Molecular Characterization of the Host Defense Activity of the Barrier to Autointegration Factor against Vaccinia Virus

Abstract

The barrier to autointegration factor (BAF) is an essential cellular protein with functions in mitotic nuclear reassembly, retroviral preintegration complex stability, and transcriptional regulation. Molecular properties of BAF include the ability to bind double-stranded DNA in a sequence-independent manner, homodimerize, and bind proteins containing a LEM domain. These capabilities allow BAF to compact DNA and assemble higher-order nucleoprotein complexes, the nature of which is poorly understood. Recently, it was revealed that BAF also acts as a potent host defense against poxviral DNA replication in the cytoplasm. Here, we extend these observations by examining the molecular mechanism through which BAF acts as a host defense against vaccinia virus replication and cytoplasmic DNA in general. Interestingly, BAF rapidly relocalizes to transfected DNA from a variety of sources, demonstrating that BAF's activity as a host defense factor is not limited to poxviral infection. BAF's relocalization to cytoplasmic foreign DNA is highly dependent upon its DNA binding and dimerization properties but does not appear to require its LEM domain binding activity. However, the LEM domain protein emerin is recruited to cytoplasmic DNA in a BAF-dependent manner during both transfection and vaccinia virus infection. Finally, we demonstrate that the DNA binding and dimerization capabilities of BAF are essential for its function as an antipoxviral effector, while the presence of emerin is not required. Together, these data provide further mechanistic insight into which of BAF's molecular properties are employed by cells to impair the replication of poxviruses or respond to foreign DNA in general.

INTRODUCTION

The barrier to autointegration factor (BAF) is a highly conserved 10-kDa dimeric DNA binding protein found in both the cytoplasm and the nucleus of many cell types (18, 19, 42). Attempts to deplete or knock out BAF in Caenorhabditis elegans and Drosophila melanogaster resulted in lethal phenotypes early in embryogenesis, indicating that BAF is essential in animal models (8, 20, 45). Extensive in vitro biochemical and structural studies of BAF have advanced our understanding of the unique DNA binding properties, protein-protein interactions, and functions of BAF. Each BAF monomer folds into five helixes containing both a helix-hairpin-helix DNA binding domain and a dimerization surface (41). BAF interacts with the minor grove of double-stranded DNA (dsDNA), with contacts made along the phosphate backbone of both strands, allowing BAF to bind specifically to dsDNA but independent of the sequence (4, 45). Upon dimerization, the DNA binding domains from each monomer are positioned at opposite ends of the protein, allowing a single BAF dimer to bind two parallel strands of DNA and form a cross bridge (4, 45). When combined, the sequence-independent DNA binding and cross-bridging abilities of BAF give it the unique capacity to compact and/or aggregate DNA into larger nucleoprotein structures (37). While the molecular nature of these BAF-DNA complexes is not yet completely understood, they are likely intimately linked to BAF's functions in both the nucleus and cytoplasm.

Nuclear BAF has been found to play an important role in the late stages of mitosis by assisting with the recruitment of the nuclear envelope to chromatin (8, 10, 20, 31). BAF achieves this via its ability to bind simultaneously to both DNA and inner nuclear membrane proteins containing a BAF interaction domain referred to as a Lap2 emerin Man1 (LEM) domain, named for the founding proteins in which this domain was characterized (7). Most LEM domain proteins reside within the inner nuclear membrane, although some LEM domain proteins are also found in the cytoplasm or nucleoplasm (33, 38). In addition to LEM domain proteins, BAF has been found to interact with histones, transcriptional regulators, and proteins involved in DNA repair (16, 23, 24, 42). This suggests that in addition to its mitotic role, BAF may modulate gene expression and DNA damage responses in the nucleus as well. The cellular control of BAF's various functions occurs in part through posttranslational modification. Specifically, the cellular vaccinia-related kinases (VRKs) have been found to phosphorylate BAF's conserved N terminus in multiple species (9, 26), leading to the potent inhibition of BAF's DNA binding ability and diminished LEM domain binding (2, 26). The importance of BAF phosphorylation is clear from studies of cells expressing the BAF-M₁A₂A₃A₄Q₅ N-terminal mutant, which is rendered unphosphorylatable through the alteration of the phosphoacceptor residues 2, 3, and 4 to alanine. Indeed, decreased BAF phosphorylation has been observed to cause defects in the association of BAF with chromatin and the aberrant formation of the nuclear envelope, resulting in a delay of cell cycle progression (9, 15).

In the cytoplasm, BAF was first characterized as a host component of retroviral preintegration complexes capable of blocking the suicidal autointegration of viral cDNA in vitro (17). However, the importance of BAF during retroviral infection in cultured cells is controversial, with recent studies demonstrating only a modest difference in viral integration in cells depleted of BAF (35, 44). In contrast to those studies suggesting that BAF may be beneficial for retroviruses, we have recently demonstrated a novel role for BAF as an antiviral host defense protein. Specifically, during infection with vaccinia virus, the prototypical poxvirus, BAF is capable of localizing to cytoplasmic DNA replication factories and strongly inhibiting viral genome replication (43). However, vaccinia virus expresses a viral kinase known as B1, which exhibits a high level of similarity to cellular VRKs. Importantly, like VRKs, B1 is also capable of phosphorylating and repressing BAF's DNA binding ability (26). These data indicate that poxviruses have usurped a conserved cellular signaling pathway in order to allow viral DNA replication to proceed. These data also reveal that BAF can act as an intrinsic immune effector against viral infection, which we hypothesize occurs via its ability to compact and aggregate viral DNA.

Despite these significant advances in the characterization of BAF, much remains to be determined regarding its function and regulation as a host defense protein. Specifically, it is not known whether BAF's response during vaccinia virus infection is triggered exclusively by that virus or whether BAF is capable of responding to foreign DNA in general. A more in-depth analysis of how BAF's interaction with DNA as well as other protein partners contributes to its antiviral properties also warrants investigation. The experiments described in this study were designed to address those knowledge gaps and define the molecular mechanism though which BAF acts as an antipoxviral factor. Here, we demonstrate that double-stranded DNA alone is sufficient to trigger the rapid relocalization of cytoplasmic BAF and that BAF's DNA binding and dimerization domains are critical for BAF's response. Furthermore, we developed the first cell models for the stable expression of BAF mutant proteins, which has allowed us to characterize the molecular interactions needed for BAF's response to foreign DNA. This has also led to our observation that BAF facilitates the recruitment of the LEM domain protein emerin to cytoplasmic DNA during transfection and infection. Finally, using both stable cell lines and recombinant vaccinia viruses, we determined that BAF's DNA binding domain and dimerization are needed to inhibit vaccinia virus growth, which is consistent with our hypothesis that BAF acts as a host defense mechanism by compacting and/or aggregating viral DNA.

MATERIALS AND METHODS

Cell culture and reagents.

All chemicals were purchased from Fisher Scientific or Sigma-Aldrich unless otherwise stated. Human thymidine kinase-negative (TK⁻) 143B osteosarcoma cells and African green monkey BSC40 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Atlanta Biologicals) and penicillin-streptomycin. Flp-In CV1 cells were purchased from Invitrogen and maintained in DMEM–10% FBS and 100 μg/ml zeocin (Invitrogen) prior to stable transfection, after which zeocin was replaced with 100 μg/ml hygromycin (Invitrogen).

Virus preparation and growth assays.

Wild-type (wt) vaccinia virus (WR strain) and ts2 virus were propagated in BSC40 cells at 37°C and 31.5°C, respectively. Viral stocks were prepared from cytoplasmic lysates of infected cells by ultracentrifugation through 36% sucrose cushions at 18,000 rpm in an SW41 rotor for 90 min and resuspension of the viral pellet in 1 mM Tris (pH 9). Quantitation of viral yield was performed by titration on BSC40 cells.

Virus yield assays following 24, 48, or 72 h of infection were performed following infection at the multiplicities of infection (MOIs) indicated in the figure legends. The day prior to infection, equal numbers of cells of each cell line were plated at confluence in a 35-mm-diameter dish. At the time points given, cells were harvested into 10 mM Tris (pH 9) and freeze-thawed three times prior to titration on BSC40 cells at 31.5°C for plaque assays. Error bars represent the standard errors for the experimental set for each figure.

qPCR.

Viral DNA was quantified by quantitative PCR (qPCR) using methods similar to those previously described (32). Briefly, DNA was extracted from infected cells with a Qiagen blood kit, treated with RNase A, and used in qPCR mixtures with SYBR green PCR master mix (Applied Biosystems) with 300 nM primers C11UP (5′-AAACACACACTGAGAAACAGCATAAA-3′) and C11DN (5′-ACTATCGGCGAATGATCTGATTATC-3′), specific for the vaccinia virus C11 gene, using a StepOnePlus real-time PCR machine (Applied Biosystems). Amplification was performed during 40 cycles of 95°C for 15 s and 60°C for 60 s. The differences in DNA quantities shown in Fig. 1B are relative to those found with CV1-CAT cells infected at an MOI of 0.1. Relative fold differences were calculated according to a method described previously by Pfaffl (29), using the equation, fold difference = E^ΔCT, where E is the relative primer amplification efficiency of the C11 primers calculated from a standard curve produced from qPCR of a serial dilution of viral DNA and C_T is the cycle threshold.

Fig. 1.

BAF inhibits ts2 growth in a dose-dependent manner. (A) Immunoblot analysis of BAF expression in cell lines overexpressing or depleted of BAF. The migration of epitope-tagged 3×Flag-BAF is indicated at the left by an arrowhead, and the migration of endogenous BAF is indicated by an arrow. Equivalent numbers of cells were loaded per lane. Total amounts of BAF in each cell line were quantified by using a Bio-Rad Chemidoc XRS instrument and are shown relative to amounts of control cell lines at the bottom. (B) Viral yield obtained following CV1 cell infection with wild-type or ts2 vaccinia virus (MOI = 0.01) after 24, 48, and 72 h at 37°C. (C) Viral yield in cells with various amounts of BAF. Cell lysates were made 24 h after infection with ts2 virus at 37°C using the cell line and MOI shown. For panels C and D, virus production was then quantified by a plaque titration assay on BSC40 cells at 32°C. Data were obtained from two independent experiments, titrated in duplicate. Error bars represent the standard errors of the means. (D) Viral DNA accumulation in cells with various amounts of BAF. Total DNA was isolated 24 h after infection with ts2 virus at 37°C using the cell line and MOI shown and then quantified by qPCR. Data were obtained from two independent experiments, PCR amplified in duplicate.

Mutagenesis and cloning of BAF expression vectors.

pcDNA5/FRT/TO/CAT, pcDNA5/FRT/TO/3XFLAG-BAF, pJS4-RFP-BAF, and pJS4-RFP-BAF-MAAAQ vectors were described previously (43). To construct vectors introducing the single-amino-acid mutations K6A, G47E, and K53E, wild-type BAF or the BAF-MAAAQ sequence was used as a template for overlap PCR mutagenesis using outside primers and one set of the following internal mutagenesis primers: WTBAF-K6AUP (5′-GACAACCTCCCAAGCGCACCGAGACTTCGTG-3′) and WTBAF-K6ADN (5′-CACGAAGTCTCGGTGCGCTTGGGAGGTTGTC-3′), MAAAQBAF-K6AUP (5′-GCAGCCGCCCAAGCGCACCGAGACTTCGTG-3′) and MAAAQBAF-K6ADN (5′-CACGAAGTCTCGGTGCGCTTGGGCGGCTGC-3′), BAF-G47EUP (5′-GACAAGGCCTATGTTGTCCTTGAACAGTTTCTGGTGCTAAAGAAAG-3′) and BAF-G47EDN (5′-CTTTGTTTAGCACCAGAAACTGTTCAAGGACAACATAGGCCTTGTC-3′), or BAF-K53EUP (5′-GTTTCTGGTGCTAGAGAAAGATGAAGACC-3′) and BAF-K53EDN (5′-GGTCTTCATCTTTCTCTAGCACCAGAAAC-3′). Outside primers were specific for the expression vector and included KpnBamFlag (5′-GAGGGTACCGGATCCGCCACCATGGACTACAAAGACC-3′) and BAF-DNBam (5′-GCAGGATCCTCACAAGAAGGCGTCGCAC-3′) for the pcDNA5/FRT/TO insertion at the BamHI site (underlined in this primer set). Alternatively, FLAG-BAF-UPXho (5′-CAGCTCGAGGCCACCATGGACTACAAAGACC-3′) and BAF-DNBam were used, which places an XhoI site (shown in italics) upstream and a BamHI site downstream of the open reading frame (ORF) for the pJS4-RFP insertion at these sites. The introduction of each of these BAF mutations into selected clones was verified by DNA sequencing.

Production of stable cell lines.

The stable integration of chloramphenicol acetyltransferase (CAT), 3×Flag-BAF, or BAF mutants was performed by using the Flp-In system (Invitrogen) according to methods described by the manufacturer. This system employs CV1 Flp cells containing a single integrated copy of the pFRT/lacZeo plasmid, which possesses an Flp recombination target (FRT) recombination site. Briefly, these cells were cotransfected with the pcDNA5/FRT/TO/3XFLAG-BAF wild-type or mutant vector of choice and pOG44, a vector expressing the Flp recombinase. Stable cell lines were selected by growth in 200 μg/ml hygromycin for 3 weeks and 100 μg/ml hygromycin thereafter.

BAF-depleted CV1 cells were derived by stably transducing cells with a lentivirus expressing a BAF-specific short hairpin RNA (shRNA) or scrambled shRNA as previously described (43), with the exception that selection was performed with 15 μg/ml of puromycin. To deplete emerin from cells, lentiviruses were produced as previously described (43), using vector pLL3.7 expressing emerin-specific (5′-GACCUGUCCUAUUAUCCUA-3′) shRNA.

Immunofluorescence.

Cells were plated onto chamber slides (Lab-Tek) 24 h prior to transfection or infection. Transfection was performed by using 1 μg of nucleic acid per milliliter of medium and 2 μl of Lipofectamine 2000 (Invitrogen) reagent according to the manufacturer's protocol. At 7 h posttransfection, cells were fixed for 15 min at room temperature using 4% paraformaldehyde in phosphate-buffered saline (PBS) (10 mM Na₂HPO₄·7H₂O, 1 mM KH₂PO₄, 2 mM KCl, 140 mM NaCl [pH 7.4]). Infections were performed (MOI = 5) using a temperature shift protocol optimized to visualize BAF and emerin at ts2 viral DNA replication factories. Specifically, infections were initiated at an initial temperature of 31.5°C for 9 h, followed by a shift to 39.7°C for 7 h prior to fixation with paraformaldehyde. BrdU (5-bromo-2′-deoxyuridine) (25 μg/ml) was added at 4 h postinfection (hpi) to label replicating DNA and allow for detection via the G3G4 anti-BrdU antibody (Developmental Studies Hybridoma Bank, University of Iowa).

Following the fixation of either infected or transfected cells, membranes were permeabilized by using 0.1% saponin or 0.2% Triton X-100 (TX-100), as indicated in the figure legends, in PBS for 5 min at room temperature. Cells were then incubated with mouse anti-Flag M2 antibody (Sigma), rabbit anti-emerin antibody (catalog number sc15378; Santa Cruz Biotechnology), or rabbit anti-Lap2α (catalog number ab5162; Abcam) in PBS plus 0.05% saponin. This was followed by Alexa Fluor 594- or 488-conjugated goat anti-mouse or anti-rabbit antibody (Molecular Probes) at a 1:500 dilution in PBS plus 0.05% saponin. DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI). To detect endogenous BAF in CV1-CAT cells, an additional antigen retrieval step was included, and cells were treated with 0.5% SDS–PBS for 5 min after fixation in paraformaldehyde (PFA) and permeabilization with TX-100 and then washed extensively prior to overnight incubation with rabbit anti-BAF primary antibody (catalog number sc33787; Santa Cruz Biotechnology). Proteins were observed by indirect fluorescence on an inverted (Olympus IX 81) confocal microscope. Images were pseudocolored using ImageJ software. The quantitation shown in Fig. 7 was done by counting the number of cells with 0, 1 to 2, 3 to 4, or 5+ speckles, and the numbers of speckles were averaged from two independent experiments. Cells from 3 random fields were counted for each cell line and included 100 to 120 total cells in each case.

Fig. 7.

Characterization of the ability of BAF mutant proteins to relocalize during DNA transfection. (A) Lysates produced using equivalent cell numbers were subjected to immunoblot analysis to verify the stable expression of BAF mutant proteins in CV1 cells (left). Doubling rates of these cells were calculated by plating equal numbers of cells and counting cells present after 96 h of growth (right). Error bars represent the standard deviations obtained after averaging three independent cell counts. WCL, whole-cell lysate. (B) CV1 cells expressing 3×Flag-BAF or BAF mutants transfected with 1 μg pUC19 DNA per ml of medium. Cells were fixed 7 h later, permeabilized by using 0.1% Triton X-100, and processed for immunofluorescence imaging using an M2 anti-Flag primary antibody and Alexa Fluor 594 secondary antibody. The representative images shown were taken by using a confocal microscope at a ×60 magnification (top). Scale bars, 50 μm. The numbers of cells with 0, 1 to 2, 3 to 4, or 5 or more puncta per cell were quantified from 3 independent fields of view and are shown as a percentage of the total number of cells counted (bottom). Bars represent average counts from two independent experiments, and error bars represent standard deviations.

Nucleic acid.

pUC19 and all other plasmid DNAs described in this study were isolated from JM109 cells and purified by using an endotoxin-free Qiagen kit. The isolation of vaccinia virus DNA was performed as described previously (40). Briefly, viral cores were first treated with β-mercaptoethanol, 1% SDS, and 1 mg/ml proteinase K and then gently extracted with phenol-chloroform and chloroform prior to ethanol precipitation of the DNA. Commercial sources of nucleic acid were used for Escherichia coli K-12 DNA (Invivogen), M13 single-stranded DNA (ssDNA) (New England BioLabs) and poly(I-C) dsRNA mimic (Invivogen).

DNA binding and immunoprecipitation assays.

Cytoplasmic lysates from equal amounts of CV1 cells were made 24 h after they had been transiently transfected to express the 3×Flag-BAF wild-type and mutant constructs. For immunoprecipitation assays, a construct expressing green fluorescent protein (GFP)-tagged wild-type BAF was also cotransfected at this time. Lysis was performed for 10 min on ice in lysis buffer (50 mM Tris [pH 7.4], 75 mM NaCl, 1 mM EDTA, 4% sucrose, and 0.5% Triton X-100) supplemented with fresh 1 mM phenylmethylsulfonyl fluoride (PMSF) and protease and phosphatase inhibitors (Complete Mini and PhosStop tablets; Roche), after which nuclei were removed following centrifugation at 800 × g. For DNA binding assays, 50 μl of native dsDNA cellulose beads (Amersham) was added to the lysates and incubated overnight at 4°C with end-over-end rotation. For immunoprecipitation assays, M2 anti-Flag beads were added to the lysates and incubated for 2 h at 4°C with end-over-end rotation. Following incubation, either DNA cellulose or M2 beads were washed three times with lysis buffer, and bound proteins were eluted by the addition of SDS sample buffer. Proteins were visualized by immunoblot analysis with anti-BAF antibody (generously provided by Paula Traktman), anti-Flag antibody (Sigma-Aldrich), or anti-GFP antibody (Santa Cruz Biotechnology). For both DNA binding and immunoprecipitation studies, the immunoblots shown are representative of three independent replicates. The dashed line in the immunoblot (see Fig. 6C, top) marks where an irrelevant lane was removed to join lanes from the same blot.

Fig. 6.

DNA binding and dimerization abilities of BAF mutant proteins. (A) Schematic of BAF mutants and their primary functional impacts based on this study and previous studies. (B) Lysates from CV1 cells transiently expressing the indicated 3×Flag-BAF protein and GFP or GFP-BAF were subjected to immunoblot analysis using an anti-GFP or an anti-FLAG antibody (bottom) or immunoprecipitated (IP) with agarose conjugated to M2 Flag antibody prior to immunoblot analysis using anti-GFP antibody (top). (C) Lysates from CV1 cells transiently expressing the indicated 3×Flag-BAF protein were incubated with DNA cellulose. Bound protein was detected and quantitatively compared with the amount loaded onto the DNA cellulose by immunoblot analysis (top). The relative binding efficiencies were calculated from three independent experiments and are shown (bottom). Error bars represent the standard errors of the means.

Attempts to produce recombinant viruses expressing BAF mutants.

To verify the expression of 3×Flag-BAF-MAAAQ, 3×Flag-BAF-MAAAQ/K6A, 3×Flag-BAF-MAAAQ/G47E, and 3×Flag-BAF-MAAAQ/K53E from pJS4-RFP, BSC40 cells were infected with wild-type vaccinia virus (MOI = 5) and transfected with each of the pJS4 constructs expressing the BAF mutants at 1 hpi. Lysates were harvested 18 h later, and expression was verified by immunoblot analysis of whole-cell lysates using an anti-Flag antibody.

To produce recombinant virus, linearized pJS4-RFP, and red fluorescent protein (RFP) plus 3×Flag-BAF-MAAAQ, 3×Flag-BAF-MAAAQ/K6A, 3×Flag-BAF-MAAAQ/G47E, and 3×Flag-BAF-MAAAQ/K53E, plasmid DNA was introduced into BSC40 cells previously infected with wild-type vaccinia virus (MOI = 0.03), and the recombinational insertion of the transcriptional cassette into the viral TK locus was allowed to proceed during 2 days of incubation. One passage was then performed in human TK⁻ cells plus 25 μg/ml BrdU to select for TK-deficient virus, after which the viral yield was then determined by titration on BSC40 cells. Confluent monolayers of BSC40 cells on 150-mm dishes were then infected with 2,000 PFU per plate, and all plaques were examined for RFP expression by immunofluorescence. Using this method, transfection with a vector expressing RFP alone routinely yielded 30% of plaques being red.

RESULTS

BAF inhibits the growth of B1-deficient ts2 virus in a dose-dependent manner.

The goal of this study was to further understand the molecular mechanisms through which BAF relocalizes to vaccinia virus replication factories and inhibits the viral growth of the B1-deficient, temperature-sensitive ts2 virus. While we first observed this inhibition in HEK293 cells (43), we have also observed the same phenotype with the African green monkey kidney CV1 cell line. We have since found that in comparison to the HEK293 line, CV1 cells exhibit a superior ability to form plaques when infected with vaccinia virus and are highly amenable to immunofluorescence imaging; therefore, we utilized them as our primary cell model system in this study. To characterize the impact of different levels of BAF on the replication of vaccinia virus, we first established four CV1 stable cell lines which express various amounts of BAF, as shown by immunoblotting (Fig. 1A). These cell lines include the control line CV1-CAT, expressing chloramphenicol acetyltransferase, and a cell line overexpressing epitope-tagged 3×Flag-BAF, which brings the total amount of BAF in the cell to ∼500% of the amount of the endogenous BAF protein. Both 3×Flag-BAF and endogenous BAF migrate as two bands, although the slower, phosphorylated form is most visible for 3×Flag-BAF because of its higher expression level. Also included were two cell lines transduced with replication-incompetent lentiviral vectors to express either a BAF-specific or scrambled shRNA. The expression of the BAF shRNA results in the depletion of the protein in this cell line to 15% of control levels, while the scrambled sequence has no impact on BAF levels.

Using these cell lines, we sought to extend our previous analysis of BAF's antiviral activity against vaccinia virus by examining BAF's impact on both viral production and DNA replication at a range of MOIs and time points. For these studies, all infections were performed at a temperature of 37°C. This was done to remove the undue stress placed on cells by the higher temperatures commonly used for studies of temperature-sensitive vaccinia viruses, which we found reduces even wild-type vaccinia virus growth in CV1 cells (data not shown) and may have unknown effects on BAF's regulation. First, we examined wild-type and ts2 growth rates over 72 h following infection at a low MOI (0.01). In the course of these studies, we found that the yields in CV1-shScram and CV1-CAT cells were indistinguishable; thus, the results for the CV1-shScram cells will not be shown. As expected, no differences were seen between yields of wild-type virus in the three cell lines (Fig. 1B, solid lines). In contrast, while ts2 virus could still replicate in all 3 cell lines at this temperature, there was an inverse correlation between viral yield and BAF expression (dotted lines). The CV1-CAT cells produced 29.6-fold less ts2 virus at 48 hpi than the CV1-shBAF cells, and the viral yield in cells overexpressing BAF was 1,230-fold less than that in BAF-depleted cells. To determine whether the impact of BAF was dependent on the MOI, we next examined the ts2 viral yield at 24 h at a range of virus concentrations. We found that the ts2 viral yield in CV1-shBAF cells plateaued rapidly and that when the amount of input virus was increased, the gap in yields between CV1-shBAF and CV1-3×Flag-BAF cells shrunk from 770-fold at an MOI of 0.1 to 32-fold at an MOI of 5 (Fig. 1C). The same trend was observed when the amounts of viral DNA produced in each infection were compared by qPCR. Specifically, at an MOI of 0.1, the DNA yield differed by 3 orders of magnitude, but at an MOI of 5, the difference was 50-fold (Fig. 1D). In sum, these data provide a more thorough characterization of BAF's impact on ts2 viral DNA yield and progeny virus than was initially described. These data demonstrate that BAF possesses potent antiviral activity at a range of infectious doses but is most effective at low MOIs. The fact that 3×Flag-BAF overexpression reduced viral growth even further than endogenous BAF in control cells also indicates that BAF remains capable of acting as a host defense mechanism against vaccinia virus infection when epitope tagged.

Presence of double-stranded DNA is sufficient for BAF relocalization to cytoplasmic depots.

Based on our previous studies, it is clear that BAF's antipoxviral activity correlates with its relocalization to vaccinia virus DNA replication factories (43). Specifically, BAF's relocalization occurs during ts2 virus infection at nonpermissive temperatures, during which a block in DNA replication occurs. In comparison, no relocalization is seen during wild-type viral infection, indicating that the phosphorylation of BAF at its N terminus by active B1 kinase blocks BAF's recruitment to viral factories. Thus, the determinants of BAF's relocalization are key to an understanding of the mechanism through which BAF responds to poxviral infection; however, they remain to be fully understood.

Our working hypothesis at the onset of these studies was that BAF binds to the viral DNA at vaccinia virus factories; however, viral replication factories are comprised of not only viral DNA but also RNA and viral and cellular proteins. Therefore, BAF's recruitment to factories could be driven by protein-protein as well as protein-DNA interactions. As an example of this, BAF's interaction with retroviral preintegration nucleoprotein complexes occurs not only via BAF's interaction with viral cDNA but also via its interaction with the viral matrix protein (18).The fact that BAF also interacts with other DNA binding proteins in nuclear lysates (24) also raises the possibility that BAF's recruitment is indirect. To distinguish between these possibilities, we first tested whether BAF can relocalize to DNA in the absence of viral proteins by examining its response to plasmid DNA delivered by Lipofectamine-mediated transfection. For these studies, we employed both the CV1-CAT cell line for the examination of the endogenous BAF and the CV1-3×Flag-BAF cell line, which allow for greater sensitivity through the use of the anti-Flag antibody. Like endogenous BAF, 3×Flag-BAF is present in both the nucleus and the cytoplasm of these cells, with no evidence of aggregation or punctum formation in cells treated with Lipofectamine 2000 alone (Fig. 2A and E). However, upon the transfection of plasmid DNA, the BAF localization shifted, now concentrating in depots or puncta easily visible by 6 to 7 h posttransfection (arrowheads in Fig. 2B and F) but detectable in as little as 2 to 3 h (data not shown). The BAF puncta colocalized very well with DAPI (Fig. 2D), consistent with BAF binding to the plasmid DNA. Strikingly, these puncta were observed in >90% of the cells examined, as quantified and discussed in more depth in the legend of Fig. 7.

Fig. 2.

BAF relocalizes to discrete puncta during plasmid transfection. CV1-CAT (A and B) or CV1-3×Flag-BAF cells (C to H) were mock transfected with Lipofectamine 2000 alone or with 1 μg pUC19 DNA per ml of medium. Cells were fixed 7 h later. CV1-CAT cells were permeabilized first with 0.1% Triton X-100 and then with 0.5% SDS. CV1-3×Flag-BAF cells were permeabilized only by using 0.1% saponin. Cells were processed for immunofluorescence imaging using an anti-BAF (A and B) or M2 anti-Flag (C to H) primary antibody, Alexa Fluor secondary antibody, and DAPI. The representative images shown were taken by using a confocal microscope at a ×60 magnification. Arrowheads mark puncta containing endogenous BAF (B) and sites of 3×Flag-BAF–DAPI colocalization (D and F). Scale bars, 10 μm.

Next, we examined the specificity of BAF for the type of transfected nucleic acid. In vitro, BAF has been shown to exhibit highly specific binding to double-stranded DNA (dsDNA) (in a sequence-independent manner), with little or no affinity for single-stranded DNA (ssDNA) or dsRNA (12, 45), but this specificity has not been tested in live cells. Therefore, we examined endogenous and 3×Flag-BAF localizations during transfection with equal amounts of bacterial DNA, purified vaccinia virus DNA, M13 ssDNA, or the dsRNA mimic poly(I-C). Interestingly, the puncta formed during either bacterial or vaccinia virus DNA transfection were indistinguishable in size or number (Fig. 3A to D). Indeed, BAF relocalization was observed with other dsDNAs, including other plasmids and synthetic DNA oligonucleotides as well (data not shown). However, no relocalization was observed during transfection with either M13 ssDNA or poly(I-C) (Fig. 3E to H). Together, these data demonstrate that dsDNA devoid of viral proteins is sufficient to cause BAF's relocalization. Furthermore, BAF punctum formation requires dsDNA, providing the first evidence that BAF's behavior in live cells is consistent with BAF's nucleic acid binding specificity characterized in vitro.

Fig. 3.

BAF relocalizes to dsDNA but not ssDNA or dsRNA. CV1-CAT (A, C, E, and G) or CV1-3×Flag-BAF cells (B, D, F, and H) were transfected with identical amounts (1 μg/ml) of purified vaccinia virus (VacV) DNA (A and B), E. coli K-12 DNA (C and D), M13 single-stranded DNA (E and F), or poly(I-C) double-stranded RNA (G and H). Cells were fixed 7 h later. CV1-CAT cells were permeabilized first with 0.1% Triton X-100–PBS and then with 0.5% SDS–PBS. CV1-3×Flag-BAF cells were permeabilized only by using 0.1% saponin–PBS. Cells were processed for immunofluorescence imaging using an anti-BAF (A, C, E, and G) or M2 anti-Flag (B, D, F, and H) primary antibody, Alexa Fluor secondary antibody, and DAPI. The representative images shown were taken by using a confocal microscope at a ×60 magnification. Arrowheads mark puncta containing endogenous BAF. Scale bars, 10 μm.

Emerin colocalizes with cytoplasmic BAF puncta during transfection or infection in a BAF-dependent manner.

Next, we wanted to investigate whether other proteins are recruited by BAF to cytoplasmic DNA, which may then lead to the formation of a more stable nucleoprotein complex. As BAF is capable of recruiting LEM domain proteins such as emerin or Lap2α upon binding to nuclear DNA late in mitosis or during retroviral infection (6, 11, 38), we selected these two proteins for our studies. As was previously reported for other cell lines, we found that in untreated CV1-CAT cells, emerin is found at the nuclear rim and in the endoplasmic reticulum (ER) (Fig. 4A), while Lap2α exhibits diffuse staining in the nucleus and more weak staining in the cytoplasm (Fig. 4B). However, 7 h after dsDNA transfection, both proteins relocalize to cytoplasmic depots as well (Fig. 4C and D), with emerin adopting ring-shaped structures. To determine whether the relocalization of these proteins was affected by BAF, we performed parallel immunofluorescence studies with CV1-shBAF, CV1-CAT, and CV1-3×Flag-BAF cells. In mock-transfected cells, no difference in the localization of either emerin or Lap2α was seen between cell lines (data not shown). Interestingly, in dsDNA-transfected CV1-3×Flag-BAF cells, both emerin and Lap2α colocalized very well with 3×Flag-BAF (compare Fig. 4G and I, or H and J). A comparison of the Lap2α relocalizations in the 3 cells lines revealed little difference in the numbers of puncta (compare Fig. 4D, F, and H), suggesting that while Lap2α localizes to the same sites as BAF, it does so in a BAF-independent manner. In contrast, emerin “rings” were more prevalent in cells overexpressing BAF and rare in the BAF-depleted cells (compare Fig. 4C, E, and G). This indicates that emerin is likely recruited to cytoplasmic DNA by BAF, as it is to cellular DNA during telophase (11, 34).

Fig. 4.

Emerin and Lap2α colocalize with cytoplasmic BAF puncta during transfection. CV1-shBAF (E and F), CV1-CAT (A to D), or CV1-3×Flag-BAF (G to J) cells were mock transfected with Lipofectamine 2000 alone or with 1 μg pUC19 DNA per ml of medium as indicated. Cells were fixed 7 h later. Cells were permeabilized by using 0.1% saponin–PBS. Cells were processed for immunofluorescence imaging using an M2 anti-Flag (I and J), anti-emerin (A, C, E, and G), or anti-Lap2α (B, D, F, and H) primary antibody, Alexa Fluor secondary antibody, and DAPI. The representative images shown were taken by using a confocal microscope at a ×60 magnification. Scale bars, 10 μm.

Next, we tested whether these LEM domain proteins are also recruited to ts2 DNA replication sites, utilizing a variation of the temperature shift protocol initially used to observe BAF at these sites (43). For these studies, we first infected CV1-CAT cells with wild-type or ts2 virus at 32°C and allowed the infection to proceed for 9 h in order to permit DNA replication factories to efficiently form. At 4 hpi, BrdU was added to the infected cultures to label replicating DNA and allow for a later detection of viral factories using a BrdU-specific antibody. Cultures were then shifted to a nonpermissive temperature (40°C) for 7 h prior to fixation and downstream processing for immunofluorescent imaging of BrdU, emerin, and Lap2α localization. In contrast to our data for transfected DNA, we observed no clear relocalization of Lap2α to BrdU-labeled replication sites during either wild-type or ts2 virus infection (data not shown). Interestingly, when we then examined emerin's localization, we found that it surrounded ts2 replication factories (Fig. 5F, G, and H). However, emerin did not form these distinctive rings around wild-type replication factories (Fig. 5C, D, and E). As BAF is also known to relocalize to ts2 replication sites but not wild-type sites, these data further support a model in which the LEM domain protein emerin can be recruited by BAF to cytoplasmic nucleoprotein complexes.

Fig. 5.

Emerin localizes with DNA replication sites during ts2 infection. CV1-CAT cells were left untreated or infected with wild-type or ts2 virus at 32°C. At 4 hpi, cells were treated with 25 μg/ml BrdU to label replicating DNA, and at 9 hpi, cells were shifted to 40°C. Cells were fixed 7 h later (16 hpi). Cells were permeabilized by using 0.1% saponin–PBS. Cells were processed for immunofluorescence imaging using M2 anti-emerin and anti-BrdU primary antibodies, Alexa Fluor secondary antibody, and DAPI. The representative independent and overlaid images shown were taken by using a confocal microscope at a ×60 magnification. Scale bars, 10 μm.

Characterization of the DNA binding and dimerization abilities of BAF mutant proteins.

Next, we sought to examine the relative importance of DNA binding, homodimerization, or LEM domain protein interactions for BAF's relocalization to foreign DNA and antipoxviral functions. It was our goal to utilize BAF mutants specifically disrupting each of these three activities, drawing on previous biochemical and structural studies of BAF's various interactions, which revealed critical residues mediating these three events (4, 5, 12, 31). The three BAF mutants that we have selected for these studies replace lysine 6 with alanine (K6A), glycine 47 with glutamic acid (G47E), and lysine 53 with glutamic acid (K53E) (Fig. 6A). The K6A mutation is in helix 1 and disrupts a lysine mediating multiple contacts with the phosphate backbone of DNA, sharply reducing BAF's affinity for DNA (12, 31). The G47E mutation is at the center of dimerization helix 3 and was shown previously to increase the rate at which BAF homodimers exchange monomers (31). K53E is also in helix 3 and was shown to abolish binding to the LEM domain protein emerin while leaving dimerization largely unaffected (31).

Upon the introduction of these three mutations into 3×Flag-BAF, we examined the ability of each mutant protein to homodimerize or bind DNA in vitro. This allowed us to determine whether the BAF mutants behaved similarly when expressed in cells compared to observations made previously using purified proteins (12, 31). To examine dimerization, we transiently coexpressed 3×Flag-BAF variants with GFP or GFP-BAF (wild-type BAF fused at its N terminus to GFP). Blots of lysates from these transfections verified that the various Flag- or GFP-tagged proteins were expressed (Fig. 6B, bottom). We then immunoprecipitated the Flag-tagged proteins using M2 anti-Flag agarose and performed immunoblot analysis with a GFP-specific antibody to determine whether GFP-BAF was coprecipitated with the 3×Flag-BAF variants. We found that 3×Flag-wt BAF coprecipitated with GFP-BAF (Fig. 6B, lane 2) but not GFP alone (lane 1), verifying the BAF homomeric interaction. Both the BAF-K6A and -K53E mutants also retrieved GFP-wt BAF, demonstrating that these mutations do not impede dimerization (lanes 3 and 5). However, no GFP-BAF was coprecipitated with the BAF-G47E protein (lane 4), consistent with the previous finding that this mutation blocks dimerization.

We next examined the ability of each of these proteins to bind to DNA. Cytoplasmic lysates from transiently expressing cells were incubated with DNA cellulose, and the bound proteins were analyzed by immunoblotting and quantified (Fig. 6C). While both wild-type BAF and the BAF-K53E protein bound efficiently to the DNA, only ∼10% of the BAF-K6A and -G47E proteins loaded was retained on the DNA cellulose. This demonstrates that both the DNA binding face of BAF and its ability to bind DNA as a dimer strongly impact its affinity for DNA.

Punctum formation requires BAF's DNA binding and dimerization properties.

Next, we examined the impact of each of these mutations on BAF's ability to relocalize to transfected DNA. To perform this study, we first developed stable cell lines expressing each of the 3×Flag-BAF variants and verified that the mutant proteins were being expressed. As shown in the immunoblot in Fig. 7A (left), the BAF-K6A and -G47E proteins were expressed at levels similar to those of the wild-type protein. However, the levels of the BAF-K53E protein were less than 10% of those of wild-type BAF. We considered that this may be an artifact of the clone that we selected for expansion; however, examination of multiple clones in CV1 as well as in HEK293 cells showed a consistent diminution of BAF-K53E protein levels (data not shown), which was taken into consideration during the interpretation of the results discussed below. The doubling rate of each of these cell lines was similar to that of the control cells (Fig. 7A, right), indicating that the overexpression of the mutants did not grossly impact cell growth.

Using these cell lines, we then compared the intracellular localizations and abilities of each of these BAF proteins to relocalize during plasmid transfection. We also scored the numbers of puncta per cell to more precisely assess the differences observed between mutants. In untransfected cells (data not shown), we observed that the overall localization of the BAF-K6A and -K53E proteins was evenly distributed between the cytoplasm and nucleus, mirroring that of the wild-type BAF protein. Interestingly, the BAF-G47E protein was more concentrated in the nucleus in untransfected cells but could still be detected in the cytoplasm as well. In regard to punctum formation, shown in the micrographs in Fig. 7B (top), BAF relocalization during plasmid transfection was greatest with the wild-type and BAF-K53E proteins and sharply reduced with the other two mutants. Specifically, puncta could be observed in more than 90% of the cells expressing wild-type BAF, with almost 40% of cells containing 5 or more puncta (Fig. 7B, bottom). Ninety percent of the BAF-K53E-expressing cells also contained puncta, although these were weaker in intensity than the puncta found in cells expressing wild-type BAF. In contrast, BAF relocalization was much less common in cells expressing the BAF-K6A and -G47E proteins, where 35% and 44% of the cells, respectively, had no puncta at all. When puncta were present, the majority of cells had only 1 to 2 puncta, further indicating that the ability of these proteins to relocalize to foreign DNA was reduced. Taking the data in Fig. 6 and 7 together, BAF's capacity to bind and dimerize on DNA closely parallels its ability to relocalize during transfection, thus strongly indicating that these properties underlie the molecular mechanism of BAF's response to foreign DNA. It also supports the model that BAF is binding directly to DNA rather than being recruited by another cytoplasmic DNA binding factor.

BAF's DNA binding and dimerization properties are critical for its antipoxviral activity.

We next examined the ability of each of these mutant proteins to act as a host defense against vaccinia virus replication using two complementary approaches. First, we compared the growths of the ts2 virus in each of the CV1 stable cell lines by quantifying viral yields at 48 hpi at a low MOI (0.01). This was done to permit multiple rounds of viral replication to occur, allowing us to detect even subtle differences in ts2 growth rates between the cell lines. Absolute viral titers from experiments done at both 37°C and 40°C are shown in Fig. 8A. In this experiment, the overexpression of the wild-type protein led to 160- and 75-fold decreases in viral yield compared to that of the control CAT cells at 37°C and 40°C, respectively. However, the BAF-K6A and BAF-G47E proteins were far less effective than wild-type BAF at inhibiting ts2 growth, as cells overexpressing those proteins exhibited at most a 3-fold decrease in viral yield compared to the control cells, even after multiple rounds of viral replication. The inhibition of ts2 growth by the BAF-K53E protein was also much less than that by wild-type BAF, differing from the control cells by only 4- to 5-fold at either temperature. However, this moderate inhibition of vaccinia virus replication may indicate that the BAF-K53E mutant retains BAF's antiviral activity, considering the fact that the BAF-K53E protein is expressed at only 10% of the levels of the other BAF proteins.

Fig. 8.

Impact of BAF mutants and emerin depletion on ts2 viral yield. Equal numbers of cells of the indicated CV1 stable cell lines (A) or shRNA-expressing lines (B) were infected with ts2 virus (MOI = 0.01) and incubated at 37°C or 40°C. Cells were harvested at 48 hpi, and viral yield was assessed. Bars represent average yields from at least three independent experiments, titrated in independent duplicates. Error bars represent standard deviations. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Next, we sought to investigate whether emerin, shown earlier to be recruited to ts2 replication sites, has an impact on the ts2 viral yield. To address this, we stably expressed an shRNA targeting emerin in CV1-CAT cells. As shown in the immunoblot in Fig. 8B (top), these cells exhibit undetectable levels of the emerin protein in comparison to the control cells. We then assayed for ts2 viral production following a 48-h infection in control and emerin-depleted cell lines at both 37°C and 40°C. This assay (Fig. 8B, bottom) demonstrated that the emerin-depleted cells exhibited no significant increase in ts2 yields in comparison to CV1-shScram controls at either temperature. Together with our above-described data showing that the K53E mutation does not impinge on BAF's host defense activity, these data provide further evidence that while emerin is recruited to viral replication sites (Fig. 5), its presence is not critical for BAF's inhibition of the poxvirus life cycle.

Second, we asked whether recombinant viruses could survive if the virus itself expressed each of the BAF mutant proteins. The basis for this approach was originally described in our previous report (43), wherein a recombinant virus expressing wild-type BAF was viable, most likely because it could be phosphorylated by the B1 kinase. However, a recombinant virus expressing a BAF mutant in which the three phosphorylation sites had been replaced with alanine residues (BAF-MAAAQ) could not be isolated. These results demonstrated that the unphosphorylatable BAF-MAAAQ protein is a dominant active form of the protein and is refractory to inactivation by B1. In the present study, we next sought to determine if altering BAF's other required functions could inactivate BAF-MAAAQ and allow it to be expressed in a recombinant virus.

Our approach was to perform recombination-mediated insertion using plasmid pJS4-RFP. This plasmid contains 2 divergent vaccinia virus promoters, one of which drives the expression of RFP (red fluorescent protein) and the second of which expresses 3×Flag-BAF-MAAAQ, -BAF-MAAAQ/K6A, -BAF-MAAAQ/G47E, or -BAF-MAAAQ/K53E. Sequences derived from each half of the vaccinia virus thymidine kinase (TK) locus flank this cassette and enable its stable insertion into the viral genome via homologous recombination. This recombinational insertion leads to the inactivation of the TK locus, which in turn renders the virus resistant to bromodeoxyuridine (BrdU). To this end, cells were infected with wild-type virus and transfected with a linearized plasmid, the expression of which was verified by RFP fluorescence in the initial transfection (data not shown). The expression of the BAF variants 18 h after infection/transfection was also verified by immunoblot analysis (Fig. 9A), demonstrating that all of the proteins were expressed. Next, the virus harvested from these infections/transfections was passaged once in the presence of BrdU to select for TK⁻ viruses and then titrated on BSC40 cells to determine total viral yields. Confluent monolayers of BSC40 cells were then infected with approximately 2,000 PFU of each virus and examined for the presence of red plaques by fluorescence microscopy, which indicated that a recombinant virus had been produced. Recombinant viruses expressing RFP alone or in addition to the BAF MAAAQ/K6A and MAAAQ/G47E mutants were readily observed in ∼30% of the plaques examined and could be expanded into larger viral stocks following plaque purification (Fig. 9B). Upon expansion, the fitnesses of the viruses expressing RFP alone or RFP with BAF-MAAAQ/K6A or -MAAAQ/G47E were compared in a 48-h growth assay. As shown in Fig. 9C, although these viruses expressed the BAF protein expected, each virus grew identically to controls, indicating these mutants have no detectable impact on the viral life cycle. To confirm that we could not isolate either BAF-MAAAQ or -MAAAQ/K53E, we performed six independent attempts using three batches of purified plasmid. Although parallel infections/transfections using a vector expressing RFP alone were performed as a positive control during each of these attempts and resulted in recombinant virus in each case, no red plaques were observed for infections/transfections using BAF-MAAAQ or -MAAAQ/K53E. Together, these data demonstrate that mutations affecting DNA binding or dimerization render BAF inactive as an antipoxviral protein. However, the introduction of the K53E mutation, shown previously to inhibit BAF's binding to LEM domain proteins, does not impair BAF's host defense capability sufficiently to allow its expression from the viral genome.

Fig. 9.

Impact of mutant BAF proteins expressed from recombinant vaccinia virus. (A) Cells infected with wild-type vaccinia virus were transfected with plasmid pJS4-RFP encoding the unphosphorylatable BAF mutant protein. Cells were harvested at 18 hpi, and lysates from equal numbers of cells were subjected to immunoblot analysis to confirm the expression of each BAF protein. (B) Linearized plasmid pJS4-RFP encoding each of the unphosphorylatable BAF mutant proteins was introduced into BSC40 cells previously infected with wild-type vaccinia virus (MOI = 0.03), and the recombinational insertion of the transcriptional cassette into the viral TK locus was allowed to proceed during 2 days of incubation (43; this study). One passage was performed in human TK⁻ cells with or without BrdU to select for TK-deficient virus. The viral yield was determined, and 2,000 plaques were examined for RFP expression by immunofluorescence. (C) BSC40 cells were infected with virus expressing RFP alone or with the BAF-MAAAQ/K6A or -MAAAQ/G47E mutant (MOI = 0.01) and placed at 37°C. Cells were harvested at 48 hpi, and lysates were used to verify the expression of the BAF mutants by immunoblot analysis (top) or to quantify the viral yield (bottom). Bars shown represent average yields in two independent experiments, each titrated in duplicates. Error bars represent standard errors of means.

DISCUSSION

Poxviruses such as vaccinia virus encode numerous homologs of cellular proteins as part of their strategy to evade or inactivate the host immune system (1, 13, 21, 28, 30). An understanding of the virus-host interactions occurring during vaccinia virus infection therefore provides insights into how critical aspects of the host immune system can be targeted by a human pathogen. It was recently discovered that the B1 kinase, which is highly homologous to the cellular VRK proteins (3, 25), should be counted among the subversion measures of vaccinia virus against antiviral responses. The studies supporting this conclusion began with the discovery that, like VRKs, B1 can phosphorylate the N terminus of BAF both in vitro and in cultured cells. It was next observed that during infection with the B1-deficient ts2 virus, but not the wild-type virus expressing active B1, BAF relocalizes to viral DNA factories. The fact that phosphorylation inhibits BAF's ability to bind DNA and LEM domain proteins in vitro indicated that wild-type vaccinia virus blocks BAF relocalization by interfering with one or more of its molecular interactions. These data, combined with the fact that the depletion of BAF could rescue the DNA replication defect of the ts2 virus, revealed for the first time that BAF has the capacity to act as an antipoxviral host defense protein (43).

Based on these observations, we wanted to better understand the molecular mechanism of BAF's response to vaccinia virus infection and set out with three main goals for the present study. The first goal was to define the molecular determinants needed for BAF's relocalization in the cytoplasm. Although BAF is known to bind DNA with a high affinity, poxviral replication factories are comprised of DNA, RNA, and viral and cellular proteins (14, 27), any of which may play a role in BAF's relocalization. Indeed, the viral matrix protein appears to participate in BAF recruitment to retroviral preintegration complexes (PICs) during HIV infection (18). Therefore, we examined whether DNA alone was sufficient to cause BAF's relocalization. Upon the examination of either endogenous BAF or epitope-tagged BAF, we found that BAF relocalized during the transfection of multiple types of double-stranded DNA obtained from independent sources, including plasmid and purified vaccinia virus DNA. (Importantly, we found that stable cell lines were required for these studies. Punctate cytoplasmic staining was often observed with transiently expressed BAF, presumably due to the relocalization of the BAF protein to the expression plasmid remaining in the cytoplasm [M. S. Wiebe, unpublished data].) It is intriguing that the puncta formed by 3×Flag-BAF appear larger than those formed by endogenous BAF in our microscopy studies. It is unclear whether we saw a bona fide difference in complex size or a difference in the sensitivities of the two antibodies employed. In either case, it is clear that both tagged BAF and untagged BAF possess the ability to relocalize in response to transfection. From these data, we conclude that while viral DNA replication factories contain numerous components, DNA alone is sufficient to trigger BAF's relocalization. Interestingly, the relocalization of BAF was not observed during transfection with other nucleic acids, including single-stranded DNA or poly(I-C), which precisely correlates with BAF's nucleic acid binding specificity observed using purified protein in vitro (12, 45). Together, this provides the first evidence that cytoplasmic DNA is sufficient for causing the relocalization of BAF and strongly suggests that BAF's recruitment to dsDNA is mediated through direct interactions between BAF and DNA.

The second goal was to examine whether known BAF-interacting proteins also join it at sites of cytoplasmic DNA and might therefore cooperate with BAF in its antiviral function. For these studies, we have focused on two LEM domain proteins, Lap2α and emerin, both of which are recruited by BAF to host DNA during postmitotic nuclear reassembly. Interestingly, we discovered that both proteins colocalize with BAF at cytoplasmic depots during transfection. The relocalization of emerin appears to be BAF dependent, while Lap2α is independent of BAF. One possible explanation for this difference between these two LEM domain proteins is that Lap2α was observed previously to bind DNA on its own (38), while no DNA binding function has been ascribed to emerin. In contrast to these transfection studies, during viral infection, Lap2α exhibited no clear localization to poxviral DNA replication sites, indicating that this protein is in some way specific for transfected DNA. However, during ts2 virus infection, we clearly observed emerin staining in a ring surrounding the DNA replication sites, while emerin was not found at viral factories during wild-type infection in this study and previous studies (39). These data suggest that an active B1 kinase helps ensure that nuclear envelope proteins are not recruited to cytoplasmic DNA introduced by poxvirus infection as these proteins would be during mitotic telophase.

The third goal was to determine which properties of BAF are required for its relocalization to DNA. For these studies, we employed three BAF mutants containing single-amino-acid changes, the properties of which have been characterized previously in multiple studies using recombinant proteins. Specifically, the K6A mutation lies on the DNA binding face of BAF and interferes directly with protein-DNA interactions while leaving dimerization unaffected (5, 12, 31). The BAF-G47E mutant was chosen because it targets a residue within the dimerization helix that leads to higher rates of monomer exchange (31). Finally, K53E was selected because it was the only mutation identified to date that impeded LEM domain binding without interfering with DNA binding or the dimerization of BAF (31). In light of the fact that each of these mutant properties had been tested using protein expressed in bacteria or through in vitro translation, we first examined how they behaved when expressed in mammalian cells. In our dimerization assay, in which we tested the ability of each BAF mutant to coimmunoprecipitate GFP-BAF, only the BAF-G47E protein lacked this ability, confirming that this mutation affects dimerization in vivo as well as in vitro. When we next assayed the efficiency of binding to DNA cellulose, the amount of the BAF-K53E protein bound was ∼80% of that of wild-type BAF, which correlates with data from previous studies showing that the K53E mutation has little effect on the DNA binding of BAF (31). In contrast, the binding efficiencies of both the BAF-K6A and -G47E proteins were only about 10% of that of wild-type BAF. While this result was expected for BAF-K6A, which disrupts a critical protein-DNA contact (4), the result for the BAF-G47E protein was somewhat surprising, as it was previously reported to bind DNA cellulose comparably to wild-type BAF (31). One experimental difference that may contribute to this discrepancy is the source of the protein; in that prior study, BAF was expressed using in vitro translation. It is possible that upon expression in cells, the BAF-G47E protein is posttranslationally modified or complexed with other proteins in a manner which impairs DNA binding. Another model can be suggested based on a previous study by Skoko et al. (37). In those experiments, the inactivation of one BAF monomer via phosphorylation decreased BAF's DNA binding affinity by several orders of magnitude, indicating that BAF monomers synergistically cooperate in DNA binding upon dimerization. As our BAF-G47E protein cannot dimerize, such synergism would be lost, and this is therefore also a strong possible reason for the decreased DNA binding that we observed with this mutant.

Upon the completion of these molecular analyses of the BAF mutant proteins, we next examined their abilities to relocalize to puncta upon dsDNA transfection. To perform these studies, we needed to construct cell lines stably expressing each of these proteins. We selected the Flp-In CV1 cell system (Invitrogen) to derive these stable lines because of the distinct advantage that it has in comparing variants of the same protein. Specifically, through the action of the Flp recombinase and a previously incorporated FRT recombination site, a recombination of the transgene occurs at the same locus in each cell stably integrated. This ensures that a similar level of gene transcription occurs with each cell line produced, allowing a more direct comparison of protein function and regulation between mutants. Using this system, we derived cells expressing very similar levels of wild-type BAF, BAF-K6A, and BAF-G47E, as determined by immunoblot analysis. However, the BAF-K53E protein was present at much lower levels in stable cells. This reduction in protein levels for the BAF-K53E protein was highly reproducible, as it was observed in 4 independently derived CV1 cell lines and was also found when the protein was stably expressed in HEK293 Flp-In cells (data not shown). While this suggests that the K53E mutation renders BAF less stable, the protein remained detectable in our immunofluorescence assays and therefore could still be used in our protein relocalization studies.

Our subsequent comparison of the formation of puncta between wild-type BAF and the three mutant proteins revealed that the BAF-K6A and -G47E proteins were far less effective at forming puncta than the other two proteins. This correlation between the ability of the BAF mutants to bind DNA cellulose and their capacity to relocalize to transfected dsDNA strengthens our support for a model in which BAF is directly binding the cytoplasmic DNA during its relocalization. Importantly, these data also indicate that the binding of DNA by BAF is an essential step in punctum formation. The fact that somewhat fewer puncta were seen with the BAF-K53E protein may indicate that the recruitment of other proteins, such as LEM domain proteins, plays a nonessential, but still substantive, role in stabilizing BAF-DNA cytoplasmic complexes. Indeed, the LEM domain protein Lap2α was found previously to enhance the salt stability of BAF-DNA complexes in vitro (38), supporting this possibility. However, the reduced stability of the BAF-K53E protein could also lead to impaired punctum formation in our model system. Further studies of the role of LEM domain proteins and their involvement in BAF's response to foreign DNA should help distinguish between these possibilities.

Finally, our fourth goal was to determine which properties of BAF are required for its inhibition of vaccinia virus replication. To achieve this goal, we first compared the impacts of BAF protein overexpression on ts2 replication by quantifying the viral yield obtained from infections of each of our stable cell lines. In regard to the BAF-K53E protein, in spite of the fact that it is expressed at only 10% of wild-type levels, this protein was still able to inhibit the ts2 virus yield 10-fold at 40°C. Thus, while the aberrant expression of BAF-K53E makes its impact more difficult to evaluate, these data suggest to us that LEM domain binding is not a critical step in BAF's repression of vaccinia virus replication. Our data showing that the depletion of emerin has no effect on the viral yield further support this model. In contrast, the BAF-K6A or -G47E protein had little, if any, impact on viral growth, despite the fact that these proteins were expressed at levels equivalent to those of wild-type 3×Flag-BAF. This demonstrates that DNA binding and dimerization are necessary for BAF's antipoxviral activity.

To complement these studies, we also attempted to produce recombinant wild-type vaccinia viruses expressing BAF mutant proteins from the viral genome. This approach using the wild-type virus required that the mutations be introduced into the BAF-MAAAQ variant, which is unphosphorylatable by B1. This BAF protein is a “dominant active” BAF and cannot be expressed from the viral genome, indicating that it is lethal for the virus (43). Predicated on this background, we sought to investigate whether the K6A, G47E, or K53E mutation could suppress the host defense activity of BAF-MAAAQ sufficiently to allow it to be expressed by a recombinant virus. We found that while virus expressing the BAF-MAAAQ/K53E mutant could not be made, both the K6A and G47E mutations now rendered BAF-MAAAQ inactive, thus allowing the isolation of the virus. For the production of BAF-MAAAQ/K53E, we also attempted to isolate the virus from cells expressing BAF-specific shRNA, which should have significantly reduced the levels of the mutant protein. However, we could not generate this virus even in the presence of BAF shRNA (data not shown), providing evidence that BAF-MAAAQ/K53E retains its antiviral potency even when it is likely depleted. Taken together, these data indicate that although BAF's ability to recruit LEM domain proteins might contribute to its function as a component of retroviral preintegration complexes (38), the binding of LEM domains is likely to be less important for BAF's impact on other viruses such as vaccinia virus. Additionally, these data confirm that BAF's DNA binding and dimerization properties are the most essential components of BAF's host defense activity, as suggested by our studies using stable cell lines.

In summary, the observation that vaccinia virus employs the B1 kinase to subvert BAF unveiled a novel function for this DNA binding protein as an intrinsic defense against poxviruses. Indeed, as shown in Fig. 1, cells depleted of BAF supported over 1,000-fold-higher levels of ts2 growth after 48 h than cells expressing high levels of BAF. We have now extended these studies by dissecting the molecular mechanism of BAF's response to vaccinia virus infection and foreign DNA in general. We have discovered that BAF's presence in the cytosol and its ability to bind and cross bridge DNA with a high affinity are needed for its antipoxviral activity. The fact that these properties are also needed for BAF's rapid response to transfected DNA is intriguing, as it suggests that BAF may have an impact on other dsDNA pathogens as well. A role for DNA cross bridging also suggests that BAF utilizes DNA compaction and/or aggregation as it inhibits the replication of foreign DNA. Such a strategy is reminiscent of chromatin condensation, a known intrinsic barrier to the expression of foreign DNA in the nucleus (22, 36). It is therefore tempting to speculate that the existence of BAF in the cytosol is an effort by the cell to extend the use of DNA condensation as an intrinsic barrier to foreign DNA beyond the nucleus.