|Home | About | Journals | Submit | Contact Us | Français|
While hepatitis C virus (HCV) has been shown to readily escape from virus-specific T and B cell responses, its effects on natural killer (NK) cells are less clear. Based on two previous reports that recombinant, truncated HCV E2 protein inhibits NK cell functions via crosslinking of CD81, it is now widely believed that HCV impairs NK cells as a means to establish persistence. However, the relevance of these findings has not been verified with HCV E2 expressed as part of intact virions. Here we employed a new cell culture system generating infectious HCV particles with genotype 1a and 2a structural proteins, and analyzed direct and indirect effects of HCV on human NK cells. Antibody-mediated crosslinking of CD16 stimulated and antibody-mediated crosslinking of CD81 inhibited NK cell activation and IFN-γ production. However, infectious HCV itself had no effect even at titers that far exceeded HCV RNA and protein concentrations in the blood of infected patients. Consistent with these results, anti-CD81 but not HCV inhibited NK cell cytotoxicity. These results were independent of the presence or absence of HCV-binding antibodies and independent of the presence or absence of other peripheral blood mononuclear cell populations.
HCV 1a or 2a envelope proteins do not modulate NK cell function when expressed as a part of infectious HCV particles. Without direct inhibition by HCV, NK cells may become activated by cytokines in acute HCV infection and contribute to infection outcome and disease pathogenesis.
Natural killer (NK) cells are an important component of the innate immune response against many viruses due to their ability to lyse virus-infected cells and to secrete cytokines that inhibit viral replication and activate and recruit cells of the adaptive immune response. The role of NK cells during the early antiviral immune response has been well described in mice. Murine cytomegalovirus (MCMV) activates myeloid dendritic cells through the engagement of toll-like receptor 9, which results in interferon (IFN)-α and interleukin (IL)-12 production and activation of NK cells (1). Activated NK cells lyse virus-infected cells and produce IFN-γ (2) and thereby limit MCMV replication in liver and spleen (3). Thus, cytokine production and cytotoxicity of NK cells contribute to the control of viral spread prior to the development of adaptive immune responses.
Viruses, however, have also developed mechanisms to escape from the antiviral response of NK cells and to establish persistent infection (4). Human cytomegalovirus prevents NK-cellmediated lysis of infected fibroblasts by down-regulating the expression of molecules such as MICA (5) and CD155 (6), which are ligands for activating NK cell receptors. Hepatitis C virus (HCV) is characterized by an extraordinary ability to establish viral persistence, which it achieves by evading intracellular responses of infected hepatocytes (7) and humoral and cellular responses of the adaptive immune system (8). However, HCV is also induces type I interferons, which are detectable in the liver within the first weeks of infection (9) and are known as potent stimulators of NK cells. Activated NK cells have been shown to recognize and lyse HCV-replicon containing hepatoma cells in vitro in a perforin- and granzyme-dependent manner (10). It has therefore been postulated that HCV may have develop strategies to persist in the face of early NK cell responses.
This concept appeared to be validated by two reports that a plate-bound, truncated, recombinant form of the HCV envelope 2 (E2) glycoprotein binds to the tetraspanin CD81 expressed by NK cells (11, 12). In those studies, plate-bound HCV E2 protein crosslinks CD81 molecules on the surface of NK cells and reduces the capacity of NK cells to respond to Fcγ receptor IIIA (CD16) ligation with IFN-γ secretion and cytotoxicity. Based on these findings, it has been stated in several reviews that HCV E2 impairs NK cell function via direct contact as a means to establish persistence (13–15). The observed delay of adaptive immune responses (8) could be a consequence of an inhibitory effect of HCV on NK cell function early in the course of infection, because IFN-γ secretion by NK cells promotes downstream adaptive immune responses (16).
However, it is yet to be established whether those effects observed with high concentration (1 – 10 µg/ml) of recombinant, plate-bound, HCV E2 protein (11, 12) can be reproduced with intact virions. In this scenario, the virus surface itself would provide the platform for repetitive patterns of E2 proteins and thereby facilitate crosslinking of CD81 molecules. However, it is also possible that the configuration of the HCV E2 protein on the surface of infectious virions differs from that of truncated recombinant E2 used in the in vitro assays (11, 12) and its concentration in the blood of infected patients is likely much lower.
The recently developed HCV culture system (17–20) enabled us to test the effect of infectious HCV particles on NK cells under controlled in vitro conditions. Specifically, we asked whether acute exposure to HCV affected activation, IFN-γ production and cytotoxicity of NK cells from healthy controls, who had never been exposed to HCV. NK cells, either purified or in the context of other cell populations were exposed to cell-culture produced hepatitis C virus for up to 18 hours, in the presence or absence of virion-specific antibodies and at HCV RNA and HCV protein concentrations similar to or higher than those observed in the blood during acute HCV infection (21).
Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats or blood samples of healthy blood donors as described (22) and used fresh unless indicated otherwise. Sera of individuals who had cleared persistent HCV genotype 2 infection were preserved at −80°C. All subjects gave written informed consent for research testing under protocols approved by Institutional Review Boards of the National Institutes of Health.
Anti-CD3-FITC(UCHT1), anti-CD3-PE-Cy5(UCHT1), anti-CD3-Alexa700(UCHT1), anti-CD14-APC(M5E2), anti-CD16-Pacific Blue(3G8), anti-CD19-PE-Cy5 (HIB19), anti-CD25-FITC(M-A251), anti-CD69-PE(FN50), anti-IFN-γ-PE-Cy7(B27) (all from BD Pharmingen, San Jose, CA), anti-CD56-PE(AF12–7H3), anti-CD56-APC(AF12–7H3) (all from Miltenyi Biotec, Berglisch Gladbach, Germany), and anti-CD14-PE-Cy5(TÜK4) (Serotec, Raleigh, NC) were used for flow cytometry. Anti-CD16(3G8) and anti-CD81(JS-81) (all from BD Pharmingen) were used for NK cell stimulation and inhibition, respectively.
HCV JFH1 strain (genotype 2a) (18) and the chimeric H-NS2/NS3-J virus, which expressed HCV core to NS2 proteins of genotype 1a (H77) sequence and the remaining nonstructural proteins of JFH1 sequence (23), were produced as described (24) by transfecting Huh7.5 cells (Apath, St. Louis, MO) with linearized RNA from plasmids (provided by Dr. T. Wakita, National Institute of Infectious Diseases, Tokyo, Japan and Dr. S. Lemon, University of Texas Medical Branch, Galveston, TX, respectively). Transfected Huh7.5 cells were cultured in DMEM with 10% fetal bovine serum (FBS; US Bio-Technologies, Pottstown, PA), 10 mM HEPES, 100 IU/ml penicillin, 100 µg/ml streptomycin and 2 mM L-glutamine (Mediatech, Herndon, VA). Cell culture supernatants at the peak of HCV production (typically days 12 through 16 after transfection) were used to infect naïve Huh7.5.1 cells (provided by Dr. F. V. Chisari, Scripps Research Institute, La Jolla, CA) in 10 cm2 culture dishes at a multiplicity of infection of 0.006. The infected Huh7.5.1 cells were transferred into T175 flasks after 1–2 days and passaged every 3.5 days. The viability of HCV-infected Huh7.5.1 cells was assessed by trypan blue staining and lactate dehydrogenase release using the CytoTox96 assay (Promega Corp., Madison, WI).
Filtered (0.45 µm) supernatant of HCV-infected Huh7.5.1 cells containing HCV was used for NK cell experiments. HCV RNA, infectious titer and HCV core protein concentration were determined exactly as previously described (24). To determine the HCV E2 concentration cell culture supernatants were incubated with 0.1% NP-40 for 15 minutes at 4°C followed by a Galanthus nivalis antigen (GNA) lectin capture EIA (25). Soluble JFH1 E2 of known concentration (provided by Dr. J. McKeating, University of Birmingham, UK) was included as calibrant. In experiments reported in figure 1–figure 4 and supplementary figure 1 the JFH1 infectious titer was 4 × 103 ffu/ml (5 × 107 RNA copies/ml; 0.03 nmol HCVcore/l), comparable to those in the blood of infected patients (26). Supernatant from uninfected Huh7.5.1 cells was used as control. To achieve an even higher viral titer (Fig. 5), 500 ml supernatant from a separate batch of infected Huh7.5.1 cells was concentrated using a Pellicon XL filter system (Millipore, Billerica, MA) with a pressure pump. This resulted in an increase of the JFH1 infectious titer to 2.39 × 105 ffu/ml (2 × 108 HCV RNA copies/ml, 2.99 nmol HCVcore/l, 12 ng HCV E2/ml). The infectious titer of the chimeric H-NS2/NS3-J virus was 2.93 × 105 ffu/ml (2 × 108 HCV RNA copies/ml, 3.4 nmol HCVcore/l, 37 ng HCV E2/ml). Concentrated supernatant from uninfected Huh7.5.1 cells served as control.
96-well EIA plates (Dynex Technologies, Chantilly, VA) were coated with 1 µg/ml GNA lectin (Sigma, St. Louis, MO) overnight at 4°C and incubated with 1.8 × 107 RNA copies HCV/well for 2h at room temperature. Fifty microliter HCV-RNA-negative serum of individuals who had cleared HCV genotype 2 or normal healthy donors were added at a 1:20 dilution. After 1h at room temperature, plates were washed three times with PBS and the captured antibodies were detected with mouse anti-human IgG-horseradish peroxidase (Serotec) and tetramethylbenzidine super sensitive 1-C (BioFX Laboratories, Owings Mills, MD) (25). Optical density was determined at 450 nm with a microplate reader (Bio-Rad, Hercules, CA).
Huh7.5.1 cells were incubated with 5 × 107 RNA copies/ml HCV in the presence of 20% human serum. The percentage of HCV-infected Huh7.5.1 cells was assessed by flow cytometry using an anti-HCVcore antibody and a secondary, Alexa488-conjugated anti-mouse IgG (Invitrogen). Percent neutralization was calculated as (control infectivity – experimental infectivity) / control infectivity × 100.
NK cells were negatively isolated from PBMCs with the NK Cell Isolation Kit and autoMACS separator (Miltenyi Biotec) after lysis of red blood cells with ACK lysing buffer (Biosource, Rockville, MD). The purity of isolated CD3−CD56+ NK cells as assessed with anti-CD3-FITC, anti-CD56-PE, and anti-CD14-APC antibodies by flow cytometry was greater than 91% in all experiments.
106 cells/ml NK cells were cultured for 18h in RPMI medium containing 100 IU/ml penicillin, 100 µg/ml streptomycin and 2 mM L-glutamine (Mediatech) and either 10% FBS or 20% human serum (complete medium) in (i) 96-well flatbottom culture plates (Corning Inc., Corning, NY) with the indicated concentration of HCV in Huh7.5.1 supernatant or with supernatant from uninfected Huh7.5.1 cells or in (24) 96-well flat-bottom EIA plates (Immulux HB; Dynex Technologies) that had been coated with 1 µg/ml anti-CD16 and/or 10 µg/ml anti-CD81 in carbonate buffer (pH 9.5) overnight at 4°C. Human IL-12 (1.2 ng/ml, PeproTech, Rocky Hill, NJ), which is typically secreted by monocytes and DCs, was added in all stimulation conditions. As previously described (27, 28) and shown in figure 1, this suboptimal concentration IL-12 alone did not induce significant activation of resting NK cells but primed them to respond to activating stimuli.
To examine possible inhibitory effects of HCV and anti-CD81 on NK cell stimulation, NK cells were incubated with the indicated concentration of HCV or plate-bound anti-CD81 2h before, at the same time or 2h after stimulation with plate-bound anti-CD16 or 50 µg/ml poly(I:C) (InvivoGen, San Diego, CA) and 1.2 ng/ml recombinant human IL-12.
After 18h culture supernatants were frozen at −20°C and later tested with the Human IFN-γ Quantikine EIA Kit (R&D systems, Minneapolis, MN). Cells were washed, stained with anti-CD25-FITC, anti-CD69-PE, anti-CD3-PE-Cy5, and anti-CD56-APC antibodies and analyzed on a flow cytometer (FACSCalibur; BD Biosciences, San Jose, CA) using CellQuest v3.3 (BD Biosciences) and FlowJo v6.4.7 software (Tree Star, Ashland, OR).
Thawed PBMCs were incubated in complete medium at 37°C in a CO2 incubator for 8h prior to NK cell isolation. Two thousand 51Cr-labeled (GE Healthcare, Piscataway, NJ) MHC class I negative K562 cells/well (ATCC, Manassas, VA) were plated in 96-well round-bottom plates (Nunc, Roskilde, Denmark) and in parallel in 96-well round-bottom EIA plates (Immulux HB; Dynex Technologies) coated or not coated with 10 µg/ml anti-CD81. NK cells were added at effector to target (E:T) ratios of 30:1, 15:1, 7.5:1, and 3.75:1 with or without 5 × 107 RNA copies/ml HCV in complete RPMI medium supplemented with 20% human serum with or without HCV JFH1-neutralizing antibodies. Virus was incubated with serum for 30 minutes prior to the use in NK cell assays. The amount of 51Cr released in the culture supernatant was determined after 3h, 5h and 7h using a scintillation counter (PerkinElmer, Wellesley, MA). Percent specific cytotoxicity was calculated as (experimental release – spontaneous release) / (maximum release – spontaneous release), and the mean cytotoxicity of at least three cultures per E:T ratio was calculated. Spontaneous release was less than 30% of maximum release in all assays.
PBMCs were stimulated at 2.5 × 106 cells/ml as described for NK cells. Brefeldin A (1 µl/ml, Golgi Plug; BD Biosciences) was added after 1h. PBMCs were stained with 50 µg/ml ethidium monoazide (EMA; Invitrogen Molecular Probes), anti-CD3-Alexa700, anti-CD14-PE-Cy5, anti-CD16-Pacific Blue, anti-CD19-PE-Cy5, and anti-CD56-APC antibodies after 18h. After fixation and permeabilization with BD Cytofix/Cytoperm solution (BD Biosciences), cells were stained with anti-IFN-γ-PE-Cy7, anti-CD25-FITC, and anti-CD69-PE antibodies and analyzed on a flow cytometer (LSR II; BD Biosciences) using FACSDiva v4.1.2 (BD Biosciences) and FlowJo v6.4.7 software. NK cells were identified by gating on singlets (FSC-A vs. FSC-H plot), lymphocytes (FSC-A vs. SSC-A plot) and finally on CD3−CD56+ NK cells after exclusion of CD14+, CD19+, and EMA+ (dead) cells.
Paired Student t test analysis was performed with GraphPad Prism (San Diego, CA). A two-sided P value of less than 0.05 was considered statistically significant.
Freshly isolated, peripheral blood NK cells of healthy blood donors responded to stimulation by plate-bound anti-CD16 in the presence of IL-12 with a significant upregulation of the activation markers CD25 (IL-2 receptor α chain, Fig. 1A,B) and CD69 (Fig. 1A,C) and secretion of IFN-γ (Fig. 1D). Most responding NK cells were CD56dim (Fig. 2A). This correlated well with the observation that nearly all CD56dim NK cells are CD16 (Fcγ receptor IIIA)bright (not shown) and that CD56bright NK cells are typically CD16− or CD16dim. Similar results were observed using poly(I:C) for stimulation (not shown). In contrast, neither HCV, derived from the culture supernatant of HCV JFH1 infected Huh7.5.1 cells, nor plate-bound anti-CD81 stimulated CD25 and CD69 expression (Fig. 2B–D) or IFN-γ production (Fig. 2E). These results show that IL-12-primed primary human NK cells can be activated by anti-CD16, but neither by HCV nor by anti-CD81.
Crosslinking of CD81 by plate-bound anti-CD81 resulted in significant inhibition of activation marker expression (Fig. 2A–D) and IFN-γ secretion (Fig. 2E) by NK cells in response to anti-CD16. Inhibition of NK cell activation was observed for all NK cells, but stronger for the CD56dim than CD56bright subset (Fig. 2A).
Two previous reports (11, 12) described crosslinking of CD81 on NK cells by high concentration of plate-bound, recombinant HCV E2 protein, which resulted in inhibition of NK cell responsiveness to anti-CD16 (11). To examine whether HCV E2 also inhibits NK cells when expressed as a part of intact, infectious HCV particles, we stimulated NK cells with anti-CD16 and IL-12 in the presence or absence of HCV. Infectious HCV JFH1 was harvested as supernatant of JFH1-infected Huh7.5.1 cells and used at 5 × 107 RNA copies/ml, a concentration 10 to 100-fold higher than typically observed in chronic HCV infection (26) and comparable to HCV RNA titers in acute HCV infection (29). However, in contrast to previous reports with high concentrations of plate-bound, recombinant HCV E2 protein (11, 12), infectious HCV JFH1 did neither inhibit the expression of the activation markers CD25 and CD69 (Fig. 2A–D) nor the production of IFN-γ (Fig. 2E) by anti-CD16-stimulated NK cells. These results held true, whether NK cells were incubated with HCV 2h prior to (Fig. 2C–E), at the same time or 2h after stimulation with anti-CD16 (not shown). Similar results were also obtained using poly(I:C) rather than anti-CD16 for NK cell stimulation (not shown). In contrast, plate-bound anti-CD81 inhibited activation marker expression and IFN-γ secretion of NK cells over a wide concentration range and as low as 40 ng/ml (not shown) and irrespective of whether it was administered 2h prior to, at the same time or 2h after NK cell activation. These results demonstrate that HCV E2 as a part of intact, infectious HCV JFH1 virions does not efficiently crosslink CD81 and does not inhibit NK cell function.
To investigate the possibility that HCV affects NK cells indirectly via interaction with other cell populations, HCV was directly added to PBMCs. As shown in supplementary Fig. 1, HCV did not induce CD69 expression (Suppl. Fig. 1A,B) and IFN-γ production by NK cells (Suppl. Fig. 1A,C). Likewise, HCV did not inhibit CD69 expression and IFN-γ production of anti-CD16 stimulated NK cells. Thus, infectious HCV JFH1 virions affect NK cell function neither directly nor indirectly via interaction with other peripheral blood cells.
Another potential mechanism for HCV to interact with NK cells is via antibodies bound to its surface. In infected patients, antibodies against HCV structural proteins cover HCV virions and thereby create repetitive patterns of extruding Fc fragments, which may engage and crosslink CD16 (FcγRIIIA) on NK cells. To evaluate whether HCV modulates NK cell function via this mechanism, we had to identify antibodies against genotype 2a JFH1 structural proteins. For this purpose, we screened sera from 11 subjects who had previously recovered from HCV genotype 2 infection and who tested HCV RNA negative by RT-PCR. As shown in Table 1, sera of subjects 4 and 6 contained antibodies that bound to HCV JFH1 in an EIA and neutralized in vitro infection of Huh7.5 cells by 79.4% and 88.9%, respectively.
The NK cell activation (Fig. 1) and inhibition (Fig. 2) experiments were then repeated in the presence of serum from subject 4 or control serum from a healthy uninfected subject. As shown in Fig. 3, HCV JFH1 did not stimulate NK cell activation and IFN-γ production and did not inhibit anti-CD16-mediated activation of NK cells in the presence of antibodies. In addition to activation marker expression and IFN-γ secretion, we also investigated cytotoxicity as a readout of NK cell activation and effector function. As shown in Fig. 4A, HCV JFH1 did not significantly enhance or inhibit NK cell cytotoxicity in the presence of virus-specific antibodies. In contrast, crosslinking of CD81 on NK cells by anti-CD81 substantially inhibited NK cell cytotoxicity (p<0.05, Fig. 4B).
Because a plate-bound, truncated HCV E2 protein of genotype 1a sequence was used to inhibit NK cell activation and function in the two previously published reports (11, 12) we asked whether a potential effect of HCV on NK cells could be genotype-specific. For this purpose we used the recently described chimeric H-NS2/NS3-J virus that expressed the core, E1, E2, p7 and NS2 proteins of the HCV genotype 1a (H77) sequence and the remaining nonstructural proteins of genotype 2a (JFH1) sequence [Fig. 5A, (23)]. The HNS2/NS3-J virus replicated well in infected Huh7.5.1 cells, but HCV RNA titers in the cell culture supernatant reached maximum titers 3 days later than the JFH virus (Fig. 5B) consistent with a 3.5-day delayed cytopathic effect evidenced by an increase in lactate dehydrogenase release (Fig. 5C) and a decrease in the percentage of viable Huh7.5.1 cells (Fig. 5D).
Parallel to testing the effect of H-NS2/NS3-J virus on NK cells, we repeated some of the previous JFH1 experiments at a very high concentration which typically exceeds that of chronically HCV-infected patients (26). For this purpose, supernatant of JFH1-infected Huh7.5.1 cells was concentrated to 2 × 108 RNA copies/ml. H-NS2/NS3-J virus was used at the same titer and concentrated supernatant from uninfected Huh7.5.1 cells served as control. As shown in Fig. 5E–G, anti-CD16 stimulated NK cell activation and IFN-γ production. In contrast, neither high titer HCV H-NS2/NS3-J virus nor high titer HCV JFH1 virus modulated NK cell activation or effector functions (Fig. 5H–J).
In this study, we examined the hypothesis that NK cell functions are inhibited by short-term exposure to HCV E2. Previous reports used high concentrations of truncated, recombinant, platebound, HCV E2 protein to crosslink CD81 on NK cells (11, 12) and reported inhibition of NK cell activation and IFN-γ production. The recently developed tissue-culture system for HCV allowed us to study the effects of HCV E2 as part of a complete, infectious HCV particle. Consistent with the previous reports (11, 12), crosslinking of CD81 with plate-bound monoclonal antibodies inhibited NK cell activation and effector functions in our experimental setting. However, HCV virions with either genotype 2a or 1a structural proteins did not affect NK cell activation or effector functions. These results suggest that HCV E2 does not efficiently crosslink CD81 on the surface of NK cells when it is displayed as part of infectious virions. Likewise, HCV-antibody complexes did not affect NK cell functions via crosslinking of FcγRIIIA (CD16) receptors.
The discrepancies between our findings and the previous reports (11, 12) are most likely due to differences in the configuration of natural E2 on the surface of HCV virions as compared to that of plate-bound, recombinant E2. In addition, differences in the HCV E2 concentration may play a role. Whereas very high concentrations of 1 and 10 µg/ml HCV E2 protein were used previously, the E2 concentrations in our study were much lower even though RNA titers were comparable to those in the blood of HCV-infected patients (21).
We also considered potential indirect effects of HCV on NK cells that may involve other PBMC subpopulations, such as dendritic cells and monocytes. Dendritic cell (24) and monocyte functions (21) have been shown to be affected by HCV and HCV core protein, respectively. To investigate whether HCV affected NK cell function via an immune network with other cell populations, HCV was added to the complete PBMC population and NK cell function was assessed by multicolor flow cytometry. However, HCV virions did not inhibit NK cells in this setting. Likewise, the presence of HCV-specific antibodies on the surface of virions did not modulate NK cell function.
A potential limitation of the current and previous studies on NK cells in HCV infection is the use of peripheral blood NK cells. NK cells constitute a large population of liver-resident lymphocytes and have never been studied in acute infection. Virions could for example become enriched on liver resident cells such as sinusoidal lining cells which express HCV-binding proteins DC-SIGN and L-SIGN. Also, hepatocytes may upregulate MHC class I in response to viral proteins and thereby avoid recognition by NK cells (30). The specific inflammatory micromilieu, and specifically cytokines released by virus-infected hepatocytes may play an important role in this context. These effects cannot yet be analyzed in vitro, because the HCV-permissible cell lines Huh7.5 and Huh7.5.1 are impaired in their type I IFN response (31). It also remains possible that other NK cell functions other than activation, IFN-γ production and cytotoxicity are altered.
It is important to note that our results are most relevant for the early acute phase of HCV infection because we analyzed direct short-term exposure of NK cells to HCV, mostly in the absence of HCV-specific antibodies. In this context, Khakoo et al reported in an immunogenetic study (32) that specific KIR/HLA compound haplotypes, which have been shown to be associated with more rapid and more vigorous NK cell activation and effector function (33), are associated with a higher likelihood of HCV clearance and lower likelihood of chronic hepatitis C. As regards to chronic HCV infection, there are multiple reports that NK cells are activated and display effector functions (34–37). The questions how the early innate immune response impacts the development of adaptive immunity and to which extent activated NK cells contribute to the pathogenesis of liver disease in chronic infection therefore require further analysis in detailed prospective studies.
We thank Drs. Takashi Wakita, Stanley Lemon and MinKyung Yi for providing HCV expression constructs, Dr. Jane McKeating for advice and reagents for the E2 EIA, and Drs. Takanobu Kato, Sukanya Raghuraman and Eui-Cheol Shin for discussion.
This study was supported by the NIDDK, NIH intramural research program. GA was supported by grant AH173/1-1 from Deutsche Forschungsgemeinschaft (DFG), Bonn, Germany.