Potential conflict of interest: Nothing to report.
The mouse monoclonal antibody (MAb) AP33, recognizing a 12 amino acid linear epitope in the hepatitis C virus (HCV) E2 glycoprotein, potently neutralizes retroviral pseudoparticles (HCVpp) carrying genetically diverse HCV envelope glycoproteins. Consequently, this antibody and its epitope are highly relevant to vaccine design and immunotherapeutic development. The rational design of immunogens capable of inducing antibodies that target the AP33 epitope will benefit from a better understanding of this region. We have used complementary approaches, which include random peptide phage display mapping and alanine scanning mutagenesis, to identify residues in the HCV E2 protein critical for MAb AP33 binding. Four residues crucial for MAb binding were identified, which are highly conserved in HCV E2 sequences. Three residues within E2 were shown to be critical for binding to the rat MAb 3/11, which previously was shown to recognize the same 12 amino acid E2 epitope as MAb AP33 antibody, although only two of these were shared with MAb AP33. MAb AP33 bound to a panel of functional E2 proteins representative of genotypes 1-6 with higher affinity than MAb 3/11. Similarly, MAb AP33 was consistently more efficient at neutralizing infectivity by diverse HCVpp than MAb 3/11. Importantly, MAb AP33 was also able to neutralize the cell culture infectious HCV clone JFH-1. In conclusion, these data identify important protective determinants and will greatly assist the development of vaccine candidates based on the AP33 epitope. (HEPATOLOGY 2006;43:492–601.)
Hepatitis C virus (HCV) infection is a major cause of severe chronic liver disease, including cirrhosis and hepatocellular carcinoma.1 With an estimated 3 to 4 million new infections each year,1 an urgent need exists for the development of an effective prophylactic vaccine.
HCV, a member of the Flaviviridae family, has a 9-kb genome encoding a polyprotein precursor that is cleaved to yield three structural proteins, core, E1, and E2, together with at least six non-structural proteins. E1 and E2 are highly glycosylated membrane-anchored proteins that mediate viral entry.2–6 E2 has been shown to bind to a number of cell surface molecules.7–10 Whereas the exact mechanism of viral entry is unknown, mounting evidence indicates that CD81 and SR-BI are key molecules.2, 4–6, 11–14
In an infected individual, HCV exists as a viral quasispecies.15 HCV can be classified into at least six major genotypes that exhibit extensive genetic variability, particularly in E1 and E2.16 E1 and E2 are the principle targets for neutralizing antibodies, and identification of protective epitopes conserved across different strains of HCV is a major challenge for vaccine design.17 A number of antibodies that are capable of blocking E2 binding to cells or cell receptors have been described,18–23 some of which neutralize HCV entry in animal or in vitro models.6, 24–26 The first hypervariable region of E2 contains potent neutralizing epitopes, and antibodies raised against this region are capable of preventing infection. However, antibodies to HVR1 tend to be strain specific and have limited cross-neutralizing potency. Most antibodies that demonstrate broad neutralization of cell binding and/or infection are directed against conformational epitopes within E2.27–31 Induction of antibodies recognizing conformational epitopes is a challenging task, as these are difficult to mimic, and protein subunit vaccines are more likely to generate strain-restricted responses due to the immuno-dominance of the variable regions.32 However, we have recently shown that the mouse monoclonal antibody (MAb) AP33 can potently neutralize retroviral pseudo-particles (HCVpp) bearing genetically diverse E1E2 glycoproteins.33 The epitope recognized by MAb AP33 has been mapped to a 12 amino acid region corresponding to residues 412 to 423 located immediately downstream of HVR1.21, 22, 34–37 This region is also recognized by rat MAb 3/11.21 Like AP33, MAb 3/11 is capable of neutralizing HCVpp carrying strain H77 E1E2, although whether its neutralization capacity extends to genetically diverse E1E2 molecules is unknown.3 Identification of a highly conserved linear neutralizing epitope is a significant milestone in the development of a cross-reactive vaccine. Rational development of an AP33-epitope–based vaccine will require a thorough understanding of the interaction between MAb AP33 and its epitope. In the current report, we have finely mapped the amino acid residues recognized by MAb AP33 and demonstrated that MAb 3/11 recognizes a distinct yet overlapping epitope. Importantly, we also show that MAb AP33 has a higher neutralizing ability than MAb 3/11, highlighting variability in the specificity and quality of the antibody response against this region of the E2 glycoprotein.
The anti-E2 monoclonal antibodies AP33, ALP98, 3/11, and H53 have been previously described.21, 22, 38 The dengue type 2–specific MAb 46D (hybridoma 3H5-1. ATCC No: HB-46) and the CD81-specific MAb JS-81 (Pharmingen, Oxford, UK) were used as control antibodies in the HCV JFH-1 cell culture neutralization assay. MAbs from non-commercial sources were purified from hybridoma supernatants using a protein G column (GE Healthcare, UK).
The branched peptides used in immunoassays corresponded to the AP33 epitope QLINTNGSWHVN22 encompassing amino acids 412-423 of the genotype 1a Glasgow strain,36 the control peptide VNLHDFRSDEIE, and a peptide HLANHQGKWRLH, which represented the most frequent peptide selected by MAb AP33 from a random peptide phage display library (below).
Human hepatoma cells Huh-7,39 Huh-Lunet cells, and human epithelial kidney (HEK) 293FT cells (Invitrogen, Paisley, UK) were grown in Dulbecco's modified Eagle's medium, GIBCO BRL, Paisley, UK) supplemented with 10% fetal calf serum, 5% non-essential amino acids, and 200 mmol/L L-glutamine (Sigma, Dorset, UK).
Enrichment of Random Peptide Phage Display Libraries Using AP33 and 3/11.
A commercially available 12-mer, M13 gene 3–based random peptide phage display library (New England Biolabs, Hitchin, Hertfordshire, UK) was enriched for MAb AP33- and MAb 3/11- specific peptides by three to four rounds of affinity selection according to the manufacturer's instructions. After the final round of enrichment, individual phage clones were isolated, and the sequence of the peptide insert deduced following automated DNA sequencing using the −96 sequencing primer supplied by the manufacturer. Deduced peptide insert amino acid sequences, together with the sequence corresponding to amino acids 412-423 of the H77c strain, were aligned using ClustalX,40 followed by manual adjustment.
To determine the reactivity of selected phage with the selecting antibody, a phage capture enzyme immunoassay was performed. Approximately 1 × 1011 phage particles were added to microtiter plate wells coated with 10 μg/mL of MAb AP33, 3/11, or ALP98. Bound phage were detected by anti-fd antibody followed by an alkaline phosphatase conjugated anti-mouse antibody and Sigmafast p-nitrophenol phosphate (p-NPP) substrate (all from Sigma). Absorbance values were determined at 405 nm.
Peptide Enzyme Immunoassay.
Branched peptide (500 ng/mL) was coated onto wells of a microtiter plate as described above and then used to capture test MAb. Bound MAb was detected using an alkaline-phosphatase–conjugated anti-human IgG antibody diluted 1/1,000 (Sigma), followed by p-NPP substrate. Absorbance was measured at 405 nm.
Plasmid pCR3.1 (Invitrogen) containing E1E2 from the infectious H77c strain of HCV was mutated using the QuickchangeII mutagenesis kit (Stratagene, Amsterdam, Netherlands), according to the manufacturer's protocol. Primers were designed using the Stratagene primer design program (www.bioinformatics.org/primerx).
GNA Capture ELISA.
To detect MAb binding to E2 glycoprotein, an ELISA was performed essentially as previously described.36 Briefly, E1E2 glycoproteins from the clarified lysates of transfected HEK 293FT cells were captured onto GNA (Galanthus nivalis) lectin (Sigma)-coated microtiter plates then detected by MAb AP33 or 3/11, followed by an anti-species IgG-alkaline phosphatase conjugate (Sigma) and p-NPP substrate. Absorbance values were determined at 405 nm.
Reactivity of MAbs to alanine replacement mutant E1E2 was detected in the same way, except that E1E2 normalized for the amount of E2 protein according to reactivity to H53 MAb was used.
To assess the effect of E1E2 denaturation on MAb binding, E1E2 was denatured as previously described,41 then reactivity to MAbs determined using the GNA capture enzyme immunoassay.
Competition assays were carried out using the peptide and sE2/E1E2 ELISAs described, except that immobilized target antigens were used to capture a mixture of biotinylated target MAb and un-biotinylated competing MAb or negative control MAb.
HCVpp Infection and Neutralization Assays.
Full-length E1E2 (representing amino acid residues 170 to 746 of the HCV open reading frame referenced to strain H77c42) clones were generated and their nucleotide sequence determined as previously described.13 HCVpp were produced essentially as previously described2 by co-transfection of plasmids expressing the full-length E1E2, murine leukemia virus (MLV) Gag-Pol and the MLV transfer vector carrying the luciferase gene under the control of human cytomegalovirus promoter. Antibody-mediated neutralization of HCVpp was carried out as previously described.33 The neutralizing activity was expressed as % infectivity compared with “no antibody” control.
Production of Cell Culture Infectious HCV and Neutralization Assays.
Cell culture infectious HCV (HCVcc) were generated as previously described using the plasmid pFK-Luc-JFH1. pFK-Luc-JFH1ΔE1-E2 was also used in the analyses to control for non-specific transduction of reporter activity.5 Neutralization assays were performed by infecting Huh7-Lunet cells43, 44 with virus stocks that were pre-incubated with or without antibodies at room temperature for 1 hour before application. Four hours post-infection, the inoculum was removed and replaced with culture medium. Seventy-two hours later, infectivity conferred by the virus inoculum was determined using a luciferase assay, as previously described.5
Sequence Analysis of the VH and VL Regions of MAbs AP33 and 3/11.
cDNA was generated from mRNA obtained from approximately 1 × 106 hybridoma cells using Thermoscript (Invitrogen) with the poly-dT oligonucleotide primer included in the kit. Two microliters of resulting cDNA was used as template in polymerase chain reaction designed to amplify the variable regions of the light and heavy chains. Heavy chain amplification was achieved as previously described,45 using the sense primer VH1BACK (5′-AGG TSM ARC TGC AGS AGT CWG G-3′) with antisense primer VH1FoR-2 (5′-GGG GCC AAG GGA CCA CGG TCA CCG TCT CCT CA-3 ′). Light chain amplification was achieved as previously described,46 using the primers Mk (5 ′-GGG AGC TCG AYA TTG TGM TSA CMC ARW CTA MCA-3 ′) with reverse primer Kc (5 ′-GGT GCA TGC GGA TAC AGT TGG TGC AGC ATC-3 ′). Polymerase chain reaction products were cloned into the pGEM-T vector (Promega, Madison, WI) and two clones for each heavy and light chain were sequenced using the T7 forward and M13 reverse primer (Promega) and the ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit (Perkin Elmer Applied Biosystems, Boston, MA), according to the manufacturer's protocol. Framework (FWR), and complementarity determining (CDR) regions were identified using the IgBLAST (http://www.ncbi.nlm.nih.gov/igblast/) and Web Antibody Modelling tool (http://www.bioinf.org.uk/abs/) with subsequent manual adjustment.
Enrichment of Random Peptide Phage Display Libraries Identifies E2 Residues Associated With MAbs AP33 and 3/11 Binding.
Both MAb AP33 and 3/11 have previously been shown to bind a 12mer peptide corresponding to amino acids 412-423, immediately downstream of HVR1, and were raised after immunization with soluble E2 from the Gla and H strains of HCV, respectively.21, 22 To determine whether these antibodies might recognize overlapping, yet distinct epitopes encompassed within this region, each antibody was used to enrich random peptide phage display libraries to identify critical residues involved in mediating MAb binding. After three to four rounds of affinity selection against each antibody, individual phage clones were isolated and their reactivity to the selecting antibody assessed by phage capture ELISA (Fig. 1). Nineteen and 40 clones selected by AP33 and 3/11, respectively, were analyzed, and of these 17 and 35 were reactive to the selecting antibody. None of the AP33-selected clones reacted with 3/11 and vice versa, suggesting that these antibodies recognized distinct amino acid residues within this region. None of the enriched clones reacted with the control antibody ALP98 (not shown).
DNA sequence analysis of random peptide inserts present in selected antibody-reactive phage clones was performed. Amino acid sequences aligned to the region of H77 E2 corresponding to the putative MAb AP33 and 3/11 epitopes are presented in Fig. 2. Phage selected by MAb AP33 could be divided into three groups based on the amino acid sequence of the random peptide insert. Multiple alignment of the phage sequences showed absolute conservation of four amino acids, namely L413 (Leucine corresponding to position 413 of the H77 E1E2 sequence), N415, G418, and W420.
Considerably more sequence diversity was evident in the phage clones selected by MAb 3/11 (Fig. 2B). One group of clones was identical to the corresponding region of H77c at positions 420 (W), 421 (H), and 423 (N) of the H77c polyprotein. The remaining phage clone peptide inserts had varying degrees of similarity to the H77 E2 region, although commonly shared residues were the tryptophan or histidine at positions 420 and 421. One sequence present in clones isolated during both rounds of selection (e.g., 3/11.27), had extensive homology to the 412-423 region, including positions 415 and 418-421. This phage peptide was unreactive to MAb AP33 in ELISA, even though it contained three of the four putative contact residues recognized by MAb AP33 and was reactive to MAb 3/11.
To verify that the phage-displayed peptides were specific for each antibody, a branched peptide representing the most prevalent random peptide sequence (clone AP33.12) was synthesized and used in ELISA. MAb AP33, but not MAb 3/11 reacted with this peptide (Fig. 3).
Reactivity of MAb AP33 and 3/11 to Alanine Replacement Mutant E1E2 Proteins.
To further delineate the MAb reactive region, a panel of H77 E1E2 mutant clones was generated, in which each residue of the putative MAb AP33 and 3/11 epitope was substituted in turn by alanine. Mutants were expressed in HEK 293FT cells and reactivity of the resulting proteins to MAb AP33 and 3/11 assessed using a GNA capture ELISA (Fig. 4). Binding of AP33 was reduced by more than 75% compared with wild-type for mutants L413A, N415A, G418A, and W420A, indicating that these residues were critical for binding. Mutation T416A and N417A also reduced AP33 binding, although this effect was not as marked as the other four mutations. Substitution of glutamine by alanine at position 412 (Q412A) consistently enhanced binding of AP33 to E1E2 by approximately 50% compared with the wild-type H77 protein. By contrast, this substitution had no effect on MAb 3/11 binding. Alanine replacement of the remaining five residues had negligible effect on AP33 recognition.
Consistent and significant reduction of binding by MAb 3/11 compared with wild-type H77 protein was observed for mutant N415A, W420A, and H421A, highlighting the importance of these residues in binding by MAb 3/11. Substitution of the isoleucine at position 422 resulted in moderate enhancement of 3/11 binding. Alanine replacement of the remaining residues either had no effect or resulted in moderate reductions in binding compared with wild-type.
Comparison of Binding Affinities and HCVpp Neutralization Efficiencies of MAb AP33 and MAb 3/11.
The fine epitope mapping experiments indicated that MAbs AP33 and 3/11 were recognizing different contact residues within the E2 protein, so we went on to assess whether these differences might translate into differences in binding affinity or neutralizing potency. Binding curves of reactivity of each MAb against E1E2 and 412-423 branched peptide, together with a MAb competition assay, are presented in Fig. 5. The concentration of MAb AP33 required to obtain 50% binding to the 412-423 branched peptide was more than 10-fold lower than that required for MAb 3/11 (Fig. 5A). Similarly, in biotinylated MAb binding assays, AP33 was more efficient at competing for binding than 3/11 (Fig. 5B). Compared with MAb 3/11, competition with MAb AP33 resulted in greater reduction in binding by both biotinylated MAb AP33 and 3/11. The MAbs also exhibited marked differences in affinity to E1E2 representative of diverse HCV genotypes (Fig. 5C). Concentrations of MAb AP33 needed to obtain 50% binding to the E1E2 proteins ranged from approximately 1 × 101 to 1 × 103 ng/mL, whereas 50% binding was achievable using 3/11 at concentrations ranging from approximately 1 × 102 to 1 × 104 ng/mL. Together, these data indicate that AP33 has a higher affinity for E2 than MAb 3/11.
Similarly, a comparison of the ability of MAbs AP33 and 3/11 to neutralize HCVpp carrying E1E2 representative of genotypes 1 to 6 (Fig. 6A) showed that, whereas both antibodies were capable of broad neutralization, neutralization potency of AP33 was consistently greater than MAb 3/11 (P < .001, Wilcoxon's matched pairs test). When used at a concentration of 50 μg/mL, MAb AP33 was able to neutralize HCVpp infectivity by between 80% and 99%. By contrast, the same concentration of MAb 3/11 only resulted in 10% to 80% neutralization. As has been previously reported,33 MAb AP33 failed to neutralize HCVpp carrying E1E2 from the genotype 5 strain UKN.5.14.4; similarly, MAb 3/11 was also unable to neutralize HCVpp carrying this E1E2 clone. This isolate has a 4–amino acid change (QLIQNGSSWHIN) in the E2 region corresponding to residues 412 - 423. This mutation alters two of the residues critical for AP33 (N415 and G418) recognition and one (N415) for 3/11. Consistent with this, our data show that both MAbs AP33 and 3/11 fail to react with UKN5.14.4 E2 (data not shown).33 Both MAbs also fail to neutralize UKN5.14.4 HCVpp.
MAb AP33 Neutralizes HCVcc Infectivity.
We also assessed the neutralizing capability of MAb AP33 using the recently described JFH-1 HCVcc system. MAb AP33 was able to reduce HCVcc infectivity by greater than 80% when added to the JFH-1 virus inoculum at a concentration of 50 μg/mL (Fig. 6B).
The Epitopes for AP33 and 3/11 Are Sensitive to E1E2 Conformation.
To examine the requirement for correct E2 conformation on the binding of MAb AP33 and 3/11, we compared binding with native and denatured reduced E1E2. For comparison, the MAb ALP98, known to bind to a linear epitope in E2, and MAb H53, a well-characterized conformation-sensitive anti-E2 antibody, were included in the analysis. After denaturation, MAb ALP98 binding was reduced by approximately 10%, whereas both MAbs AP33 and 3/11 demonstrated a 50% inhibition of binding (Fig. 7). Consistent with its conformational nature, H53 binding was completely abrogated by denaturation. This indicates that, despite recognizing a linear region of E2, optimal recognition by both MAbs is affected by the conformation of E1E2.
Sequence Analysis of the VL and VH Regions of MAb AP33 and 3/11.
The deduced amino acid sequences corresponding to the light and heavy chain variable regions of MAbs AP33 and 3/11 is presented in Fig. 8. The most striking feature of this comparison was that the heavy chain CDR3 for MAb AP33 contained 10 amino acids, whereas the MAb 3/11 heavy chain CDR3 contained only three.
We have previously shown that the MAb AP33 can potently neutralize HCVpp reconstituted with E1E2 molecules from diverse genotypes of HCV. Similarly, MAb 3/11 has previously been shown to neutralize HCVpp reconstituted with the autologous H77 E1E2 protein. The current study highlights that two monoclonal antibodies directed to the same highly conserved region of E2 have different neutralization potencies and binding affinities. We show that MAb 3/11, although having a similar spectrum of activity, has a significantly lower affinity and lower neutralizing potency than MAb AP33.
There was evidence that MAb recognition and neutralization varied across genotypes. However, such comparisons should be treated with caution. First, differences in the relative proportion of native and incorrectly folded aggregates in the various E1E2 preparations will affect epitope exposure and reactivity. Similarly, differential incorporation of E1E2 into HCVpp will result in altered neutralization sensitivities. Therefore, KD and IC50 estimates were not attempted to avoid potentially misleading cross-genotype comparisons.
Previous mapping studies have shown that the epitopes for both antibodies lie within a region of E2 encompassing amino acids 412-423.21, 22 Using random peptide phage library enrichment and antibody probing of a panel of 412-423 point mutated E1E2 glycoproteins, we have identified four key residues critical for MAb AP33. These residues were discontinuous, spanning an 8–amino acid region of the putative AP33 epitope. Probing the same panel of point mutated E1E2 clones with MAb 3/11 identified three residues critical for binding, two of which were also critical for AP33 binding. When MAb 3/11 was used to selectively enrich a 12mer random peptide phage display library, the diversity observed in the amino acid sequences of the selected phage was greater than that observed for AP33. However, all but one of the reactive phage contained one or more of the three amino acids identified as being critical for MAb 3/11 binding when probing the panel of mutant E1E2s. These analyses indicate that MAbs AP33 and 3/11 recognize the same region of E2, but their epitopes are distinct. In addition, these data also suggest that AP33 binding to E2 is highly dependent on all of the four critical residues being present in both E2 and the reactive phage peptides. By contrast, binding of MAb 3/11 to phage enriched by this antibody was not dependent on all of the residues identified in the mutant E1E2 analysis being present. Clustering and overlap of neutralizing epitopes has been reported for other viruses, most notably HIV, where a membrane proximal region in gp41 has been shown to elicit human antibodies with different specificities and varying degrees of neutralizing potency.47 Similarly, we show that MAbs AP33 and 3/11 also have very distinct phenotypes. A better understanding, at the molecular level, of the mechanism of binding by each antibody will help elucidate the reasons for these differences, as well as determine whether the residues identified as being important for binding are contact residues or provide a molecular environment suitable for binding. Surface plasmon resonance and crystallography experiments are underway to address these issues.
Alanine replacement of the glutamine residue at position 412 enhanced AP33 binding. This amino acid lies immediately upstream of the asparagine residue critical for AP33 binding. Interestingly, this replacement had no effect on 3/11 binding, even though mutagenesis experiments implicated this residue in MAb 3/11 binding. Similarly, alanine replacement of the isoleucine residue at 423 enhanced MAb 3/11 binding yet had no effect on AP33 binding. Why these mutations have a differential enhancing effect is unclear. Enhanced binding might result through abrogation of a potentially inhibitory effect of a charged or polar residue or removal of steric hindrance caused by a relatively more bulky side chain. These substitutions also might lead to a localized conformational change, improving the exposure of the contact residues. We also showed, using protein denaturation experiments, that MAb binding is sensitive to conformational changes, at least in the context of the E1E2 complex.
Region 412-423 contains one potential N-glycosylation site at position 417, which appears to be used.48 Alanine substitution at this site resulted in reduced MAb AP33 binding. What effect glycosylation via this site has on antibody binding or its effects on immunogenicity of this region is unclear. Glycosylation may result in immune shielding, yet the ability of MAbs AP33 and 3/11 to bind to E2 suggests that even when glycosylated their epitopes are still accessible. In addition, reactivity to peptides suggests that glycosylation might not be necessary for induction of AP33-like antibodies. These findings are of significant relevance if this epitope is to be included in future vaccine design.
We have previously shown that the 412-423 region of E2 is highly conserved across all genotypes.33 Re-analysis of sequences available from the Los Alamos HCV sequence database shows that the residues critical for MAb AP33 and 3/11 binding are conserved to an even greater degree, with sequence variation of less than 0.5% to 2.5% at these sites (data not shown). Such high conservation suggests that these residues might be important in maintaining envelope protein conformation or function. Similarly, it would also suggest that these sites are not targeted by neutralizing antibodies in vivo. Indeed, we have recently shown that the prevalence of antibodies reactive to region 412-423 is low in natural infection (Tarr et al., manuscript in preparation). Focussing the human immune response on the exact epitope recognized by MAb AP33 may be important in vaccine design. In this context, the availability of phage mimotopes may be beneficial, as previous studies have shown that such mimotopes are capable of inducing antibodies of similar phenotype to the selecting antibody.49
Finally, we compared the sequence of the MAb AP33 and 3/11 VH and VL chains. The striking difference between the two antibodies was in the length of the heavy chain CDR3. The heavy chain CDR3 region is thought to have a critical role in antigen binding and demonstrates higher levels of sequence and structural variability than the other variable loops. MAb 3/11 contained arginine and aspartate at positions 94 and 101, respectively, which are capable of forming a salt bridge and subsequent bulged50 or kinked51 structure. However, MAb 3/11 has a very large deletion within this region and is therefore unlikely to adopt a complex conformation. MAb AP33 lacks the arginine at position 94 and is therefore likely to take on a non-bulged structure.51 Given the large difference between the two MAbs at the sequence level, a better understanding of the molecular interactions between the antibodies and their epitope will help explain the underlying reasons for their phenotypical differences.
In summary, these data show that the region of E2 encompassing residues 412-423 contains more than one overlapping and distinct neutralizing epitope. We also show that antibodies recognizing this region may differ significantly in their ability to bind to E2 and also in their ability to neutralize HCV infection; an important finding if this region is to be the focus of future vaccine candidates.
The authors thank Jane McKeating and Jean Dubuisson for the gift of MAbs 3/11 and for MAb H53 and the MLV luciferase construct, respectively. Takaji Wakita and Francois-Loic Cosset for provision of the JFH-1 clone and retroviral pseudotype constructs, respectively.