Potential conflict of interest: Nothing to report.
Both strong antigenic avidity and acquisition of proper effector functions contribute to the efficacy of antiviral T cell responses. To correlate these parameters with the outcome of hepatitis C virus (HCV) infection, we characterized HCV-specific CD8 T cell lines isolated after immunomagnetic sorting of peripheral blood mononuclear cells from human leukocyte antigen A*02 (HLA-A*02) individuals with various HCV serological statuses, using recombinant HLA-A*0201 multimers loaded with three immunodominant HCV genotype 1-derived epitopes. CD8 T cells specific for these three epitopes were derived from most HLA-A*0201 individuals, regardless of their HCV serology or clinical outcome. Donors recovered from genotype 1 HCV infection were enriched for high-avidity T cells with enhanced interferon gamma (IFN-γ), tumor necrosis factor alpha, and cytotoxic T lymphocyte responses, when compared with seronegative donors and seropositive patients infected with irrelevant HCV genotypes. Patients chronically infected with genotype 1 strain yielded almost exclusively low-avidity T cells, whose hyporesponsiveness was primarily attributable to low T cell receptor (TCR) avidity rather than intrinsic functional defects. Conclusion: This study suggests that strong IFN-γ responses associated with efficient viral clearance primarily result from Ag-driven selection/survival of HCV-specific T cells expressing high-avidity TCR. It also suggests a link between the quality of the initial HCV-specific T cell repertoire and susceptibility to chronic infection. (HEPATOLOGY 2008.)
Efficient control of viral infections in various animal and human models depends on a wide array of mechanisms affecting swiftness and strength of antiviral cellular and humoral immune responses. Diversity of such mechanisms is highlighted by analysis of immune responses elicited by hepatitis C virus (HCV) in humans. Control of acute HCV infection correlates with strong, sustained, and broad virus-specific cellular immunity mediated by both CD4 and CD8 T cells. By contrast, HCV persistence has been associated with transient T cell responses in infected donors.1 Two nonmutually exclusive mechanisms could explain inefficient immune control of HCV in chronically infected patients. Mutations of several HCV epitopes, described both in chimpanzees and humans, can lead to decreased viral recognition by CD8 T cells.2–5 Functional decline of HCV-reactive T cells resulting from progressive loss of interleukin-2 (IL-2) secretion, proliferation, interferon gamma (IFN-γ) secretion, or cytotoxicity may also account for inefficient viral clearance by T cells directed against nonmutated epitopes.6–8 Mechanisms contributing to this progressive functional defect remain unclear but could involve (1) lack of CD4 T cell help,7, 9 (2) defective antigen-presenting cell function,10 (3) impairment of HCV-specific immunity by regulatory T cells,11 (4) expression of inhibitory receptors,12, 13 or (5) T cell exhaustion caused by chronic antigenic stimulation. The overall weak T cell responses observed along with viral persistence could also result from “holes” in the HCV-specific T cell repertoire or a preexisting functional defect of HCV-specific T cells before infection.
Analysis of the mechanisms underlying HCV immune control has been hampered by the difficulty of getting large numbers of HCV-specific T cells, not only from patients with different clinical status but also from seronegative donors. To address this point, we applied a protocol allowing efficient isolation and unbiased amplification of HCV-specific CD8 T cells from seropositive patients with different clinical status and healthy seronegative donors not presumably at risk of infection.14, 15 In-depth comparative analysis of their functional features indicates that differential selection of HCV-specific T cells expressing high-avidity T cell receptor (TCR) along with HCV infection correlates with virological outcome.
Ag, antigen; CTL, cytotoxic T lymphocyte; HCMV, human cytomegalovirus; HCV, hepatitis C virus; HLA, human leukocyte antigen; IFN, interferon; IL-2, interleukin-2; mAb, mononuclear antibody; PBL, peripheral blood leukocyte; PE, phycoerythrin; pMHC, peptide major histocompatibility complex; TCR, T cell receptor; TNF, tumor necrosis factor.
Patients and Methods
Donors and Samples.
Samples of peripheral blood were obtained from 26 anti-HCV–seropositive individuals: 17 HCV RNA–positive chronically infected patients with (n = 3) or without hepatocarcinoma (n = 14) and nine HCV RNA–negative long-term recovered donors. No HCV–chronically infected patients had received antiviral therapy at least within the 6 months before specimen collection. Subjects R1, R2, R3, and R9 spontaneously resolved HCV viremia, whereas subjects R4 to R8 were treated with IFN/ribavirin therapy and subsequently resolved HCV viremia. Blood samples were also obtained from 10 healthy donors; all were anti-HCV seronegative and without any HCV risk factors. This latter cohort was also seronegative for human cytomegalovirus (HCMV). Peripheral blood mononuclear cells were isolated from fresh heparinized blood by Ficoll density gradient centrifugation (LMS, Eurobio). Human leukocyte antigen (HLA) class I genotyping was performed by Etablissement Français du Sang (Nantes, France). All subjects were HLA-A*0201–positive. The presence of HCV RNA was evaluated by qualitative polymerase chain reaction in the serum of donors (HCV Roche Amplicor assay, detection limit of 100 HCV RNA copies/mL of plasma). All patients included gave informed consent to participate in the study, which was approved by a local ethical committee and regulatory agencies.
RNA extraction, reverse transcription polymerase chain reaction, and nested polymerase chain reaction were performed as previously described.16 Primers were designed according to HCV Sequence Database (http://hcv.lanl.gov) to an amplified NS3 region containing NS31073–1081 epitopes from genotypes 1a-1b, 2c-2k, and 3a (primers sequences and predominant epitope sequences are shown, respectively, in Supplementary Tables S1 and S2). Amplicons were sequenced using sequencer 3730 (Applied Biosystems) and Big Dye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems).
Peptides and Generation of Recombinant Peptide Major Histocompatibility Complexes.
The following HLA-A*0201–restricted peptides were used: NS31073–1081 (CINGVCWTV), NS31406–1415 (KLVALGINAV), and NS4b1807–1816 (LLFNILGGWV) deriving from the nonstructural subtype 1a viral proteins, NA231–239 (CVNGSCFTV) deriving from influenza virus IV, and pp65495–503 (NLVPMVATV) deriving from HCMV. Peptides were synthesized at greater than 90% purity by ProImmune (UK). Soluble peptide major histocompatibility complex (pMHC) monomers were synthesized as previously described.14, 15 The pMHC monomers used in this study were mutated HLA-A*0201 carrying an Ala to Val substitution in the alpha 3 domain at position 245, which enhances both sorting and staining efficiency.14
Immunomagnetic Cell Sorting Using pMHC-Loaded Microbeads and T Cell Expansion.
T cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS), 1 mM L-glutamine, and 150 IU/mL recombinant IL-2. Immunomagnetic sorting of antigen-specific T cells using recombinant pMHC complexes loaded onto streptavidin-coated magnetic microbeads was performed on CD8 sorted T lymphocytes derived from peripheral blood mononuclear cells as described.14, 15 Cells were then expanded in vitro under nonspecific stimulation in IL-2 (150 IU/mL)/CM supplemented with leukoagglutinin at 1 μg/mL, irradiated peripheral blood leukocytes (PBLs), and B lymphoblastoid cells as previously described,17 and maintained for 3 weeks without restimulation before analysis. Purity of antigen-specific T cells was checked by multimer staining.
Staining with Phycoerythrin-Labeled pMHC Complexes and Antibodies.
pMHC monomers were oligomerized with either phycoerythrin (PE)-labeled streptavidin (BioSource, Belgium) to form high-order oligomers referred to as “multimers,” or with a tetramer grade PE-labeled streptavidin (Molecular Probes, France) to form low-order oligomers referred to as “tetramers,” at a molar ratio of 4: 0.8 during 1 hour 20 minutes at 4°C.15 The degree of oligomerization of these two streptavidin-PE/pMHC conjugates was evaluated by gel filtration as previously described.15
T cells were stained as described,15 with 10 μg/mL PE-labeled pMHC complexes or the following monoclonal antibody (mAb) directed against CD3 (Immunotech), PD-1 (BD Biosciences, MIH4 clone) and PD-L1 (eBiosciences) at 4°C during 30 minutes. Data analysis was performed using CellQuest Pro software (Becton Dickinson, France).
Cytotoxicity was measured as described,15 using 51Cr-labeled transporter associated with antigen processing–deficient HLA-A*0201–positive T2 cells pulsed with various concentrations of HCV peptides (effector/target ratio: 10/1). For redirected killing assays, murine mastocytoma P815 cells were 51Cr-labeled and pulsed with various concentrations of anti-CD3 (OKT3).
For cytokine intracellular staining, 2.5 × 105 lymphocytes were incubated with T2 cells loaded with 100 nM peptides (E/T ratio: 1/1). Protein-secretion inhibitor brefeldin-A (1 mM) was added after 2 hours' incubation. Intracellular staining assays were performed as previously described18 with anti-IFN-γ or anti-tumor necrosis factor alpha (TNF-α) mAb (100 ng/mL, BD Pharmingen, Biosciences).
For degranulation assays, CD107a-specific mAb (anti-CD107a-FITC, BD-Pharmingen) was added to effector NS31073-specific CD8 T cells at the beginning of the incubation time with T2 cells loaded with relevant (100 nM) or irrelevant peptides. Staining with anti-Vβ mAbs was performed after a 4-hour incubation.
Data were analyzed using GraphPad Prism 4.0 software (San Diego, CA). Nonparametric Kruskal-Wallis test was performed to compare measurements from the different cohorts.
Isolation of Highly Purified HCV-Specific T Cell Lines from HCV-Seropositive and HCV-Seronegative Donors.
We previously described efficient isolation and nonbiased amplification of human antigen (Ag)-specific CD8 T cells through immunomagnetic sorting with mutated recombinant HLA class I/peptide complexes showing decreased affinity for the CD8 coreceptor (see Patients and Methods).14, 15 To compare HCV-specific T cells in seronegative versus seropositive donors with different clinical outcomes, we applied this strategy to generate in large numbers, and without any antigen specific stimulation, highly purified CD8 T cells directed against three previously described immunodominant HLA-A*0201–restricted HCV epitopes (NS31073, NS31406, and NS41807) derived from the HCV genotype 1 strain. Although HCV multimer-positive cells were not or barely detectable within unsorted CD8 T cells, highly purified HCV-specific T cell lines were generated from almost all seropositive genotype 1 donors regardless of their clinical status (Fig. 1A and Table 1). Strikingly, highly purified HCV genotype 1–specific T cells were also efficiently obtained from most seropositive non–genotype 1 patients (ie, recovered or infected with viral strains that do not express the relevant epitopes, Supplementary Table S2) and nonexposed seronegative controls (Fig. 1A and Table 1). Specificity of these cell lines was supported by lack of staining by irrelevant HLA-A2 multimers or by recognition of target cells loaded with relevant, but not irrelevant, antigenic peptides (Fig. 1B). These cell lines were polyclonal as shown in Supplementary Table S3.
Table 1. Sorting Efficiency of HCV-Specific CD8 T Cells
HCV Serological Status
Donor Status (Genotype)
% NS31073 Multimer+ CD8+ T Cells
% NS31406 Multimer+ CD8+ T Cells
% NS4b1807 Multimer+ CD8+ T Cells
Sorted with NS31073/A2 Magnetic Beads
Sorted with NS31406/A2 Magnetic Beads
Sorted with NS4b1807/A2 Magnetic Beads
The percentages of NS31073-specific, NS31406-specific, and NS41807-specific T cells stained with multimers before and after immunomagnetic sorting in HCV-seropositive spontaneously recovered donors (R), chronically infected patients (C and HCC), and seronegative healthy donors (H) are shown. The background staining with irrelevant multimers was subtracted from the percentage of cells stained with relevant multimers.
Abbreviation: ND, not determined.
Percentage of specific T cells below 0.1%: in this case, it was not possible to distinguish a multimer-positive subset from the staining background. CD8 T cells from which specific T cells were obtained are bolded.
Two successive immunomagnetic sortings were performed to achieve the indicated purity of pMHC-specific T cells.
Differential Ag-Responsiveness of HCV-Specific T Cells from Recovered Versus Chronically Infected Patients Expressing the Relevant Viral Genotype.
Whereas all but one NS31073-specific T cell line derived from recovered donors efficiently lysed peptide-loaded target cells, 10 of 16 cell lines from chronic patients were not or poorly cytotoxic at high peptide concentrations (100 nM; P < 0.05; Fig. 2A, upper panel). Similarly, Ag-stimulated NS31073-specific T cell lines from recovered donors displayed stronger IFN-γ and TNF-α responses than those from chronic patients (P < 0.01; Fig. 2A, middle and lower panels).
To test whether such differences were linked to an in vivo Ag-driven process, we compared the functional behavior of NS31073-specific T cells from individuals infected with or recovered from viral genotypes expressing (genotype 1) or not (genotypes 2 and 3) the relevant epitope (Supplementary Table S2). Cytolytic and cytokine responses of NS31073-specific T cells significantly differed between genotype 1 recovered and chronically infected patients (P < 0.05 for cytotoxicity and TNF-α, P < 0.01 for IFN-γ; Fig. 2B, dashed lines) but not between non–genotype 1 individuals (Fig. 2B, dotted lines).
Surprisingly, one cell line deriving from a genotype 1–infected donor (#C15, Fig. 2 arrows) yielded strong cytotoxic T lymphocyte (CTL) and cytokine responses. Sequencing of the NS31073 region of HCV isolates showed that the wild-type epitope was left unmutated in chronic patients carrying hyporesponsive T cells, whereas the predominant viral sequence in patient C15 carried one mutation at position 6 (Cys to Phe substitution) associated with strongly decreased T cell response (Supplementary Fig. S1 and Supplementary Table S2).
Importantly, similar functional differences between HCV-seropositive patients were confirmed when analyzing T cell responses directed against another HCV epitope (NS31406; Supplementary Fig. S2A). Moreover, this overall differential responsiveness between infected and recovered donors only applied to HCV-specific T cell responses, because T cells directed against a dominant HLA-A*0201–restricted HCMV epitope (pp65495) displayed similar cytotoxic and cytokine responses when isolated from recovered or chronic patients (data not shown).
TCR Avidity Underlies Differential Responsiveness Between NS31073-Specific T Cells Isolated from HCV Genotype 1 Recovered and Chronically Infected Donors.
NS31073-specific and NS31406-specific T cells from patients recovered or infected with genotype 1 virus yielded similar CTL and cytokine responses after stimulation by phorbol 12-myristate 13-acetate (PMA)/calcium ionophore or with grading doses of anti-CD3 mAb (Fig. 3). Therefore, these functional differences were not accounted for by intrinsic defects of effector functions or deficient TCR signaling. Moreover, we failed to detect either PD1 or PD1L on any of the NS31073-specific T cell lines derived from genotype 1 donors (not shown).
These observations suggested instead that differences in overall functional avidity underlie differential responsiveness of NS31073-specific and NS31406-specific T cells isolated from genotype 1 patients. Indeed, analysis of CTL responses to graded doses of antigenic peptide revealed significant differences between genotype 1 recovered and infected patients (P < 0.05) but not between patients with nongenotype 1 virus infection or recovery (results expressed as half maximal effective concentration in Fig. 4A and Supplementary Fig. S2B).
To evaluate the contribution of TCR affinity/avidity in these functional differences, we examined the ability of NS31073-specific T cell lines from genotype 1 patients to bind multimers (that is, pMHC-oligomers of high valency multimerized with aggregated streptavidin) versus tetramers (that is, pMHC-oligomers of low valency multimerized with “tetra-grade” streptavidin), because these tools allow discrimination between high-avidity versus low-avidity TCR/CD8 complexes.15, 19 Whereas NS31073-specific T cells from recovered donors were stained by both multimers and tetramers, most cell lines from chronic patients bound only to multimers (P < 0.01, Fig. 4B, C). Collectively these results indicate an overall higher TCR affinity/avidity of cell lines derived from genotype 1 recovered donors compared with genotype 1 infected patients.
Avidity Differences Between NS31073-Specific T Cell Lines from Genotype 1 Recovered or Chronically Infected Patients Also Apply to TCR Vβ Subsets and T Cell Clones.
Because NS31073-specific T cells expressed a diverse TCR Vβ repertoire (Supplementary Table S3), we assessed functional avidity of each Vβ subset by analyzing Ag-induced surface translocation of CD107a.20 Most Vβ subsets identified in T cell lines from genotype 1 recovered donors strongly up-regulated CD107a after specific Ag stimulation, unlike Vβ subsets from chronic patient-derived cell lines (Fig. 5). Therefore, the overall functional behavior of NS31073-specific polyclonal cell lines reflects that of most clonotypes found in these cell lines. Accordingly, T cell clones derived from an NS31073-specific T cell line from patient #C11 were hyporesponsive, unlike specific T cell clones derived from patient #R8 (Supplementary Fig. S3).
Functional Analysis of Specific T Cell Lines from Seronegative Healthy Donors.
Although IFN-γ responses of NS31073-specific and NS41406-specific T cells from seronegative donors were significantly weaker than those from genotype 1 recovered donors (Fig. 2B middle panel and S2), their overall cytotoxic and TNF-α responses or antigenic avidities did not differ significantly from genotype 1 donor T cells (Figs. 2B, 4A). However, the magnitude of antigenic avidities and cytotoxic and TNF responses was broader in NS31073-specific T cells from seronegative donors than in genotype 1 subjects (Figs. 2B, 4A). This suggests that seronegative-derived NS3-specific T cells are heterogeneous in terms of high-avidity T cell frequencies. Accordingly, some Vβ subsets identified among several seronegative-derived NS31073-specific cell lines showed strong CD107a up-regulation after Ag stimulation (Fig. 5). In line with these results, staining ratios between tetramers and multimers of seronegative donor-derived NS31073-specific T cells lines ranged between those of T cells derived from genotype 1 infected and recovered subjects (Fig. 4B, C). Collectively, these results show that high/intermediate avidity subsets directed against HCV are found within polyclonal T cell lines derived from some seronegative donors, but in any case represent a small fraction only of most of these polyclonal T cell lines (Fig. 5 and Supplementary Table S3), consistent with their overall low antigenic avidity.
NS31073-Specific, NS31406-Specific, and NS41807-Specific T Cells from Seronegative Donors Could Be Isolated from Both the CD45RA+CD27+ and CD45RA−CD27+ Subsets.
To determine whether HCV-specific T cells isolated from healthy donors were derived from naive or memory subset, PBLs from HCV seronegative healthy donors were stained with CD45RA-specific and CD27-specific mAbs and sorted into CD45RA−CD27+ and CD45RA+CD27+ T cell subsets. Although NS31073-specific T cells were isolated from presumably memory (CD45RA−) T cells in one donor, NS31073-specific, NS31406-specific, and NS41807-specific T cells were enriched within CD45RA+ CD27+ subsets, consistent with a naive phenotype (Fig. 6), and in one donor could not be isolated from the CD45RA− T cell compartment, thus ruling out in this case contamination by cells derived from the memory compartment. Importantly, NS31073-specific T cell lines derived from seronegative donors were not or poorly reactive against the homologous Flu-derived NA231 peptide in cytotoxicity and cytokine release assays (not shown), thus not favoring the possibility that NS31073-specific T cells isolated from seronegative healthy donors correspond to T cells directed against a cross-reactive Ag. Moreover, all HCV-seronegative donors included in this assay were also HCMV-seronegative, and we failed to isolate T cells directed against a dominant HLA-A*0201–restricted HCMV epitope (pp65495) in these donors. Collectively, our results suggested an increased frequency of T cells directed against these HCV epitopes in seronegative donors, possibly reflecting a naive repertoire bias linked to positive selection.
We performed an in-depth functional analysis of HCV-specific T cells in nonexposed versus infected individuals to better understand the mechanisms that influence progression to chronic infection versus viral control. Because frequency of HCV-specific CD8 PBLs was below the detection threshold in most individuals, CD8 T cell subsets reactive against several immunodominant HCV epitopes were sorted and expanded before analysis. Availability of highly purified HCV-specific T cells from almost all donors then allowed a systematic comparison of their functional properties. NS31073-specific and NS31406-specific T cell populations derived from recovered donors yielded strong cytolytic and proinflammatory cytokine responses after antigenic stimulation, in particular in terms of IFN-γ production. By contrast, and in line with previous studies, most T cell lines derived from chronically infected patients were hyporesponsive. Importantly, these functional differences between recovered and chronically infected patients only applied to cell lines derived from individuals exposed to genotype 1 HCV strains, that is, those that carried the relevant epitopes, thus suggesting a direct link between this functional behavior and antigen exposure in vivo.
The availability of T cells with the same fine specificity from a large panel of individuals also provided an opportunity to address the immunological mechanisms underlying these differential responses. Our results do not favor an intrinsic or acquired functional defect of these populations, but suggest instead that specific T cells from genotype 1 chronic patients, unlike those from recovered patients, express low-avidity TCR/CD8 complexes. Finally, HCV-specific T cells derived from HCV-seronegative donors displayed a broad range of antigenic avidities, and low frequencies of intermediate-avidity T cells were identified in some but not all donors. Collectively our results suggest that enrichment for high-avidity Ag-specific T cells along HCV infection is associated and most probably required for efficient viral clearance. By contrast, inefficient selection/survival of high-avidity clonotypes during infection or a preexisting lack of such T cells before infection may lead to persistent viremia.
We are aware of the technical limitations of this study. In particular, analysis of sorted T cells directed against defined pMHC complexes does not give any indication about ex vivo frequencies of specific T cells (which are in many patients too low to be accurately assessed, Table 1) and thus precludes quantitative analysis of our results. Besides, we cannot formally rule out that particular HCV-specific T cells are lost during the in vitro expansion process, even though our stimulation protocol using polyclonal activators and high amounts of recombinant IL-217, 21 should allow efficient proliferation of the full HCV-specific T cell repertoire from both HCV-recovered and chronic patients.17, 21 Accordingly our HCV-specific T cell lines displayed heterogeneous functional properties that closely mimicked those of ex vivo specific T cells.6, 12, 13, 22–24 Finally, because this study focused on peripheral blood mononuclear cell–derived T cell lines, we cannot rule out a specific trapping of high-avidity T cells within the liver in chronically infected patients. Although an in-depth analysis of liver-derived HCV-specific T cells would be required to directly address this issue, the assumption that high-avidity T cells are enriched in patients having recovered from HCV infection would better fit with studies reporting a correlation between selection of high-avidity T cells and clearance of several other viral infections.25, 26 This is also in agreement with a recent report suggesting higher functional avidity of NS3-specific T cells derived from recovered donors than chronically infected patients,27 although this study did not allow direct comparison of CTL responses recognizing the very same pMHC complex nor a direct evaluation of TCR/pMHC-avidity.
The rather common occurrence of HCV-specific T cells in most healthy seronegative donors raises questions about the origin and possible contribution of these T cells to the early control of HCV infection. T cells directed against NS31073, NS31406, and NS41807/HLA-A*0201 complexes were readily sorted out from most HCV seronegative donors, unlike those directed against several immunodominant CD8 epitopes from HCMV or human immunodeficiency virus in seronegative donors (Supplementary Table S4). This could reflect prior exposure to HCV along cryptic infection or transient seroconversion.28, 29 Alternatively, because all donors included in our study were not presumably at risk, HCV-specific T cells might also represent memory T cells cross-reactive to other environmental agents.30, 31 Lack of in vitro cross-reactivity of HCV-specific T cells to previously described Flu epitopes, which is in agreement with a recent study showing that the NA231 Flu epitope was a weak agonist for NS31073-specific T cell lines,32 and enrichment for HCV-specific T cells within the naive compartment in several donors argue against this possibility, but suggest instead a repertoire bias directly linked to positive selection or peripheral homeostatic processes. In this regard, the wide range of avidity of HCV-specific responses is reminiscent of features of other human naive subsets directed against melan-A/HLA-A2 complex.33 As previously suggested,34 particular structural features of NS31073, NS31406, and NS41807/HLA-A*0201 complexes could underlie the increased frequency of the corresponding specific T cells in the naive repertoire.
With respect to the functional relevance of these observations in the context of HCV infection, one should note that although some seronegative donor-derived cell lines yielded strong CTL and TNF-α responses, their IFN-γ responses remained in all instances much weaker than those from genotype 1 recovered patients, as a likely consequence of their lower functional avidity. Because IFN-γ, unlike CTL or TNF-α, correlates with efficient control of chronic viral infections,25, 26 preexistence of such HCV-specific T cells with intermediate functional avidity in seronegative donors is unlikely to confer any protection against viral infection and modulate susceptibility to chronic HCV infection. Further longitudinal studies in donors at risk of infection are needed to address this important issue and analyze the mechanisms responsible for altered clonal selection in chronic HCV infection.
The authors thank Dr. Emmanuel Scotet for helpful discussion and careful reading of the manuscript.