Functional Neutralization of HIV-1 Vif Protein by Intracellular Immunization Inhibits Reverse Transcription and Viral Replication*

Human immunodeficiency virus type 1 (HIV-1)-encoded Vif protein is important for viral replication and infectivity. Vif is a cytoplasmic protein that acts during virus assembly by an unknown mechanism, enhancing viral infectivity. The action of Vif in producer cells is essential for the completion of proviral DNA synthesis following virus entry. Therefore, Vif is considered to be an important alternative therapeutic target for inhibition of viral infectivity at the level of viral assembly and reverse transcription. To gain insight into this process, we developed a Vif-specific single-chain antibody and expressed it intracellularly in the cytoplasm. This intrabody efficiently bound Vif protein and neutralized its infectivity-enhancing function. Intrabody-expressing cells were shown to be highly refractory to challenge with different strains of HIV-1 and HIV-1-infected cells. Inhibition of Vif by intrabody expression in the donor cell produced viral particles that do not complete reverse transcription in the recipient cell. The anti-Vif scFv was shown to be specific for Vif protein because its function was observed only in nonpermissive cells (H9, CEM, and U38). Moreover, transduction of peripheral blood mononuclear cells with an HIV-derived retroviral vector expressing Vif intrabody was shown to confer resistance to laboratory-adapted and primary HIV strains. This study provides biochemical evidence for the role of Vif in the HIV-1 lifecycle and validates Vif as a target for the control of HIV-1 infection.

Human immunodeficiency virus type 1 (HIV-1) 1 is a complex retrovirus that contains a number of accessory genes not present in other retroviruses. One of the critical determinants of HIV-1 infectivity in vivo is the Vif protein. Vif (virion infectivity factor) is essential for the establishment of productive infection of HIV-1 in peripheral blood lymphocytes and macrophages in vitro and for pathogenesis in animal models of AIDS (1)(2)(3)(4)(5)(6)(7). In cell culture, vif-defective HIV-1 is able to replicate in some T-lymphoblastoid cell lines termed permissive (CEM-SS, SupT1, C8166, and Jurkat), whereas Vif is required in other cell lines, such as H9, U38, or MT-2, termed nonpermissive (2,4,6,8,9). Vif acts during late steps of the viral life cycle to increase the infectivity of HIV-1 virus particles as much as 100 -1000-fold. Vif plays a critical role in the cell-free transmission of HIV-1 and also appears to be important for cell-tocell virus transmission (4 -6). Previous studies have shown that Vif acts to enhance viral infectivity during the production of virus particles, most likely by affecting virus assembly or maturation (10 -14).
Vif is a phosphorylated 23-kDa protein in multimer form that is abundantly expressed in the cytoplasm of infected cells (15)(16)(17)(18)(22)(23)(24)(25). Vif has been shown to interact or co-purify with membranes (15,21), intermediate filaments (22), HIV-1 Gag (23,11), and, most recently, viral RNA (16,19,26). It was also proposed that Vif counteracts the tyrosine kinase Hck as an inhibitor of HIV-1 replication (27). Virus particles produced in the absence of a functional Vif protein can bind and penetrate susceptible cells but are defective in their ability to synthesize proviral DNA, most likely indirectly as a result of an effect on a component of the virus core (6,24,25,28). Virus uncoating, reverse transcription, transport of the preintegration complex to the nucleus, and subsequent integration of the viral DNA into the host genome are necessary steps that must occur for infection to be established (30).
Intrabodies are single-chain variable region antibody fragments expressed and confined intracellularly, where they can bind to viral proteins and other targets (29,31,(33)(34)(35)(36)(37). The early events of the viral life cycle are potential therapeutic targets that could be inhibited using anti-HIV-1 Vif intrabodybased strategies and may therefore represent a new therapeutic approach to inhibit HIV-1 reverse transcription. Therefore, strategies that prevent or limit expression of Vif are expected to be beneficial in the treatment of HIV-1 disease (29,39).
Here we have investigated the potential benefit of intracellular neutralization of Vif activity (36,37,40). We developed a Vif-specific single-chain antibody from immunized rabbits that was expressed intracellularly and confined to the cytoplasm. Cell lines and primary cells expressing the Vif intrabody were shown to be highly refractory to challenge with the HIV-1 virus or HIV-1-infected cells. Furthermore, replication of an HIV virus constructed to express anti-Vif scFv in cis was highly reduced after challenge with wild-type HIV-1. The formation of completed reverse transcripts was reduced when cells were infected with HIV-1 virus produced from intrabody-producing cells. These results suggest that gene therapy approaches, which deliver Vif intracellular antibodies, may represent a new therapeutic strategy for inhibiting HIV reverse transcription.

Cells, Viruses, and Reagents
Cells-H9, U38, Jurkat, SupT1, and CEM cells were grown in RPMI 1640 medium containing 10% fetal bovine serum and antibiotics. All of the cell cultures were maintained at 37°C in 5% CO 2 . Human PBMC from healthy donors were activated for 48 h with phytohemagglutinin (PHA) and were then cultivated in RPMI 1640 medium supplemented with 20% fetal calf serum and 50 units of interleukin-2/ml (Roche Molecular Biochemicals). HeLa and 293T cells were grown in Dulbecco's modified minimal essential medium supplemented with 10% fetal calf serum. Transduced cells were usually maintained in the presence of puromycin (0.5 g/ml). The tissue culture media and reagents were from BioWhitaker.
Plasmids-Plasmid pSVLvif expresses the vif gene of the HXB2 clone of HIV-1 under the control of the SV40 promoter (15). Plasmids for the trans-complementation assay were described previously (15). Plasmid pComb3X is derived from pComb3H (33). The coding sequence for anti-thyroglobulin scFv is cloned in plasmid pHEN obtained from Griffiths et al. (41). The envelope plasmid pMD.G, the packaging plasmid pCMVR8.91, and the vector plasmid pRRL.SIN have been described previously and were kindly provided by Didier Trono (42)(43)(44). pHXBn-PLAP-IRES-Nϩ was obtained from the AIDS Research and Reference Reagent Program.
Antibodies-Rabbit polyclonal Vif was described previously (15). For detection of Vif protein, rhodamine-conjugated sheep anti-rabbit immunoglobulin was used as a secondary antibody (Pierce). HRP-conjugated anti-M13 phage antibody was obtained from Amersham Biosciences. FITC-conjugated, HRP-conjugated, and nonconjugated anti-HA high affinity antibody was purchased from Roche.
Proteins-Vif protein was derived from HIV-1 HXB2 strain and purified as described (45,46). HIV-1 protease was obtained from the AIDS Research and Reference Reagent Program.

DNA Transfections
To produce large amounts of HIV-1 particles, 5 ϫ 10 6 293T cells were transfected by FuGENE (Roche) with 2 g of wild-type HIV-1 NL4 -3 or mutant pHIV-1 NL4 -3 ⌬vif. After 24 h Jurkat and H9 cells were cocultured during 48 h in the presence of 50 g/ml of dextran. The cells were maintained in RPMI medium plus 15% fetal calf serum. p24 antigen concentration was determined as recommended by the manufacturer (Innotest). For immunofluorescence and co-immunoprecipitation experiments, HeLa cells were co-transfected by the FuGENE reagent (Roche) according to the manufacturer's protocol with 2 g of pSVLvif and 2-fold molar excess of plasmid encoding intrabodies or with control plasmid, pcDNA3.1/Neo containing no insert. For co-immunoprecipitation, cell lysis was performed as described by Anderson (47). For immunoprecipitation, magnetic protein A beads were used as described by the manufacturer (Miltenyi Biotec).

Replication Complementation Assay
A transient complementation assay was used to provide a quantitative measure of the ability of wild-type or mutant Vif proteins to complement a single round of HIV-1 replication in trans. Briefly, H9 cells (10 7 ) were co-transfected by FuGENE (Roche), with 2 g of either pHXB⌬envCAT or pHXB⌬Avr⌬envCAT, 2 g pSVIIIenv, and plasmids encoding scFv as described (49). The ability of the intrabody to inhibit a single round of infection was measured by assaying for chloramphenicol acetyltransferase (CAT) activity in the transfected culture at 9 or 10 days after transfection. CAT assay was performed by ELISA (Roche).

Immunofluorescence Staining
HeLa cells transfected by FuGENE with 2 g of plasmids were fixed in phosphate-buffered saline (PBS) with 4% paraformaldehyde for 10 min at room temperature, permeabilized with PBS plus 0.1% Triton X-100 for 5 min, and washed with PBS plus 2% fetal calf serum before staining. The fixed cells were incubated with Vif antiserum (1:200) for 90 min at 37°C, incubated with rhodamine-conjugated goat anti-rabbit immunoglobulin antibody (1:200) for 20 min at 37°C, washed, and mounted on glass slides. For immunostaining of scFv, direct immunofluorescence was performed with FITC-conjugated anti-HA monoclonal antibody (1:40). For double immunofluorescence staining of intrabodies and Vif protein, FITC-conjugated anti-HA monoclonal antibody was used in combination with Vif antiserum at similar concentrations. The imaging setup consists of an Olympus IX-50 inverted microscope, Ludl BioPoint filter wheels and a 12-bit V-scan cool CCD (Photonic Science). Integrated control of filter wheel and image acquisition is achieved by Image-Pro Plus 4.0 and Scope-Pro 3.1 (Media Cybernetics). The settings for image acquisition (camera exposure time, filters, time interval, and storing modes) were determined by custom-made macros. The images were collected with Olympus 40ϫ or 100ϫ plan apo objectives (numerical apertures ϭ 0.95 and 1.4, respectively).

Rabbit Immunization
A rabbit (New Zealand White) was treated with four subcutaneous injections containing 50 g of purified Vif protein in a 1-ml emulsion of adjuvant (Ribi Immunochem Research, Hamilton, MT). The injections were administered at 2-3-week intervals. Five days after the final boost, spleen and bone marrow were harvested and used for total RNA preparation (48,50).

RNA Isolation, cDNA Synthesis, Rabbit Antibody Library Construction, and Sfi Cloning
Total RNA was prepared from rabbit bone marrow and spleen using TRI reagent from the Molecular Research Center (Cincinnati, OH) according the manufacturer's protocol and was further purified by lithium chloride precipitation. First strand cDNA was synthesized using the SUPERSCRIPT Preamplification System with oligo(dT) priming (Invitrogen). The protocol and the oligonucleotide primers for the construction of chimeric rabbit antibody libraries, where rabbit VL and VH sequences are combined with human C and CH1 sequences, have been described previously (32). Final PCR fragments encoding a library of antibody fragments were Sfi-cut, purified, and cloned into pComb3X vector. PComb3X contains a suppressor stop codon and sequences encoding peptide tags for purification (His 6 ) and detection (HA) (50).

Expression and Purification of Fab Fragments
PCR fragments encoding Fab were generated by overlap extension PCR and cloned into pComb3X vector. To express Fab, phagemid DNA was transformed into the nonsuppressor Escherichia coli strain TOP 10F (50). Fab was purified from the concentrated supernatants of induced cultures by affinity chromatography. Purified Fab fragments were analyzed by SDS-PAGE followed by Coomassie Blue staining and Western blot with HRP-conjugated anti-HA monoclonal antibody. Protein concentration was determined by measuring the optical density at 280 nm by the classic Bradford method.

DNA Constructs
For the conversion of a Vif-specific Fab into a scFv, specific oligonucleotides primers were used to amplify V H and V L gene segments from purified phagemid DNA isolated from J4, a Fab fragment specific for the Vif protein. This Fab was isolated from an immunized rabbit by using the phage display approach (50). The following primers were used: V L , RSCVK1, 5Ј-GGGCCCAGGCGGCCGAGCTCGTGMTGACC-CAGACTCCA-3Ј, and RKB9J0-B, 5Ј-GGAAGATCTAGAGGAACCACC-TAGGATCTCCAGCTCGGTCCC-3Ј; and V H , RSCVH3, 5Ј-GGTGGTT-CCTCTAGATCTTCCCAGTCGYTGGAGGAGTCCGGG-3Ј, and HSCG-1234, 5Ј-CCTGGCCGGCCTGGCCACTAGTGACCGATGGGCCCTTG-GTGGARGC-3Ј. The purified PCR products were assembled by another PCR using the following primers: RSC-F, 5Ј-GAGGAGGAGGAGGAG-GAGGCGGGGCCCAGGCGGCCGAGCTC-3Ј, and RSC-B, 5Ј-GAGGA-GGAGGAGGAGGAGCCTGGCCGGCCTGGCCACTAGTG-3Ј. The resulting overlap PCR product encodes an scFv in which the N-terminal V L region is linked with the V H region through a 7-amino acid peptide linker (GGSSRSS) and a 18-amino acid peptide linker (SSGGGGSGG-GGGGSSRSS), named scFv 4B and 4BL, respectively. The DNA fragment was gel-purified, digested with the restriction endonuclease SfiI, and cloned into the appropriately cut phagemid vector pComb3X, a variant of pComb3H. The binding activity of the expressed scFv was confirmed, and the genes encoding the scFv were transferred to pCDNA3.1/Neo (Invitrogen), pBabePuro vectors, and pRRL.SIN. A methionine initiation codon sequence was added into the 4BL and 4B scFv by PCR amplification. The primers used for cloning in pCDNA3.1 were scFv5, 5Ј-GGCATGGGGGCCCAGGCGGCCCAGCTC-3Ј, and scFv3, 5Ј-GCCACCACCCTCCTAAGAAGC-3Ј. Similar primers were used for cloning anti-Vif scFv in pBabePuro and pRRL.SIN: Vector/scFv5, 5Ј-C-GCGGATCCGCGGGCATGGGGGCCCAGGCGGCCGAGCTC-3Ј, and Vector/scFv3, 5Ј-ACGCGTCGACGTCGGATATCGCGGCCGCGGAAG-CCACCACCCTCCTAAGAAGC-3Ј. Downstream of the sequence, we maintained a sequence encoding the HA tag sequence (YPYDVPDYA) followed by a stop codon. Control scFv anti-thyroglobulin was amplified from pHEN2 and Sfi-cloned in pComb3X. A methionine was introduced at the N terminus, and the DNA fragment with a HA-tag at the C terminus was cloned into the appropriately digested vector DNAs. The modified pcDNA3.1/Neo plasmid encoding 4BL, 4B, and anti-thyroglobulin was designated pI4BL, pI4B, and pThyr. Anti-Vif scFv I4BL and anti-thyroglobulin was cloned in pHXBnPLAP-IRES-Nϩ by inserting the gene in place of PLAP-IRES-nef. The primers used for cloning were scFv5/HIV, 5Ј-ATAAGAATGCGGCCGCTAAACTATATGGGGGC-CCAGGCGGCCGAGCTC-3Ј, and scFv3/HIV, 5Ј-CCGCTCGAGC-GGGCCACCACCCTCCTAAGAAGC-3Ј.

In Vitro Studies of Anti-Vif Antibody Fragments
The methods for bacterial expression of anti-Vif scFv were described before (50). Briefly, scFv were cloned in pComb3X that allows the expression in the supernatant and purification with a Ni 2ϩ -NTA column caused by the presence of a His 6 tag at the C-terminal end. After induction of scFv expression by growth in 1 mM isopropyl-␤-D-thiogalactopyranoside for 20 h at 30°C, analysis of supernatants indicated that both anti-Vif scFv were expressed. For assays of expression of antibody fractions, 1 ml of culture supernatant was concentrated (Amicon 10) and separated by 12% PAGE. For purification of ScFv and Fab proteins, 100 ml of bacterial supernatant were concentrated to 2 ml (Centricon 30), and Ni 2ϩ columns were used (CLONTECH). Relative binding affinities of anti-Vif ScFv were determined via ELISA after coating of the wells with 200 ng of purified recombinant HIV-1 Vif protein. Purified anti-Vif scFv and Fab were diluted at various concentrations starting at 5 g/ml and added to the wells for further incubation. After washing the wells with PBS, horseradish-conjugated anti-HA monoclonal antibody was applied for detection. Recognition of Vif by anti-Vif scFv and Fab was next demonstrated by Western blot analysis. Vif protein and HIV-1 protease (200 ng) were separated by PAGE and transferred to nitrocellulose membranes. After blocking, the proteins were probed with anti-Vif Fab and scFv antibody fragments and then with horseradish-conjugated anti-HA monoclonal antibody as a secondary antibody. The proteins were visualized using the ECL system (Amersham Biosciences).

Retroviral Gene Transfer: Generation of Transduced H9 and Jurkat Cells
The amphotropic packaging cell line PhoenixAmpho was transfected with pBabe Puro plasmids encoding the anti-Vif scFv inserts (51). For control purposes pBabe Puro plasmids encoding scFv specific to thyroglobulin were also used (41). These packaging cell lines were used to generate virus-containing medium for two rounds of infection of H9 and Jurkat cells in the presence of 10 g/ml Polybrene (Calbiochem). Two days after the last infection, transduced H9 and Jurkat cells were selected in puromycin (0.5 g/ml) by limiting dilution cloning. After 15 days of selection, analysis of cells for anti-Vif expression and infectability was started. The untransduced parental H9 and Jurkat cell lines were named H9-P and Jurkat-P, respectively, and the H9 and Jurkat cells transduced to express anti-Vif intrabodies were named H9-I4BL, H9-I4B, Jurkat-I4BL, and Jurkat-I4B. The cell lines transduced to express the control intrabody anti-thyroglobulin were named H9-thyr and Jurkat-thyro.

HIV-1 Infection of H9 and Jurkat Cells
Transduced and untransduced H9 and Jurkat cells were infected at a multiplicity of infection of 0.01-0.05 for 6 h at 37°C, washed, and then cultured for up to 30 days. To monitor infection, aliquots were taken from the cultures at the indicated time points, and the p24 levels were determined in an HIV-1 ELISA (Innotest).

PCR Analysis of HIV-1 Reverse Transcription
DNase-I-treated stocks of HIV-1 NL4 -3 or mutant pHIV-1 NL4 -3 ⌬vif produced from parental Jurkat and H9 cells and cells expressing anti-Vif scFv were used to challenge 1 ϫ 10 6 H9 cells and PBMC in a volume of 1 ml. For infection, a stock of 2 ng of soluble p24 antigen was used. The PCR procedure to analyze reverse transcription was described previously (28,52). The primers used in this assay (M667/M661-LTR/ gag, nucleotides 496 -695) amplify the late DNA products in reverse transcription. Similar reaction conditions were used, except only the 24 h time point was analyzed. The products of all PCR reactions were electrophoresed through 1.5% agarose gels, transferred to Hybond Nϩ membranes (Amersham Biosciences), subjected to Southern hybridization by using random-primed radiolabeled DNA probes (Amersham Biosciences), and visualized by autoradiography. The DNA used for radiolabeling were obtained following PCR amplification of segments of pHIV-1 NL4 -3 or a human ␤-globin expression vector with the relevant primer pairs.

HIV Vector Stock Preparation
The stocks were prepared as described previously by transient cotransfection of three plasmids into 293T cells: pRRL.SIN.I4BL, the envelope plasmid pMD.G, and the packaging plasmid pCMV⌬R8.91 (42)(43)(44). The p24 concentration was determined by antigen ELISA (Innotest). Vector production and gene delivery were done in a biosafety level 2 environment. The in vitro transduction experiments were done in six-well plates (Nunc). Filtered vector-containing medium was added 24 h after seeding PBMC (2 ϫ 10 5 cells/well). When applicable, transduction of H9 and Jurkat cells was also performed. The transduction protocol was repeated two more times in the following 48 h. For the transduction, 100 ng of p24 was used. The multiplicities of infection were estimated by assuming that 1 ng of p24 corresponds to 1000 -5000 transducing units.

Selection of Vif-specific Antibody Fragments from a Phage
Display Chimeric Rabbit Fab Library-A chimeric rabbit/human Fab library was generated from cDNA derived from spleen and bone marrow RNA of an immune rabbit that had a strong immune response to HIV-1 Vif (50). The phagemid vector pComb3X was used, and an antibody library of Ϸ 9 ϫ 10 7 independent clones was constructed and selected. Several Fab clones that bound to purified Vif protein were isolated and studied. The relative binding affinities reported for each of the Fab clones are summarized in Fig. 1. The analyzed clones showed some sequence variation with conservation of sequence within some complementarity determining regions (data not shown). One clone, J4, bound strongly to Vif protein in ELISA and was chosen for further characterization and expression.

Anti-Vif scFv Gene Expression in E. coli and in Vitro
Interaction with HIV-1 Vif-The antibody fragment J4, that binds purified Vif protein was derived from a Fab phage display library. To express it in the cytoplasm in a lower molecular weight form, J4 Fab was converted to a single-chain fragment. Anti-Vif scFv was constructed by PCR amplification of VL and VH fragments and covalently linked with a short peptide linker of 7 amino acids (4B, GGSSRSS) and a long peptide linker of 18 amino acids (4BL, SSGGGGSGGGGGGSSRSS). It was expected that the short linker peptide would result in a dimeric scFv protein (33). To determine whether the binding efficiency of recombinant scFv to the HIV-1 Vif would correlate with those of the parental Fab clone, both scFv 4BL and 4B were expressed in E. coli, and binding to HIV-1 Vif was assessed by ELISA and Western blotting. Both 4B and 4BL anti-Vif scFv, together with J4 Fab, were expressed in bacterial cell culture supernatant and purified by Ni 2ϩ affinity columns. As the data in Fig. 2 demonstrate, 4B and 4BL scFv have different binding patterns. Binding of anti-Vif scFv prepared with a long linker (4BL) binds Vif protein in ELISA at similar levels to Fab J4. In contrast, anti-Vif scFv prepared with a short linker (4B) demonstrates comparatively a very low affinity. In contrast, nearly background signals indicate very low nonspecific binding of scFv 4BL to control proteins HIV-1 protease and bovine serum albumin, a profile similar to that of Fab J4.
The specificity of binding of anti-Vif scFv 4BL was next investigated by immunoblotting and immunofluorescence (Fig.  3). Vif protein and HIV-1 protease were separated by SDS-PAGE and probed with anti-Vif Fab J4 and recombinant anti-Vif scFv 4BL. In the Western blot the major band of 23 kDa corresponding to HIV-1 Vif was recognized strongly by the scFv 4BL and Fab J4. The recombinant antibody fragment, as expected from the ELISA results, did not recognize HIV-1 protease. Immunofluorescence studies also show a specific recognition of Vif protein by anti-Vif Fab and scFv 4BL (Fig. 3). Thus, in opposition to scFv 4B, Vif protein is specifically recognized in vitro by scFv with a long linker.
Cellular Expression of Anti-Vif scFv Co-immunoprecipitates Vif Protein and Co-localizes with HIV-1 Vif in the Cytoplasm-To express the antibody fragments 4BL and 4B in the cytoplasm of human T-lymphoid cells, an initiation codon was appended to the N terminus of the proteins, and an HA epitope was maintained at the C terminus. Expression plasmids were constructed in pCDNA3.1Neo (pI4BL and pI4B). Because 4B scFv was unable to bind Vif protein in ELISA, we anticipated that its expression in the cell would not effectively trap Vif in the cytoplasm. As a control for the activity of scFv, the antibody fragment anti-thyroglobulin (41) was cloned in the same eukaryotic vector and expressed in the human T-lymphoid cells.
The efficacy of intrabody expression for binding HIV-1 Vif protein in the cell was examined by co-immunoprecipitation and fluorescence microscopy studies (Fig. 4). HeLa cells were co-transfected with intrabodies and pSVLvif and then lysed 48 h after transfections. The intrabody fragment I4BL was immunoprecipitated with an anti-HA high affinity monoclonal antibody and protein A magnetic beads. Cell lysis and immunoprecipitation were carried out using mild detergent and salt conditions to allow co-immunoprecipitation of binding proteins. Upon immunoprecipitation of cells transfected with pSVLvif and pI4B, no Vif protein was detected by Western blot, whereas co-transfection of pSVLvif with pI4BL resulted in co-immunoprecipitation of Vif protein (Fig. 4). This result is consistent with previous data in ELISA showing Vif binding. As a control, anti-thyroglobulin scFv expressed in HeLa cells did not bind Vif because this protein did not co-immunoprecipitate with the antibody fragments (data not shown).
To evaluate the cellular localization of the antibody fragment I4BL expressed together with the antigen Vif, intracellular staining of co-transfected HeLa cells using an antibody specific to the HA tag and a polyclonal serum against Vif was performed in HeLa cells (Fig. 4). 48 h after co-transfection with pI4BL and pSVLvif, HeLa cells were fixed in paraformaldehyde and observed by fluorescence microscopy. Microscopy showed a predominant punctate pattern of co-localization of these two proteins in the cytoplasm, with nucleoplasm exclusion (Fig. 4). In addition, Vif was consistently detected in the nucleus, and scFv was only observed in the cytoplasm in the majority of cells evaluated. When Vif and scFv 4BL are expressed alone, its pattern of localization is different from the co-localization staining. In contrast, when anti-thyroglobulin scFv was expressed together with anti-Vif scFv, no obvious co-localization was detected (data not shown). These results demonstrate that Vif protein is specifically recognized in the cytoplasm by the intracellular expression of anti-Vif scFv and that its interaction co-localizes Vif and scFv 4BL throughout the cytoplasm.
I4BL Inhibits Vif Function in a trans-complementation Assay-To provide a qualitative and quantitative measure of the biological activity of the anti-Vif scFv, a transient complementation assay in nonpermissive cells was used to examine the abilities of the anti-Vif scFv and control scFv to inhibit a single round of HIV-1 replication in trans (27). In a previous study, we showed that the expression of Vif in trans in this assay restores the efficiency of a single round of replication of a vif-defective virus to the wild-type level (15). The Vif expresser plasmid is co-transfected with a vif-positive (pHXB⌬envCAT) or vif-negative env-defective (pHXB⌬Avr⌬envCAT) HXB2 provirus plasmid that expresses CAT and an HIV-1 envelope expresser plasmid. The HIV-1 virus particles produced in this assay result in only a single round of infection, because the packaged viral genome is defective for Env production. The efficiency of a single round of virus replication is quantitated by measuring the level of CAT enzyme activity in the infected cultures after 9 -10 days, the minimum time for a detectable signal above background. The ability of the anti-Vif scFv to inhibit a single round of HIV-1 replication in trans was examined. In the absence of Vif, replication of the vif-negative virus was ϳ10-fold lower than that of the vif-positive virus (Fig. 5). Coexpression of I4BL scFv in H9 cells reduced trans-complementation to 15% of the wild-type level, as shown in Fig. 5. In contrast, the expression of I4B scFv, which does not bind Vif, caused a much less significant reduction in trans-complementation to 85% of the wild-type level. Expression of anti-thyroglobulin scFv caused no significant reduction in trans-complementation. Thus, the biological activities of the anti-Vif scFv correlated directly with their Vif binding activities in ELISA and co-immunoprecipitation assays, indicating that I4BL scFv is likely to inhibit HIV-1.
H9 Cells Expressing Vif-specific Intrabody I4BL Are Protected from HIV-1 Infection-The experiments described above examined the biological activity of anti-Vif scFv in a transient assay under conditions in which most virus transmission occurs by cell-to-cell spread (15,49). To investigate whether anti-Vif scFv can inhibit Vif function during several rounds of FIG. 5. Neutralization of Vif function in a trans-complementation assay. The values shown by the black and white bars represent the percentages of replication complementation in Jurkat and H9 cells, respectively, relative to the value obtained for the wild type. The cells were co-transfected with a wild-type Vif protein, scFv expressor plasmids, either pHXB⌬envCAT or pHXB⌬Avr⌬envCAT, and pSVIIIenv as described under "Experimental Procedures." Replication complementation for each testing plasmid was determined by the level of chloramphenicol acetyltransferase activity in cells co-transfected with pHXB⌬Avr⌬envCAT relative to that obtained with pHXB⌬envCAT. Background levels obtained when pSVIIIenv was not co-transfected were 8 Ϯ 2%. The results shown are the means Ϯ S.E. of three independent experiments.  2-4). Western blot was performed with anti-Vif polyclonal serum and HRP-conjugated anti-rabbit antibody. Co-immunoprecipitation of Vif is observed with scFv 4BL but is absent when scFv 4B and scFv anti-thyroglobulin are expressed (lanes 3 and 4).
HIV-1 replication, permissive (Jurkat) and nonpermissive (H9) cell lines expressing intrabodies were generated. For these experiments recombinant murine leukemia virus (MLV)-derived retroviral vectors encoding anti-Vif intrabody I4BL and control intrabody anti-thyroglobulin were used to transduce the human T cell lines H9 and Jurkat. Transduced cells were established through puromycin selection and limiting dilution cloning. Retroviral transduced cell lines were analyzed for intrabody and anti-thyroglobulin expression by Western blot (Fig. 6). Intrabodies were also detected by staining permeabilized transduced H9 and Jurkat cells with an anti-HA antibody (data not shown). The resultant H9-I4BL and Jurkat-I4BL cell lines showed homogeneous and stable expression of intrabody after puromycin selection during at least 3 months in cell culture. Subsequent studies on virus replication were performed in these intrabody-expressing cell lines.
Parental and transduced permissive and nonpermissive cell lines were challenged with HIV-1 utilizing the highly cytopathic viral strain HIV-1 NL4 -3 . Infectivity assays were performed with an MOI of 0.05 or 0.1 to mimic natural infections (35). Parental nontransduced H9 and Jurkat cell lines supported vigorous replication of HIV-1, as shown by the initial increases in HIV-1 p24 antigen, which peaked at ϳ14 -20 days. In contrast, low levels of HIV-1 p24 antigen were observed in the supernatants of H9-I4BL and Jurkat-I4BL as well as parental cell lines infected with HIV-1 NL4 -3 ⌬vif at early time points after infection. On days 14 -20 post-infection, H9-I4BL cells showed ϳ94 -98% inhibition of HIV-1 p24 antigen production compared with the parental H9 cells. Similar results were obtained in H9 cells infected with HIV-1 NL4 -3 ⌬Vif. In contrast, at the same time points of viral replication, Jurkat-I4BL cells showed similar levels of HIV-1 p24 antigen production compared with parental Jurkat cells. These results indicate that HIV-1 replication and infectivity was dramatically reduced in nonpermissive H9 cells by expression of a Vif-specific intrabody and were consistent in multiple independent experiments. In addition, expression of control anti-thyroglobulin scFv in the same cells had no significant inhibitory effect on HIV-1 replication, indicating that anti-Vif scFv is specific for Vif inhibition in other nonpermissive cell lines (Fig. 6).
Previous studies have shown that intracellular antibody expression has no obvious negative effects on cell viability or proliferation. Nevertheless, we quantified cell proliferation and cell viability of transduced H9-I4BL and Jurkat-I4BL cells compared with parental cells. The assay consists of a colorimetric assay based on the cleavage of the tetrazolium salt WST-1 by mitochondrial dehydrogenases in viable cells (Roche). The kinetics of the metabolism of WST-1 reagent showed that H9-I4BL and Jurkat-I4BL cells have similar proliferation curves compared with parental cell lines (Fig. 6). The evaluated time points were the same as those used for p24 antigen detection. Therefore, expression of anti-Vif singlechain antibody stably expressed in the cell inhibits HIV-1 replication in multiple rounds of infection.
Proviral DNA Synthesis Is Impaired for Virus Produced from H9 Cells Expressing I4BL Intrabody-The synthesis of viral DNA in H9 cells exposed to HIV-1 NL4 -3 produced from H9-I4BL and Jurkat-I4BL cells was analyzed by PCR, with prim-  5 and 6). The lysates of Jurkat and H9 cells not expressing scFv were also used in Western blot as negative controls (lanes 1 and 2). ers specific for products of reverse transcription, as described previously (6,28,52). Reverse transcription of retroviral RNA includes a series of steps, beginning with the synthesis of the minus strand strong stop and continuing with the elongation of this minus strand, with the final product constituting the completed double-stranded DNA (30). The M667/M661 primer pairs flank the primer-binding site and specifically detect HIV-1 DNA within completed or nearly completed reverse transcripts. To determine whether proviral DNA synthesis is impaired, Jurkat-I4BL and H9-I4BL were co-cultured with 293T cells transfected with HIV-1 NL4 -3 , and virus was isolated 9 days later. H9 cells were incubated with the same amounts of virus normalized by TCID50, and aliquots of the infected cultures were harvested and lysed. Total cellular lysates were then subjected to PCR analysis. The results in Fig. 7 showed that late reverse transcripts were significantly reduced in H9 cells infected with HIV-1 NL4 -3 derived from H9 intrabody-expressing cells (H9-I4BL). Completed or nearly completed reverse transcripts were reduced to 5% of the viral infection originated with HIV-1 NL4 -3 produced in H9 cells. PCR analysis was also performed in H9 chronically infected cells after 12 days, and the results were similar. These results show that infection of H9 cells with virus derived from Jurkat-I4BL showed similar infection levels compared with viruses produced in the absence of intrabody expression (Fig. 7). In contrast, we demonstrate that HIV-1 NL4 -3 virus derived from intrabody-producing cells is defective for reverse transcription.
Inhibition of HIV-1 Viral Infectivity by Anti-Vif I4BL Is Specific for Nonpermissive Cells-Vif modulates HIV-1 infection in cultured T-cell lines in a cell-dependent manner (2,4,6,8,9). To evaluate the specificity of viral inhibition in different cell lines, a recombinant HIV-1 expressing I4BL or anti-thyroglobulin scFv in cis was generated. This recombinant HIV-1 was derived from pHXBnPLAP-IRES-Nϩ by replacing the human PLAP gene by I4BL or anti-thyroglobulin scFv, generating pHXB-I4BL and pHXB-thyro. In this experiment, 293T cells were transfected, and a high titer supernatant of HXB-I4BL, HXB-thyro and HXB2⌬vif was obtained. HIV-1 supernatants normalized for the same TCID50 were used to infect permissive (Jurkat and SupT1), semi-permissive (CEM), and nonpermissive cells (H9, U38, and PBMC), and their ability to replicate in these cells was assessed by performing standard virus growth curves. Three patterns of results were obtained (Fig. 8). All of the nonpermissive cells tested in this experiment did not support the spread of HXB-I4BL. Similar results were obtained with HXB2⌬vif. In contrast, permissive cell lines supported replication of all viruses used in this experiment. In CEM cells, replication of HXB-I4BL exhibited an intermediate behavior, although in some experiments its replication was higher than expected. As a control, HXB-thyro was able to replicate at wild-type levels in all cells tested (Fig. 8). HXB-I4BL was also co-cultivated with HIV-1 NL4 -3 in nonpermissive cells, and infection was monitored by p24 antigen production. In this coculture experiment MOI of HIV-1 NL4 -3 were similar and five times higher than HXB-I4BL. Compared with infection of HIV-1 NL4 -3 alone in nonpermissive cells, p24 antigen production was dramatically reduced from 94 to 97% at similar MOI and from 87 to 95% with higher MOI (data not shown). Thus, the HXB-I4BL can control replication of wild-type virus. These results show that the specificity of Vif inhibition by the intrabody correlates with the cellular requirements for Vif function, confirming the specificity of anti-Vif scFv.

Inhibition of HIV Replication by Anti-Vif scFv in Transduced
Human PBMC-To evaluate the potential of anti-Vif scFv for clinical application, PHA-stimulated human PBMC were transduced with a retroviral vector that expresses anti-Vif scFv I4BL. Lentiviruses are attractive vectors for gene therapy because of their ability to integrate into nondividing cells (42,43,44). The retroviral vectors used in this study are derived from the HIV SIN system described previously by Naldini et al. (42). This delivery system has been shown to be highly effective in transduction of primary cells and offers significant advantages for its predicted biosafety (42). Intrabodies were cloned in the self-inactivating vector pRRL.SIN and expressed under the cytomegalovirus promoter. Replication-defective retroviral particles were generated by transient co-transfection of 293T human kidney cells with the combination of three plasmids as described under "Experimental Procedures." The p24 antigen concentration was determined, and the amount corresponding to 100 ng was used to transduce PBMC. The transduction protocol consisted of the daily infection of stimulated human PBMC with fresh 100 ng of viral p24 supernatants for 3 days, leading to transduction efficiencies, as measured by anti-HA immunofluorescence, ranging from 40 to 48% for three different experiments. The efficiency of transduction was significantly higher then previous experiments with MLV vectors, which may reflect the higher capability of HIV-based vectors for infecting human cells.
PBMC were challenged with HIV-1 NL4 -3 and two clinical isolates of HIV-1 2 in the range of MOI 0.05-0.1. Fig. 9 shows a dramatic inhibition of HIV-1 NL4 -3 production by anti-Vif I4BL transduction. Similar values of inhibition are obtained for the HIV-1 clinical isolates. In contrast, HIV-1 replication in PBMC transduced with anti-Vif I4B and anti-thyroglobulin showed less inhibition or no inhibition, respectively. These results show the potential for anti-Vif scFv for inhibiting HIV-1 replication in primary cells by ex vivo transduction.  7. Proviral DNA synthesis is impaired for virus produced from H9 cells expressing I4BL intrabody. The total cell lysates were prepared from aliquots of 5 ϫ 10 4 cells and were subjected to PCR analysis with a primer pair specific for late products of reverse transcription. A, H9 cells were challenged with DNase-treated stocks of HIV-1 NL4 -3 derived from H9 and Jurkat parental cells (lanes 1) and cells expressing anti-Vif scFv I4BL (lanes 2). HIV-1 NL4 -3 ⌬vif was also produced from H9 and Jurkat cell lines and used as a control (lanes 3). B, PHA-stimulated PBMC were challenged with similar viral stocks as in A. C, H9 cells infected as in A were subjected to PCR analysis with control primers for ␤-globin. A dilution series of H9 and PBMC was also subjected to PCR amplification with the same primer pairs as in A and B (right hand panels). These reactions were performed within the semiquantitative range. All of the reaction products were visualized by Southern hybridization and then by autoradiography.

DISCUSSION
The action of Vif is essential for the completion of proviral DNA synthesis after virus entry, most likely as a result of its effect during virus production (6, 25, 28). ScFv-based strategies directed against Vif may be an effective approach to inhibit these two crucial steps of the viral replication cycle (36 -38). The genetic approach described here abrogates viral infectivity and reverse transcription by functional deletion of Vif, thereby inhibiting HIV-1 replication prior to integration of viral DNA in the host genome. We showed that anti-Vif scFv binds to Vif in the cytoplasm. We also showed that scFv co-immunoprecipitates Vif, probably because of its high affinity. Anti-Vif scFv expressed transiently, constitutively, or by retroviral transduction abrogated viral replication in primary human T-lymphocytes. Wild-type viral particles produced from nonpermissive cells expressing anti-Vif scFv have a defect in infectivity at the level of reverse transcription because of neutralization of Vif in trans. These results confirm previous studies demonstrating that Vif acts in the producer cell and affects virus assembly. Consistent with this model, the expression of anti-Vif scFv in target cells had no effect on the infectivity of wild-type HIV-1.
The transducing vector used in this study was derived from HIV-1 (42,43,44), which can transduce dividing and nondividing human T-lymphocytes. Comparative studies using a vector derived from MLV demonstrated that the HIV-derived vector is more efficient at protecting human T-lymphocytes from HIV infection (data not shown). The effectiveness of neutralization was assessed by challenging intrabody-transduced PBMC with an HIV-1 laboratory-adapted strain and HIV-1 primary isolates. The Vif epitope recognized by the 4BL intrabody may be dominant, because it neutralizes both laboratory adapted and primary HIV-1 strains. Effective intracellular immunization must be designed for highly divergent proteins that share only a few peptide motifs because HIV-1 rapidly evolves compensatory mutations in response to selective pressures. Extended in vitro challenge with HIV-1 is necessary to determine whether Vif will acquire escape mutations associated with resistance.
Vif is predominantly localized in the cytoplasm of infected cells. A small fraction is localized in the nucleus, reflecting a possible role of Vif in this compartment. Our results show that scFv expressed in the cytoplasm co-localizes with Vif. Thus, the result of our study predicts that the cytoplasm is where Vif functions during HIV-1 replication.
The requirement for Vif differs among cell lines (2,4,6,8,9). Recent studies suggest that Vif enhances infectivity by overcoming an inhibitory factor present in nonpermissive cells (53,54). When anti-Vif scFv was cloned in HIV-1 in place of the nef gene, viral replication occurred only in permissive cells, whereas in nonpermissive cells replication was similar to vifnegative HIV-1. The activity of the anti-Vif intrabody was shown to be cell-specific, because in cis and in trans it had inhibitory effects only in nonpermissive cells. Anti-Vif scFv will help to validate candidates for HIV-1 cell-specific inhibitors that can be neutralized by Vif (20,54).
The results presented here extend previous studies demonstrating that high affinity single-chain antibodies can be stably expressed in the cytoplasm (31,34,36,35,37). Folding of the scFv in the reducing environment of the cytoplasm can form functional binding sites as demonstrated by co-immunoprecipitation. Furthermore, the toxicity of the scFv was low as assessed by cell viability in intrabody-expressing human T-lymphocytes. We constructed a single-chain antibody by ligating VL with VH using a short and long linker. It was anticipated that a short linker peptide would result in a dimer scFv with two sites for antigen binding (33). Nevertheless, we observed a lack of binding of the 4B antibody. The reason for this result might be reduced stability or folding yield at the growth temperature. The short linker peptide may inhibit the coherent structure of the framework regions, thereby altering the binding affinity.
A large panel of scFvs against Vif is being constructed to map epitope-binding regions and putative functional sites. Using a panel of anti-Vif scFvs directed to specific intracellular sites, it may be possible to gain novel insights into Vif function. In addition, this strategy will facilitate the devel-opment of rational combinatorial gene therapies that simultaneously target different functional motifs.