host conditioning reinvigorates exhausted Ova-specific T cells and promotes the formation of T cells that show markers of liver residency. Accordingly, CpG ODN potentiates increased CXCR5+, TCF-1+ and CD62L+ T cell numbers in the liver. Finally, CpG ODN treatment facilitates the enrichment of GZMB in CXCR5+ and CXCR5-Ova-specific CD8+ T cells, which fosters antigen clearance. In conclusion, CXCR5+ Ova-specific CD8+ T cells may be the subset of cells preferentially targeted by CpG ODN treatment to mediate reinvigoration of exhausted T cells.

Mice and Cell Isolation
The design of the transgenic OvaXCre mice deployed for this study is depicted in Figure S1 and was previously described (23,24,29). The OvaXCre mice were administered with 50 µg Tamoxifen (Tam; Ratiopharm, Ulm, Germany) to induce high Ova antigen, as described elsewhere (24). Tam treated mice were used two weeks post treatment for all experiments. All experiments were performed with mice 6-20 weeks of age. The mice were bred in-house in individually ventilated cages and maintained under specific pathogen-free conditions. OT-1 cells used for adoptive transfer were isolated from the spleen of Thy1.1 OT-1 or dsRED OT-1 mice, as previously described (24). These OT-1 mice were generated on a C57Bl/6 background. Isolation of single cell suspension of liver nonparenchyma cells were performed as described elsewhere (24) but without in situ liver digestion step.

Adoptive OT-1 Transfer and Vaccination of Mice
For adoptive transfer Thy1.1+ OT-1 and dsRED+ OT-1 cells were purified using negative CD8a+ T cells isolation kit (Miltenyi Biotec, Germany), according to the manufacturer's protocol. The adoptive transfer procedure was described previously (24) and in all experiments, 3-5 x 10 6 cells were transferred per high antigen mouse. Vaccination was performed with an adenoviral vector encoding OVA antigen (AdOva). The AdOva vector (30)(31)(32) and the vaccination protocol were previously described (30). In addition, administration of CpG ODN was performed as described elsewhere (30). Briefly, CpG ODN was reconstituted to 0.2mg/ml in 0.9 NaCl and 100ml was i.v injected into recipient mice on days 13, 16, 20, 24, and 28 post OT-1 transfer.

Serum Isolation for CXCL13 ELISA Assay
Serum was isolated from 100 µl of blood collected from each mouse. CXCL13 was quantified using the murine BLC (CXCL13) ELISA kit (Thermo Scientific) according to the manufacturer's protocol. The absorbance was measured with TriStar ELISA plate reader (Berthold Technologies, Germany) at 450 nm. The concentration of serum CXCL13 was interpolated from the standard curve plotted.

RNA Isolation and qRT-PCR
RNA isolation was performed as previously described (24) and 2 µg of total RNA were reverse transcribed using the RevertAid First strand cDNA synthesis kit (ThermoFischer Scientific), according to the manufacturer's protocol. The qRT-PCR for Ova was performed using primer pairs outlined elsewhere (24) and normalized to the expression of albumin.

Flow Cytometry
OT-1 cells were identified based on Thy1.1+ or dsRED expression. Vaccination induced Ova specific T cells were identified by Ova pentamer staining according to a previously published protocol (30). To stain for extracellular markers, Fc receptors on single cell suspension of liver non-parenchyma cells were blocked with anti-mouse CD16/32 (produced in-house) and stained with indicated antibodies (Supplementary  Table), cells were fixed and permeabilized with FoxP3/transcription factor staining kit (eBiosciences), according to the manufacturer's protocol. For intracellular staining (Supplementary Table), single cells were re-stimulated 5 h ex vivo in RPMI complete media (RPMI media, 5% FCS, 1% Pen/Strep, 1% Glutamine, 1mM HEPES) supplemented with PMA/Ionomycin (10 ng/ml PMA and 1µg/ ml Ionomycin) with Brefeldin A for the last 2 h of re-stimulation. Prior to intracellular staining, the stimulated cells were extracellularly stained with antibodies, fixed and permeabilized using Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer's protocol. For CD107a staining, CD107a was reconstituted with liver non-parenchyma cells and ex vivo restimulated with PMA/Ionomycin as previously described. Data were collected on LSRII or LSR Fortessa (BD Biosciences) and analyzed using FlowJo software (v10.6.0; BD Pharmingen). In this study, all flow cytometry gates were defined for the expression of positive markers of interest using fluorescence minus one (FMO) controls on the same cell batch. The gating strategy employed to identify Ova-antigen specific CD8+ T cells is shown in Figure S1C, D.

Mitochondria Staining
Specific mitochondria sensitive dyes were reconstituted in RPMI media as follows: 50  In vitro proliferation and conversion assays were performed on CXCR5+ and CXCR5-CD8+ T cells isolated on D9 post OT-1 transfer into high antigen mice. The sorted cells were labelled with 5µM CFSE and stimulated in vitro with Dynabead anti-CD3/28 (ThermoFischer Scientific), supplemented with 100 U/ ml IL-2 (PeproTech, Germany), based on a protocol previously published for progenitor and terminally exhausted T cells in tumor (33). Similar labelling of CXCR5+ and CXCR5-CD8+ T cells with CFSE was confirmed with naïve OT-1 cells. Approximately, 6 x 10 4 CXCR5+ or CXCR5-CD8+ T cells were stimulated for 48 h at 37°C/5% CO 2 . The cells were washed, stained and analyzed by flow cytometry as previously described.

Statistical Analysis
All graphs were plotted with, and statistical differences were calculated in GraphPad using Mann-Whitney's test. Each data point (dot) on the graph represents a biological replicate and the error bars were computed as: Mean ± SEM. The p values ≤0.05 (*), ≤0.01 (**), or ≤0.001 (***) are representative of significant differences between two groups and non-significant (p >0.05) differences were indicated as 'ns'.

Exhausted Intrahepatic Ova-Specific CD8+ T Cells Harbor CXCR5+ CD8+ T Cells
To investigate the dynamic regulation of exhausted CD8+ T cell subsets in the liver, we employed the OvaXCre mouse model upon induction with Tam, a condition in which about 50% of hepatocytes express the intracellular, non-secreted model antigen Ova (24) (see Figure S1A for the description of the model). In this condition (so-called high antigen) adoptive transfer of antigen specific T cells or vaccination results in functionally impaired T cells (29,30). Upon vaccination with an adenoviral vector encoding ovalbumin (AdOva) or upon adoptive transfer of antigen specific CD8+ T cells, we observed accumulation of Ova-antigen specific T cells in the liver ( Figure  S1B, D). The frequency of Ova-specific CD8+ T cells in the liver increased over time and plateaued to ca. 8%-10% from D14 ( Figure S1D) while the absolute numbers increased until D9 post vaccination and decreased thereafter ( Figure S1E). These Ova-specific CD8+ T cells were previously shown to be functionally impaired, characterized by severe exhaustion (29,30). Of note, the antigen specific CD8+ T cells in the liver failed to eradicate the Ova antigen load ( Figure S1F). These data demonstrate that the presence of persistent antigen in the liver induces functional impairment and subsequent depletion of Ovaspecific CD8+ T cells.
We asked whether Ova antigen recognition in the livera non-lymphoid organwould facilitate the generation and retention of CXCR5+ CD8+ T cells in the liver, as previously demonstrated in the spleen after chronic LCMV infection (10,11). Upon AdOva vaccination of high antigen mice, we profoundly observed~20%-30% of Ova-specific CXCR5+ CD8+ T cells in the liver, which plateaued from D9 ( Figures 1A, B). The absolute number of CXCR5+ CD8+ T cells were relatively maintained for at least 42 days ( Figure 1C). To ascertain the identity of CXCR5+ CD8+ T cells in the liver we assessed Bcl6, a transcription factor that regulates CXCR5 expression. We observed that~42%-66% of CXCR5+ CD8+ T cells expressed Bcl6 and the expression was largely enriched in CXCR5+ CD8+ T cells ( Figures 1D, E, S2A). In contrast, Bcl6 expression in CXCR5and endogenous naïve CD8+ T cells was comparable ( Figures  1D, E, S2A). The presence of antigen specific T cells in liver was accompanied by increased serum concentrations of CXCL13 in AdOva vaccinated mice if compared to AdOva vaccinated wildtype mice ( Figure 1F). Immune histological analysis further showed an increased frequency of CXCL13 expressing cells in liver ( Figure S1G). These data, therefore, suggest that CXCL13 is induced upon antigen recognition in the liver and may foster the retention of CXCR5+ CD8+ T cells in the liver. We confirmed the tendency of CXCR5+ and CXCR5-Ova-specific CD8+ T cells to reside in the liver by assessing the canonical liver residency marker LFA-1 (34) as well as tissue resident markers CXCR6 and CX3CR1 (35)(36)(37). Of note, almost all the CXCR5+ and CXCR5-CD8+ T cell subsets expressed CX3CR1 and CXCR6 ( Figure  S2B), and LFA-1 was present on both subsets of T cells ( Figure  1G, S2C). Instructively, we also observed higher LFA-1 median fluorescence intensity (MFI) on CXCR5+ CD8+ T cells ( Figure  1H). LFA-1 has been shown to contribute to T cell proliferation and activation (38,39). Therefore, we asked if CXCR5+ CD8+ T cells receive higher signaling events through their T cell receptor. In this regard, we quantified Nur77, a transcription factor that is upregulated in response to T cell receptor signaling. We observed a higher frequency of Nur77+ cells in the CXCR5+ compared to CXCR5-CD8+ T cell subset ( Figure 1I, S2D). The assessment of Nur77 MFI showed enriched expression in CXCR5+ CD8+ T cells compared to CXCR5-CD8+ T cells as well as naïve CD8+ T cells ( Figure 1J), indicating higher T cell receptor signaling strength in   CXCR5+ CD8+ T cells. Collectively, these data suggest that CXCR5+ Ova-specific CD8+ T cells may be liver resident, exhibiting increased signaling through the T cell receptor.

Intrahepatic CXCR5+ Ova-Specific CD8+ T Cells Show Enhanced Activation and Memory Properties Which Are Accompanied With Reduced Exhaustion
We asked if the different level of Nur77 expression may be associated with distinct activation, exhaustion and memory characteristics of CXCR5+ and CXCR5-Ova-specific CD8+ T cells. We observed higher frequency and MFI of the activation marker CD69 on CXCR5+ CD8+ T cells on D21 post AdOva vaccination (Figures 2A, B, S3A). Although CXCR5+ and CXCR5-CD8+ T cell subsets expressed the activation marker CD44, CD44 MFI was significantly higher on CXCR5+ CD8+ T cells ( Figure 2C). Likewise, the frequency of the activation marker KLRG-1 was higher in CXCR5+ CD8+ T cells ( Figure  2D, S3B). Further, we probed the memory markers CD62L, CD127, and CCR7 expression on antigen specific CXCR5+ and CXCR5-CD8+ T cell subsets D21 post AdOva vaccination. Higher frequencies of CD127 and CCR7 were present on . In addition, we observed exclusive expression of CD62L on CXCR5+ CD8+ T cells ( Figures 2G, S3F).
To evaluate the extent of exhaustion, we assessed the exhaustion markers CD244, TIM-3, Lag-3, and PD-1 on Ovaspecific CXCR5+ and CXCR5-CD8+ T cells from the liver D21 post AdOva vaccination. We observed higher frequencies of CD244, Lag-3, TIM-3, and PD-1 on CXCR5-CD8+ T cells ( Figures 2H-K, S4A-D). The MFIs of these exhaustion markers were not different between CXCR5+ and CXCR5-CD8+ T cells ( Figure S4E-H). Also, the analysis of multiple exhaustion markers revealed a higher frequency of cells expressing both PD1 and Lag-3 as well as TIM-3 and CD244 in the CXCR5-CD8+ T cell population as opposed to the CXCR5+ CD8+ T cell population ( Figure S4I). Putting together, these data indicate that in the liver microenvironment, the CXCR5+ Ova-specific CD8+ T cell subset possesses enhanced activation and memory properties, which is characterized by lower exhaustion marker expression.

CXCR5 Expression Identifies a Subpopulation of Exhausted Ova-Specific CD8+ T Cells Exhibiting Residual Cytotoxic Function in the Liver
To address if the differential expression of exhaustion markers on intrahepatic T cells potentially affects T cell activity, we analyzed the frequency of effector cytokines IFN-g in Ova-specific CXCR5+ and CXCR5-CD8+ T cells D21 post AdOva vaccination. CXCR5+ CD8+ T cells showed a minor increase in the frequency of IFN-g expression ( Figure 3A) as well a low but significant increase of IFN-g MFI ( Figure S5A). We thus asked if the CXCR5+ Ovaspecific CD8+ T cells also upregulate the cytotoxic molecules GZMB and CD107a. Notably, we quantified a higher frequency of GZMB in CXCR5+ CD8+ T cells ( Figure 3B). Moreover, further analysis showed that almost all CXCR5+ CD8+ T cells expressed CD107a compared to ca. 60% in CXCR5-CD8+ T cells ( Figure  3C, S5B). In addition, the MFI of CD107a was elevated in the CXCR5+ CD8+ T cell compartment ( Figure 3D).
We evaluated the consequences of the enriched expression of GZMB and CD107a in Ova-specific CXCR5+ CD8+ T cells compared to CXCR5-CD8+ T cells by determining the killing capacity of the cells in vitro. To this end, on D14 post transfer of naïve dsRED+ OT-1 cells into high antigen mice, CXCR5+ and CXCR5-OT-1 cells were sorted by flow cytometry, and cocultured with OVA-peptide pulsed EL4 target cells. We injected mice intravenously with the s+16a nanobody, isolated cells from the liver and sorted for CXCR5+ and CXCR5-OT-1 cells by FACS. The s+16a nanobody preserves the vitality and promotes higher recovery of liver resident T cells (40,41). We chose D14 post transfer because of the documented evidence that D3 post OT-1 transfer, the exhaustion trajectory has already been initiated (29). Effector-target cell ratios of 2:1 and 5:1 were exploited for this purpose. We observed enhanced killing of target EL4 cells by CXCR5+ OT-1 cells compared to CXCR5-OT-1 cells in both conditions, albeit to different degrees ( Figure  3E). Taking together, these data demonstrate that Ova-specific CD8+ T cells responding to chronic Ova-antigen in the liver comprise CXCR5+ cells, which are less exhausted and possess enhanced residual cytotoxic functions.

Exhausted CXCR5+ Ova-Specific CD8+ T Cells Demonstrate Efficient Mitochondria and Nutrient Acquisition Function
The activation induced differentiation of T cells is, in part, profoundly regulated by the metabolism. A recent study suggested that the purinergic receptor P2X7R drives the metabolic fitness and long-term maintenance of CD8+ T cells in acute LCMV infection (42). We addressed the question of whether P2X7R is expressed on exhausted Ova-specific CD8+ T cells in the liver. Interestingly, we observed that P2X7R was mainly expressed on Ova-specific CXCR5+ CD8+ T cells as opposed to CXCR5-and naive CD8+ T cells ( Figure 4A, S6A). Further, we quantified a higher P2X7R MFI on CXCR5+ CD8+ T cells ( Figure 4B), suggesting that Ova-specific CXCR5+ CD8+ T cells may be exhausted but long-lived memory cells. The preferential expression of P2X7R on CXCR5+ CD8+ T cells ( Figures 4A, B) coupled to enhanced memory properties (Figure 2), led to the hypothesis that efficient mitochondrial function may be restricted largely to the Ova-specific CXCR5+ CD8+ T cell subset. We tested this hypothesis by assessing different aspects of mitochondria function in Ova-specific CD8+ T cells, D21 post AdOva vaccination. Using Mitotracker Green dye (MTG) staining, we observed that the mass of the mitochondria was increased in Ova-specific CXCR5+ CD8+ T cells compared to the CXCR5-CD8+ T cell subset. More so, we observed slightly higher mitochondria mass in CXCR5-CD8+ T cells compared to naïve endogenous T cells ( Figure 4C). Since memory T cells have higher mitochondria mass compared to effector T cells (43,44), these data are in agreement with the enhanced memory marker expression ( Figure 2). Subsequently, we quantified the mitochondria potential of Ova-specific CXCR5+ and CXCR5-CD8+ T cells, using Tetramethylrhodamine ethyl ester (TMRE) and Mitotracker Deep Red (MTDR). We observed a higher frequency of MTDR and TMRE in the CXCR5+ T cell compartment ( Figures 4D, E, S6B). Moreover, the MFI of TMRE was reduced in CXCR5-CD8+ T cells ( Figure 4F), suggesting reduced mitochondria membrane potential in this subset. Reduced mitochondria activity may be a result of excessive production of mitochondria reactive oxygen species (mROS), which disrupts mitochondria function (45). Upon probing mROS in antigen specific T cells, we observed~60% of Ova-specific CXCR5-CD8+ T cells while a minority of CXCR5+ CD8+ T cells expressed mROS ( Figure 4G). Interestingly, mROS MFI was higher in CXCR5-CD8+ T cells, albeit not significantly ( Figure 4H). Effector cytokine production is largely coupled to the bioenergetics and nutrient acquisition efficiency of T cells (46,47). These processes are extensively and tightly regulated by the mitochondria function. Therefore, we asked if the efficient mitochondria function and the improved cytotoxic function may be associated with elevated nutrient acquisition in Ovaspecific CXCR5+ T cells. In this regard, Ova-antigen specific T cells isolated from the liver D21 post AdOva vaccination were phenotyped for the glucose 1 transporter (Glut-1). We observed a higher frequency of Glut-1 expression in CXCR5+ CD8+ T cells compared to the CXCR5-CD8+ T cell subset ( Figure 4I, S6C).
Quantification of Glut-1 MFI further indicated higher protein levels in CXCR5+ CD8+ T cells ( Figure 4J). To evaluate the functionality of Glut-1 expression by CXCR5+ and CXCR5-CD8 + T cells we determined the uptake of fluorescent glucose (2-NBDG). In concord to higher Glut-1 expression, higher frequency and MFI of 2-NBDG uptake were observed from Ova-specific CXCR5+ CD8+ T cells (Figures 4K, L). Collectively, these data suggest that improved mitochondria function of Ova-antigen specific CD8+ T cells in the liver is intimately linked to memory-like CXCR5+ CD8+ T cells, which exhibit efficient nutrient uptake.

Exhausted CXCR5+ and CXCR5-Ova-Specific CD8+ T Cells Display Differential Maintenance and Proliferation
We hypothesized that the improved metabolic properties of Ovaspecific CXCR5+ CD8+ T cells may facilitate rapid proliferation and, possibly, other energy demanding cellular processes. Thus, we examined the cell cycle marker Ki67 expression in Ovaspecific CD8+ T cells. We observed that all the Ova-specific CXCR5+ and CXCR5-CD8+ T cells were positive for Ki67, however, a careful examination showed a higher frequency of CXCR5+ CD8+ T cells expressing high levels of Ki67 (Ki67 hi ), and an increased MFI of Ki67 in CXCR5+ CD8+ T cell subset ( Figures 5A, B, S6C). Next, we analyzed the expression of the costimulatory molecule CD28 within antigen specific CD8+ T cells. T cell receptor signaling alongside the co-stimulatory molecule CD28 has been shown to promote T cell function and proliferation (48,49). Congruent to the increased Ki67 expression, we observed higher CD28 on CXCR5+ CD8+ T cells compared to CXCR5-CD8+ T cells ( Figure 5C, S6D). Analysis of CD28 MFI showed 50% reduction on Ova-specific CXCR5-CD8+ T cells ( Figure 5D). We asked if the expression levels of Ki67 and CD28 are functionally relevant for CXCR5+ CD8+ T cell proliferation. To test the proliferation, differentiation and maintenance capacity of Ova-specific CXCR5+ and CXCR5-CD8+ T cells, we transferred dsRED+ OT-I cells to high antigen mice ( Figure 5E). On D9 post OT-1 transfer, mice were intravenously injected with s+16a nanobody prior to liver non-parenchyma cell isolation. The CXCR5+ and CXCR5-dsRED+ OT-1 cells were FACS sorted and labeled with the proliferation tracker dye CFSE ( Figure 5E). D9 was selected because of the recovery of highest numbers of Ova-specific CD8+ T cells ( Figure S1D). Upon in vitro stimulation with anti-CD3/ CD28 coupled Dynabeads, the CXCR5-OT-1 cells remained largely CXCR5-. In contrast, the CXCR5+ OT-1 cells differentiated into CXCR5-OT-1 cells but also maintained a small fraction of CXCR5+ OT-1 cells ( Figure 5F). Interestingly, we observed higher survival of CXCR5+ OT-1 cells upon stimulation ( Figures 5G, H). Furthermore, we assessed the proliferation capacity by evaluating the dilution of CFSE in surviving CXCR5+ and CXCR5-OT-1 cells ( Figure 5I). We observed a rather slow rate of cell division in CXCR5-OT-1 cells while the CXCR5+ OT-1 cells underwent rapid cell division ( Figure 5I) as reflected by a higher frequency of CXCR5+ OT-1 cells in >4 cell division cycles. These data, therefore, suggest that CXCR5+ CD8+ T cells have a survival and proliferative advantage over CXCR5-CD8+ T cells, and CXCR5+ CD8+ T cells represent a progenitor T cell subset capable of generating CXCR5-CD8+ T cells.
To evaluate if the higher in vitro proliferation and survival properties are also displayed in vivo, we transferred naive dsRED+ OT-1 cells into high antigen mice ( Figure 5J). D9 post T cell transfer, CXCR5+ and CXCR5-OT-1 cells were FACS sorted from liver post s+16a nanobody treatment and transferred into low antigen recipient mice ( Figure 5J). D6 post CXCR5+ and CXCR5-OT-1 cell transfer, non-parenchyma cells were isolated from liver and analyzed by flow cytometry for dsRED+ OT-1 cells. Although not statistically significant, we observed higher frequencies and absolute numbers of dsRED+OT-1 cells in the liver of CXCR5+ OT-1 cell recipient mice compared to the CXCR5-OT-1 cell recipient mice (Figures 5K, L). In agreement with the in vitro analysis, CXCR5-OT-1 cell remained CXCR5while the CXCR5+ OT-1 cells partially differentiated to CXCR5-OT-1 cells ( Figure 5M). In sum, these data suggest that Ovaspecific CXCR5+ CD8+ T cells have a proliferation, selfmaintenance and memory recall advantage over the CXCR5-CD8+ T cell subset in the liver.

Host Conditioning With CpG ODN Reinvigorates Exhausted Ova-Specific CD8+ T Cells and Promotes Intrahepatic T Cell Formation
We previously reported that host conditioning with CpG ODN during the onset of Ova-specific CD8+ T cell responsei.e., prior to the development of T cell exhaustionprecludes the formation of an exhausted phenotype (30). We asked if host conditioning with CpG ODN also has a beneficial effect on exhausted Ova-specific CD8+ T cells in the liver. To this end, we transferred naïve OT-1 cells into high antigen mice. On day 13 post adoptive OT-1 transfer, when T cells are exhausted (29), we treated the mice with CpG ODN over a period of 15 days ( Figure  6A). On day 14 post the last CpG ODN application (D42 post adoptive transfer of cells), the mice were sacrificed and the antigen load in liver was determined. Strikingly, CpG ODN treated mice showed efficient antigen clearance from the liver ( Figure 6B). Contrarily, control groups that received OT-1 cells without CpG ODN or CpG ODN without OT-1 cells did not clear the antigen load and showed Ova levels comparable to high antigen mice ( Figure 6B). These data suggest that CpG ODN mediates the reprogramming of exhausted OT-1 cells, resulting in improved cytotoxic activity. To understand how CpG ODN mediated the reprogramming of exhausted Ova-antigen specific T cells on the cellular level, we analyzed the OT-1 cells in the liver of CpG ODN treated mice. We observed a significant increase in the frequency and absolute number of OT-1 cells upon CpG ODN conditioning (Figures 6C, D), suggesting improved maintenance of antigen specific CD8+ T cells. Moreover, the frequency of GZMB producing OT-1 cells was enriched in CpG ODN recipient mice ( Figure 6E), suggesting that the functionality of T cells was improved. Interestingly, we observed higher levels of the activation marker CD69 expression on OT-1 cells in CpG ODN treatment ( Figure 6F). T cells expressing CD69 in the absence of antigen were previously shown to possess improved tissue residency capacity (50,51). To determine whether in our model the CD69+ CD8+ T cells express other markers of tissue resident memory T cells, we stained for CXCR6 and CX3CR1. Of note, OT-1 cells isolated from the group of high antigen mice treated with CpG ODN homogeneously expressed CXCR6 and CX3CR1 as opposed to OT-1 cells from control mice without CpG ODN treatment ( Figure 6G), suggesting that CD69+ OT-1 cells after antigen clearance are possibly tissue resident T cells. Taking together, these data suggest that CpG ODN may: (a) rewire exhausted OT-1 cell function and/or (b) induce extensive proliferation in the less exhausted CXCR5+ T cells to support cytotoxicity and promote intrahepatic memory T cell formation.

CpG ODN-Mediated Reinvigoration of Exhausted Ova-Specific T Cells Potentiates the Maintenance of Long-Lived CXCR5+ CD8+ T Cells
To determine how CpG ODN mediated reprogramming affects exhausted Ova-specific T cell subsets, we first assessed the transcription factor TCF-1, which regulates the maintenance of T cells (52) and the memory marker CD62L on antigen specific T cells (37). OT-1 cells from CpG ODN treated mice effectively down-regulated TCF-1 and CD62L expression ( Figures 7A, C), suggesting that CpG ODN potentiates the reduced expression of TCF-1 and CD62L in OT-1 cells. Still, the absolute numbers of TCF-1 and CD62L expressing cells in the pool of OT-1 cells were significantly higher (Figures 7B, D). While the frequency of CXCR5+ OT-1 cells was reduced ( Figure 7E), notably, the absolute numbers of CXCR5+ OT-1 cells were increased upon CpG ODN treatment ( Figure 7F) which is in agreement with the TCF-1 and CD62L analysis. Further, we assessed the expression of TCF-1 in CXCR5+ and CXCR5-OT-1 cell subsets in the absence of CpG ODN treatment. We observed maximum TCF-1 expression in the CXCR5+ OT-1 cell subsets ( Figure 7G   is in accordance to the previous notion that TCF-1 may potentially drive the long term-survival of antigen specific CXCR5+ CD8+ T cells (14,15). We further asked if the CpG ODN treatment would affect the expression of TCF-1 in CXCR5+ OT-1 cells. As expected, we observed a decreased frequency of TCF-1 in both CXCR5+ and CXCR5-OT-1 cells after CpG ODN conditioning ( Figure 7G), suggesting that the inflammatory milieu created by CpG ODN may be involved in dampening TCF-1 expression as previously shown in context of vaccination or L. monocytogenes infection (53). Importantly, the frequency of TCF-1 expression was increased in CXCR5+ OT-1 cells both with or without CpG ODN conditioning ( Figure 7G). Finally, we assessed the impact of CpG ODN treatment on GZMB expression in antigen specific CXCR5+ and CXCR5-OT-1 cells. With CpG ODN treatment, we observed 2-fold increased levels of GZMB expression in both CXCR5+ and CXCR5-OT-1 cells ( Figure 7H). Taking together, these data suggest that although host conditioning with CpG ODN drives the depletion of TCF-1 in Ova-specific CXCR5+ CD8+ T cells, it promotes the maintenance of long-lived memory T cells and potentiates the enrichment of GZMB in both CXCR5+ and CXCR5-Ova-specific CD8+ T cells.

DISCUSSION
The in-situ regulation of exhausted T cells in chronic liver infections remains unclear in many aspects, largely due to the lack of small animal models which would recapitulate HBV and HCV infection as well as the respective consequences including liver fibrosis, liver cancer, and liver failure. In recent chronic LCMV infection studies, lymphoid associated CXCR5+ CD8+ T cells were shown to facilitate the long-term maintenance of exhausted T cells, providing the basis for PD-1 mediated reinvigoration of exhausted T cells (10,11). Although previous studies gave valuable insights (54)(55)(56), the dynamic in-situ function of CXCR5+ CD8+ T cells in chronic HBV and HCV infection remains to be fully explored. In this study, we have identified a canonical CXCR5+ CD8+ T cell subset within the pool of exhausted Ova-specific T cells responding to persistent Ova antigen in the liver. We observed that the frequencies and numbers of CXCR5+ CD8+ T cells are relatively stable in the presence of persistent antigen. Effectively, CXCR5+ CD8+ T cells show potent cytotoxic function, possibly by virtue of their reduced exhaustion state coupled to improved T cell receptor signaling. Although cytotoxic, the Ova-specific CXCR5+ CD8+ T cells are memory-like with enhanced mitochondria function and nutrient uptake potential, suggesting they may utilize oxidative phosphorylation (OXPHOS) and glycolysis for energy metabolism. Importantly, we observed a rapid proliferation burst in CXCR5+ CD8+ T cells in vitro, and a better recall response to cognate antigen in vivo. The CXCR5+ CD8+ T cells from the liver maintain themselves and generate CXCR5-CD8+ T cells whereas the CXCR5-CD8+ T cells lack the ability to generate CXCR5+ CD8+ T cells. Importantly, mice conditioning with CpG ODN reinvigorates the exhausted Ova-specific CD8+ T cell pool in the liver, potentiating persistent antigen eradication from the liver. Upon reinvigoration, we observed higher numbers of Ova-specific CD8+ T cells and an increased number of resident memory T cells in the liver.
The hierarchical differentiation of T cells during chronic viral infection fosters heterogeneity, which drives the long-term maintenance of exhausted T cells (3,4). This heterogeneity is reflected by progenitor CXCR5+ and terminally exhausted CXCR5-CD8+ T cell subsets (3,33). The progenitor CXCR5+ CD8+ T cells were initially identified in context of LCMV infection and were reported to localize exclusively in lymphoid organs (10,57). However, we have herein identified memory-like CXCR5+ Ova-specific CD8+ T cells in the liverthe site of antigen recognition. This suggests that the site of antigen recognition may be the fate determinant of CXCR5+ CD8+ T cells' tissue localization. The CXCR5 chemoattractant CXCL13 is enriched in lymphoid organs (16), suggesting that the liver is not the natural homeostatic niche for CXCR5+ CD8+ T cells. Of note, a recent study documented that high levels of CXCL13 in livers of chronically infected HBV patients facilitate intrahepatic recruitment of CXCR5+ CD8+ T cells (54). In the study herein, we show that at the onset of antigen recognition, CXCL13 levels increase in serum. While the particular cell type expressing CXCL13 in course of intrahepatic T cell responses remains to be elucidated, our results indicate that CXCL13 may co-operate with antigen recognition to facilitate the retention of Ovaspecific CXCR5+ CD8+ T cells in the liver. The retained CXCR5+ CD8+ T cells in the liver were relatively stable over time, indicating that the CXCR5+ CD8+ T cells may undergo constitutive self-maintenance and renewal, as previously suggested (11). Notably, upon transplantation into low antigen recipient mice the CXCR5+ CD8+ T cells largely differentiate to CXCR5-CD8+ T cells, while a small fraction of CXCR5+ CD8+ T cells was maintained. Therefore, the CXCR5+ CD8+ T cells may be referred to as a progenitor exhausted Ova-specific T cell subset in the liver.
The upregulation of LFA-1 has been shown to foster the retention of T cells in the liver (34). Likewise, the expression of CXCR6 and CX3CR1 (35)(36)(37) is also known to identify tissue resident T cells. We observed the expression of CXCR6, CX3CR1, and LFA-1 on Ova-specific CXCR5+ and CXCR5-CD8+ T cells, suggesting that these may be tissue resident cells. Tissue residency of T cells is regulated by complex transcriptional signatures (58,59). Accordingly, the expression of these surface markers may not be sufficient to define the intrahepatic T cells as tissue resident. Thus, detailed transcriptional profiling as well as experimental parabiosis studies would be required to fully confirm their tissue residency.
The strength of T cell receptor signaling is a rheostat for Nur77 expression (60) and signaling through the T cell receptor is shown to be reduced in exhausted T cells (8). Thus, the level of Nur77 expression herein may demonstrate the degree of exhaustion at play in T cells. Indeed, we observed reduced frequencies of exhaustion markers on CXCR5+ CD8+ T cells, coupled to higher Nur77 expression, which is indicative of a dampened state of exhaustion. Accordingly, CXCR5+ CD8+ T    Highly Functional Intrahepatic CXCR5+ T Cells cells were more cytolytic. CXCR5+ CD8+ T cells have been previously shown to provide the proliferative burst upon PD-1 blockade therapy (10). Herein, we have demonstrated that upon in vitro stimulation and adoptive transfer, intrahepatic Ovaspecific CXCR5+ CD8+ T cells underwent enhanced proliferation. These data suggest that in chronic liver infection, PD-1 therapy may preferably target and foster CXCR5+ CD8+ T cell proliferation and function to enhance the functionality of the pool of exhausted CD8+ T cells, and thereby facilitate disease remission. The results obtained in our OVA model mice are in agreement with a recent study that was based on an AAV-HBV mouse model (54). Upon hydrodynamic injection of recombinant adeno-associated viral vectors encoding the HBV 1.2 replicon, CXCR5+ CD8+ T cells were observed in liver as well as in spleen and blood. Consistent with the OVA model herein, antigen specific CXCR5+ CD8+ T cells from the AAV-HBV model exhibited improved cytotoxic functions. The findings obtained from these independent mouse models hint towards a crucial role of these cells for a potent immune response in liver.
The extensive proliferation of the progenitor CXCR5+ CD8+ T cells coupled to the upregulation of Nur77 strongly suggest that they may preferably be amenable to immunomodulation. Indeed, CpG ODN reinvigorated exhausted T cells, with an enhanced GZMB expression in CXCR5+ CD8+ T cells, which highlights the notion that exhausted Ova-specific CD8+ T cells are not anergic. The CpG ODN mediated reinvigoration translates into enrichment of Ova-specific CD8+ T cells in the liver that consist of higher numbers of progenitor CXCR5+ CD8+ T cells. Interestingly, CpG ODN treatment downmodulates the expression of TCF-1 in CXCR5+ CD8+ T cells. This observation indicates that CpG ODN may drive the dampening of TCF-1, which controls memory T cell formation. Thus, CpG ODN may facilitate effector T cell differentiation. Indeed, a recent study showed that inflammatory cytokines suppress TCF-1 in T cells and promote the formation of effector cells (53). CpG ODN is known to mediate the upregulation of inflammatory cytokines and co-stimulatory molecules CD80/86 and OX40 on myeloid cells, which promotes improved T cell function (9). Signaling through OX40L and CD28, the receptor for CD80/86, facilitates enhanced proliferation of T cells (61). Therefore, exhausted Ova-specific T cell reinvigoration may partly be initiated by co-stimulatory signaling, which facilitates exhausted T cell proliferation and the recovery of high Ova-specific T cells. In fact, the success of PD-1 immunotherapy in chronic infection and cancer is dependent on CD28 signaling (62). Thus, CpG ODN treatment may preferentially induce extensive proliferation in CXCR5+ CD8+ T cells, as they mainly expressed CD28, which possibly promoted the rapid proliferation of CXCR5+ CD8+ T cells in vitro. Together, our data are in line with the hypothesis that CpG ODN reprogram exhausted T cells, thereby reinvigorating their cytotoxic capacity. In addition, the reinvigoration of exhausted T cells could be a result of targeted CpG ODN-mediated proliferation of the less exhausted and more functional CXCR5+ CD8+ T cells. Future studies would be needed to decipher the underlying mechanism facilitating CpG mediated exhausted CD8+ T cell revival.
The stemness properties of exhausted T cells are maintained by Bcl6 and TCF-1 co-operation via restraining type I interferon pro-exhaustion mechanisms (15). In addition, the regulation of exhausted T cell stemness is mediated by a complex transcriptional network, with TCF-1 playing a cardinal role (63). The restriction of Bcl6 and TCF-1 to CXCR5+ CD8+ T cells suggests that the TCF-1-Bcl6 axis may play a critical role in regulating the multi-potency, cytotoxicity, pro-survival and exhaustion states of CXCR5+ CD8+ T cells. Consistently, CXCR5+ CD8+ T cells are more cytotoxic and upregulate multiple memory associated markers. Of interest is the enriched expression of CD127 and the purinergic receptor P2X7R, which may predispose CXCR5+ CD8+ T cells to longterm survival (64,65). Accordingly, CXCR5+ CD8+ T cells show better recall response and maintenance in vitro and in vivo.
The improved recall response may be intimately linked to the rapid proliferation and efficient mitochondria function of CXCR5+ Ova-specific CD8+ T cells. P2X7R was recently implicated in CD8+ T cell metabolism and this is mediated mechanistically via AMPK stimulation, OXPHOS, increased glucose and fatty acid uptake (42). Fused and elongated mitochondria networks promotes efficient electron transport chain activity, facilitating robust OXPHOS (66,67). The mitochondria mass, which shapes memory T cell development (68), is larger in the CXCR5+ CD8+ T cells, suggesting efficient electron transport chain activity which may foster enhanced OXPHOS. Efficient electron transport chain activity regulates mROS expression (66,68), which may compromise mitochondria integrity (69,70). In this regard, the fragmented mitochondria in CXCR5-CD8+ T cells may be driving increased mROS, resulting in reduced mitochondria potential. Moreover, upregulation of Bcl6 dampens glycolysis (66), supporting the possible use of OXPHOS by CXCR5+ CD8+ T cells. Importantly, OXPHOS promotes increased spare respiratory activity, the ability to produce energy under stressful conditions (66, 67)a hallmark of memory cells. Aside OXPHOS, CXCR5+ Ovaspecific CD8+ T cells upregulate glucose intake, possibly to fuel glycolysis needed for rapid proliferation and cytotoxic function. Indeed, glucose is fed into the glycolytic pathway to fuel the cellular processes of T cells (66,67). Therefore, we propose that CXCR5+ Ova-specific CD8+ T cells may depend on OXPHOS and glycolysis, which may be a unique feature of tissue resident memory T cells. These features support the bifunctional memory and cytotoxic nature of CXCR5+ CD8+ T cells.
In summary, based on the OVA mouse model that reflects the early stages of chronic hepatotropic viral infection, we have shown that the intrahepatic pool of exhausted Ova-specific CD8+ T cells consists of a specialized subpopulation of CXCR5+ CD8+ T cells. These CXCR5+ CD8+ T cells retain improved cytotoxic, metabolic, and proliferative functions upon chronic antigen stimulation. Accordingly, CXCR5+ CD8+ T cells may provide a novel option for therapeutic strategies to chronic liver infections in humans. However, experimental models that mimic the unique immunological characteristics upon HBV and HCV infection as well as the development of late stage liver diseases, would greatly facilitate our understanding of CXCR5+ CD8+ T cells for the treatment of infected patients. Of note, recent emerging strategies eliminating mouse-specific HCV restriction factors (19) might pave the way towards the development of such models.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The animal study was reviewed and approved by Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, Dezernat 33. Written informed consent was obtained from the owners for the participation of their animals in this study.