HDV-induced IFN activates the metabolism of amino acids and glutathione, but downregulates pathways feeding the TCA cycle

To analyse the impact of IFN signalling on liver metabolism, differentiated HepaRG or HepaRG^NTCP cells were infected with HDV, treated with the interferon signalling inhibitor ruxolitinib (rux) or not and harvested at 6-7 days post infection (dpi), when HDV antigen expression was at maximal levels (Fig. 1A). HDV RNA and protein levels were not sensitive to treatment with rux, which blocks JAK1 and JAK2 signalling (Fig. 1A). As few data in the literature are available on metabolic changes induced by interferon and HDV, a global RNAseq analysis was performed in the presence or absence of rux (Fig. 1B; Fig. S1A). Approximately 3000 genes were differentially regulated by HDV. Infection upregulated type I IFN signalling, as previously reported [18], an effect that was reversed by rux treatment (Fig. 1B). Serine, cysteine and asparagine synthesis were induced, while expression of genes involved in the TCA cycle and fatty acid (FA) metabolism was downregulated (Fig. 1B). Major pathways such as glycolysis, gluconeogenesis and purine metabolism were not affected. An upregulation of de novo serine biosynthesis genes (PHGDH, PSAT1 and PSPH), as well as serine dehydratase (SDS) responsible for serine catabolism into pyruvate, and serine hydroxymethyltransferase 2 (SHMT2), which links serine metabolism to the folate cycle was detected. However, KEGG pathway analysis showed a slight downregulation of transcripts of genes involved in folate metabolism. Serine can also be utilized for cysteine synthesis via the transsulfuration pathway, consistent with the observed increase in cystathionine gamma-lyase (CTH) transcripts. Synthesis of asparagine was activated, indicated by upregulation of asparagine synthetase (ASNS) and glutamic-oxaloacetic transaminase 1 (GOT1). In line with the transcriptome data, the levels of serine and asparagine, as well as GSH, for which cysteine is a precursor, were higher in HDV-infected cultures (Fig. 1C). Transcript levels of aconitase (ACO1) and isocitrate dehydrogenase 2 (IDH2) responsible for α-ketoglutarate (αKG) formation in the TCA cycle, were lower in HDV-infected cells (Fig. 1B). The carbon backbone used for TCA cycle metabolites originates from three main sources: glutamine, pyruvate, and fatty acids. Pyruvate can enter the TCA cycle either directly via pyruvate carboxylase (PC)-mediated conversion to oxaloacetate, or be used for acetyl-CoA synthesis by pyruvate dehydrogenase (PDH). PC expression was decreased in HDV-infected cells. PDH activity is regulated by pyruvate dehydrogenase kinase (PDK); no changes in PDH phosphorylation were observed (Fig. S1B). Acetyl-CoA is also the endpoint metabolite of FA oxidation (FAO). Transcriptome analysis revealed downregulated expression of genes involved in saturated and polyunsaturated FA (SFA and PUFA, respectively) synthesis as well as degradation (Fig. 1B). All these data were verified by RT-qPCR and western blotting in additional infections using HepaRG^NTCP as well as fully differentiated HepaRG cells (Fig. S1C, D). Consistent with the fact that HDV is not sensitive to rux treatment, various metabolic inhibitors did not affect HDV replication (Fig. S1E).

A Left panel: Representative immunoblot analysis of mock and HDV-infected HepaRG^NTCP cells treated or not with 1 μM ruxolitinib (rux) for the last 4 days of infection and analysed at the indicated dpi (7 dpi for rux treatment). Middle panel: Relative HDV RNA levels in HepaRG^NTCP cells treated or not with 1 μM rux for the last 4 days of infection at 7dpi. Right panel: HDV-positive cells were visualized via immunofluorescence by HDAgs (green) and nuclear DAPI (blue) staining at 7dpi. n = 2. B A simplified metabolic scheme with HDV-affected pathways; an RNAseq-based heatmap depicting relative expression levels of selected genes normalized to average mock condition. Saturated fatty acids (SFA); polyunsaturated fatty acids (PUFA). C Serine, asparagine and reduced gluthatione (GSH) quantification in mock and HDV-infected HepaRG^NTCP cells. n = 3. Statistical analysis: for C data are shown as mean +/- SD. For serine and GSH t-test and for asparagine Mann-Whitney test were used. ****p < 0.0001; ***p < 0.001.

The role of IFN in HDV-induced metabolic changes was furthermore validated with BX795, which blocks IFN signalling via TBK1 (Fig. 2A, B), as well as neutralizing antibodies to IFN (Fig. 2C). Anti-IFN-β reversed the metabolic phenotype of infected cells for most targets, suggesting that IFN-β is secreted and is responsible for most, if not all, metabolic changes that were induced by HDV. Treatment with conditioned cell medium harvested from HDV-infected cells at 7 dpi and dialysed or not (2-kDa cut-off) or with recombinant IFN-β induced very similar metabolic changes, confirming that IFN-β is the main mediator of metabolic host cell responses to HDV in vitro (Fig. 3A–C). Incubation with recombinant IFN-α or IFN-λ showed similar effects, pointing to a shared metabolic phenotype induced by type I and III IFNs in hepatocytes (data not shown). Taken together, these results suggest that HDV-induced IFN-β promotes asparagine, serine and cysteine/GSH synthesis and modifies TCA cycle activity.

A Relative mRNA (left panel) and protein (right panel) levels in HepaRG^NTCP cells treated or not with 1 μM rux for the last 4 days of infection. Normalized to non-treated mock cells, n = 3. The non-treated panel of the immunoblot is also used in Fig. S1C and Fig. 4D. B Relative mRNA levels in HepaRG^NTCP cells treated or not with 3 μM BX795 for the last 4 days of infection. Normalized to non-treated mock cells, n = 3. C Relative mRNA levels in HepaRG^NTCP cells treated with 1 μg/mL IFN-β-neutralizing antibodies starting 1 dpi. Normalized to non-treated mock cells, n = 3. Statistical analysis: data are shown as mean +/- SD. t-test (for HDV levels) or one-way ANOVA or one-way ANOVA on ranks with Tukey multiple comparison test were used. ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.

A Relative mRNA levels in HepaRG^NTCP cells treated with conditioned medium (CM) for 3 days. Normalized to CM mock cells, n = 3. B Relative mRNA levels in HepaRG^NTCP cells treated with conditioned medium, dialyzed (CMd) or not (CM), for 3 days. Normalized to CM mock cells, n = 3. C Relative mRNA levels in HepaRG^NTCP cells treated or not with 1 ng/mL of IFN-β for 3 days. Normalized to non-treated cells, n = 3. Statistical analysis: data are shown as mean +/− SD. t-test or Mann-Whitney test were used. ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.

To evaluate the activity of the TCA cycle in more detail, uniformly ¹³C-labelled glucose or glutamine was added to HepaRG^NTCP cultures and isotopes were traced (Fig. 4A). Upon HDV infection the fractions of ¹³C carbon for several metabolites such as malate, citrate, succinate and αketoglutarate (αKG), derived from glucose and glutamine tracers were generally lower, confirming reduced activity of the TCA cycle (av. by 25%) (Fig. 4A). In line with the reduced isotope tracing data, the total amount of NADH, a product of the TCA cycle, was reduced in HDV-infected cultures (Fig. S2A). Consistent with reduced availability of NADH and succinate, which are key substrates for oxidative phosphorylation, HDV reduced basal oxygen consumption rates (OCR) ~ by 30% (Fig. 4B), and almost completely eliminated spare respiratory capacity (SRC) (Fig. 4B), while the extracellular acidification rate, a measure of glycolytic activity, was not altered (Fig. S2B). The effects of HDV on respiration were at least partially reversed in the presence of rux (Fig. 4C). Furthermore, recombinant IFN-β also reduced SRC in a similar fashion (Fig. S2C) suggesting that the effects of HDV on TCA cycle activity are in great part IFN mediated. Staining with mitochondria-specific dyes did not reveal significant differences in mass and structure or the membrane potential of mitochondria (Fig. S2D,E). The total expression levels of the ETC complexes were not altered either (Fig. S2F), suggesting that reduced availability of TCA cycle-derived substrates for oxidative phosphorylation results in decreased basal respiration and SRC.

A Isotope tracing. Scheme: distribution of ¹³C carbons from tracers (¹³C₆ glucose, red; ¹³C₅ glutamine, violet) in TCA cycle metabolites. Graphs: relative isotopologue fractions measured in mock and HDV-infected HepaRG^NTCP cells treated with glucose-¹³C₆ or glutamine-¹³C₅ for the last 10 h or 3 h of infection, respectively. Isotopologues used for the measurement from glucose/glutamine: citrate M + 2/M + 4; αKG M + 2/M + 4; succinate M + 2/M + 4; malate M + 2/M + 4. Mean +/− SD, normalized to mock cells, n = 4. B Oxygen consumption rates (OCR) were measured in mock and HDV-infected HepaRG^NTCP cells. A representative experiment (left panel, mean +/− SEM) and relative OCR (right panel, mean +/− SD, normalized to basal OCR in mock cells, n = 5) are shown. C OCR was measured in mock and HDV-infected HepaRG^NTCP cells treated or not with 1 μM rux for the last 4 days of infection, normalized to basal OCR in mock cells, n = 2. Statistical analysis: for A, (B, right panel) and C data are shown as mean +/− SD. For (B, left panel) data are shown as mean +/−SEM. For (A) Mann-Whitney or t-test were used. For (B, right panel and C) one-way ANOVA with Dunn’s multiple comparison test was used. ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.

Pathways contributing most to TCA cycle activity are glycolysis, FAO, and glutaminolysis. Levels of AMPK phosphorylation, one of the key regulators of FAO, remained unaffected (Fig. S2G). Glutaminolysis seemed also unlikely to be involved as isotope tracing using ¹³C-labelled glutamine did not reveal reduction of its activity (data not shown). While PDH activity was not altered by HDV/IFN (Fig. S1B), PC expression was downregulated (Fig. S1C). Thus, changes to pyruvate import cannot be excluded.

In summary, these results show that HDV-induced interferon signalling may lead to metabolic stress related to inhibition of TCA cycle activity that is accompanied by reduced basal respiration and abolished ability of cells to increase respiration.

Metabolic stress response factors ATF4 and mTORC1 are activated downstream of IFN-β production

Amongst the IFN-target genes that were identified, many are well-known targets of the ATF4 transcription factor as for example those related to serine, asparagine and GSH metabolism [19]. ATF4 is a master regulator of the integrated stress response (ISR) and is implicated in life-death decisions during many types of stress. Indeed, ATF4 mRNA and protein levels were increased in HDV-infected cultures in an IFN-dependent manner (Fig. 5A). To investigate a potential role of ATF4 in the metabolic changes induced by HDV/IFN-β, ATF4 expression was silenced using shRNAs (Figs. 5B, S3A). ATF4 knock-down did not affect HDV replication or expression of IFN-dependent OAS1 (Fig. 5B). However, upregulation of PHGDH, ASNS and CTH was confirmed to be ATF4-dependent, while reduction of GLS1 and PC mRNA levels seemed to involve another mechanism. The best-known mechanism of ATF4 induction is phosphorylation-driven inactivation of eukaryotic translation initiation factor 2α (eIF2α), leading to the initiation of ATF4 translation (Fig. S3B). eIF2α mediates stress responses in the settings of unfolded protein accumulation, dsRNA detection, dysregulation of heavy metal metabolism, and amino acid starvation. The type of stress defines which kinase phosphorylates eIF2α. However, neither inhibitors of PKR (C16), PERK (GSK2606414), eIF2α (ISRIB), nor the addition of essential and non-essential amino acids to block GCN2 changed the induction of ATF4 targets PHGDH and ASNS in HDV-infected cultures (Fig. S3C, D). Another known ATF4 regulator is the mTORC1 pathway. This pathway was found activated in HDV-infected cells (S6 phosphorylation, Fig. 5C), however, mTORC1 inhibitor torin 1 did not abolish ATF4 induction, excluding this pathway as a regulator of ATF4 (Fig. 5C).

A Relative mRNA (top panel) and protein (bottom panel) levels in HepaRG^NTCP cells treated or not with 1 μM rux for the last 4 days of infection. Normalized to non-treated mock cells, n = 3. The Actin panel of the immunoblot is also used in Fig. S2E. B Relative mRNA levels in shScrambled (Scr) and shATF4 HepaRG^NTCP cells infected or not with HDV. Normalized to scrambled mock cells, n = 5. C Relative mRNA (left panel) and protein (right panel) levels in HepaRG^NTCP cells treated or not with 200 nM of torin 1 for the last 24 h of infection. Normalized to non-treated mock cells, n = 3. The non-treated panel of the immunoblot is also used in Fig. S1C and Fig. 2A. Statistical analysis: data are shown as mean +/− SD. For A–C Mann-Whitney or t-test or one-way ANOVA or one-way ANOVA on ranks with Tukey multiple comparison test were used. ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.

In summary, HDV-associated metabolic changes occurring downstream of IFN-β production led to metabolic stress and activation of pro-survival metabolic factors ATF4 and mTORC1. ATF4 activation is mTORC1-independent, and another known ATF4 regulator, eIF2α is also not involved in the IFN-β/ATF4 axis, suggesting induction of ATF4 in this setting by a yet unknown mechanism.

HDV-induced IFN-β signalling increases the resistance to oxidant-induced cell death and reduces functions specific for differentiated hepatocytes

Based on the finding that GSH levels were induced by HDV/IFN, and ATF4 being a stress response factor, we explored whether ATF4 induction had phenotypic effects, such as protection from stress stimuli and in particular, oxidative stress. HDV significantly increased the resistance to toxic effects of oxidants such as hydrogen peroxide (H₂O₂), tert-butyl hydroperoxide (tBHP) and cumene hydroperoxide (data not shown; Fig. 6A, tBHP data). Trypan blue staining was used as a readout for cell viability in order to avoid approaches that may be impacted by the HDV/IFN-induced metabolic changes, such as neutral red uptake or LDH release. The resistance of HDV-infected cells to oxidants was found to be IFN/ATF4-mediated, as the protective effects were reversed by rux treatment (Fig. 6A). Compounds inducing redox imbalance, such as ETC inhibitors rotenone, sodium azide and oligomycin A, or drugs triggering genotoxic and oxidative stress, such as doxorubicin, were equally toxic for mock and HDV-infected cells (Fig. S4A).

A Trypan blue staining of mock and HDV-infected HepaRG^NTCP cells treated or not with 200 μM tBHP for 24 hours; 1 μM rux for 4 days; 10 μM BSO for 29 hours. B HDV/mock ratio of total glutathione (GSH) levels. HDV-infected HepaRG^NTCP cells pre-treated or not with 10 μM BSO for 5 hours, and then treated or not with 200 μM tBHP for 2 hours. Normalized to non-treated mock cells, n = 4. C Relative albumin secretion (n = 6); CYP3A4 activity (n = 3); CYP1A2 activity (n = 3) in HepaRG^NTCP cells infected with HDV. Normalized to mock cells. Statistical analysis: for B and C data are shown as mean +/− SD. For B and Albumin Mann-Whitney test was used. For CYP3A4 and CYP1A2 t-test was used. ***p < 0.001; **p < 0.01; *p < 0.05.

To reveal the mechanism underlying the resistance to oxidative stress the levels of key antioxidant molecules GSH and GSSG were measured (Figs. 6B, S4B). Both, GSH and GSSG levels were higher in HDV-infected cells pointing to increased synthesis and oxidation. Pre-treatment with the inhibitor of GSH synthesis BSO restored the sensitivity of HDV-infected cells to tBHP (Fig. 6A). Importantly, treatment with BSO or tBHP slightly increased the GSH ratio between HDV/mock, while the combined treatment BSO + tBHP further aggravated the effect (Fig. 6B). Thus, the IFN response rendered HDV-infected cells resistant to oxidative stress. To check if this was a response to a shifted redox balance, intracellular reactive oxygen species were measured. A small decrease in DCFH2DA fluorescence in HDV-infected cells was detected using flow cytometry (Fig. S4C). However, based on the transcriptome analysis, levels of ROS scavenging enzymes (peroxiredoxins, glutathione peroxidases, catalase) were also slightly reduced. As the redox balance is of particular importance in mitochondria, superoxide anion production was measured in this organelle using MitoSOX dye (Fig. S4C). A slight, but reproducible reduction in superoxide production in mitochondria of HDV-infected cells was detected. Since addition of BSO did not reverse the effect (Fig. S4C), this observation may reflect reduced respiration activity and not increased GSH synthesis. Levels of NADP⁺/NADPH, crucial for mitochondrial antioxidant enzyme recycling, were not altered (Fig. S4D). The Nrf2 pathway did not seem to be involved either, as the Nrf2-transcriptional targets Nqo1 and HO1 were not induced (Fig. S4E). Therefore, the resistance of HDV-infected hepatocytes to pro-oxidants likely results from increased GSH biosynthesis.

As an increase of the GSH pool is known to accompany hepatocyte proliferation [20], the differentiation status of HDV-infected cultures was assessed. Infected hepatocyte cultures secreted approximately 25% less albumin into the supernatant (Fig. 6C) and activities of cytochromes CYP3A4 and CYP1A2 were reduced (Fig. 6C). De-differentiation was confirmed in HDV-infected PHH by measuring CYP2C9 activity and mRNA levels of several hepatocyte markers (Fig. S4F, G).

Taken together, IFN-β production upon HDV infection increased GSH synthesis and led to a significant resistance to oxidant-induced toxicity. Moreover, the ability of cultures to maintain key physiological functions – such as albumin secretion and cytochrome activity – was reduced.