Main

Conventional T and B cells have a crucial role in HBV infection1,2,3,4,5. In contrast, the contribution of cells at the interface between innate and adaptive immunity, such as NKT cells, remains controversial6. NKT cells respond in a T cell receptor (TCR)-restricted manner to lipid antigens presented by CD1d on professional and nonprofessional antigen-presenting cells and show pronounced cytokine secretion within hours of cognate antigen recognition, which enables broad effects on activation of other innate (NK) and adaptive (T and B) immune cells7,8. Given the central role of NKT cells in both direct and indirect modulation of the immune system, NKT cells have been shown to be crucial in the defense against a variety of microbial pathogens8.

Analysis of liver gene expression in chimpanzees 2 weeks after HBV infection has shown evidence for a lack of induction of immune-related genes, suggesting that HBV acts a stealth virus that escapes innate immune responses during early infection9. However, studies in HBV-infected humans and chimpanzees have demonstrated the presence of HBV-specific T and B cells within weeks of infection, consistent with successful priming of an adaptive immune response1,10,11. These observations suggest that HBV might be susceptible to recognition by the immune system directly after infection. Accordingly, recent studies in animal models of Hepadnaviridae infection and in HBV-infected humans have shown activation of NKT cells at very early time points following infection10,11,12. Thus, infection of woodchucks with woodchuck hepatitis virus led to hepatic NKT cell infiltration within 48 h, which correlated with interferon-γ (IFN-γ) secretion and temporary suppression of viral replication12. These findings are consistent with the observation that pharmacological stimulation of invariant NKT (iNKT) cells, a subset of NKT cells defined by expression of the Vα14-Jα18 TCR in mice and Vα24-Jα18 TCR in humans, leads to rapid IFN-γ–dependent inhibition of viral replication in transgenic mice expressing replication-competent HBV genomes13. Similarly, a study of humans during the incubation phase of HBV infection showed increased levels of peripheral NK cells consistent with innate immune activation early after HBV infection11. Most notably, a recent report on two humans with HBV demonstrated very early activation of peripheral natural T cells, a population of cells that phenotypically resemble classical NKT cells. Natural T cell activation preceded activation of NK and conventional T cells and was associated with subsequent control of HBV infection10. These studies demonstrate a correlation between viral control and NKT cell activation. To investigate whether NKT cells are an important checkpoint that contributes to control of HBV infection, we studied various in vitro and in vivo models of HBV infection.

Results

Early activation of NKT cells in response to Ad-HBV

To study NKT cell responses, we investigated a mouse model that overcomes the nonpermissiveness of mouse hepatocytes to HBV through adenoviral delivery of a replication-competent HBV genome under the control of its endogenous promoters14,15,16. Injection of 1 × 109 HBV-expressing adenoviral particles (Ad-HBV), a dose shown by us (Supplementary Fig. 1a) and others16 to induce an immune response against HBV but not the adenoviral carrier, led to HBV replication (Supplementary Fig. 1b) followed by a rapid drop in hepatic HBV DNA and serum HBV surface antigen (HBsAg) that preceded hepatitis (Supplementary Fig. 1b,c). A control adenovirus expressing β-galactosidase (Ad-LacZ) did not lead to hepatitis, confirming HBV-dependent inflammation (Supplementary Fig. 1a)14,15,16. As Ad-HBV infection in mice resembles the course of natural HBV infection in humans17, we further studied NKT cell responses in this model.

Notably, the entire population of liver but not splenic iNKT cells showed activation and IFN-γ secretion as early as 1 d after Ad-HBV but not Ad-LacZ administration and thus before histological inflammation and the rise in serum alanine aminotransferase (ALT) levels (Fig. 1a–c and Supplementary Fig. 2). Similar to iNKT cells, hepatic but not splenic noninvariant NKT cells, an NKT cell population expressing a rather diverse set of TCRs that were detected by a reporter model developed for these studies (see Supplementary Data and Supplementary Fig. 3), showed pronounced activation and IFN-γ secretion in response to Ad-HBV (Fig. 1d–f and Supplementary Fig. 4). In contrast, activation of NK and T cells was not observed until 3 d after Ad-HBV challenge (Fig. 1g–i and data not shown) and thus followed activation of NKT cells, similar to observations in patients with HBV10.

Figure 1: NKT cells become activated in response to Ad-HBV and contribute to adaptive immune responses.
figure 1

(a–f) Expression of the indicated markers on hepatic iNKT cells (a–c, αGalCer–CD1d-tetramer+CD3+) and non-invariant NKT cells (d–f, 4Get × Jα18 KO) 2 d after intravenous administration of PBS (no virus), Ad-HBV or Ad-LacZ as determined by flow cytometry. In a–c, H-Mttp−/− (H-Mttp KO) and wild-type mice are shown. (g–i) NK (CD3−NK1.1+) and conventional (αGalCer–CD1d-tetramer−) CD8+ and CD4+ T cell activation 5 d after virus injection as described above. Shown is the mean fluorescence intensity (MFI). For all panels, data are means ± s.e.m. of 4–6 mice per group. Results are representative of three independent experiments.

Full size image

NKT cells contribute to the immune response against HBV

To investigate whether NKT cells contribute to innate and adaptive immune responses against HBV, we analyzed activation of liver mononuclear cells (LMNCs) in response to Ad-HBV in mice lacking invariant NKT cells (Jα18-deficient mice) or all NKT cells (CD1d-deficient mice). In wild-type mice, hepatic NK, CD4+ and CD8+ T cells showed strong activation and IFN-γ secretion in response to infection with Ad-HBV, but not Ad-LacZ (Fig. 1g–i and data not shown). In contrast, CD1d-deficient mice showed significantly diminished NK, CD4+ and CD8+ T cell activation, whereas Jα18 deficiency predominantly affected CD4+ T cell activation (Fig. 1g–i). Viral transduction of hepatocytes was similar in all mouse strains (Supplementary Fig. 5).

NKT-dependent activation was also observed for HBV-specific CD8+ T cell responses among liver mononuclear cells (LMNCs) against envelope (HBsAg, S, amino acids 190–197; VWLSVIWM) and HBV core (HBcAg, C, amino acids 93–100; MGLKFRQL) antigens analyzed ex vivo but not in response to phytohemagglutinin (Fig. 2a and Supplementary Fig. 6a,b). Analysis of LMNC responses to pools of peptides spanning the entire HBV envelope18 revealed pronounced defects in the magnitude but not diversity of HBV-specific CD8+ T cell responses in HBV envelope–transgenic × Rag1−/− × CD1d-deficient mice (HBVEnv Rag1−/− CD1d-deficient mice) that received, by adoptive transfer, wild-type splenocytes (Fig. 2b and Supplementary Fig. 6c). The adaptive immune defects extended to B cells, as levels of HBsAg-specific antibodies were lower in Ad-HBV–challenged CD1d-deficient and Jα18-deficient mice compared to wild-type mice(Fig. 2c) and in HBVEnv Rag1−/− CD1d-deficient mice compared to HBVEnv Rag1−/− mice (Fig. 2d), resulting in lack of sustained HBsAg clearance from serum (Fig. 2e). These data show that NKT cells contribute to activation of NK, T and B cells in different HBV mouse models.

Figure 2: NKT cells contribute to control of Ad-HBV and prevent chronic hepatitis.
figure 2

(a) IFN-γ secretion by LMNCs after in vitro stimulation with HBsAg S190–197 peptide (VWLSVIWM) 14 d after injection of the indicated viruses. (b) IFN-γ–secreting cells as detected by enzyme-linked immunospot (ELISPOT) among LMNCs obtained 8 d after wild-type splenocyte transfer into HBV envelope × Rag1−/− (HBVEnv Rag1−/−) and HBVEnv Rag1−/− CD1d-deficient mice (C57BL/6 background) and re-stimulation in the presence of the indicated HBV envelope peptide pools. (c) Anti-HBsAg level in wild-type mice 30 days after injection of the indicated viruses. (d,e) HBsAg-specific antibody (HBsAb) titers (d) and percentage of HBsAg-positive mice (e) at the indicated time after transfer of wild-type splenocytes into the indicated HBsAg-transgenic mouse strains. MIU, milli–international units. (f–j) Serum ALT (f), IFN-γ mRNA in liver tissue as determined by qPCR (g, day 5 after virus injection), histological grading of liver inflammation44 (h, day 30 after virus injection), serum HBsAg (i) and serum HBV DNA (j, day 7 after virus injection). In f and i, significance levels as indicated by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001) reflect from top to bottom Jα18 KO, CD1d KO, H-Mttp KO versus control wild-type mice at each relevant time point. Mean ± s.e.m. is shown in all panels (6–8 mice per group). Results are representative of two independent experiments.

Full size image

In accordance with a central role of NKT cells in HBV-induced hepatitis19,20, CD1d-deficient and Jα18-deficient mice showed reduced peak levels of serum ALT (Fig. 2f), liver IFN-γ (Fig. 2g) and hepatic immune cell infiltration (data not shown). However, CD1d-deficient and Jα18-deficient mice developed chronic low-grade inflammation that persisted at least to day 30 (Fig. 2f,h), the end of the observation period. This was associated with delayed clearance of serum HBsAg (Fig. 2i) and serum and liver HBV DNA (Fig. 2j and Supplementary Fig. 7d,e), consistent with defects in viral control leading to chronic inflammation. Consistent with a crucial role of CD8+ T cells in the antiviral response against HBV1, NKT cells were necessary but not sufficient for viral control, as Tap1−/− mice lacking CD8+ T cells also showed severe defects in HBV clearance (Supplementary Fig. 7). These results show that NKT cells contribute to the initiation of antiviral immune responses against HBV and to viral control and prevention of chronic viral replication in the Ad-HBV model.

HBV-infected human and mouse hepatocytes activate NKT cells

To identify the cell type responsible for NKT cell activation upon Ad-HBV challenge, we studied activation of NKT cell hybridomas in response to T cell–depleted liver and spleen mononuclear cells and primary hepatocytes obtained from Ad-HBV–infected mice. Ad-HBV–transduced hepatocytes but not T cell–depleted mononuclear cells induced activation of NKT cells (Fig. 3a, Supplementary Fig. 8a and data not shown). We observed similar activation of NKT cells upon in vitro exposure of primary hepatocytes to Ad-HBV (Supplementary Fig. 8b; for purity and transduction of primary hepatocytes see Supplementary Fig. 9).

Figure 3: NKT activation in response to HBV is mediated by hepatocytes and dependent on hepatocyte CD1d and MTP.
figure 3

(a) Activation of invariant (24.7, 24.8, 24.9, DN32.D3) and noninvariant (14S.6, 14S.7, 14S.10, 14S.15) NKT hybridomas in response to coculture with wild-type primary mouse hepatocytes 48 h after Ad-HBV transduction. (b) Activation of invariant 24.7 and noninvariant 14S.6 in response to primary hepatocytes of wild-type and CD1d-KO mice 48 h after virus administration. Viral transduction was similar in hepatocytes from both strains (see Supplementary Fig. 9). (c) Activation of the MHC class I–restricted T cell hybridoma RF33.70 in response to SIINFEKL (1 μg ml−1) presented by primary hepatocytes 48 h post infection. (d) Activation of human iNKT cell clones J24L.17 and J3N.5 and Jα24Vβ11-positive human iNKT cells (huNKT) expanded from peripheral blood in response to human hepatocytes infected with Ad-LacZ, Ad-HBV, or a primary HBV isolate. (e) IFN-γ secretion by J24L.17 NKT cells, as determined 72 h after HBV infection of primary human hepatocytes and 48 h after treatment with CD1d-blocking antibodies and MTP inhibitors (BMS212122, BMS197636) or a structurally related negative compound. (f,g) Activation of noninvariant 14S.6 (f) and invariant 24.7 (g) by primary hepatocytes from the indicated mouse strains 48 h after infection with the indicated viruses or vehicle. Cytokine secretion in a–g was determined by ELISA 24 h after coculture. Mean ± s.e.m. of triplicate cultures are shown. Results are representative of three independent experiments.

Full size image

NKT activation by Ad-HBV–infected hepatocytes was CD1d restricted (Fig. 3b) and limited to NKT cells, as demonstrated by unaffected major histocompatibility complex (MHC) class I presentation and CD8+ T cell activation (Fig. 3c). Activation was specific for a subset of invariant (24.7 cell line) and noninvariant (14S.6 cell line) mouse NKT cell hybridomas (Fig. 3a) and could also be demonstrated with human primary hepatocytes infected with Ad-HBV or a primary (nonadenoviral) HBV isolate (Fig. 3d,e). These data reveal an unanticipated role of hepatocytes in HBV-dependent NKT cell activation.

NKT cell activation is dependent on lipid transfer

Microsomal triglyceride transfer protein (MTP) is an ER-resident protein that transfers endogenous phospholipids onto CD1d and is crucial for CD1d function21,22,23,24,25. To delineate the role of hepatocyte MTP and CD1d in Ad-HBV infection, we generated mice with hepatocyte-specific deletion of Mttp, which codes for MTP (H-Mttp−/−26, Supplementary Fig. 10a). MTP deficiency led to severe defects in the presentation of endogenous and exogenous lipid antigens that were specific for hepatocytes and limited to CD1d (Supplementary Fig. 10b–e).

Consistent with a key role of MTP in HBV-induced and CD1d-restricted antigen presentation, primary hepatocytes from H-Mttp−/− mice showed impaired activation of NKT cells in response to Ad-HBV in vitro (Fig. 3f,g and Supplementary Fig. 10). We made similar observations upon chemical inhibition of MTP (Supplementary Fig. 11) and in human hepatocytes infected with a primary HBV isolate (Fig. 3e). These results were confirmed by in vivo experiments, in which NKT cells from Ad-HBV–challenged H-Mttp−/− mice showed reduced activation and impaired IFN-γ secretion (Fig. 1a–c). Accordingly, NKT cell–dependent activation of NK cells and HBV-specific T and B cells was impaired in H-Mttp−/− mice (Fig. 1g–i, Fig. 2a,c and Supplementary Fig. 6a). H-Mttp−/− mice showed diminished acute hepatitis, delayed viral control and chronic low-grade inflammation upon Ad-HBV challenge (Fig. 2 and Supplementary Fig. 7). These data confirm that hepatocytes are the antigen-presenting cells responsible for NKT cell activation and show that this effect is dependent upon hepatocyte expression of MTP and CD1d.

Alterations in hepatocyte lipids upon Ad-HBV infection

Ad-HBV–induced NKT cell activation was not the consequence of altered CD1d expression or trafficking (Supplementary Fig. 12a,b). Similarly, iNKT cell activation by hepatocytes loaded with the exogenous model lipid α-galactosylceramide (αGalCer) was not affected by Ad-HBV (Supplementary Fig. 12c). Given that nucleocapsid-free subviral HBsAg particles, which are produced at considerably higher numbers compared to complete HBV virions27, bud at ER and Golgi membranes28 and selectively recruit ER lipids for their envelopes29, we investigated whether HBV-induced NKT activation is the consequence of alterations in endogenous ER-acquired CD1d lipids. We studied activation of NKT hybridomas in response to plate-bound CD1d loaded with lipids obtained from Ad-HBV–infected primary mouse hepatocytes. We found that even uninfected primary hepatocytes contained antigenic CD1d lipids that were enriched in microsomal ER preparations (Supplementary Fig. 13a–e). However, microsomes from Ad-HBV– but not Ad-LacZ–transduced hepatocytes showed significantly increased activation of invariant and noninvariant NKT cell hybridomas (Fig. 4a,b) compared to microsomes from uninfected hepatocytes. Microsomal lipids obtained from CD1d-deficient and H-Mttp−/− hepatocytes showed similar NKT cell activation (Supplementary Fig. 13f and data not shown), indicating that antigenic lipids were unable to be loaded onto (H-Mttp−/−) or presented by (CD1d-deficient mice) CD1d. Accordingly, purified MTP enhanced presentation of immunogenic Ad-HBV–dependent lipids (Fig. 4c).

Figure 4: Microsomal lipids of Ad-HBV-infected hepatocytes contain NKT cell-activating lipid antigens.
figure 4

(a–c) IL-2 release of invariant 24.7 (a) and noninvariant 14S.6 (b) NKT cells in response to platebound CD1d loaded with microsomal lipids of hepatocytes infected with Ad-LacZ and Ad-HBV. In c, microsomal lipids were loaded in the presence or absence of purified MTP. ***P = 0.0001 of HBV versus HBV + MTP. (d,e) Activation of primary invariant (d) and noninvariant (e) NKT cells in response to plate-bound CD1d presenting microsomal lipids obtained from hepatocytes transduced with the indicated viruses. Invariant and noninvariant NKT cells were as sorted GFP+ αGalCer–CD1d-tetramer+ 4Get (d) and GFP+ 4Get × Jα18 KO liver mononuclear cells (e), respectively. Flow cytometric analysis of CD69 (MFI, left; histogram, right) gated on GFP+CD3+ cells is shown. Numbers in histograms indicate the percentage of cells that shift in expression of CD69 compared to microsomal lipids of uninfected cells (PBS). IL-12 neutralization demonstrates absence of indirect cytokine-mediated NKT cell activation as expected in an antigen-presenting cell–free assay. (f,g) Activation of noninvariant 14S.6 (f) and invariant 24.7 (g) NKT cells in response to plate-bound CD1d presenting chloroform-, acetone- or methanol-soluble lipids obtained 2 d after infection with Ad-LacZ and Ad-HBV. IL-2 secretion was determined by ELISA. Mean ± s.e.m. of triplicate cultures is shown. Results are representative of two (d,e) or three (a–c,f,g) independent experiments.

Full size image

Only a subset of NKT cell hybridomas recognized Ad-HBV–infected hepatocytes (Fig. 3a). Given that the TCRαβ repertoire of these hybridomas is unlikely to reflect that of liver NKT cells in vivo30, we investigated the response of primary liver-derived NKT cells to hepatocyte microsomal lipids. A significant population of sorted noninvariant liver NKT cells recognized Ad-HBV–induced alterations in microsomal lipids, as determined by upregulation of CD69 expression (Fig. 4d,e). CD69 upregulation was less pronounced for sorted iNKT cells, suggesting that iNKT cell responses to Ad-HBV are either restricted to a subpopulation of these cells or that iNKT cells are broadly but less strongly activated by Ad-HBV (Fig. 4d,e). In conclusion, exposure to Ad-HBV is associated with increased antigenicity of ER lipids, which are recognized by NKT cells.

Ad-HBV induces antigenic lysophosphatidylethanolamine

To characterize Ad-HBV–dependent lipid alterations, we separated microsomal lipids obtained from hepatocytes according to solubility31. Ad-HBV–induced stimulatory fractions for noninvariant and invariant NKT cells were both recovered in phospholipid-enriched methanol eluents, but subfractionation indicated that the two NKT cell subtypes were activated by different lipids (Fig. 4f,g). To characterize the particular lipids induced by Ad-HBV and stimulatory for NKT cells, we screened methanol fractions by HPLC mass spectrometry (HPLC-MS). Previous reports demonstrated selective recruitment of ER lipids by HBV subviral particles and underrepresentation of phosphatidylethanolamine (PE) species in the HBV envelope29. Comparison of chromatograms corresponding to the mass and collisional spectra of diacyl PE showed strong enrichment of PE and lysophosphatidylethanolamine (lysoPE) after Ad-HBV exposure (Supplementary Fig. 14). Lysophospholipids have recently been shown to be antigenic endogenous CD1d ligands in human cells in vitro32,33. Therefore, we rescreened stimulatory fractions at masses corresponding to monoacyl PE species with C16, C18:2, C18:1 and C18 monacyl forms and found in all cases that these lipids were upregulated after Ad-HBV administration in vivo (Fig. 5a). Thus, there is an overlap among the types of lipids upregulated in response to Ad-HBV and lysophosopholipids that can stimulate NKT cells.

Figure 5: HBV-mediated activation of noninvariant NKT cells is dependent on lysophospholipids and sPLA2.
figure 5

(a) Methanol fraction 2 of lipid extracts of microsomes obtained from Ad-HBV and Ad-LacZ-infected hepatocytes as analyzed by MS. Ion chromatograms corresponding in mass to the indicated lyso-PE species are shown. (b) IL-2 secretion of noninvariant 14S.6 and invariant 24.7 NKT cells in response to lipids presented by plate-bound CD1d is shown. PA, phosphatidic acid; PG, phosphatidylglycerol; PI, phosphatidylinositol. The dash separates the carbon chain length for both carbon chains in case of diacylated lipids. (c) IL-2 secretion by the 14S.6 noninvariant NKT cell hybridoma in response to lysophosphatidylethanolamine (LPE) and lysophosphatidic acid (LPA) presented by plate-bound CD1d and loaded in the presence or absence of purified MTP. ***P = 0.005, *P = 0.04 of LPE versus LPE + MTP. (d) Activation of primary noninvariant (left, middle) and iNKT (right) cells in response to plate-bound CD1d presenting the indicated lipids. NKT cells were obtained as sorted invariant GFP+ αGalCer–CD1d-tetramer+ 4Get (right) and noninvariant GFP+ 4Get × Jα18-KO liver NKT cells (left, middle). Flow cytometric analysis of CD69 gated on GFP+ CD3+ cells is shown. (e,f) Invariant 24.7 and noninvariant 14S.6 NKT cell responses to primary hepatocytes infected as indicated and treated with the sPLA2 inhibitor c(2NapA)LS(2NapA)R before addition of NKT cells. (g) Expression of PLA2 transcripts in mouse liver tissue after in vivo exposure to Ad-HBV. (h) Analysis of PLA2 RNA expression in primary human hepatocytes 24 h after in vitro infection. Mean ± s.e.m. of triplicate cultures are shown. ND, not detectable. (i,j) IL-2 release by NKT cells in response to primary hepatocytes infected with the indicated viruses before transfection with siRNAs as indicated and coculture with NKT cells. Results are representative of two (a,d) or three (b,c,e–j) independent experiments. All data are expressed as means ± s.e.m.

Full size image

To further test initial conclusions derived from complex cellular lipid mixtures generated in vivo, we tested activation of mouse NKT hybridomas in response to various purified phospholipids in vitro. 14S.6, a noninvariant NKT cell hybridoma that responded to Ad-HBV (Fig. 3a), showed activation in response to lysoPE and lysophosphatidylcholine (lysoPC) species containing hydrocarbon chains with lengths of 16 and 18 carbons (Fig. 5b). As hepatocyte lysoPE but not lysoPC abundance was increased in response to Ad-HBV (Supplementary Fig. 14), only lysoPE is implicated in the response to Ad-HBV. NKT cell activation by lysoPE was, among hybridomas, specific for 14S.6 and not observed with other iNKT hybridomas (24.7, 24.8, 24.9 or DN32.D3) or noninvariant NKT hybridomas (14S.7, 14S.10 or 14S.15; Fig. 5b and data not shown). Lyso-PE–induced NKT activation required plate-bound CD1d (data not shown), and purified MTP facilitated lysoPE presentation (Fig. 5c). These results extended to primary NKT cells, as >50% of sorted noninvariant liver NKT cells obtained from IL-4–IRES-GFP–enhanced transcript (4Get) × Jα18-deficient mice liver NKT cells showed upregulation of CD69 expression in response to lysoPE but not PE (Fig. 5d). Noninvariant NKT cell activation was not limited to cells expressing Vβ14, the TCRβ chain shared by the noninvariant NKT hybridoma (14S.6) that recognized lysoPE (Fig. 5d). This suggests that lysoPE broadly activates noninvariant NKT cells in the liver and further reiterates the limited overlap between currently available NKT hybridomas and primary NKT cells in vivo. In contrast, iNKT cells did not show activation in response to lysoPE (Fig. 5d).

NKT cell activation is dependent on secretory phospholipases

Secretory phospholipase A2 (sPLA2) enzymes contribute to the generation of lysophospholipids recognized by NKT cells and are active within secretory compartments33,34. We therefore investigated the role of sPLA2 in Ad-HBV–dependent NKT cell activation. Chemical inhibitors of sPLA2 but not cytoplasmic PLA2 (cPLA2) prevented HBV-dependent activation of noninvariant NKT cells without affecting MHC class I presentation (Fig. 5e and Supplementary Fig. 15a–g). sPLA2 inhibitors also did not affect iNKT cells, in accordance with the lack of lysoPE-dependent iNKT activation (Fig. 5f).

As Ad-HBV–transduction of hepatocytes leads to increased abundance of PE (Supplementary Fig. 14b), the substrate for PLA2-mediated production of lysoPE, we investigated hepatocyte gene expression of sPLA2 (gene names listed in Fig. 5g), cPLA2 (Pla2g4a) and calcium-independent PLA2 (iPLA2, Pla2g6a) enzymes. Ad-HBV led to rapid and selective upregulation of sPLA2 enzymes Pla2g2c and Pla2g5 but not other sPLA2, cPLA2 and iPLA2 enzymes in vivo (Fig. 5g and Supplementary Fig. 16a). These results are in accordance with increased hepatic PLA2 group II and V expression in patients with viral hepatitis35,36. Primary mouse hepatocytes also showed Ad-HBV–induced Pla2g2c upregulation, confirming hepatocyte-specific effects and demonstrating that sPLA2 regulation is not secondary to hepatic inflammation but a direct consequence of exposure to Ad-HBV (Supplementary Fig. 16b). Primary human hepatocytes exposed to Ad-HBV or a primary HBV isolate also showed upregulation of sPLA2, specifically of PLA2G2A (Fig. 5h). As lysoPE activates transcription of PLA2G2A37, it most likely supports its own production through a positive-feedback loop involving sPLA2. Thus, HBV leads to upregulation of sPLA2 group II in mouse and human hepatocytes. Specific subgroups of sPLA2 differ between humans and mice in accordance with differences in the genomic organization of sPLA2 loci (PLA2G2C is a pseudogene in humans38; Pla2g2a is inactivated by a frameshift in C57BL/6J mice).

Consistent with a crucial role for sPLA2 in HBV-dependent generation of lysophospholipids as major antigens recognized by noninvariant NKT cells, siRNA directed against Pla2g2c and Pla2g5 (Supplementary Fig. 17a) but not the genes encoding other sPLA2, cPLA2 and iPLA2 enzymes prevented Ad-HBV–induced activation of noninvariant NKT but not invariant NKT cells or MHC class I–restricted T cells (Fig. 5i,j and Supplementary Fig. 17b and data not shown).

HBsAg contributes to NKT cell activation

To investigate the requisite HBV element for NKT cell activation, we inactivated individual HBV open reading frames and studied the effects on NKT activation. Although all viral mutants showed similar transduction rates (Supplementary Figs. 5 and 9), only deletion of the small HBsAg (HBsAg S−), which is required for viral secretion, led to impaired activation of noninvariant NKT cells (Fig. 6a). We observed a similar but nonsignificant trend for iNKT cells (Fig. 6b). Ad-HBV S− led to impaired hepatocyte sPLA2 upregulation and impaired activation of noninvariant NKT cells by microsomal hepatocyte lipids (Fig. 6c and Supplementary Fig. 16c).

Figure 6: Expression of HBsAg and cytokines contribute to NKT cell activation.
figure 6

(a,b) IL-2 release by NKT cells in response to primary hepatocytes infected in vitro with the indicated Ad-HBV mutants deficient in the large (L−), middle (M−), and small (S−) HBsAg or carrying a point mutation in the myristylation site of the large HBsAg (Myr−), which were cocultured 24 h after infection with noninvariant (14S.6, a) and invariant (24.7, b) NKT cell hybridomas. NS, not statistically significant. (c) NKT cell IL-2 release in response to microsomal lipids of primary hepatocytes obtained 48 h after infection with the indicated viruses and presented by plate-bound CD1d to the noninvariant NKT cell hybridoma 14S.6. (d–f) NKT cell IL-2 release in response to primary hepatocytes of the indicated mouse strains treated with MTP inhibitors or 9-fluorenyl carboxylic acid as a negative control compound for 48 h. Hepatocytes were cocultured with noninvariant 14S.6 NKT cells (d), invariant 24.7 NKT cells (e) or MHC class I–restricted RF33.70 T cells (f). (g,h) NKT cell IL-2 release in response to microsomal lipids of the indicated mouse strains loaded onto plate-bound CD1d. Presentation to invariant 24.7 (g) and noninvariant 14S.6 (h) is shown. IL-2 secretion by T cell hybridomas in a–h was determined by ELISA. (i,j) Expression of CD69 as determined by flow cytometric analysis of GFP+ αGalCer–CD1d-tetramer+ CD3+ invariant NKT cells (i) and GFP+ αGalCer–CD1d-tetramer− CD3+ noninvariant NKT cells obtained from 4Get mice injected with the indicated viruses, LPS (indirect NKT cell activation) or αGalCer (direct iNKT cell activation) in the presence or absence of antibody-mediated IL-12 neutralization. Liver mononuclear cells were obtained 6 h (αGalCer) and 24 h (viruses, LPS) after stimulation. (j). Representative histograms and dot plots are shown. Bar graphs show mean ± s.e.m. Results are representative of two (d–j) or three (a–c) independent experiments.

Full size image

To extend these findings, we examined mice transgenically expressing HBsAg19. HBsAg-transgenic mouse hepatocytes led to CD1d-dependent activation of noninvariant and invariant NKT cell hybridomas but not MHC class I–restricted T cell hybridomas (Fig. 6d–f). In addition, microsomal lipids of HBsAg-transgenic hepatocytes activated NKT cells (Fig. 6g,h). In vivo administration of Ad-HBV S− resulted in prolonged chronic necroinflammation and delayed viral control in association with sustained viral replication and impaired hepatic IFN-γ induction (Supplementary Fig. 18). These studies show that the small HBsAg contributes to Ad-HBV–mediated activation of NKT cells, which is in accordance with a temporal association between HBsAg expression and early NKT cell activation in human HBV infection10.

iNKT cell activation in vivo is cytokine dependent

Primary liver iNKT cells were not activated by lysoPE (Fig. 5d) and showed less robust activation in response to Ad-HBV microsomal lipids (Fig. 4d). We therefore investigated whether cytokine-mediated indirect activation contributes to Ad-HBV–induced iNKT cell activation in vivo. Indeed, neutralization of interleukin-12 (IL-12), a cytokine essential for indirect NKT cell activation39,40,41, largely prevented CD69 upregulation and IFN-γ secretion by iNKT cells but had negligible effects on noninvariant NKT cells (Fig. 6i,j and Supplementary Fig. 19). These results suggest that noninvariant NKT cells are directly activated upon Ad-HBV exposure by CD1d-restricted presentation of lysoPE and other uncharacterized hepatocyte antigens. This, in turn, may lead to cytokine-dependent activation of iNKT cells, presumably through activation of liver dendritic cells and macrophages39,40. Notably, IL-12–mediated indirect activation of liver iNKT cells requires hepatocyte MTP (Fig. 1a–c), suggesting that cytokine-dependent iNKT activation is downstream of CD1d-restricted activation of noninvariant NKT cells and NKT-dependent dendritic cell maturation42,43.

Discussion

The findings presented here show that immune responses provided by NKT cells contribute not only to HBV-induced hepatitis19 but also to viral control. Although HBV-induced NKT cell activation is associated with acute hepatitis, NKT cells contribute considerably to the emergence of antiviral B and T cell immunity. In the absence of this NKT cell response, adaptive immunity and viral control are diminished and chronic, low-grade hepatitis ensues. We show that the cell type responsible for orchestrating these HBV-induced responses is previously unappreciated and, notably, the infected hepatocyte itself and not a professional antigen-presenting cell. Within the hepatocyte, HBV induces the production of ER lipids, including lysophospholipids derived from PE through the action of HBV-induced secretory phospholipases. CD1d-restricted presentation of lysophospholipids leads to direct activation of noninvariant NKT cells and subsequent IL-12–mediated indirect iNKT cell activation, confirming that both direct and indirect mechanisms contribute to NKT cell activation in vivo39,40.

The generation of antigenic lipids is partially dependent on the presence of the HBV surface antigen, the principal HBV structural component responsible for assembling PC-rich particles in the ER, and possibly results from an imbalanced lipid milieu within the ER29. As such, the hepatocyte uses secretory phospholipases, MTP and CD1d to sense the presence of HBV and alert NKT cells through the display of modified self-lipids on the cell surface of the hepatocyte to trigger an adaptive immune response.

Our findings of an NKT cell response soon after HBV exposure are in accordance with other recent observations in humans and animal models of HBV10,11,12 and suggest that NKT cells are part of an early, important sensing system that activates the immune response, leading to effective priming of HBV-specific adaptive immune cells that are required for viral clearance. Our data suggest that although HBV acts as a stealth virus during a prolonged phase preceding viral control9, it is susceptible to a distinct type of immune recognition directly following infection, which is important for subsequent immune clearance. Further studies in HBV-infected humans are required to confirm a central role of NKT cells in human HBV infection.

Methods

Mice.

CD1d-deficient, Jα18-deficient and Tap1−/− mice have been described previously45,46,47. H-Mttp−/− mice with an hepatocyte-specific deletion of the Mttp gene encoding microsomal triglyceride transfer protein (MTP) were generated by crossing Alb-Cre (B6.Cg-Tg(Alb-cre)21Mgn/J) mice48 that express Cre recombinase under control of the hepatocyte-specific albumin promoter with Mttpfl/fl mice that contain loxP sites flanking Mttp exon 1 (ref. 49). To generate mice that allow for specific detection of noninvariant NKT cells, 4Get mice expressing GFP via an internal ribosome entry site (IRES) in the IL-4 transcript (IL-4–IRES-GFP–enhanced transcript (4Get) mice)50,51 were crossed with CD1d-deficient and Jα18-deficient mice (see Supplementary Data for further information). All mice were maintained on C57BL/6J background and were backcrossed for at least eight generations. HBV-Env mice expressing the entire HBV envelope (subtype ayw) under control of the albumin promoter52 were crossed with Rag1−/− and CD1d-deficient mice as described previously19.

Primary cells and cell lines.

Primary human hepatocytes were obtained from Invitrogen (Carlsbad, CA). Primary mouse hepatocytes were extracted as described previously53. Purity and transduction rates of primary hepatocytes are shown in Supplementary Figure 9. A description of all cell lines can be found in the Supplementary Methods.

Lipids and chemical inhibitors.

Information on lipids and chemical inhibitors can be found in the Supplementary Methods.

Viral constructs and virus administration.

The parental plasmid for HBV constructs was pGEM7-HBV1.3 containing a 1.3-fold-overlength genome of HBV, subtype ayw, with a 5′ terminal redundancy encompassing enhancers I and II, the origin of replication (direct repeats DR1 and DR2), the pregenomic/core promoter regions, the transcription initiation site of the pregenomic RNA, the unique polyadenylation site and the entire X open reading frame. To generate HBV mutants, point mutations were introduced using the QuickChange II site-directed mutagenesis kit and protocol (Agilent Technologies, Santa Clara, CA) at the nineteenth codon of the HBsAg pre-S1 gene from TTG to TAG (large (L) HBsAg-deficient viral mutant; L−) or at the initiation codon of pre-S2 and S from ATG to GTG (middle (M) HBsAg-deficient viral mutant, M−; small (S) HBsAg-deficient viral mutant, S−). The pre-S1 myristylation-defective (myr−) mutation was generated by changing the second codon of the pre-S1 gene from GGG (glycine) to GCG (alanine).

Additional information on virus generation and application can be found in the Supplementary Methods. Viral transduction rates in vivo and in vitro are shown in Supplementary Figures 5 and 9.

Determination of ALT, HBsAg and HBV DNA levels and liver histology.

Details can be found in the Supplementary Methods.

Flow cytometry.

Flow cytometry was performed as described previously25. Details can be found in the Supplementary Methods.

Preparation and separation of cellular and microsomal lipids.

For extraction of microsomal lipids, primary hepatocytes were obtained as described above, microsomes were extracted according to methods used by Ernster et al.54 and lipids were extracted following the Folch protocol55. Total hepatocyte lipids were extracted in a similar manner. Separation of lipids according to solubility was done as described in ref. 31. Details can be found in the Supplementary Methods.

Liquid chromatography–mass spectrometry.

Lipid extracts were dried down, weighed and then resuspended in 95% 60:40 hexanes/isopropanol and 5% methanol (vol/vol) before mass spectrometry analyses. Twenty micrograms of lipid from each fraction was injected onto a 150 mm × 2.0 mm monochrom 3 diol column (Varian) and eluted with a gradient program in which solvent A consisted of methanol (0.1% formic acid and 0.05% ammonium hydroxide, wt/vol) and solvent B consisted of 60% hexanes and 40% isopropanol (vol/vol) (0.1% formic acid and 0.05% ammonium hydroxide wt/vol). The gradient was run from 95% to 85% solvent B from 0 to 6.6 min, to 0% solvent B until 16.2 min, and then increased back to 95% solvent B from 22.8 to 26 min. The HPLC system was an Agilent 1200 series with a 6520 quadrupole accurate mass time of flight (Q-TOF) mass spectrometer (Agilent Technologies, Santa Clara, CA), with the voltage 3.5 kV, source temperature 325 °C, drying gas 5 l/min, nebulizer pressure 30 p.s.i., running in the negative ion mode. Collision experiments were carried out by subjecting target ions to 25 eV in the collision cell. Data analysis was performed using Agilent Qualitative Analysis Mass Hunter Software Version B.03.01.

Antigen presentation.

Antigen presentation assays were performed in 96-well flat-bottom plates using 2 × 104 hepatocytes, 5 × 104 RMA-S/d cells, 1 × 105 splenocytes or hepatic mononuclear cells and, as responders, 1 × 105 NKT cells or NKT cell hybridoma cells. Cytokine secretion was determined by ELISA after 16 h of coculture (BD Biosciences). In some experiments, MTP inhibitors were added at a final concentration of 10 μM (BMS212122, BMS200150) and 100 nM (BMS197636). Alternatively, purified monoclonal CD1d-blocking antibodies (1B1 clone, 19G11 clone) were used at a final concentration of 10 μg/ml.

For cell-free antigen-presentation assays, monomeric mouse CD1d (NIH Tetramer Core Facility) was loaded onto 96-well flat-bottom plates (0.25 μg per well), unbound CD1d was washed off and lipids were added at a molar ratio of 40:1. Unbound lipids were then washed off, NKT cells were added and cytokine secretion was determined by ELISA as described above. In some experiments, MTP purified from bovine liver (M.M.H.) was added at a final concentration of 500 ng/ml. In addition, in some experiments, GFP+ αGalCer–CD1d-tetramer+ 4Get invariant NKT cells and GFP+ 4Get × Jα18-deficient noninvariant NKT cells were sorted from liver mononuclear cells using a BD Biosciences FACSAria II cell sorter and were used as responders in antigen presentation assays. Flow cytometric determination of cell surface CD69 and intracellular IFN-γ expression on GFP+CD3+ NKT cells was studied as readout. For in vitro neutralization of IL-12, a monoclonal anti–IL-12 p40 antibody (clone C17.8, eBioscience) was used at a final concentration of 10 μg/ml.

Details of ELISPOT assays can be found in the Supplementary Methods.

Protein extraction and western blotting.

Protein extraction and western blotting were performed as described previously25. A detailed description can be found in the Supplementary Methods.

Real-time PCR.

Real-time PCR was performed as described previously25. Primer sequences are listed in Supplementary Table 1. A detailed description can be found in the Supplementary Methods.

RNA interference.

Inhibition of phospholipase expression was achieved by FlexiTube siRNA (Qiagen Inc., Hilden, Germany). siRNA was transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Downregulation of PLA2 expression was investigated by SYBR-Green quantitative PCR 48 and 72 h after transfection. For antigen presentation assays, primary hepatocytes were infected with Ad-HBV and control viruses and were transfected with siRNA 24 h after infection. NKT cells were added 48 h after siRNA transfection, and cytokine secretion was determined after 16 h of coculture.

Statistical analyses.

Data are expressed as means ± s.e.m. Statistical testing was done using the unpaired Student's t test. For comparisons of more than two groups, one-way analysis of variance was performed and was followed by Dunnett's correction. Statistical analyses were done using GraphPad Prism (GraphPad Software, Inc.).

Additional methods.

Detailed methodology is described in the Supplementary Methods.