Programmed cell death is thought to function as an innate antiviral defense mechanism. This hypothesis is supported by the fact that many viruses have evolved cell death-inhibiting proteins (1). However, in only a few instances has the importance of these proteins for viral fitness been determinedin vivo (2–6). Murine cytomegalovirus (MCMV), a betaherpesvirus, has a protracted replication cycle and causes persistent infections in its host. MCMV expresses several cell death inhibitors, which are presumably needed to keep the host cell alive and promote viral replication and dissemination (7,8). Since MCMV's natural host, the mouse, is an easily available and genetically manipulable organism, the MCMV-infected mouse model can serve as an excellent system to study thein vivo relevance of different cell death pathways.

In MCMV-infected cells, the extrinsic, death receptor-induced apoptosis pathway is blocked by the MCMV protein M36, which inhibits caspase-8 activation (9). An MCMV ΔM36 mutant was strongly attenuatedin vitro andin vivo (2). MCMVs carrying mutations in M45 showed an even stronger attenuationin vivo with barely detectable organ titers (5,10). The viral M45 protein inhibits both death receptor-induced and virus-induced programmed necrosis by interacting with receptor-interacting protein 1 (RIP1) and RIP3 (11,12). Thein vivo attenuation of an M45 mutant MCMV was abrogated in RIP3-deficient mice, which elegantly confirmed the importance of RIP3-dependent necrosis as a host defense mechanism (10).

Apoptosis can also be induced by different intracellular stress signals via the intrinsic mitochondrial pathway (13). These stress signals activate the proapoptotic Bcl-2 family members BAX and BAK. Activated BAX and BAK oligomerize within the outer mitochondrial membrane and increase membrane permeability, resulting in the release of cytochromec and other proapoptotic factors into the cytosol. Cytochromec then induces the formation of a so-called apoptosome complex and subsequent activation of caspases (13). MCMV inhibits the intrinsic apoptosis pathway at mitochondria with two separate viral proteins: a BAX-specific inhibitor, encoded by open reading frame (ORF) m38.5 (14–17), and a BAK-specific inhibitor, encoded by ORF m41.1 (18). The mitochondrion-localized m41.1 protein is expressed from the m41 locus and functions as viral inhibitor of BAK oligomerization (18). In this study, we investigated the role of m41.1/viral inhibitor of BAK oligomerization (vIBO) for MCMV replication and dissemination in mice.

Since the previously constructed m41.1 mutants (18) were not suitable forin vivo infection experiments, we constructed a new m41.1 knockout (m41.1ko) virus and a corresponding revertant. Mutagenesis was done by homologous recombination using the bacterial artificial chromosome (BAC) clone of MCMV strain K181 (19). First, the viral m157 gene was deleted by replacing it with a zeocin resistance (zeo) marker, as described previously (20) (Fig. 1A). This deletion was necessary in order to infect C57BL/6 mice, which are resistant to MCMV infection due to expression of the NK cell receptor Ly49H (21–24). Resistance can be overcome by removal of m157, which encodes a Ly49H ligand (20). In this Δm157 wild-type-like (wt-l) virus genome, the m41 locus was first replaced by agalK-kan cassette using kanamycin for positive selection. Using 2-deoxy-galactose for negative selection,galK-kan was then replaced by an m41 sequence in which all three m41.1 ATG codons were mutated to ACG, leaving the protein sequence of the overlapping m41 ORF unchanged (18). Using the same two-step replacement strategy (which is described in detail elsewhere [25]), the revertant (rev) was constructed (Fig. 1A). All BACs were analyzed by restriction digestion with XbaI (Fig. 1B), as well as SnaBI and ScaI (not shown), and the entire m41 locus was sequenced to ensure that only the intended changes were introduced (Fig. 1C). Recombinant viruses were reconstituted by transfecting BAC DNA into 10.1 fibroblasts (26) using Polyfect transfection reagent (Qiagen). Viral DNA was purified from virions as described previously (27) and analyzed by restriction digestion. The restriction patterns of the wt-l, m41.1ko, and rev viruses were as predicted (data not shown). To verify that m41.1 was knocked out and expression of the unrelated m41 protein had remained unchanged, lysates of infected cells were analyzed for m41 expression by immunoblotting using the m41-specific antibody 2A6 and for m41.1 expression by reverse transcription (RT)-PCR (Fig. 1D andE).

Before using the newly constructed K181-derived viruses forin vivo experiments, we functionally characterized them in cell culture-based assays. First we infected 10.1 fibroblasts at a multiplicity of infection (MOI) of 5. Six hours postinfection (hpi), cells were stimulated with 200 nM staurosporine (STS) (Sigma) to induce apoptosis. Cell viability was measured 24 hpi by MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulphophenyl)-2H-tetrazolium] assay (CellTiter 96 AQ_ueous cell proliferation assay; Promega). As shown inFig. 2A, cells infected with the m41.1ko virus were more sensitive to STS-induced apoptosis than the parental wt-l or rev viruses. Next we infected RAW264.7 macrophages (ATCC TIB-71), which are more sensitive than fibroblasts to infection-induced proapoptotic stress. Two days postinfection (dpi), macrophages infected with m41.1ko virus showed clearly reduced viability compared to cells infected with the control viruses (Fig. 2B). To confirm that the observed phenotype was BAK dependent, BAK and BAX knockout cells (28) were infected for 72 h. Cell viability was then analyzed by MTS assay. As shown inFig. 2C andD, BAK deficiency but not BAX deficiency compensated for the loss of m41.1. Infection of 10.1 cells with m41.1ko resulted in caspase-3 activation (Fig. 2E), confirming that m41.1 is required for apoptosis inhibition. In addition, the replication kinetics of the three viruses were determined on 10.1 fibroblasts and RAW264.7 macrophages. Replication of the m41.1ko virus was indistinguishable from that of the control viruses in 10.1 fibroblasts but significantly reduced in macrophages (Fig. 3A andB), consistent with previously published results (18). The m41.1ko virus also replicated to lower titers in primary bone marrow-derived macrophages (BMDMs) of C57BL/6J but not BAK-deficient mice (B6.129-Bak1^tm1Thsn/J) (29) mice, indicating that the replication defect is BAK dependent (Fig. 3C andD).

Manzur and colleagues showed that the viral BAX inhibitor m38.5 is important for MCMV replication and dissemination in mice (16). Therefore, we wanted to analyze if BAK inhibition by m41.1 is also of importance for viral replication and spread. To do this, we infected 8- to 10-week-old C57BL/6J andBak1^−/− mice with 2 × 10⁵ PFU of wt-l, m41.1ko, or rev virus intravenously or intraperitoneally (Fig. 4A toE). Mice were sacrificed on days 3 and 14 after intraperitoneal infection and, due to a faster viral dissemination, on days 3 and 10 after intravenous infection. Spleen, salivary gland (SG), and lung titers were determined by plaque assay. On day 3 postinfection, viral titers in the spleen did not differ significantly independent of the mouse strain used (Fig. 4A andD). However, the m41.1ko strain reached moderately (10- to 50-fold) but significantly reduced SG titers compared to wt-l and rev viruses in C57BL/6J mice (Fig. 4B andE). InBak1^−/− mice, the SG titers were similar for all three viruses, providing strong support for the notion that BAK inhibition is necessary for optimal replication and dissemination to SGs. The m41.1ko virus also reached lower titers in lungs of C57BL/6J mice (Fig. 4C), although lung titers generally showed more variability. To rule out that a spontaneous reversion was responsible for the modest attenuation of m41.1ko virusesin vivo, we sequenced the m41 locus of m41.1ko viruses isolated from SGs and lungs of C57BL/6J mice. All sequenced isolates carried the mutant m41.1 sequence (data not shown).

Previous studies with MCMV ΔM36 and ΔM45 mutants have demonstrated a severely attenuated phenotype of these virusesin vivo (2,5). In contrast, MCMV mutants with an inactivated m38.5 (16) or m41.1 (this study) gene were only modestly attenuated in mice, reaching about 10- to 100-fold-lower SG titers. As these organs are important for MCMV persistence and transmission, even a modest attenuation is likely to be of biological importance. One should also keep in mind that the m38.5 and m41.1 gene products target two related proapoptotic mitochondrial proteins, BAX and BAK, that have been proposed to act in a largely redundant fashion (30). In fact, an m38.5 m41.1 double-knockout (dko) virus has shown a more prominent phenotype in cultured cells than the single-gene knockouts (18). Hence it might be necessary to test a dko virus in the mouse to fully appreciate the role of the intrinsic, mitochondrially mediated apoptosis pathway as an antiviral defense mechanism. However, MCMV with its highly specific inhibitors offers the unique opportunity to study the contributions of BAX and BAK to infection control individually.

This work was supported by Deutsche Forschungsgemeinschaft grant BR1730/3 to W.B.

We thank Mijo Golemac for help with animal experiments, which were performed at the University of Rijeka according to national and institutional animal care and ethical guidelines and were approved by local ethical committees. All mice were raised under specific-pathogen-free conditions at the Central Animal Facility of the Medical Faculty, University of Rijeka, Croatia.