Feline Immunodeficiency Virus OrfA Is Distinct from Other Lentivirus Transactivators

Abstract

The feline immunodeficiency virus (FIV) accessory factor, OrfA, facilitates transactivation of transcription directed by elements of the viral long terminal repeat (LTR). In order to map OrfA domains required for this transactivation, we used N- and C-terminal deletion constructs of the protein, expressed in a Gal4-based transactivation system. The results demonstrated that FIV OrfA, unlike other lentiviral transactivators such as visna virus Tat, is unable to transactivate from minimal promoter-based reporters and requires additional elements of the viral LTR. Stable CrFK-based cell lines were prepared that expressed OrfA to readily detectable levels and in which we were able to demonstrate 32-fold transactivation of an LTR-chloramphenicol acetyltransferase construct. Transactivation was heavily dependent on the presence of an ATF site within the viral LTR. Changing the translation initiation codon context substantially increased the level of production of OrfA from a bicistronic message that also encodes Rev. In the presence of a more favorable context sequence, the upstream expression of OrfA increased 21-fold, with only a 0.5-fold drop in downstream Rev expression. This suggests that Rev translation may occur via an internal ribosomal entry site rather than by leaky scanning.

Accessory factors constitute important elements of the lentiviral genome that distinguish this group from simple retroviruses. These factors facilitate the growth of lentiviruses in cell environments that are typically not conducive for the growth of simple retroviruses, such as nondividing cells of the monocyte/macrophage lineage. Accessory factors are typically found in the cells rather than packaged in the virus, a characteristic that fits well with their role in modulating viral replication. The body of evidence also points to the ability of these factors to alter viral pathogenesis in vivo by altering specific biochemical pathways to facilitate virus survival and propagation within a particular host cell. The consensus model for most of the accessory proteins is that they serve as tethering proteins that in some circumstances may interact with the cellular transcription-translation machinery in several unique ways.

One of the most interesting groups of accessory proteins is the transactivators, which are responsible for increasing the level of viral mRNA synthesis in the host cell. In human immunodeficiency virus type 1 (HIV-1), the Tat protein acts on the transactivation response (TAR) element of the viral long terminal repeat (LTR) to facilitate a dramatic increase in the levels of viral transcripts (5). Similarly, equine infectious anemia virus (EIAV) Tat acts on a Tar-like element of the equine lentivirus LTR to activate transcription (11). The Tax protein of human T-lymphotrophic virus type 1 (HTLV-1), on the other hand, mediates transactivation via direct interaction with members of the ATF/CREB family of transcription factors binding to three 21-bp repeat regulatory elements present in the viral LTR (1, 4, 54, 61). A similar protein-protein interaction involving the binding of visna virus Tat to AP-1 transcription factors, Fos and Jun, brings it proximal to the TATA box-binding protein (TBP) and mediates transactivation (35). In feline immunodeficiency virus (FIV), a short open reading frame referred to as OrfA (or Orf2) is required for productive growth of the virus in T lymphocytes, the primary target for virus infection (59), and facilitates the transactivation in FIV (18, 49, 59). Transactivation in the feline lentivirus resembles that in visna virus in magnitude and is significantly different from that of HIV. Both FIV and visna virus lack a TAR element to act as a target for transactivation (5, 23, 49). Moreover, the basal transcriptional levels in visna virus and FIV are very high compared to HIV (23, 49). Visna virus Tat has an acidic domain between amino acid residues 13 and 38 required for high-level transactivation (10), as well as a pattern of critical hydrophobic residues reminiscent of other acidic activation domains (35). FIV, on the other hand, has a large hydrophobic region in the N-terminal half with acidic residues distributed within the same region, followed by a well-defined stretch of hydrophobic residues.

In previous studies, we have shown that FIV OrfA increases LTR-driven gene expression 14- to 30-fold over basal transcription rates (18). The sites in the LTR important for transactivation in vitro were localized to the 3′ end of U3 and include AP-1, C/EBP, and ATF sites. The results indicated that the removal of all three sites was required to totally abrogate OrfA-mediated transactivation in transient-transfection assays (18). The purpose of the present study was to expand on these findings by using a series of Gal4-binding constructs tethered to subgenomic fragments of FIV OrfA to activate transcription from a minimal promoter, a technique similar to what has been done with visna virus Tat (10). In addition, OrfA-expressing stable cell lines were prepared to further analyze the transcription, translation, and function of OrfA. Surprisingly, the findings show that the mechanism of OrfA-driven transactivation is totally distinct from that of visna virus Tat, in spite of the apparent parallels and evolutionary relatedness between these two lentiviruses.

MATERIALS AND METHODS

Cell lines and viruses.

Crandell feline kidney cells (CrFK [21]) were obtained from the American Type Culture Collection and were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and other supplements as described earlier (33). Briefly cells were cultured in DMEM (Gibco-BRL, Rockville, Md.) containing 10% heat-inactivated FBS (Gemini Bioproducts, Calabasas, Calif.), 200 μM l-glutamine (Sigma, St. Louis, Mo.), 1× nonessential amino acids (NEA; Sigma), 1× sodium pyruvate (Sigma), 1× MEM-Vitamins (Sigma), 5.5 × 10⁻⁵ M β-mercaptoethanol (Gibco-BRL) and 50 μg of gentamicin sulfate (Gemini Bioproducts)/ml.

For the studies described here, two molecular clones, FIV-34TF10 and FIV-PPR, derived from the Petaluma isolate (53) and the San Diego isolate (39), respectively, were used. Infection was monitored by a micro-reverse transcriptase (RT) activity assay as described previously (17).

Sequence comparison.

We used the CLUSTALW algorithm (56) within the MacVector 6.5 software (Oxford Molecular, Madison, Wis.) for calculating the percent identity and the percent similarity of OrfA from different strains of FIV. OrfA sequences were aligned in two stages; the first was a pairwise alignment by using a full dynamic programming algorithm based on the Blosum 30 matrix, an open gap penalty of 10.0 and an extended gap penalty of 0.1. In the second stage, multiple alignments were carried out by using identical penalties (mentioned above) with a cutoff for sequences that were <40% identical.

Synthetic oligonucleotide primers.

Sequences of all oligonucleotides used for the present study are available from the authors on request. The primers used for various amplifications are described under the respective sections.

PCR amplification.

PCRs were carried out in 100 μl containing 200 μM deoxynucleoside triphosphates (Promega), 1× KlenTaq PCR buffer (Clontech, Palo Alto, Calif.), 100 ng of template DNA, 700 ng of each of the 5′ and 3′ primers, and 0.5× KlenTaq polymerase mix (Clontech). Reactions were carried out in a Perkin-Elmer Cetus thermocycler with 5-min presoak at 94°C; followed by 35 cycles of 94°C for 15 s, 58°C for 30 s, and 72°C for 60 s; followed by a final 10-min soak at 72°C.

Bacterial expression of OrfA.

A construct was prepared for overexpression of OrfA protein in bacteria by PCR amplification of the sequence corresponding to nucleotides 5992 to 6225 of FIV-PPR, with a 5′ primer encoding an NdeI site and 3′ primer encoding an EcoRI site to facilitate cloning. The amplified DNA was cloned into pET21a(+) and pET28a(+) vectors (Novagen, Madison, Wis.). The latter vector has an N-terminal six-histidine tag, in-frame with OrfA and separated by a thrombin cleavage site to aid in nickel-affinity purification. Clones were screened primarily by colony PCR, followed by restriction analysis and sequencing according to standard protocols (46, 47). Competent BL21(DE3) cells (50, 51) were transformed with the pET21OrfA and pET28OrfA constructs, and representative transformants were randomly checked by colony PCR. Cultures of these two clones were grown to an A₆₀₀ of 0.6 and then induced with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 2 to 4 h. Inclusion bodies were purified, and the protein was extracted as described earlier (12, 31), followed by ion-exchange and size exclusion chromatography. OrfA expressed from pET28OrfA was further purified by affinity chromatography on a nickel-nitrilotriacetic acid column prior to gel filtration and size exclusion chromatography. All of the steps were monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) as described elsewhere (46).

Antibodies.

To raise OrfA-specific antibodies, we used the method of Summers and Szewczyk (52) to purify Western blotted and transferred inclusion body proteins. Bands corresponding to the overexpressed 9-kDa OrfA were eluted from Immobilon-P membranes (Millipore) in a high-pH elution buffer, concentrated through an Ultrafree-15 centrifugal filter device (Millipore) with a molecular weight cutoff of 5 kDa, quantitated, and then injected into New Zealand White rabbits. Postimmunization bleeds were tested by Western blot against the bacterially expressed OrfA protein and in immunoprecipitations against ³⁵S-labeled OrfA protein transcribed and translated in vitro, as described below. Affinity-purified JL-8 monoclonal antibody (Clontech) was used for both Western blot analyses, as well as immunoprecipitation of green fluorescent protein (GFP), in these studies as per the manufacturer's recommendations.

Western blot analysis.

To test the reactivity of the OrfA protein under various conditions, we tested the protein in Western blot analysis as described earlier (46). Typically, antibodies were used at a dilution of 1:100 in 3% bovine serum albumin (fraction V) suspended in 1× phosphate-buffered saline (PBS).

In vitro transcriptions-translations and immunoprecipitations.

Expression of all cDNAs cloned under the human cytomegalovirus (CMV) promoter in pCR3 (Invitrogen, Carlsbad, Calif.) were tested in vitro by using the proximal T7 promoter in coupled transcription-translation assays with a rabbit reticulocyte system (Promega). In vitro-translated lysate (15 to 20 μl) was used in immunoprecipitations of OrfA or Gal4-OrfA constructs with 2.5 μl of anti-OrfA polyclonal rabbit serum, followed by precipitation with protein A- and protein G-agarose beads.

Gal4-OrfA deletion constructs.

We fused the N- or C-terminally deleted Orf2 in frame with the Gal4 DNA-binding domain (see Fig. 3A and B). Internal deletions of the PPR OrfA gene in-frame with the Gal4 DNA-binding domain were generated by PCR-ligation-PCR (2). Briefly, the first PCR amplification was carried out on pCR3-Gal4 with G1/HindIII+G5/EcoRI and on ARPPR with T1/EcoRI+LA13, T3/EcoRI+LA13, T5/EcoRI+LA13, and T7/EcoRI+LA13 (see Fig. 3A). The G1/HindIII+G5/EcoRI and the T1 to T7/EcoRI+LA13 amplifications after cleanup were ligated to each other through the EcoRI site. The ligation products served as the templates for the second round of PCR. Amplification on the G1/HindIII-G5/EcoRI-T1/EcoRI-LA13 template was carried out with G1/HindIII+FT2/XbaI, G1/HindIII+FT4/XbaI, G1/HindIII+FT6/XbaI, and G1/HindIII+FT8/XbaI; on G1/HindIII-G5/EcoRI-T3/EcoRI-LA13 with G1/HindIII+FT8/XbaI; on G1/HindIII-G5/EcoRI-T5/EcoRI-LA13 with G1/HindIII+FT8/XbaI; and on G1/HindIII-G5/EcoRI-T7/EcoRI-LA13 with G1/HindIII+FT8/XbaI. The amplifications were restricted with HindIII and XbaI and ligated into pCR3 to yield, respectively, Gal4-OrfA_1-32, Gal4-OrfA_1-42, Gal4-OrfA_1-53, Gal4-OrfA_1-77, Gal4-OrfA_34-77, Gal4-OrfA_44-77, and Gal4-OrfA_55-77 (see Fig. 3B). The constructs were verified by sequencing and then expressed in vitro in a rabbit reticulocyte system (Promega) and immunoprecipitated with anti-OrfA antibody as described above.

FIG. 3.

Gal4-OrfA assay. (A) PCR strategy followed for generating the Gal4-OrfA deletion clones by PCR-ligation-PCR (see Materials and Methods for details). (B) Diagram of the deletion clones of OrfA fused in frame with the Gal4 DNA-binding domain generated by the strategy describe in panel A. (C) Autoradiogram of the various deletion constructs expressed in vitro under the T7 promoter, immunoprecipitated with anti-OrfA polyclonal serum, and resolved by SDS-10 to 20% PAGE. (D) Transactivation of the G₅-E₁b-CAT target construct by the different Gal4-OrfA deletion constructs and appropriate controls.

OrfA deletion constructs.

To define domains within OrfA, we prepared N- and C-terminal deletions by using PCR methodologies. Briefly, we used primers 5′OrfA-1/EcoRI+3′OrfA-234/XbaI, 5′OrfA-1/EcoRI+3′OrfA-99/XbaI, 5′OrfA-1/EcoRI+3′OrfA-162/XbaI, 5′OrfA-97/EcoRI+3′OrfA-234/XbaI, and 5′OrfA-160/EcoRI+3′OrfA-234/XbaI (primer sequences available on request) to amplify specific regions of OrfA from FIV-PPR. All of the PCR samples were purified, restricted with EcoRI and XbaI, and ligated into similarly digested pCR3 to generate OrfA_1-77, OrfA_1-33, OrfA_1-54, OrfA_34-77, and OrfA_55-77 (see Fig. 4). The selected clones were sequenced on both strands, expressed in vitro, and immunoprecipitated as described above.

FIG. 4.

Analysis of domains within OrfA. (A) Diagram of the different OrfA deletion constructs used in the present study is shown. (B) Autoradiogram of the various deletion constructs expressed in vitro under the T7 promoter, immunoprecipitated with anti-OrfA polyclonal serum, and resolved by SDS-10 to 20% PAGE.

Transfections and CAT assays.

Transfection-grade DNAs were prepared by using Qiagen Midi- and/or Maxiprep kits as per the manufacturer's protocols. The Gal4-E1b-chloramphenicol acetyltransferase (CAT) construct (a kind gift from Michael Green, Howard Hughes Medical Institute, University of Massachusetts Medical Center) used in the present study has been described (34). FIV-LTR_WT-CAT (also referred to as FIV LTR-CAT in previous publications) and the various site-specific internal deletion mutants of the FIV LTR used in the present study have been described in detail before (18, 55).

We typically used 1 μg of the target, along with 10 μg of the OrfA full-length or C- and N-terminal deletion constructs or pUC18 as a filler plasmid, in calcium phosphate-mediated cotransfections of subconfluent CrFK cells, as reported previously (59). All transfections were carried out in triplicate, and the cells were incubated for 40 to 44 h prior to reporter assays. CAT activity was measured from the 20 μg of total cellular protein estimated by Lowry assay (18) by using the phase extraction assay described earlier (48).

For transfecting cell lines stably expressing OrfA, we used a slight modification of the above procedure. In brief, at 24 h after seeding, the stable cells were washed twice with DMEM supplemented with 10% FBS to remove all traces of hygromycin or G418 present in the medium, followed by the standard transfection outlined above. This allowed us to maintain identical transfection conditions for these cells. At 24 h posttransfection, the medium for the transfected stable cells was switched back to the original medium (containing the drug marker of choice) to maintain conditions optimal to OrfA production. These steps increased the transfection efficiencies of the OrfA-expressing stable cells to a significant extent.

In vitro expression of OrfA-Rev bicistronic message.

For cloning and expression of OrfA with ideal Kozak consensus sequences (26, 27, 29, 30), we used primers 5′ Kozak OrfA/EcoRI and 3′ OrfA-Rev/XbaI (sequences available on request). To maintain the original sequence around OrfA, we used the 5′ OrfA/EcoRI as a control in conjunction with the 3′ OrfA-Rev/XbaI. We also made an additional construct where the consensus Kozak sequence 5′ to the ATG was shifted by −1 bp by using 5′ Kozak (−1) OrfA/EcoRI in conjunction with the same 3′ primer. Using the three different primer sets enumerated above, we amplified the OrfA-Rev sequence from a previously described bicistronic mRNA from FIV-PPR-infected cells (pFIV1.3.4 [18]) to generate sequences encoding the bicistronic message with or without an ideal Kozak sequence preceding it. The PCR-amplified product was restricted with EcoRI and XbaI and cloned into identical sites of pCR3 (Invitrogen). The modified and control OrfA-Rev cDNAs were transcribed and translated as described earlier. Each reaction was divided into two 50-μl aliquots; α-OrfA serum was added to one half, and α-Rev serum was added to the other. As a control, identical reactions were incubated with prebleeds from the respective animals in which the anti-OrfA and anti-Rev serum had been generated (data not shown).

Generation of OrfA-expressing stable cell lines.

In order to generate the OrfA-expressing cell lines, we generated three new constructs. In the first instance, OrfA was put under the CMV promoter of the pcDNA3.1/Hygro(+) vector (Invitrogen). The second construct placed an N-terminal FLAG tag on OrfA with an optimal Kozak sequence, cloned under the CMV promoter of the pcDNA3 vector (Invitrogen) (see Fig. 6A). The last construct placed a C-terminal hemagglutinin tag on OrfA, cloned under the CMV promoter of the same vector, with an ideal Kozak sequence. The CrFK-basal cell lines were tested for their sensitivity to hygromycin B and G418 under increasing concentrations of these drugs to optimize the right concentration for stable cell line selection (data not shown). OrfA-expressing cell lines were selected on the basis of resistance to hygromycin B or G418 and screened in a CAT reporter assay by transfection with the FIV LTR-CAT construct. Based on cell growth and viability characteristics and on relative OrfA expression levels, we selected the G418-resistant C-5-9 cell line that stably expresses FLAG-tagged OrfA for detailed studies.

FIG. 6.

Generation of OrfA-expressing cell lines. (A) Schematic of the FLAG-tagged OrfA construct coexpressing the neomycin phosphotransferase gene for resistance to G418. (B) The putative OrfA product from nuclear extracts of metabolically labeled C-5-9 cells (lane 2) immunoprecipitated with anti-OrfA antiserum was compared to identically treated CrFK cells (lane 1). (C) CAT assay of the same cell line (C-5-9) used in panel B but transfected with the FIV LTR-CAT construct and assayed for relative CAT levels. As a control, a construct (shown as -LTR) having the CAT gene but lacking the LTR was used to determine background CAT levels for each of these cell lines. In addition, simultaneous cotransfections of CrFK cells with the same FIV-LTR_WT-CAT and the FLAG-OrfA constructs were done to serve as a control. The CAT activity from C-4-3, a cell line expressing higher levels of OrfA but exhibiting a high rate of cell death, is also shown.

Construction of GFP- and OrfA-coexpressing cell line.

For retroviral infection, we cloned the OrfA gene from FIV-PPR between the XhoI and EcoRI sites of Mig R1 (38). We excised out the DC-SIGN gene from a Mig R1 construct containing DC-SIGN (a kind gift of Benhur Lee, University of California at Los Angeles) at the XhoI-EcoRI sites and used the backbone vector. The recombinant retroviral stock for transducing the PPR OrfA gene in CrFK and primary cell line was generated by transient transfection of the GP2-293 packaging cell line (Clontech). Cells in six-well plates were transfected with 3.5 μg of Mig R1-OrfA or Mig R1 and 0.5 μg of pVSV-G by using Fugene as recommended (Roche, Indianapolis, Ind.). CrFK and the primary T-cell line 104-C1 were transduced with 1 ml of the retroviral supernatant by spinoculation as described before (36). GFP-expressing cells were sorted with a FACS Vantage SE I (BDIS, San Jose, Calif.) at the TSRI Flow Cytometry Core Facility to select and enrich for the GFP-expressing population. The sorted cells were scaled up and analyzed for GFP expression by flow cytometry analysis on a FACScan by using CellQuest software (BDIS).

Metabolic labeling of cells expressing OrfA.

In order to detect OrfA in cells, we labeled the control CrFK and C-5-9 cells with Tran³⁵S-label (ICN, Irvine, Calif.). Both of these cell lines were split, seeded in 150-mm tissue culture dishes, and allowed to grow to ∼70% confluence. Cells were starved for 2 h in Met⁻ Cys⁻ NEA⁻ DMEM supplemented with 10% FBS, with or without G418, and then metabolically labeled for 16 h with 0.1 mCi of Tran³⁵S-label/ml. Nuclear and cytoplasmic extracts were prepared from the labeled cells according to published protocols (3, 19, 20). For immunoprecipitations, 50 μl of the respective extracts was diluted in 500 μl of ice-cold PBS containing 0.5 mM phenylmethylsulfonyl fluoride (Sigma). Immunoprecipitations were carried out successively using a combination of protein A- and protein G-agarose. Samples were precleared by treatment with preimmune rabbit serum plus protein A/G-agarose, followed by treatment with anti-OrfA polyclonal serum from the same animal. The protein A/G-agarose-complexed immunoprecipitated products were resolved on a 10 to 20% Tris-Tricine gel (Invitrogen), fixed, treated for signal enhancement with En³Hance (NEN, Boston, Mass.), and then dried. Dried gels were exposed to Biomax MR film (Kodak, Rochester, N.Y.) at −70°C for various periods.

For detecting expression of OrfA in 104-C1 cells, we used ³⁵S-labeled methionine and cysteine in RPMI 1640 medium lacking these amino acids. Cells were labeled for 1 h and then lysed, and the cell lysate was used in immunoprecipitations as described above.

RESULTS

Comparison of OrfA from different strains of FIV.

In order to analyze the similarity of OrfA from the different strains of FIV published thus far, we compared each of them to the FIV-PPR isolate by using the CLUSTALW program. As can be seen from Fig. 1A, OrfA is highly variable, with the TM2 strain from Japan being the most diverse (60% identity) and FIV-14 OrfA being the most related (71% identity) to FIV-PPR. A striking consensus in the primary protein sequence between the different strains exists at the C-terminal end within residues 66 to 75, as well as the cysteine cluster from residues 56 to 62 (Fig. 1A). There are other regions of consensus dispersed throughout the sequence that either have the same amino acid or an amino acid of high similarity at equivalent positions. Comparison between the secondary structure formation potential of different isolates revealed that the TM2 isolate of FIV may be fundamentally distinct from the other isolates (Fig. 1B). The consensus overlap between the Chou-Fasman (13-15) and the Robson-Garnier (22, 41) algorithms revealed that TM2 OrfA had the potential to form a β-sheet structure not evident in other consensus sequences. Furthermore, the TM2 isolate had a drastically reduced consensus α-helix spanning through the OrfA protein, whereas all of the other isolates had distinctly larger stretches of helix-forming potential. However, this is only a model, and the actual structures may not fit this prediction.

FIG. 1.

OrfA sequence comparisons. (A) Published sequences of OrfA from different FIV isolates were compared by using the CLUSTALW multiple alignment algorithm. (B) Chou-Fasman and Robson-Garnier consensus secondary structure-forming potential for PPR-OrfA and TM-2 OrfA are compared to show distinctions.

OrfA purification.

OrfA was purified by the isolation of inclusion bodies from E. coli after induction by IPTG. Induction of the pET21OrfA construct did not result in the high expression of OrfA in any of the seven constructs identified by Western analysis with an anti-OrfA polyclonal serum raised against His-tagged OrfA (A. K. Waters and J. H. Elder, unpublished data). Coomassie blue staining of gels, as well as Western blot analysis (Fig. 2A), revealed the presence of both 9- and 18-kDa forms of OrfA. To further purify these proteins and verify relatedness, proteins in the inclusion bodies were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes, and bands corresponding to the 9- and 18-kDa forms were excised and eluted from the membrane. As can be seen in Fig. 2A, reanalysis of the recovered bands by SDS-PAGE revealed that the 9-kDa band generated the 18-kDa band, while a percentage of the 18-kDa band now ran as 9 kDa, a finding consistent with the notion that these were one and the same protein. These results indicate that the 9-kDa component could associate strongly with itself as an apparent dimer and resist complete breakdown during electrophoresis, even under denaturing conditions. To check the stability of the 18-kDa protein, we subjected the protein to a number of different denaturing reagents, detergents, and reductants. These included guanidine hydrochloride, dithiothreitol, CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfo-nate}, diethylamine, SDS at 0 to 12.5%, and β-mercaptoethanol. Only β-mercaptoethanol at 0.5 M was able to break the 18-kDa complex, with or without boiling (Fig. 2B). Thus, the protein formed extremely stable dimers, apparently via disulfide linkages. Caution is warranted, however, since overexpressed proteins in E. coli often form multimers (unpublished observations). Any functional relevance of dimerization remains to be determined. We were unable to detect any 18-kDa product when OrfA was expressed in vitro under the T7 promoter by coupled in vitro transcription and translation (data not shown). This could be due to the low level of protein production (150 to 300 ng [Protocols and Applications Guide, 3rd ed.; Promega Corp., Madison, Wis.]) from such systems compared to the bacterially expressed product.

FIG. 2.

Purification of OrfA. (A) Western blot (lane 1) and Coomassie blue-stained gel (lanes 2 and 3) of purified OrfA showing the presence of both the 9- and the 18-kDa forms (lane 1) and the membrane-purified 9-kDa (lane 2) and its dimeric form (lane 3) analyzed by SDS-10 to 20% PAGE. (B) Dissociation of the OrfA dimer in the presence of 0.5 M β-mercaptoethanol (lane 5) and with simultaneous boiling (lane 6) or neither of these treatments (lane 4), as detected by Western blot analysis with anti-OrfA antibody. Lanes 1 to 3 show identical treatments of the 9-kDa OrfA protein.

OrfA antibody.

The 9-kDa form of OrfA was used to immunize rabbits, and postimmunization bleeds were checked at various intervals for the ability to detect in vitro-transcribed and -translated OrfA by immunoprecipitations, as well as in Western blot analysis of lysates from virus-infected cells. Both 9- and 18-kDa forms of OrfA overexpressed in Escherichia coli were recognized strongly in Western blots (Fig. 2). In addition, the antibody immunoprecipitated OrfA was expressed in a coupled in vitro transcription-translation system (see Fig. 5B, lanes 2, 4, and 6). However, repeated attempts to detect OrfA expression in virus-infected cells failed (not shown), a finding consistent with extremely low expression levels in the latter cells or with rapid turnover.

FIG. 5.

Effect of altering the consensus Kozak sequence in the bicistronic message on the relative expression levels OrfA and Rev. (A) The placement of the Kozak sequence (middle) and its modification (bottom) in the context of the original OrfA sequence (top) is depicted in cartoon form. (B) A total of 1.5 μg each of the Kozak-OrfA-Rev (lanes 1 and 2), Kozak (−1)-OrfA-Rev (lanes 3 and 4), and OrfA-Rev (lanes 5 and 6) constructs were expressed in vitro, and 50% of the lysate were immunoprecipitated with either anti-Rev polyclonal serum (lanes 1, 3, and 5) or anti-OrfA polyclonal serum (lanes 2, 4, and 6). The autoradiogram of the immunoprecipitated products analyzed by SDS-10 to 20% PAGE is presented. The fold levels of the expression with respect to the original sequence are presented on the right of each construct in panel A. (C) Putative hairpin structure at the end of the OrfA reading frame and preceding that of Rev.

Transactivation by OrfA.

Based on earlier data (18, 59) showing that a functional OrfA is important for transactivation, we attempted to define the domains of OrfA important for the transactivation phenomenon. We followed a strategy similar to that described earlier in other systems (10, 34, 35, 42, 43), wherein the transactivation domain was determined by expression of various deletion constructs of the protein fused in frame with the Gal4(1-147) DNA-binding domain (Fig. 3A and B). The terminal deletion mutants of OrfA were fused in frame with the Gal4 DNA-binding domain, and the constructs were tested for their ability to express in vitro in a coupled transcription and translation system in rabbit reticulocytes. As can be seen in Fig. 3C, the fusion product expressed very well, and each could be immunoprecipitated by rabbit anti-OrfA antibody. Each construct (10 μg) was cotransfected into subconfluent CrFK cells with 1 μg of the E₁b-CAT construct. CAT activity from the transfected cells was measured at 48 h posttransfection as described earlier (48). As can be seen from Fig. 3D, none of the Gal4-OrfA constructs could transactivate from the Gal₅-E₁b-CAT target, whereas Gal4-VP16 or Gal4-Tat(Visna)_13-38 could transactivate more than 300-fold in CrFK cells. This indicates that OrfA is unable to facilitate transactivation by direct interaction with the TATA box, in contrast to finding with visna virus Tat.

In order to determine whether the OrfA could interact with the His-tagged human TBP (hTBP), we expressed both of these proteins as bacterial expression products and then tried reciprocal pull-down assays with a monoclonal antibody to the His₆ tag and the rabbit polyclonal serum to OrfA. We were unable to detect any association between the hTBP and OrfA by SDS-PAGE and Coomassie staining (data not shown).

The lack of transactivation from the Gal4-OrfA fusion products forced us to reconsider our strategy to demarcate domains of OrfA important for the transactivation phenomenon. Based on the amino acid residue charge, as well as property distribution, we redefined the domains that could be important and narrowed it down to four specific deletion constructs (Fig. 4A) instead of the six we had originally chosen. We considered the cysteine- and the leucine-rich areas to be of prime importance, while designing our deletion constructs based on the high degree of conservation in these regions (Fig. 1A). OrfA expression vector constructs were sequenced, and the proteins were transcribed and translated in vitro as described above. The expressed products were immunoprecipitated with anti-OrfA antibody, and the precipitates were analyzed on a 10 to 20% SDS-PAGE gel. As can be seen from Fig. 4B, we had a faithful expression of the truncated OrfA proteins in vitro. The same constructs were tested in CAT reporter assays with the UK8 LTR- or PPR LTR-driven CAT genes. We were, however, unable to see transactivation from any of the truncated OrfA constructs (data not shown).

Presence of the consensus Kozak sequence 5′ to OrfA leads to an increase in the OrfA expression.

To examine the issue of the influence of OrfA with respect to Rev translated from the same bicistronic message, we attempted sequence modifications favorable to OrfA expression. In brief, we changed the translation initiation codon context of OrfA to that favored by most eukaryotic and viral genes, CC(A/G)CCATGG (26-28) (Fig. 5A). We transcribed and translated the bicistronic message from this modified gene in vitro and immunoprecipitated the OrfA and Rev products with their respective antibodies (Fig. 5B). The radiolabeled immunoprecipitates were analyzed by SDS-PAGE and autoradiography, and the respective products were quantitated by densitometry. As can be seen in Fig. 5B, the expression of OrfA increased 21-fold by incorporating a favorable Kozak consensus sequence around the OrfA initiation methionine. Expression of Rev decreased to 0.55-fold of the level obtained by using wild-type bicistronic message. A modified construct in which an additional adenine residue was inserted 5′ to the AUG sequence gave an eightfold increase in OrfA expression relative to the wild-type construct, whereas Rev expression remained identical to that of the wild-type control. Again, results with FIV vary from those reported for EIAV (9), where a dramatic increase in EIAV Tat expression resulted in a concomitant decrease in downstream Rev expression.

Examination of the intergenic region of the OrfA-Rev bicistronic message revealed the presence of a 22-bp region that has remarkable dyad symmetry, likely to be capable of forming a stable stem-loop structure (Fig. 5C). This sequence starts immediately downstream of the OrfA stop codon UAG and is located 16 bp upstream of the initiation codon for Rev. The mRNA of the TM-2 isolate of FIV lacks the potential to form such a hairpin structure (data not shown). However, the TM-2 strain expresses Rev from a monocistronic mRNA (57), thus negating the need for such a structure.

Generation of a stable cell line expressing OrfA.

Despite having a functional antibody that recognizes the OrfA protein in Western blot analyses, as well as in vitro-generated OrfA, we failed to detect OrfA in infected cells or in purified virus (data not shown). This could be attributed to low production in cells under normal conditions associated with chronic infections, rapid clearing from cells, or masking by association with other cellular or viral components. In order to examine this issue, we generated a cell line in a CrFK background expressing OrfA under the CMV promoter and colinked to a hygromycin B resistance cassette driven by the simian virus 40 promoter (data not shown). Clones were selected by endpoint dilution and resistance to 0.5 mg of hygromycin B/ml. We also generated a cell line where the OrfA construct was fused 3′ to a FLAG tag (DYKDDDDK) (40) (Fig. 6A). These clones were also selected as described above, but by expressing the colinked resistance to neomycin-kanamycin by using G418 at 1 mg/ml. Lysates (200 μg) from different OrfA-expressing cell lines were analyzed by Western as well as immunoprecipitation with the anti-OrfA antibody. As can be seen from Fig. 6B, a faintly detectable band corresponding to ∼9 kDa was immunoprecipitated from nuclear extracts with the anti-OrfA polyclonal serum in C-5-9, a cell line resistant to G418 (lanes 2) but not from the lysates of vector-transfected CrFK cells (Fig. 6B, lane 1). Efforts to localize OrfA by immunocytochemistry in formaldehyde-, paraformaldehyde-, or methanol-fixed cells (12) failed with the same anti-OrfA serum (data not shown).

To assay for the functionality of OrfA in these cells, we transfected the C-5-9 cell line with a wild-type FIV-LTR_WT-CAT construct. OrfA activity was detected in all stable cell lines (data from two representative lines are shown in Fig. 6C). The hygromycin-resistant OrfA-expressing cell line produced consistently lower levels of transactivation (data not shown) than did the G418-resistant FLAG-tagged OrfA line (Fig. 6C), with the highest levels of transactivation being 32-fold in C-4-3 with respect to transactivation from the non-drug-resistant CrFK cell line. This cell line, however, grew very slowly and exhibited a high level of cell death (data not shown) presumably due to toxicity from higher expression of OrfA. Transient cotransfection with the FLAG-OrfA and LTR-CAT construct gave results similar to the cell line expressing FLAG-OrfA, although the CAT activity was somewhat higher in case of the former (Fig. 6C). All cell lines were screened for OrfA mRNA by RT-PCR with oligonucleotides directed specifically to this gene, as well as primers for the 18S RNA as an internal control (data not shown).

OrfA expression and LTR-CAT interactions in the stable cell line.

In order to understand the interaction between OrfA and the FIV LTR, we decided to concentrate on C-5-9, a CrFK cell line stably expressing OrfA. The prime determinant for this selection was the median expression level for OrfA and its ability to grow well in culture without dying early (as opposed to toxicity observed with high levels of OrfA expression). In initial experiments, we focused on the ability of this line to transactivate the wild-type FIV LTR relative to CrFK. Figure 7A shows a typical transfection with the appropriate controls in CrFK cells. The wild-type FIV LTR was transactivated ∼11-fold in the C-5-9 cell line compared to CrFK cells. We used FIV LTR constructs containing various site-specific mutations (18) to map effector regions of stably expressed OrfA. As shown in Fig. 7B, deletion of the ATF site caused a marked reduction in the ability of OrfA to augment transcription (66.3% reduction). A marked reduction (31.8%) was also observed with the C/EBP deletion mutant, but AP1 and NF1 deletions caused no significant reduction in transactivation (Fig. 7B).

FIG. 7.

Modulation of LTR activity by OrfA-expressing cell lines. (A) CAT assay of CrFK cells or the C-5-9 cell line transfected with FIV-LTR_WT-CAT or FIV-47LTR-CAT constructs and assayed for relative CAT levels. Simultaneous cotransfection of CrFK cells with the same FIV-LTR_WT-CAT constructs with or without FLAG-OrfA were performed as a control. (B) CAT activity in the OrfA-expressing cell line C-5-9 transfected with FIV-LTR_WT-CAT (WT) or FIV-47LTR-CAT (−47) or having various site-specific deletions (δAP1, δC/EBP, δNF1, and δATF).

Selection of cell lines transduced with Mig R1 and Mig R1-OrfA.

Expression of GFP was detected in both 104-C1 and CrFK cell lines transduced with Mig R1, as well as Mig R1-OrfA, by Western blot analysis, as seen in Fig. 8A (lanes 2, 3, 5, and 6). However, anti-OrfA serum failed to detect the 9-kDa protein from the same lysates in the same Western blot analysis (data not shown), a finding consistent with a low net OrfA concentration in these cells. We were, however, able to faintly detect OrfA in the 104-C1 cells transduced with the OrfA-GFP construct by immunoprecipitation but not in those transduced with GFP alone (Fig. 8B, lanes 1 and 2). Lysates from these cells had an abundance of GFP when immunoprecipitated with the anti-GFP monoclonal antibody (Fig. 8B, lanes 3 and 4). Given that OrfA should have been expressed at levels comparable to GFP by using this vector system, the findings suggest that a high turnover rate may be a major contributing factor to the low abundance of OrfA.

FIG. 8.

Stable expression of OrfA in CrFK and 104-C1 cells. (A) Western blot analysis of equal amounts of lysates from 104-C1 (lanes 1 to 3) or CrFK (lanes 4 to 6) either mock transduced (lanes 1 and 4) or transduced with GFP (lanes 2 and 5) and OrfA-GFP (lanes 3 and 6) with the anti-GFP monoclonal antibody (JL-8; Clontech). (B) Immunoprecipitation with anti-OrfA polyclonal serum (lanes 1 and 2) and anti-GFP antibody (lanes 3 and 4) from equal amounts of cell lysates of 104-C1 cells transduced with either GFP (lanes 1 and 3) or OrfA-GFP (lanes 2 and 4) metabolically labeled with ³⁵S-labeled methionine and cysteine. The arrows point to the positions of OrfA and GFP in the autoradiogram. (C) Micro-RT activity assay of CrFK cells transduced with either GFP or OrfA-GFP infected with 34TF10 or mock infected. The values represent RT activity within the first 5 days postinfection.

Effect of OrfA expression on the infectivity of 34TF10.

In order to determine the effect of OrfA supplied in trans on the replication of FIV-34TF10, which lacks a functional OrfA (32), we infected CrFK cells expressing GFP or OrfA and GFP with 500,000 RT units of the virus for 1 h. As can be seen from Fig. 8C, the presence of OrfA leads to an increase in virus expression in these cells in the first 5 days postinfection. At a later time (5 to 10 days postinfection), the amounts of RT from the two different cells increase to similar levels (data not shown). The PPR isolate of FIV failed to productively infect either GFP-expressing or OrfA-GFP-expressing CrFK cells, a finding consistent with previous findings that indicated that OrfA was not a factor in failure of FIV-PPR to productively infect CrFK cells (32).

DISCUSSION

The findings demonstrate several unique properties of FIV OrfA that set it apart from other viral transactivators, as well as confirming many similarities. First of all, alignment comparisons reveal substantial sequence divergence in OrfA among the known FIV clades, with variability greater than that observed between the Env genes of this virus family. However, certain domains are highly conserved, which presumably gives clues as to critical functional domains of the protein. An apparent homology exists between the cysteine cluster in the C-terminal half of OrfA and a 7-amino-acid stretch of the G-alpha interacting protein (19, 20). G-alpha interacting protein is a palmitoylated membrane-associated protein that specifically interacts with members of the Gα_i subfamily (16). Studies to determine any functional relevance of this homology are in progress.

Findings also indicate a unique mechanism by which OrfA facilitates net transactivation. In spite of ancestral relationship and parallels in their life cycles, FIV and visna virus do not share similar mechanisms of transactivation. Studies of the visna virus transactivator (10) demonstrated that the activation of visna virus Tat could be mapped by using the Gal4 DNA-binding domain tethered to various N- or C-terminal deletions of the proteins. Since visna virus is the closest lentivirus relative of FIV (53), we predicted that OrfA would act in a similar manner. We verified that a construct consisting of visna virus Tat residues 13 to 38 linked to the Gal4 DNA-binding domain [Gal4-Tat(Visna)_13-38] could strongly transactivate the basal promoter. However, cotransfection of a series of FIV OrfA-Gal4 constructs with the G₅-E₁b-CAT gave the unequivocal result that these constructs, unlike Gal4-VP16 (34) or Gal4-Tat(Visna)_13-38 (10), were unable to transactivate from a minimal promoter. We were also unable to see electrophoretic mobility shifts with either recombinant or in vitro-transcribed or -translated OrfA with various fragments of the FIV LTR (data not shown), a finding consistent with a lack of direct interaction between the FIV LTR and OrfA. In addition, we were unable to detect an interaction between in vitro-expressed or -purified His-tagged hTBP and OrfA (data not shown), as has been shown with HIV-1 Tat (24). Thus, OrfA seems to transactivate through a pathway that is mechanistically different from that published for other lentiviruses. The minimal E₁b promoter is identical in 9 of 11 bases to that of FIV, suggesting that inherent differences did not lead to our observations. It is more likely that OrfA requires additional factors in order to achieve transactivation. Current experiments have focused on nested deletions of OrfA to determine the region most important for transactivation from a FIV-LTR_WT-CAT reporter. Combinatorial analysis of these OrfA deletion mutants and deletions in the LTR (18, 55) are also under way to fine map the region where the interaction occurs and to aid in defining potential interaction partner(s).

The generation of OrfA-expressing cell lines will go a long way in aiding the delineation of OrfA function. These cell lines have provided a valuable insight regarding the interaction of expressed OrfA with the FIV LTR. Earlier studies implicated the importance of the AP1 site on the FIV LTR, and our studies reaffirm the importance of this site in maintenance of high basal level transcription from the FIV LTR. However, the present data also indicate that the ATF site is critical for OrfA-directed enhancement of transcription. Importantly, the stable expression of OrfA seems to overcome the necessity for the presence of the AP-1 site that was deemed essential in earlier studies (16, 55, 59). This dependence on the ATF site closely mirrors the model proposed for HTLV-1 Tax-mediated assembly of enhancer complexes (6). OrfA, like Tax, is unable to contact the transcriptional machinery at the minimal promoter directly. As with Tax, OrfA may require the presence of one or more proteins(s), perhaps like cyclic AMP response element (CRE)-binding protein, to bring about transcriptional transactivation of the FIV LTR. The fact that OrfA doesn't bind to synthetic oligonucleotides containing AP-1, C/EBP, or the ATF sequences in electrophoretic mobility shift assays (not shown) is consistent with the notion that secondary interactions are indeed important for OrfA to contact the viral LTR. Unlike HTLV-1, however, there is only a single CRE within the FIV LTR through which such interactions might be taking place. The data presented in the present study are in agreement with recent findings that deletions of the ATF-binding sequence resulted in restricted viral expression and replication in peripheral blood mononuclear cells and macrophages (7).

We have been unable to detect OrfA in virus-infected cells by metabolic labeling or by Western blot techniques. We have been able to immunoprecipitate OrfA from the stably transfected cells with OrfA-specific antibodies. However, we were unable to localize the protein by immunocytochemistry (with the same antibodies), in spite of demonstrating a functional presence of the protein in transactivation assays. The ability to detect the 9-kDa form by immunoprecipitation from metabolically labeled cells but not by immunocytochemistry probably reflects the extremely low levels of OrfA in the cells. The low levels of OrfA detected may represent either high turnover or low stability within the cells. Initial metabolic experiments suggest that stability might indeed be an issue, since 104-C1 cells transduced with an OrfA-GFP expression construct and labeled for 12 h failed to immunoprecipitate OrfA, whereas immunoprecipitation from cells labeled for 1 h revealed its presence. It is also likely, although unproved at this point, that association with other complexes within the cell might direct OrfA through a ubiquitin-mediated pathway for proteolysis, as has been shown for a number of transcription factors (44, 45). Further investigations are under way to determine whether OrfA is indeed being degraded. Evidence of increased toxicity in transfected cell lines expressing high levels of OrfA would explain a selection for low-level expression of the transactivator.

In FIV-PPR, OrfA is expressed as a bicistronic message with Rev as the second part of the message. A possibility is that FIV-PPR has evolved in a parallel manner by providing an RNA hairpin that could displace the ribosomal machinery after OrfA translation and prior to Rev (see Fig. 5C). Important to this line of reasoning is that, unlike FIV-PPR, FIV-TM-2 lacks the potential to form the hairpin structure preceding Rev and generates Rev from a monocistronic mRNA (57). Recent data suggest that an internal ribosomal entry site (IRES) may often be present in bicistronic messages as a way to differentially regulate the expression of two genes (8, 60) or to regulate alternate initiation from the RNA, as shown for a number of retroviruses (18, 37). In FIV-PPR a number of observations lead us to propose that OrfA and Rev expression might be regulated in this manner. First, Rev occurs in a more favorable context in FIV with respect to the −3 as well as the +4 nucleotides (with respect to the ATG) (25) than does OrfA. Second, the presence of a GCA as the codon flanking the AUG (30) of FIV Rev might strongly augment the expression of Rev as opposed to OrfA from the bicistronic message. Third, the presence of a strong dyad symmetry at the 3′ end of OrfA and 15 bp upstream of the Rev AUG suggests the potential to trap the 40S ribosomal subunit behind the base-paired structure (29), either partially or totally uncoupling the expression of the two proteins. This model of entrapment and displacement of the ribosome would be analogous to that of the oligopyrimidine motif of Moloney murine leukemia virus (58). The presence of an IRES upstream of Rev would be consistent with our findings in that enhanced upstream translation of OrfA had little influence on downstream Rev production. Studies are in progress to test directly for the presence of an IRES element upstream of Rev.

Overall, the findings presented here underscore the unique properties of the OrfA gene product of FIV but also draw parallels with transactivators in other systems that will direct future efforts to define mechanisms utilized by this interesting protein.