Hypoxia‐Inducible Factor 1 Alpha–Mediated RelB/APOBEC3B Down‐regulation Allows Hepatitis B Virus Persistence

Background and Aims Therapeutic strategies against HBV focus, among others, on the activation of the immune system to enable the infected host to eliminate HBV. Hypoxia‐inducible factor 1 alpha (HIF1α) stabilization has been associated with impaired immune responses. HBV pathogenesis triggers chronic hepatitis‐related scaring, leading inter alia to modulation of liver oxygenation and transient immune activation, both factors playing a role in HIF1α stabilization. Approach and Results We addressed whether HIF1α interferes with immune‐mediated induction of the cytidine deaminase, apolipoprotein B mRNA editing enzyme catalytic subunit 3B (APOBEC3B; A3B), and subsequent covalently closed circular DNA (cccDNA) decay. Liver biopsies of chronic HBV (CHB) patients were analyzed by immunohistochemistry and in situ hybridization. The effect of HIF1α induction/stabilization on differentiated HepaRG or mice ± HBV ± LTβR‐agonist (BS1) was assessed in vitro and in vivo. Induction of A3B and subsequent effects were analyzed by RT‐qPCR, immunoblotting, chromatin immunoprecipitation, immunocytochemistry, and mass spectrometry. Analyzing CHB highlighted that areas with high HIF1α levels and low A3B expression correlated with high HBcAg, potentially representing a reservoir for HBV survival in immune‐active patients. In vitro, HIF1α stabilization strongly impaired A3B expression and anti‐HBV effect. Interestingly, HIF1α knockdown was sufficient to rescue the inhibition of A3B up‐regulation and ‐mediated antiviral effects, whereas HIF2α knockdown had no effect. HIF1α stabilization decreased the level of v‐rel reticuloendotheliosis viral oncogene homolog B protein, but not its mRNA, which was confirmed in vivo. Noteworthy, this function of HIF1α was independent of its partner, aryl hydrocarbon receptor nuclear translocator. Conclusions In conclusion, inhibiting HIF1α expression or stabilization represents an anti‐HBV strategy in the context of immune‐mediated A3B induction. High HIF1α, mediated by hypoxia or inflammation, offers a reservoir for HBV survival in vivo and should be considered as a restricting factor in the development of immune therapies.

not its complete eradication because of the persistence of the viral DNA matrix, called covalently closed circular DNA (cccDNA). (2) Upon treatment arrest, the infection can relapse. (2) Therefore, treatments are urgently needed to progress toward a cure for chronic HBV infection.
Therapeutics developed for the treatment of HBV focus on activation of the adaptive and innate immune system. Several Toll-like receptor agonists have offered promising results both in vitro and in vivo. (3)(4)(5) Among these treatments, we and others have shown that induction of the cytidine deaminase, apolipoprotein B mRNA editing enzyme catalytic subunit 3B (APOBEC3B; A3B), upon immune-mediated lymphotoxinβ receptor (LTβR) agonization (e.g., by T cells) leads to cccDNA decay. (6,7) Most immune receptors such as LTβR are described to signal through the nuclear factor-kappa B (NF-κB) pathways. (8,9) NF-κB signaling is divided into two arms: the classical/canonical and the alternative/noncanonical pathway. (10) The canonical pathway signals through the IκB kinase (IKK) complex (inhibitor of nuclear factor κB kinase complex, consisting of NF-κB essential modulator/IKKα/IKKβ), triggering the phosphorylation and ubiquitination of nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor alpha and the release of p50/RelA (NF-κB p65 subunit) heterodimer. (10) The noncanonical pathway signals through NF-κBinducing kinase (NIK), leading to the phosphorylation of IKKα and p100, which is subjected to processing into p52 forming p52/RelB (v-rel reticuloendotheliosis viral oncogene homolog B) heterodimers that activate target genes such as immune mediators. (11) To reduce the extent of chronic inflammation and its deleterious effects, NF-κB signaling has to be tightly regulated. (12) Among the factors involved in this regulation, hypoxia-inducible factor 1 alpha (HIF1α) has been shown to (1) be stabilized or induced by and (2) regulate NF-κB signaling, (13) in addition to its canonical induction by low oxygen levels. (14) HIF1α is constantly produced and is targeted to the proteasome in the absence of stabilizing conditions. (14) Here, we identify HIF1α stabilization and the concomitant decrease of RelB protein level as a restricting factor for immune-mediated antiviral strategies against HBV.

HBV pRepaRatIoN aND INoCUla
HBV was purified and concentrated from the culture medium of HepAD38 cells by heparin columns and sucrose gradient ultracentrifugation as described. (17) dHepaRG cells were infected with 200 viral genome equivalents per cell, in medium supplemented with 4% PEG-8000 (Sigma-Aldrich). Twenty-four hours after infection, cells were washed three times with PBS.

HUMaN lIVeR SpeCIMeN
Sections of formalin-fixed, paraffin-embedded liver resections of 15 patients chronically infected with HBV were obtained from the DZIF partner site in Heidelberg/Institute of Pathology of the Medical University Heidelberg. Chronic hepatitis B (CHB) patients were all in the immune-active phase of the disease and presented F3/F4 fibrosis grading and A3 activity (METAVIR scoring). Sections were cut to be 2 or 5 µM thick. Work with patient material was approved by the Heidelberg ethics committee under the following number: S206/2005. We confirmed that informed consent was collected from all co authors for the manuscript.
Additional materials and methods information can be found in the Supporting Information.

HIF1α StaBIlIZatIoN oFFeRS a ReSeRVoIR FoR HBV IN IMMUNe-aCtIVe patIeNtS
Hypoxia has been shown to strongly modulate immune responses, both positively and negatively, depending on the cells and the immune mechanisms involved. (14) Inflammatory cytokines and/or ligands have been shown to efficiently inhibit HBV infection. (3,18,19) Thus, we wanted to decipher whether HIF1α might be involved in HBV persistence in chronically infected patients by preventing immune activation. Consecutive cuts of livers from CHB patients with end-stage CHB, also considered as an immune-active phase, were stained for HIF1α and HBcAg. Highly oxygenated/low inflammation zones, highlighted by an absence of HIF1α staining, were also low for HBcAg staining in these CHB patients (Fig. 1A,B). In contrast, zones with low oxygen level or with inflammation (i.e., strong HIF1α staining) presented an increased number of HBcAg-positive nuclei. A correlation was found between the numbers of HIF1α-and HBcAg-positive cells (Fig. 1C).
We have previously shown that, on the one hand, LTβR agonization by an agonistic antibody (BS1) leads to cccDNA decay and HBV clearance, whereas, on the other hand, LTα/β are up-regulated in CHB patients. (6,20) Therefore, induction of LTβ in CHB patients should clear the infection given its antiviral effect. To assess whether the correlation of HIF1α and HBc observed in vivo (Fig. 1C) could be attributable to lower immune response in this area, liver of CHB patients were either stained for HIF1α and A3B by in situ mRNA hybridization on consecutive slides, or by costaining of mRNA and protein. High HIF1α staining was found in areas with low A3B expression, whereas low HIF1α staining was found in areas with strong A3B expression (Fig. 1D,E and Supporting Fig. S1).
Altogether, these data highlight that in areas with high HIF1α stabilization, A3B expression is impaired, allowing viral persistence even during liver inflammation. Therefore, high HIF1α areas provide a reservoir for HBV persistence in vivo.

HIF1α, BUt Not HIF2α, IS INVolVeD IN HypoXIa-MeDIateD apoBeC3B RepReSSIoN
Hypoxia can induce the stabilization of both HIF1α and HIF2α. Although we show that HIF1α knockdown can rescue A3B expression and antiviral effects of BS1 under HIF-stabilizing conditions (Fig. 2), we aimed to investigate a potential additional role of HIF2α. Therefore, cell lines doxycycline inducible for the overexpression of wild-type HIF1α, degradationresistant HIF1α, or wild-type HIF2α were generated. Of note, only a degradation-resistant HIF1α (carrying a P402A and a P564A mutation, eliminating the sites that, when hydroxylated, target HIF1α for degradation) was detected in the overexpressing cell line (Supporting Fig. S3A). Consequently, subsequent experiments were only performed with the degradation-resistant HIF1α. Transcriptional activity and expression of mutated HIF1α and HIF2α were confirmed by RT-qPCR and immunoblotting, respectively (Supporting Fig. S3A-D). Overexpression of HIF1α or HIF2α alone inhibited A3B up-regulation induced by BS1 (Fig. 3A). However, under hypoxia, only siRNAs against HIF1α, but not HIF2α, rescued A3B up-regulation, and no cumulative effect was observed when knocking down both HIF1α and HIF2α, highlighting that HIF2α only plays a minor role in A3B inhibition under hypoxic conditions (Fig.  3B). HIF1α and HIF2α knock-down efficiencies were confirmed by RT-qPCR (Supporting Fig. S3E). Moreover, inhibition of A3B by HIF1α and rescue by HIF1α knockdown were confirmed using different HIF1α stabilizers (DMOG, CoCl 2 , and VH298; Fig.  3C,D and Supporting S3F). Of note, LTβR surface expression remained unchanged under hypoxia, with a mild increase after HIF1α knockdown, highlighting that the effect of HIF1α stabilization was not attributable to a decreased receptor expression (Supporting Fig. S4G,H). Moreover, A3B repression was not attributable to cell death under hypoxia (Supporting Fig. S3I).
Altogether, these data show that under hypoxic conditions, HIF1α-but not HIF2α-impairs the induction of A3B.

HIF1α StaBIlIZatIoN INHIBItS NF-κB-INDUCeD a3B tRaNSCRIptIoN By DeCReaSINg RelB pRoteIN eXpReSSIoN leVel
The main signaling pathways activated upon LTβR agonization are related to NF-κB, suggesting that A3B is an NF-κB target gene. To confirm this hypothesis, we used two kinase inhibitors ([N-(6-chloro-7-methoxy-9Hβ-carbolin-8-yl)-2-methylnicotinamide] dHepaRG cells were infected with HBV. Six d.p.i., cells were transfected with either 10 nM of HIF1α-targeting or control siRNAs. On the next day, cells were subjected to 1% or 20% oxygen for 3 days and treated with ±0.5 µg/mL of BS1. Transfection and treatments were repeated once. (D,E) dHepaRG cells were infected with HBV. At 10 and 13 d.p.i., cells were transfected with either 10 nM of HIF1αtargeting or control siRNAs. Cells were then treated with ±0.5 µg/mL of BS1 and with ±100 µM of DMOG. (F,G) dHepaRG cells were infected with HBV. At 10 and 13 d.p.i., cells were transfected with either 10 nM of HIF1α-targeting or control siRNAs. One day after the second transfection, cells were treated or not with 0.5 µg/mL of BS1, either under the presence of 30 µM of FG-4592 or DMSO. Six days later, (B,D,F) mRNAs and (C,E,G) DNA were extracted and analyzed by RT-qPCR and qPCR. Bars represent the mean ± SD of (B,C) one or (D-G) three independent experiments performed in quadruplicates. Data were submitted to (C,E,G) an unpaired Student t test or (B,D,F) one-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.0001. (H) dHepaRG cells were infected with HBV. At 10 d.p.i., cells were treated with ±0.5 µg/mL of BS1 and with ±100 µM of DMOG for 12 days. Episomal DNA was extracted and analyzed by Southern blotting. Abbreviations: DIG, digoxigenin; d.p.i., days postinfection; mitoDNA, mitochondrial DNA; MW, molecular weight; NT, nontreated; PF, protein-free; rcDNA, relaxed circular DNA. and [5-(p-fluorophenyl)-2-ureido] thiophene-3carboxamide) that target the IKK complex (IKKα/β). We observed that inhibition of IKKα/β reduces BS1-induced A3B in dHepaRG cells (Supporting Fig. S4A). Given that we showed that HIF1α stabilization prevents BS1-induced A3B, we anticipated that   HIF1α would inhibit NF-κB target genes. Indeed, the induction of the well-known NF-κB target genes, nfkb2 and nik, upon BS1 treatment in normoxia is highly reduced in hypoxic conditions, and this effect was confirmed for A3B (Supporting Fig. S4B-D). We also extended our analysis with other activators of NF-κB (TNFα, IL-17, and LPS) and observed the same trend on the tested NF-κB target genes. Therefore, our results indicate a hypoxia-related impairment of the NF-κB signaling pathways. Interestingly, RelB is at the crossroad of both NF-κB pathways; relb transcription is dependent on the canonical, whereas RelB protein is part of the noncanonical, NF-κB dimer, p52/RelB. (10) We confirmed that, whereas BS1 increased RelB protein expression and A3B transcription, depletion of RelB drastically reduces BS1-induced A3B expression (Supporting Fig. S5B,C). Therefore, we addressed whether the inhibitory effect of HIF1α stabilization on BS1-induced A3B upregulation was a consequence of RelB inactivation.
Cell fractionation highlighted that DMOG strongly reduces BS1-induced RelB protein in both the cytosolic and the nuclear compartments, whereas RelA expression and nuclear translocation were not strongly affected (Fig. 4A). More important, the decrease of RelB protein levels in the DMOG/BS1 condition was completely rescued in HIF1α-depleted cells (Fig. 4B). HIF1α stabilization did not repress BS1-induced RelB mRNA up-regulation (Fig. 4C). These results were confirmed using longer DMOG treatment, a different level of hypoxia, and other HIF1α stabilizers (Supporting Fig. S5D-G). By immunostaining, we also confirmed that RelA nuclear translocation remained unchanged under hypoxia (Supporting Fig. S5H,I), whereas hypoxia impaired RelB induction (Fig. 4D). Interestingly, hypoxia also prevented BS1-induced p52 (the main binding partner of RelB) recruitment to the A3B promoter (Fig. 4E).
To investigate whether our in vitro findings would also be of relevance in vivo, C57BL6/J mice were injected either with DMSO or DMOG and euthanized 6 hours postinjection. In vivo, DMOG triggered HIF1α stabilization and a strong reduction of RelB protein expression in the liver, without affecting RelB mRNA. No change was observed for RelA or p50 (Fig. 4F).
Altogether, our in vitro and in vivo results identified a strong reduction of RelB protein, but not mRNA expression, as the main driver of HIF1αinduced impairment of A3B expression.

HIF1α-MeDIateD INHIBItIoN oF RelB/a3B eXpReSSIoN IS INDepeNDeNt oF ItS tRaNSCRIptIoNal aCtIVIty
HIF1α belongs to a large family of proteins, including ARNT and AhR. (21) It has been reported that RelB can dimerize with AhR or ARNT (RelB/ AhR or RelB/ARNT), either controlling RelB protein stability and/or RelB transcriptional activity. (22,23) Moreover, crosstalks between these proteins can occur through competition for common partners (e.g., HIF1α/ARNT vs. AhR/ARNT). (24) Thus, we investigated whether such processes could control RelB activity in our model. A schematic timeline of the experiments is depicted in Fig. 5A.
In dHepaRG cells, AhR knockdown did not interfere with BS1-induced RelB expression, highlighting that AhR was dispensable for RelB stability (Fig. 5B). Interestingly, contrary to HIF1α knockdown, RelB protein levels were not rescued in ARNT-depleted cells treated with DMOG/BS1 (Fig. 5C). It was reported that ARNT represses the transcription of particular NF-κB target genes, (23) as confirmed by the elevated expression of C-X-C motif chemokine ligand 10 in ARNT-depleted cells (Supporting Fig. S6A). However, ARNT knockdown had no impact on RelB mRNA expression, whereas vascular endothelial growth factor alpha expression (a target gene of the HIF1α/ARNT heterodimer) was reduced (Supporting Fig. S6B,C). In addition, neither AhR nor ARNT knockdown rescued A3B levels in DMOG-treated cells (Fig. 5D,E). These results indicate that HIF1α/ARNT dimerization, which is necessary for the canonical function of HIF1α as a transcription factor, is not the cause of decreased RelB protein and A3B mRNA expression.
In summary, our results demonstrate that HIF1α/ RelB crosstalk prevents BS1-mediated A3B expression through an unconventional HIF1α-dependent mechanism.

HypoXIa pReVeNtS IMMUNe INDUCtIoN By DySRegUlatINg eXeCUtINg patHWayS
To investigate the global effect of hypoxia, mass spectrometry was performed on control or HIF1αtargeting siRNA-transfected dHepaRG cells treated  with or without BS1 under normoxia (NO) or hypoxia (HO). A schematic timeline of the experiment is depicted in Fig. 6A. Interestingly, whereas 418 proteins were significantly dysregulated in BS1-treated versus nontreated cells under normoxia (NO/NT vs. NO/ BS1), only two proteins were found to be dysregulated when comparing the same treatments under hypoxia (HO/NT vs. HO/BS1), indicating a global inhibition of responses to BS1 treatment (Fig. 6B). Pathways were grouped into four different clusters: I, transcription and translation; II, signal transduction and immune response; III, metabolism; and IV, DNA replication and repair. Results highlighted that BS1 treatment impaired the metabolism (e.g., drug and fatty acid metabolism) of dHepaRG cells and cellular transcriptional and translational machinery were among the most up-regulated pathways, leading to production of immune response pathway effectors (Fig. 6C).
Altogether, these data showed that hypoxia globally impaired immune responses by inhibiting cellular pathways implicated in RNA processing and surveillance, as well as protein production, independently of the target gene or the stimulus. Interestingly, HIF1α knockdown rescued A3B induction, most probably by rescuing RNA processing and ribosome pathways, although it was not sufficient to completely revert the hypoxic state of the cells.

Discussion
Development of new therapeutics against HBV have largely focused on the use of immune mediators, given that they have shown promising results both in vitro and in vivo. (3)(4)(5) We and others have previously shown that immune-mediated induction of A3B by LTβR agonization (i.e., with the LTβR agonist, BS1, or LTα 1 β 2 -expressing T cells) leads to noncytolytic degradation of nuclear HBV cccDNA, enabling longterm inhibition of HBV replication without rebound, even after treatment arrest. (6,7) HIF1α has been shown to impair immune responses. (13,25) Inflammatory signaling has been shown to induce HIF1α, which we confirmed in our current study. Moreover, HBV pathogenesis and resulting fibrotic scaring processes will influence liver oxygenation, therefore modulation of HIF1α induction and stabilization. In the liver of CHB patients in immune-active (i.e., patients who potentially could clear the infection given that they likely express high levels of cytokines), we found a positive correlation between HIF1α expression and HBcAg-positive areas. Given that A3B mRNA was low in areas with high HIF1α, it can be expected that, in vivo, HBV might escape the immune responses in areas with elevated HIF1α staining.
We hypothesized that the correlation observed between HIF1α, HBcAg, and A3B mRNA highlights that low immune responses in HIF1α-high areas allow viral persistence, creating a viral reservoir. Therefore, we can hypothesize that blocking HIF1α stabilization during the immune-active phase of CHB patients could indeed be sufficient to allow morepotent immune responses, among which is induction of A3B, and viral elimination.
In vitro, we confirmed, using 1% oxygen, DMOG, and a number of other molecules inducing HIF1α stabilization, as well as HIF1α-overexpressing cell On the next day, cells were subjected to 1% (Hypoxia) or 20% (Normoxia) oxygen for 3 days ± 0.5 µg/mL of BS1. Proteins were submitted to unbiased mass spectrometry analysis. (B) Data are presented as volcano plot of normoxia nontreated (NO/NT) versus normoxia BS1-treated (NO/BS1) comparison. Dotted line represents the limit of significance (adjusted P value, <0.05). Red dots represent the only two proteins that are still significantly dysregulated (i.e., adjusted P value, <0.05) in similar comparison under hypoxia (HO/NT vs. HO/BS1). (C-F) Pathway analysis of significantly changed proteins was conducted with preselected KEGG pathways using the ROAST algorithm. The pathways are represented for (C) NO/NT versus NO/BS1, (D) NO/NT/siCtrl versus NO/BS1/siCtrl, (E) NO/ BS1/siCtrl versus HO/BS1/siCtrl, and (F) HO/BS1/siCtrl versus HO/BS1/siHIF1α. The significantly (respectively, nonsignificant) upregulated (dark red bar; respectively, light red bar) or down-regulated (dark blue bar; respectively, light blue bar) pathways are presented as the percentage of proteins analyzed in the pathways. Of note, black bars represent the number of significantly dysregulated proteins in the pathway. Data were submitted to a LIMMA algorithm for selection of significantly changed proteins. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Abbreviations: Akt, protein kinase B; CYP450, cytochrome P450; FDR, false discovery rate; JAK, Janus kinase; KEGG, Kyoto Encyclopedia of Genes and Genomes; MAPK, mitogen-activated protein kinase; NT, nontreated; PI3K, phosphoinositide 3-kinase; PPAR, peroxisome proliferator-activated receptor; RIG-I, retinoic-acid-inducible gene I; ROAST, rotation gene set testing; STAT, signal transducer and activator of transcription.
lines, that HIF1α stabilization mediates a strong impairment of LTβR-dependent A3B induction. However, impairment of immune responses was not limited to A3B as an NF-κB target gene, neither to BS1 as an NF-κB inducer, highlighting that HIF1α modulated NF-κB and other immunesignaling pathways (e.g., IFNα/γ-induced cccDNA degradation) to prevent the induction of immune mediators. Indeed, we identified that HIF1α impairs RelB protein, but not RelB mRNA level, in vitro and in vivo. This suggests that either RelB mRNA is not properly exported from the nucleus and/or is not efficiently translated, as confirmed by our proteomic data, which showed an impairment of RNA-processing and ribosome pathways under hypoxia. Alternatively, RelB stability is subjected to posttranslational modifications associated with proteasomal/lysosomal protein degradation. (26) We also found that the inhibitory activity of HIF1α toward RelB was independent of its partner, ARNT. An ARNT-independent function of HIF1α starts to emerge, (27) and the HIF1α/RelB crosstalk we discovered could bring more insights into the immune metabolism of the liver.
The global inhibition of immune responses observed under HIF1α stabilization, with different ligands and on several targets, suggests the need to modulate HIF1α to obtain optimal immune activation and thus an antiviral response during immune therapies administration. However, it will be important to confirm the effect of HIF1α on other immune therapies and antiviral targets, as well as in vivo, in a therapeutic setup. Mass spectrometry revealed that even though HIF1α knockdown partially rescued pathways implicated in RNA and protein production and processing, it could not fully reactivate the immune response in cells. Interestingly, although the rescue of the "hypoxic state" of the proteome was only partial, it was sufficient to rescue A3B induction and thereby restore the anti-cccDNA effects of BS1 Unknown mechanism treatment. From a clinical perspective, this could have severe consequences for the outcome of immunestimulatory approaches for the treatment of CHB patients. The oxygen status of the liver microenvironment is not only important for parenchymal cells to be able to integrate external stimuli, but also for immune cells to exert their function properly. (14,25) Moreover, given that inflammation can trigger HIF1α stabilization, it will be mandatory to inhibit HIF1α to insure potent immune responses. Recently investigated HIF inhibitors have shown encouraging results in cancer therapies. (28) These molecules should be tested in the treatment of CHB, especially in patients with fibrosis, and thus with compromised liver oxygenation. In the context of immune-mediated A3B activation, a focus should be made on HIF1α inhibitors. Additionally, HIF1α inhibitors could be combined with immune therapies (3,5) to insure potent immune activation in the whole liver.
In summary, we have shown that HIF1α stabilization impairs NF-κB-mediated A3B induction, which is important for HBV cccDNA purging (Fig. 7). We believe that preventing the inhibitory activity of HIF1α toward RelB might represent a therapeutic window that should be considered as a support of combinatory immune therapies, to ensure a better efficacy of the treatment.