The DNase of Gammaherpesviruses Impairs Recognition by Virus-Specific CD8⁺ T Cells through an Additional Host Shutoff Function

Abstract

The DNase/alkaline exonuclease (AE) genes are well conserved in all herpesvirus families, but recent studies have shown that the AE proteins of gammaherpesviruses such as Epstein-Barr virus (EBV) and Kaposi's sarcoma-associated herpesvirus (KSHV) exhibit an additional function which shuts down host protein synthesis. One correlate of this additional shutoff function is that levels of cell surface HLA molecules are downregulated, raising the possibility that shutoff/AE genes of gammaherpesviruses might contribute to viral immune evasion. In this study, we show that both BGLF5 (EBV) and SOX (KSHV) shutoff/AE proteins do indeed impair the ability of virus-specific CD8⁺ T-cell clones to recognize endogenous antigen via HLA class I. Random mutagenesis of the BGLF5 gene enabled us to genetically separate the shutoff and AE functions and to demonstrate that the shutoff function was the critical factor determining whether BGLF5 mutants can impair T-cell recognition. These data provide further evidence that EBV has multiple mechanisms to modulate HLA class I-restricted T-cell responses, thus enabling the virus to replicate and persist in the immune-competent host.

Epstein-Barr virus (EBV) is a potent growth-transforming agent of B lymphocytes and a causative agent of various malignant diseases of lymphoid or epithelial cell origins. This pathogen nevertheless persists as a lifelong and largely asymptomatic infection of B lymphocytes in more than 90% of adults worldwide (29). In healthy infected individuals, EBV is confronted by potent cellular immune responses that limit but do not completely eradicate the virus. The virus-host equilibrium is achieved by EBV colonizing the B lymphoid system and establishing latent infections in long-lived memory B cells (1, 34). In this latent state the virus does not induce B-cell proliferation, but neither does it express the virally encoded antigens that are immunodominant targets for CD4 and CD8 T cells (24, 36).

When latently infected cells are reactivated into the lytic cycle to produce infectious progeny, a large number of antigens are expressed and targeted by cellular immune responses (13). Since the lytic cycle can operate for several days before the cells die (26), the ability to produce infectious progeny is potentially compromised by the immune responses. However, in common with other herpesviruses, EBV has evolved mechanisms to enhance the likelihood that the lytic cycle proceeds to completion. For example, induction of the lytic cycle is accompanied by a reduced expression of both HLA class I and class II molecules at the cell surface (11, 16, 26). Furthermore, while the protein levels of peptide transporters associated with antigen presentation (TAP-1 and TAP-2) are unaffected, their peptide-transporting function is significantly impaired during the lytic cycle (26). The small protein product of the BNLF2a early gene was recently shown to inhibit peptide transport function via binding to TAP complexes and to reduce the surface expression of HLA class I molecules, presumably by limiting the availability in the endoplasmic reticulum of peptides to form stable ternary β2-microglobulin/HLA/peptide complexes (12). While the BNLF2a protein specifically affects the HLA class I antigen-presenting pathway and does not affect expression of HLA class II, we recently reported a host protein synthesis shutoff in the lytic cycle which potentially acts as a more general immune evasion strategy by blocking HLA class I and class II synthesis and responsiveness to interferons (31). It has yet to be demonstrated, however, that the host shutoff does actually affect the ability of the infected cell to be recognized by T-cell responses.

The host shutoff function was mapped to the BGLF5 gene (31), whose protein product was originally identified as a DNase/alkaline exonuclease (AE) with well-conserved sequence homology with the AE of all herpesviruses studied (17, 31, 33). Although BGLF5 shows no evidence of RNase enzyme activity, host shutoff is achieved by a global increase of mRNA turnover (31), suggesting that the AE and host shutoff functions are distinct features of a bifunctional protein. Such bifunctionality was previously reported for the homologous ORF37 product SOX (shutoff and exonuclease) of Kaposi's sarcoma-associated virus (KSHV) (8). This AE/host shutoff bifunctionality appears to be unique to gammaherpesviruses, since betaherpesviruses do not shut off host protein synthesis and host shutoff by alphaherpesviruses is mediated by a separate vhs (virion host shutoff) protein, e.g., the herpes simplex virus type 1 (HSV-1) UL41 product (15, 31).

In this study, we primarily addressed the question whether the downregulation of HLA class I expression by BGLF5 was functionally significant with regards to recognition by CD8⁺ T cells specific for EBV lytic cycle antigens. We present evidence showing that BGLF5 does indeed impair recognition by CD8⁺ T cells. Furthermore, by performing random mutagenesis of BGLF5, we were able to genetically separate the shutoff and AE functions and to demonstrate that the impairment of T-cell recognition mapped to the shutoff function.

MATERIALS AND METHODS

Plasmids and PCR.

The EBV BGLF5 gene was PCR amplified from B95.8 EBV DNA and cloned into the EcoRI/NotI sites of pCDNA3-IRES-nls-GFP with an additional 3′ hemagglutinin (HA) tag. Similar constructs containing the KSHV-SOX (ORF37) gene or the HSV-1 AE (UL12) gene, both with an additional 3′ HA tag, were kindly provided by Britt Glaunsinger (University of California, Berkeley) and Don Ganem (University of California, San Francisco). The wild-type BGLF5, KSHV-SOX, and HSV-1 AE vectors have been described previously (31). Random mutagenesis of BGLF5 was performed by PCR with the Genemorph II random mutagenesis kit (Stratagene) according to the manufacturer's protocol using 5 μg of pCDNA3-BGLF5 template and 30 PCR cycles to generate a pool of random mutants with a mutation frequency of one to three mutations per kilobase. The mutants were cloned into the EcoRI/NotI sites of pCDNA3-IRES-nls-GFP and screened for shutoff and AE functions. Mutants with selective effects on these functions were subsequently 5′-HA tagged using PCR methods and cloned into the EcoRI/NotI sites of pCDNA3 to generate, for example, pCDNA3-HA.BGLF5-mut.13 and pCDNA3-HA.BGLF5-mut.58. All plasmids were verified by restriction digest and sequence analyses. Plasmid p509, containing the EBV BZLF1 gene (30), and plasmid pCEP4-SM, containing the EBV BSLF2/BMLF1 spliced gene (4), were kindly provided by Paul Farrell, London, England.

Cells and transfections.

The 293 epithelial cell line (American Type Culture Collection) was maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and penicillin-streptomycin antibiotics. The MJS (Mel JuSol) melanoma-derived cell line (14) was maintained in RPMI 1640 (Gibco BRL) medium supplemented with 10% fetal bovine serum and penicillin-streptomycin antibiotics. For some experiments, 293 and MJS cells were transiently transfected with plasmid DNA using Lipofectamine 2000 (Invitrogen). The Akata cell line, derived from a Japanese patient with Burkitt's lymphoma and kindly provided by Kenzo Takada (Tokyo, Japan), is an EBV⁺ B-cell line displaying a latency I type of infection in which 20 to 50% of cells can be induced into the lytic cycle by ligation of the B-cell receptor (32); the AKBM derivative of Akata carries a reporter plasmid that expresses green fluorescent protein (GFP) when the cells enter the lytic cycle (26). The lytic cycle was induced using rabbit antibodies to human immunoglobulin G (IgG) as described elsewhere (21), and aliquots of cells were taken for Western blot analysis at various time points over 48 h. EBV-specific CD8⁺ cytotoxic T cells were grown in 10% fetal calf serum in RPMI 1640 medium supplemented with 30% supernatant from the interleukin-2-producing MLA 144 cell line (25) and 50 U/ml recombinant interleukin-2 as described elsewhere (22).

Antibodies.

Rabbit sera raised against the EBV BGLF5-encoded DNase has been described previously (7). A rat monoclonal antibody (MAb), OT15Q, reactive with the p18 viral capsid antigen (VCA) product of the EBV BFRF3 gene (38), was generated by one of the authors (J. M. Middeldorp). A rabbit serum to the EBV BSLF2/BMLF1-encoded SM, or EB2, protein (3) was kindly provided by Alain Sergeant, Lyon, France. Goat antibodies to calregulin (sc6467) were purchased from Santa Cruz Biotechnology. The BZ.1 murine MAb specific for the EBV BZLF1-encoded protein was generated in the authors’ laboratory (40). Murine MAbs used to detect human cellular HLA class I were the following: W6/32 (2), which recognizes native β2-microglobulin-associated major histocompatibility complex class I (HLA-A, -B, and -C alleles) complexes; and HC10 (35), recognizing free HLA class I heavy chains. Rat MAb, 3F10, directed against the influenza virus-derived HA tag, was purchased from Roche Diagnostics.

Flow cytometry and immunofluorescence assay.

Cell surface expression of HLA class I on viable cells was determined by staining with phycoerythrin (PE)-labeled W6/32 antibodies (Serotec) and detection on a Beckman Coulter XL flow cytometer. The data were analyzed using Flowjo software (Tree Star).

Western blot assays.

Total cell lysates were denatured in reducing sample buffer (final concentrations: 2% sodium dodecyl sulfate [SDS], 72.5 mM Tris-HCl pH 6.8, 10% glycerol, 0.2 M sodium 2-mercaptoethanesulfonate, 0.002% bromophenol blue) and then sonicated and heated to 100°C for 5 min. Solubilized proteins equivalent to 2 × 10⁵ cells/20-μl sample were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) on NuPage mini-gels (4 to 12% bis-Tris mini-gels with morpholinepropanesulfonic acid buffer) from Invitrogen. Following electroblotting to polyvinylidene difluoride membranes and blocking with I-Block (Tropix; Applied Biosystems) in phosphate-buffered saline and 0.1% Tween 20 detergent, specific proteins were detected by incubating the membranes with primary antibodies. Rabbit anti-BGLF5 and anti-SM sera were incubated at a 1/6,000 dilution at 4°C overnight; the rat anti-HA MAb was incubated at 50 ng/ml at 4°C overnight; the mouse MAbs BZ.1 and HC10 and the goat antibody to calregulin were incubated at 1 μg/ml for 2 h at room temperature. Primary antibodies specifically bound to blotted proteins were detected by incubation for 30 min with appropriate alkaline phosphatase-conjugated secondary antibodies as follows: 1/10,000 dilution of anti-rabbit Ig (catalog number 170-6518; Bio-Rad); 1/10,000 anti-mouse Ig (catalog number 170-6520; Bio-Rad); 1/10,000 anti-rat Ig (catalog number A8438; Sigma); or 1/40,000 anti-goat Ig (catalog number A4062; Sigma). Bound secondary antibodies were developed using a CDP-Star detection kit (Tropix; Applied Biosystems) and then exposed to autoradiographic film.

T-cell function assays.

The two effector T-cell clones used in this study were generated as described elsewhere (22); clone GLC is specific for BMLF1 (SM protein) and restricted through HLA-A2, and clone RAK is specific for BZLF1 and restricted through HLA-B8. Targets for the GLC clone were generated by transfection of 293 cells (HLA-A2 type) with an SM expression plasmid with or without a BGLF5 expression plasmid, while targets for the RAK clone were generated by transfection of MJS cells (HLA-B8 type) with a BZLF1 expression plasmid with or without a BGLF5 expression plasmid. At 24 h posttransfection, the cells were used as targets in T-cell assays. Recognition of target cells by EBV-specific CD8⁺ effector T-cell clones was determined by enzyme-linked immunosorbent assay of gamma interferon (IFN-γ) release from activated T cells using a standard protocol described elsewhere (20). Briefly, 10⁴ effector T cells were incubated for 18 h at 37°C in V-bottom microtest plate wells with 10⁵ target cells, and then the supernatants were harvested for quantitation of IFN-γ by enzyme-linked immunosorbent assay (Endogen) in accordance with the manufacturer's recommended protocol. Specificity control targets included HLA-matched and HLA-mismatched EBV-transformed lymphoblastoid cell lines, empty vector-transfected 293 and MJS cells, empty vector-transfected 293 cells pulsed with the GLCTLVAML synthetic peptide, and empty vector-transfected MJS cells pulsed with the RAKFKQLL synthetic peptide.

DNase assays.

Proteins analyzed for DNase activity were in vitro transcribed and translated using the rabbit reticulocyte lysate system (Promega) according to the supplier's instructions. After the in vitro translation, 1 μl of the in vitro translation product was incubated with 200 ng of linearized pGEM5Zf(+) or pcDNA3 plasmid DNA in degradation assay buffer (50 mM MgCl₂, 50 mM Tris [pH 9.0], 100 μg/ml bovine serum albumin, 5 mM β-mercaptoethanol) at 37°C for 30 min, resolved on 0.8% agarose gels, and visualized by ethidium bromide staining. In some experiments, the in vitro translation products were first purified by immunoprecipitation with 12CA5 (mouse anti-HA tag) and protein G Dynalbeads (Invitrogen) before adding them to the degradation assay mixture at 37°C for 16 h. A 10-μl aliquot of in vitro translation product was also analyzed by Western blotting using 3F10 (rat anti-HA tag) antibody to verify equivalent protein expression.

RESULTS

BGLF5 is an early lytic cycle antigen that downregulates HLA class I expression.

Using the AKBM derivative of the Akata-BL line (26), synchronous induction of the lytic cycle was achieved by ligation of surface IgG, and samples were harvested at several time points over a 20-h period for analysis of EBV gene expression by Western blotting. In the experiment shown in Fig. 1A, almost 60% of cells were induced into the lytic cycle by 20 h postinduction. Blots probed with antibodies to the BZLF1 protein showed expression of this immediate-early antigen by 3 h, whereas blots probed with antibodies to the BFRF3 p18 VCA showed a delayed expression of the late antigen. BGLF5 protein was detected as early as BZLF1 but peaked later than BZLF1. These protein expression data are consistent with mRNA data reported by Yuan et al. (41), which identified BGLF5 as an early lytic cycle antigen.

FIG. 1.

BGLF5 early antigen downregulates HLA class I expression in 293 cells. (A) The kinetics of BGLF5 expression following synchronous induction of the lytic cycle in EBV-positive Akata cells was analyzed by Western blotting with rabbit polyclonal antiserum to BGLF5 (middle blot). Expression of the immediate-early BZLF1 protein (top blot) was detected with the BZ.1 mouse monoclonal antibody, and the late p18 VCA protein (bottom blot) was detected with rabbit antiserum to BFRF3. (B) 293 cells transfected with pCDNA3-BGLF5-IRES-nlsGFP or the control pCDNA3-IRES-nlsGFP vector were analyzed at 48 h after transfection for expression of HLA class I expression at the cell surface. Cells were stained with PE-conjugated W6/32 antibody and analyzed by flow cytometry. (C) 293 cells transfected with pCDNA3-BGLF5-IRES-nlsGFP or the control pCDNA3-IRES-nlsGFP vector were subjected to fluorescence-activated cell sorting at 24 h after transfection to isolate GFP⁺ cells. Total lysates from 2 × 10⁵ sorted cells were analyzed by SDS-PAGE and Western blotting using a rabbit polyclonal antiserum specific for BGLF5 or the murine HC10 MAb specific to HLA class I heavy chains.

We have previously shown that BGLF5 can downregulate HLA class I expression at the cell surface (31). Similar results are reproduced in Fig. 1B, where 293 cells were transfected with a bicistronic vector expressing BGLF5 and GFP or with a control plasmid expressing GFP only. The amount of BGLF5 protein expressed in transfected 293 cells was similar to the levels of endogenous BGLF5 expressed during the lytic cycle in EBV-positive B cells (31). At 48 h after transfection, surface HLA class I was analyzed by staining with PE-conjugated W6/32 antibody. The representative flow cytometry experiments shown in Fig. 1B and Fig. 2A illustrate the degree of downregulation by BGLF5 of HLA class I expression at the cell surface of the GFP⁺ transfected subpopulation; in replicate experiments, BGLF5 typically caused a 70 to 80% reduction in the mean fluorescence intensity of W6/32 staining. Western blot analysis of transfected populations, isolated by fluorescence-activated cell sorting for GFP⁺ cells, demonstrated that BGLF5 also reduced the total cellular levels of HLA class I heavy chains detected with the HC10 MAb (Fig. 1C).

FIG. 2.

EBV BGLF5 can inhibit the T-cell recognition of 293 cells. (A) 293 cells were transfected with pCDNA3-BGLF5-IRES-nlsGFP or the control pCDNA3-IRES-nlsGFP vector. At 48 h after transfection, HLA class I expression at the surface of transfected cells was visualized by flow cytometry. The solid line histograms depict the surface HLA class I staining with PE-conjugated W6/32 after gating for GFP⁺ cells. The dotted histogram illustrates background staining obtained with an isotype control PE-conjugated antibody. (B) 293 cells were cotransfected in 2-ml wells with 1.0 μg pCEP4-SM plasmid together with 1.0 μg control pCDNA3-IRES-nlsGFP vector or different amounts of pCDNA3-BGLF5.HA-IRES-nlsGFP (from 0.1 to 1.0 μg) bulked to a constant amount of DNA with control vector. At 24 h posttransfection, the 293 cells were cocultured with effector GLC T-cell clones for a further 18 h, and the supernatants were tested for the release of IFN-γ to measure T-cell recognition. All results are expressed as IFN-γ release (in pg/ml), and error bars indicate standard deviations of triplicate cultures. (C) Total cell lysates were generated from the above transfections, and 2 × 10⁵ cell equivalents were separated by SDS-PAGE and analyzed by Western blotting with antibodies specific for BGLF5, SM protein, or calregulin as a loading control. The loading sequence is the same as the sequence shown in panel B. Lanes: 1, vector; 2, 1.0 μg BGLF5; 3, 1.0 μg SM; 4. 1.0 μg SM plus 0.1 μg BGLF5; 5, 1.0 μg SM plus 0.2 μg BGLF5; 6, 1.0 μg SM plus 0.4 μg BGLF5; 7, 1.0 μg SM plus 1.0 μg BGLF5.

BGLF5 inhibits T-cell recognition of 293 cells.

These observations raised the possibility that BGLF5 might affect antigen presentation. To test this hypothesis, HLA-A2-positive 293 cells were cotransfected with pCEP4-SM plasmid together with control pCDNA3-IRES-nlsGFP vector or different amounts of pCDNA3-BGLF5-IRES-nlsGFP. The pCEP4-SM vector expressed SM, an EBV early lytic cycle protein that is the target of the HLA-A2-restricted GLC CD8⁺ T-cell effector clone. At 24 h posttransfection, the 293 cells were cocultured with effector T cells for a further 18 h, and culture supernatants were tested for the release of IFN-γ as a measure of T-cell recognition. The representative experiment in Fig. 2B shows that the GLC clone did not respond to vector control-transfected 293 cells but showed clear recognition of cells cotransfected with the SM vector. This recognition was inhibited in a dose-responsive manner by cotransfection of BGLF5, exceeding 90% inhibition when 1 μg pCDNA3-BGLF5 was cotransfected with the standard 1-μg dose of pCEP4-SM target vector. Similar results were obtained in three separate experiments.

Western blots of the transfected target cells (Fig. 2C) showed that BGLF5 protein expression increased with increasing amounts of transfected BGLF5 plasmid. Conversely, expression of SM protein from 1 μg transfected pCEP4-SM plasmid decreased with increasing amounts of BGLF5. Note that the blot probed for calregulin shows no obvious effect of BGLF5 on this loading control protein, because calregulin was expressed in the untransfected cells (65 to 70% in this experiment) as well as the smaller population of transfected cells.

EBV BGLF5 can inhibit T-cell recognition of MJS cells.

To demonstrate the generality of the results shown in Fig. 2, a similar set of experiments was performed with HLA-B8-positive MJS cells expressing the BZLF1 gene targeted by the RAK BZLF1-specific T-cell clone restricted through HLA-B8.

Following transfection of BGLF5 or control vector into MJS cells, the levels of HLA class I were measured by flow cytometry. BGLF5 caused a reduction in HLA class I expression at the MJS cell surface (Fig. 3A) that was reproducible, although not as marked as the reduction observed in 293 cells (cf. Fig. 2A). To test the inhibition of T-cell recognition by BGLF5, MJS cells were cotransfected with the BZLF1 vector, p509, plus either control vector or different amounts of BGLF5. At 24 h posttransfection, the MJS cells were cocultured with the BZLF1-specific RAK CD8⁺ effector clone. After another 18 h, the culture supernatants were tested for the release of IFN-γ. The T cells responded well to BZLF1-transfected MJS cells and, as in the previous set of experiments, BGLF5 caused a reproducible and dose-responsive impairment of the T-cell recognition (Fig. 3B). Similar results were obtained in replicate experiments; BGLF5 never reduced HLA expression in MJS cells to the same extent as in 293 cells, and this correlated with a weaker inhibition of T-cell recognition by BGLF5 in MJS cells.

FIG. 3.

EBV BGLF5 can inhibit T-cell recognition of MJS cells. (A) MJS cells were transfected with pcDNA3-BGLF5-IRES-nlsGFP or an empty control vector. At 48 h after transfection, HLA class I expression at the surface of transfected cells was analyzed by flow cytometry as for Fig. 2A. (B) MJS cells were cotransfected in 2-ml wells with 0.1 μg p509 plasmid (BZLF1 expression vector) together with 2.0 μg control pCDNA3 vector or different amounts of pCDNA3-BGLF5.HA-IRES-nlsGFP (from 0 μg to 2.0 μg) bulked to a constant amount of DNA with control plasmid. At 24 h posttransfection, the MJS cells were cocultured with effector RAK T-cell clones for a further 18 h, and the supernatants were tested for the release of IFN-γ as a measure of T-cell recognition, as for Fig. 2B. (C) Total cell lysates were generated from the above transfections, and 2 × 10⁵ cell equivalents were separated and analyzed by Western blotting using antibodies specific for BZLF1, HA tag (BGLF5), or calregulin as a loading control. The loading sequence is the same as the sequence shown in panel B. Lanes: 1, vector; 2, 0.1 μg BZLF1; 3, 0.1 μg BZLF1 plus 0.5 μg BGLF5; 4, 0.1 μg BZLF1 plus 1.0 μg BGLF5; 5, 0.1 μg BZLF1 plus 1.5 μg BGLF5; 6, 0.1 μg BZLF1 plus 2.0 μg BGLF5.

Western blotting of the transfected target cells (Fig. 3C) showed that BGLF5 protein expression increased with increasing amounts of transfected BGLF5 plasmid. Conversely, expression of BZLF1 protein from 1 μg transfected p509 plasmid decreased with increasing amounts of BGLF5 (Fig. 3C).

Inhibition of T-cell recognition by homologs of BGLF5.

As a first step to determining which function of BGLF5 contributes to the inhibition of antigen presentation, two AE homologs of BGLF5, from KSHV (SOX; ORF37) and HSV-1 (AE; UL12) were compared in T-cell recognition experiments with SM-transfected 293 cells. As shown in Fig. 4A, KSHV SOX had a similar effect on inhibition of T-cell recognition, as did BGLF5. In contrast, the HSV-1 AE effected little or no inhibition of T-cell recognition at equivalent doses. A similar pattern of results was obtained with T-cell recognition of BZLF1-transfected MJS cells (data not shown). From the Western blot assays performed to monitor expression of the transfected genes (Fig. 4B), it was noted that the HSV-1 AE protein was expressed at considerably higher levels than were BGLF5 and SOX proteins, suggesting that the shutoff function of the gammaherpesvirus DNases was limiting their own expression. Overall, these results indicate that the shutoff function, and not the AE function, is likely to be most important in the inhibition of antigen presentation.

FIG. 4.

Inhibition of T-cell recognition by homologs of BGLF5. 293 cells were cotransfected with 1.0 μg pCEP4-SM plasmid together with 0.2 μg or 1.0 μg control pCDNA3-IRES-nlsGFP vector or pCDNA3-BGLF5.HA-IRES-nlsGFP, pCDNA3-SOX.HA-IRES-nlsGFP (KSHV), or pCDNA3-AE.HA-IRES-nlsGFP (HSV-1). At 24 h posttransfection, the 293 cells were cocultured with the effector GLC T-cell clone, and the IFN-γ release was measured as for Fig. 2B. (B) Total cell lysates were generated from the above transfections, and 2 × 10⁵ cell equivalents were separated and analyzed by Western blotting using antibodies specific for HA tag, or for calregulin as a loading control.

The DNase and shutoff functions of BGLF5 are separable.

To confirm that the shutoff function of BGLF5 is responsible for the inhibition of antigen presentation, we attempted to separate this function from the AE function by performing random PCR-mediated mutagenesis under conditions to achieve a mutation rate in the order of one to three mutations per kilobase. In this way, we generated a pool of mutants typically containing two or three missense mutations throughout the BGLF5 gene. All the mutants were cloned into the bicistronic vector pCDNA3-IRES-nls-GFP to allow simultaneous expression of GFP and BGLF5.

More than 100 BGLF5 mutants were screened for loss or retention of shutoff function by transfecting the constructs to 293 cells and then analyzing the GFP intensity by flow cytometry (Fig. 5A). The mutants that had very high GFP intensities compared with wild-type BGLF5 were considered to have impaired shutoff function (e.g., mut.13, mut.59, and mut.64) (Fig. 5A). Conversely, the mutants that produced very low GFP intensities compared with control vector were considered to have retained their shutoff function (e.g., mut.58) (Fig. 5A). The AE function was screened by detecting the degradation of linearized DNA in an in vitro DNase assay (Fig. 5B). The lanes of the agarose gel where linearized DNA remained intact indicated those mutants that had impaired alkaline exonuclease function (e.g., mut.12, mut.14, mut.15, mut.17, mut.55, mut.58, and mut.59) (Fig. 5B). Conversely, where the linearized DNA was no longer detected, the mutants had retained part or all of their AE function (e.g., mut.13, mut.16, mut.54, mut.56, and mut.57) (Fig. 5B).

FIG. 5.

The DNase and shutoff functions of BGLF5 are separable. (A) 293 cells were transfected with pCDNA3-IRES-nlsGFP empty vector or with wild-type (wt) BGLF5 and the different indicated BGLF5 mutant insert vectors. At 48 h after transfection, GFP intensity was analyzed by flow cytometry. (B) Linearized pGEM5Zf(+) DNA was incubated with the indicated BGLF5 mutants synthesized as in vitro translation products. After the degradation reaction, the DNA was resolved by agarose gel electrophoresis and visualized by ethidium bromide staining.

Using these assays, we successfully identified two mutants that were, in at least three independent experiments for each assay, substantially impaired for either AE or shutoff activity but not both. Mutant 58 retained wild-type shutoff activity but was completely inactive as a DNase, whereas mut.13 retained most DNase activity but was substantially defective in shutoff function. We also identified several mutants that had lost both functions, e.g., mut.59 and mut.64 (Fig. 5 and data not shown), which were selected as double inactive mutant controls in further experiments.

Inhibition of antigen presentation maps to the shutoff function of BGLF5.

The single and double inactive mutants were included in T-cell assays to determine the function(s) involved in the modulation of T-cell recognition. To maximize the sensitivity of the assay, the experiments were performed in 293 cells transfected with the SM target vector, as described above (Fig. 2B). The representative results illustrated in Fig. 6A show that mut.58 (shutoff ⁺/AE⁻) had a similar effect on inhibition of T-cell recognition as that of wild-type BGLF5, whereas the shutoff-defective mutants mut.13 (shutoff⁻/AE⁺), mut.59 (shutoff⁻/AE⁻), and mut.64 (shutoff ⁻/AE⁻) were all defective in their ability to inhibit the T-cell recognition. Western blot analysis of the BGLF5 mutants in the transfected target cells revealed considerable variation in the expression levels, correlating with shutoff function. Thus, the wild-type and mut.58 transfectants exhibited much lower levels of BGLF5 protein than did the transfectants with mut.13, mut.59, and mut.64, which had impaired shutoff function (Fig. 6B), indicating that BGLF5 shutoff function also affects its own expression. Taken together, the results in Fig. 4 and 6 show that the shutoff function is critically important for the inhibition of antigen presentation.

FIG. 6.

Inhibition of antigen presentation maps to the shutoff function of BGLF5. (A) 293 cells were cotransfected with 1.0 μg pCEP4-SM plasmid together with 1.0 μg pCDNA3-IRES-nlsGFP empty vector, with wild-type (wt) BGLF5, or with the different indicated BGLF5 mutant vectors. At 24 h posttransfection, the 293 cells were cocultured with effector GLC T-cell clones for a further 18 h, and the supernatants were tested for the release of IFN-γ as for Fig. 2B. (B) Total lysates were generated from the above transfections, and 2 × 10⁵ cell equivalents were separated and analyzed by Western blotting using antibodies specific for BGLF5 or for calregulin, as a loading control. The loading sequence is the same as the sequence shown in panel A.

Table 1 shows the mutations identified in the four BGLF5 mutants used in Fig. 6. Mutant 13 (shutoff⁻/AE⁺) contained two mutations that resulted in altered amino acid sequences (231K-M and 305G-C). We were able to generate further mutants which contained only one of these mutations each, and the shutoff-defective phenotype was found to be due entirely to the K231-M mutation (Fig. 7) which lies within the third conserved motif of herpesvirus AE homologs (9, 31). Mutant 58 (shutoff ⁺/AE⁻) also contained two mutations and resulted in altered amino acid sequences (S280-L and V309-A); the S280-L mutation lies within the fourth conserved motif.

TABLE 1.

Sequences of BGLF5 mutants

Mutant	Shutoff function	AE function	Sequence changesa
13	-	+	K231-M, G305-C
58	+	-	S280-L, V309-A
59	-	-	D80-Y, V173-D, R252-K
64	-	-	V169-A, S235-P

^aIndicated are the altered amino acid sequence changes relative to the wild-type BGLF5 protein of B95.8 EBV. Underlined residues are those located in one of the seven conserved motifs of herpesvirus AE homologs (31).

FIG. 7.

(A) Linearized pGEM5Zf(+) DNA was incubated with the indicated HA-tagged BGLF5 mutants synthesized as in vitro translation products. Samples of pGEM5Zf(+) substrate taken before and after the AE degradation reaction were resolved by agarose gel electrophoresis and visualized by ethidium bromide staining. (B) To verify that equal amounts of translated product were added to each enzyme reaction mixture,10 μl of in vitro translation product was separated by SDS-PAGE and analyzed by Western blotting using antibodies specific for the HA tag. (C) Shutoff function was assayed by flow cytometry as for Fig. 5A.

DISCUSSION

This report extends our previous observation that gammaherpesvirus DNases downregulate expression of cell surface HLA molecules (31), with evidence to show that the effects of EBV BGLF5 and KSHV SOX on HLA class I molecules are sufficient to impair the recognition of lytic cycle antigens by virus-specific CD8⁺ T-cell clones (Fig. 2 and 3). The ability of gammaherpesvirus DNases to substantially impair recognition by effector T cells was not shared by the DNase of the alphaherpesvirus HSV-1 (Fig. 4). Furthermore, mutation studies on BGLF5 showed that the shutoff function rather than its AE function was critical for this immune-modulating effect (Fig. 6).

In alphaherpesviruses, the vhs function is mediated by a separate gene, UL41 (6, 15, 18), which can also interfere with recognition by CD8⁺ T cells (37). UL41 mediates protein synthesis shutoff by initiating degradation of RNA transcripts (5, 6, 10). In gammaherpesviruses, the host shutoff function is also due to a reduction in mRNA levels (8-10, 31), although the precise mechanism is unknown, since there is no evidence that BGLF5 or SOX has intrinsic RNase enzyme activities. As has been previously reported for KSHV SOX (8), we have now demonstrated by random mutagenesis that the AE and shutoff functions of BGLF5 are genetically separable, thus adding to the evidence that gammaherpesviruses have evolved bifunctional AE proteins that fulfill the same roles as the separate AE and vhs proteins in alphaherpesviruses (31). While a detailed structure-function analysis of the BGLF5 and SOX proteins is outside the scope of the present study, it is interesting that a critical mutation preferentially affecting shutoff function in BGLF5 mut.13 was mapped to the K231 residue (Table 1 and Fig. 7). This lysine lies within the third conserved motif of herpesvirus AE homologs and is even conserved in the alphaherpesvirus family member HSV-1, which does not possess shutoff function. This unexpected result emphasizes the need for further structure-function analysis if we are to understand the molecular mechanism of the shutoff induced by gammaherpesviruses.

It is notable that the degree of inhibition of T-cell responses can be so efficient when the steady-state levels of the full-length target protein remain relatively high. Thus, in the experiment illustrated in Fig. 2, more than 90% inhibition of the T-cell response was achieved with 0.4 μg BGLF5 plasmid when the level of the full-length SM target protein was reduced by only approximately 50%. However, the amount of intact target protein may not be the critical factor which determines the availability of antigenic peptides for presentation via HLA class I. It has been proposed (23) that peptides presented by major histocompatibility complex class I molecules arise mainly from proteasomal degradation of defective ribosomal products. The dramatic effect of BGLF5 on de novo protein synthesis (31) might, therefore, be more significant than the smaller effect on steady-state levels of SM and BZLF1 proteins (Fig. 2 and 3). It is also worth noting that the levels of cell surface HLA class I/peptide complexes will be influenced by a combination of other factors that might be affected by BGLF5, including de novo synthesis of HLA molecules (11, 31) and the cycling of HLA molecules. That BGLF5 showed much greater inhibition of T-cell recognition of 293/SM targets than the MJS/BZLF1 targets likely reflects the different transfection efficiencies in the two cell types; 293 cells typically gave 30 to 60% transfected cells, whereas MJS cells typically gave around 10% transfected cells and expressed lower levels of BGLF5 at the single-cell level.

It is well documented for herpesviruses in particular that immune evasion strategies targeting antigen presentation can involve multiple genes acting at different points along the antigen-processing pathway (39). Evidence is now accumulating that EBV likewise employs multiple mechanisms to target HLA class I antigen presentation. In addition to the general effects of BGLF5, the 60-amino-acid BNLF2a protein was recently shown to bind to the TAP complex and to interfere with peptide transport by inhibiting peptide-binding and ATP-binding functions of TAP (12). Furthermore, our ongoing screening of the EBV lytic cycle genes has identified a third gene able to downregulate surface HLA class I by an as-yet-unknown mechanism (J. Zuo and M. Rowe, unpublished data). There is evidence from other viruses that host shutoff function can synergize with other more specific immune-modulating genes to enhance evasion from CD8⁺ T-cell responses (18), and it is reasonable to suppose that BGLF5 might also synergize with other specific immune-modulating genes in the EBV lytic cycle. The TAP-1 and TAP-2 molecules are relatively stable molecules whose expression remains unaltered for at least 24 h after induction of the lytic cycle in B cells (26), and so BNLF2a and BGLF5 are effectively acting on different areas of the antigen-processing pathway during the early stages of the lytic cycle. A further consideration is that not all target peptides are processed via the TAP complex. For example, the SM-derived peptide that is recognized by the GLC T-cell clone (Fig. 2) can be presented by HLA-A2 in a TAP-independent manner (19). Perhaps surprisingly, BNLF2a does inhibit presentation of this peptide from endogenous protein (12), raising the possibility that this peptide is transported by unidentified TAP-related molecules which are also targeted by BNLF2a.

The BGLF5 protein is expressed with early antigen kinetics, being first detected at around 3 to 6 h after lytic cycle induction (Fig. 1). It was previously reported that both BNLF2a and BGLF5 transcripts peaked at around 8 to 24 h after induction of the lytic cycle and declined thereafter (41). This late decline in BNLF2a and BGLF5 transcripts might be due to expression of BGLF5 itself, since the shutoff function appears to affect its own mRNA (Fig. 6) as well as other viral and cellular RNAs. It is likely that an equilibrium is established between BGLF5 protein levels and mRNA turnover so that a constant low level of BGLF5 protein is established that does not completely inhibit viral mRNA expression. Interestingly, whereas BGLF5 protein levels remain constant for at least 48 h, the BNLF2a protein levels show a marked reduction within 24 h (unpublished observation).

In conclusion, we have shown that the protein synthesis shutoff function of the bifunctional BGLF5 protein can contribute to the downregulation of HLA class I expression and significantly impair antigen presentation to EBV-specific CD8⁺ T cells. It is likely that BGLF5 synergizes with other immune-modulating viral proteins, such as BNLF2a, to impair cytotoxic T-cell responses to lytic cycle genes. Since BGLF5 also downregulates surface expression of HLA class II (31), it is also possible that BGLF5 will synergize with the gp42 product of the BZLF2 gene to interfere with HLA class II immune responses (27, 28). We are currently generating recombinant EBV with BGLF5 functional knockout mutants to establish the relative contribution of the shutoff function to HLA class I and HLA class II T-cell responses.