Viruses of all classes encode factors that modulate programmed cell death of the infected host, thereby altering disease pathogenesis. However, the effects of these viral proteins on the cell death machinery also serve additional functions while the infected cell is still alive. They tweak receptor signaling pathways to promote cell proliferation, inhibit autophagy and manipulate host immunity mechanisms. Many antideath factors encoded by viruses localize to mitochondria to prevent cytochrome c release and caspase activation. These factors also have alternative functions, such as controlling mitochondrial shape changes and perhaps cellular energetics. All of these strategies prepare host cells to successfully support a virus infection, but few are understood in molecular detail. This review is dedicated to fellow Illinoian, Stan Korsmeyer, who was a tremendous scholar, leader, friend and colleague. We especially remember and appreciate Stan's remarkable contributions to the cell death field and for his support of our work.

Introduction

Viruses are obligate intracellular parasites that are adapted to employ cellular replication, transcription, translation and trafficking machineries for the purpose of successful transmission to new hosts. This invasion of normal cellular functions was once thought to entirely explain how viruses kill cells. Although the ‘lytic’ (productive) replication cycles of many viruses indeed result in cell death, these viruses also trigger cells to commit suicide by activating a pathway of programmed cell death.1, 2, 3, 4 In multicellular organisms, and in colonies of single-cell species, the activation of cell suicide in response to a virus infection is thought to have arisen during evolution as an effective early defense strategy to prevent the spread of infection.5, 6 This ancient altruistic suicide mechanism of infected cells is a crucial component of the innate host response.

Many if not all viruses have strategies to suppress and/or modulate the programmed death pathway of their host cells. Cell death induced by some viruses can exhibit the classic morphological characteristics of apoptosis, whereas others induce a multitude of other morphologies including vacuolization and membrane rupture referred to as necrosis. Even the same virus can trigger apoptotic morphology in one cell type but necrotic morphology in another. Sindbis virus infection of cortical neurons exhibits signs of apoptosis, whereas infected spinal cord motor neurons appear necrotic or autophagic and lack markers of apoptosis.7, 8, 9, 10 Elimination of virus-infected cells by natural killer cells and cytotoxic T-lymphocytes also involves programmed suicide mechanisms of the infected cell.11 In some virus infections, such as HIV and neuroadapted Sindbis virus, even uninfected cells are prone to cell death.12, 13 Viruses also take advantage of cell death inhibitors encoded by the host cell. For example, species-specific host cell factors explain why mammalian neurons die from a Sindbis virus infection while mosquito cells support long-term replication of Sindbis virus. This is not surprising given that Sindbis virus is transmitted in nature from one avian host to another by mosquitoes, and humans are accidental hosts.3, 14, 15, 16

The fact that many viruses encode genes that inhibit the cellular death program strongly supports the concept that virus-activated cellular suicide programs evolved as an effective antiviral strategy to interrupt the virus replication cycle, thereby blocking the spread of infection. E1B proteins of adenovirus and P35 of baculovirus were the earliest viral antiapoptotic genes identified. Mutant viruses lacking these genes have severe impairment of progeny virus production owing to premature death of host cells.1, 2, 17, 18 In addition to E1B-19K, E1B-55K and P35, many other types of antiapoptotic genes are encoded in viral genomes. Programmed death is viewed as the mechanism of choice to rid the host of virus-infected cells, as failure to do so often results in virus persistence, although noncytolytic mechanisms for clearing intracellular virus also exist.19, 20 However, there are additional possible roles for virus-induced apoptosis. For example, phagocytosis of apoptotic bodies was suggested to be a possible mechanism to present virus antigens and to stimulate an immune response.21 Despite the benefits of cell death for host defense, virus-induced programmed cell death can also be the cause of destruction and disease.

Viruses interfere with the cell death pathway at conceivably every possible point of the virus–host interaction interface.14, 22, 23 The responsible viral proteins have been identified in some cases and can be grouped according to where they act in the cellular death pathway (Table 1). Viral regulators of ‘death’ receptor signaling are also known to regulate the immune response functions of these same cellular receptors. Similarly, new evidence suggests that viral and cellular regulators of the mitochondrial cell death pathway also have important prosurvival functions (day-jobs) in living cells. For example, cellular antiapoptotic Bcl-2 homologs are implicated in regulating autophagy, calcium homeostasis and cellular bioenergetics.24, 25, 26, 27, 28 Thus, viral Bcl-2 homologs likely also engage in these processes, although their biochemical mechanisms remain largely unknown.

Table 1 Virus cell death/survival factors
Full size table

Death Receptor Signaling Pathways

Rerouting death receptor signals

Many viruses encode proteins that affect the function of plasma membrane cell death receptors. Death receptors constitute a subset of the tumor necrosis factor receptor (TNFR) superfamily and mediate pleotropic effects on cells.29 Their cytoplasmic protein–protein interaction ‘death domain’ (DD) is required for relaying intracellular signals in response to binding specific extracellular ligands. Binding of TNF to the extracellular domain of TNFR induces structural changes that translate to their cytoplasmic tails leading to the formation of various intracellular protein complexes, both inhibitory and excitatory. The DD domain of cytoplasmic adaptor proteins interact with the DD of TNFR and recruit additional components to affect a series of diverse biological processes, including cell death, proliferation, inflammation and responses to stress and infectious agents.30, 31 One such complex contains receptor-interacting protein (RIP) kinase, TNFR-activating factor (TRAF2) and the cellular inhibitor of apoptosis (cIAP1/2) proteins that were first identified in baculoviruses (Figure 1a).32, 33 Through additional signaling proteins that lead to the activation and translocation of NF-κB to the nucleus, this complex can trigger protection and proliferation, but this signal can potentially divert towards death by activating the E3 ubiquitin ligase activity of cIAP1 to degrade TRAF2, XIAP, cIAP2 or other inhibitors of cell death.34, 35, 36 Sindbis virus infection activates the prodeath activity of cIAP1 via caspase-8-dependent cleavage of cIAP1.37 The displacement of NF-κB-activating complexes from TNFR may result in the recruitment of distinct complexes, such as the complex facilitated by the DD-containing adapter protein Fas-associated death domain (FADD). A second DD-related death effector domain (DED) of FADD binds and recruits caspase-8 via the DD/DED-like CARD domains in the caspase-8 prodomain, thereby triggering caspase-8 activation, which ultimately activates a cascade leading to caspase-dependent apoptosis.

Figure 1
figure 1

Diagram of death receptor signaling pathways that (a) promote cell survival, proliferation and/or cell death. (b) The same molecular pathway can mediate both cell proliferation and death in a temporal manner. (c) Viral regulators of programmed cell death have additional functions

Full size image

In contrast, others argue that caspase-8 activation is required for NF-κB activation and lymphocyte proliferation, consistent with the phenotype of specific patient populations with caspase-8 deficiency.38 This interesting but provocative model appears inconsistent with the overwhelming evidence that caspase-8 mediates cell death until the issue of timing is considered. These seemingly ‘opposite’ effects of caspase-8 would be a logical strategy to essentially guarantee that proliferation of lymphocytes during infection was intrinsically linked to their cell death, thereby avoiding unchecked proliferation leading to leukemia, lymphoma or proliferative disease (Figure 1b).

Viral and Cellular FLIPs

Several γ-herpesviruses and molluscum contagiosum virus (MCV) encode viral capase-8 (FLICE) inhibitory proteins (vFLIPs). vFLIP proteins have two DED domains, mimicking the two DED/CARD domains of caspase-8/-10, and inhibit activation of these initiator caspases after death receptor ligation.39 Recently reported three-dimensional structures reveal that the viral FLIP protein, MC159 of MCV, adopts a ternary complex by simultaneously binding to the DD domain of the death receptor Fas and the DED domain of the adaptor protein FADD, but without interfering with the formation of another ternary complex between Fas, FADD and caspase-8.40 That is, the presence of vFLIP MC159, and perhaps other viral and cellular FLIPs, prevents the FADD–FADD associations required to oligomerize individual preassembled trimeric Fas–FADD–caspase-8 complexes. In the absence of vFLIP, oligomerization of these trimeric complexes leads to high levels of caspase-8 activation. The consequence of this complex formation is that FADD fails to self-associate into large complexes that serve to build an oligomerized death-inducing signaling complex (DISC). Another group recently reported that MC159 binds to TRAF3 to enhance TRAF2 binding and recruitment to the death receptor Fas.41 The identification of vFLIPs rapidly led to the identification of cellular FLIP (cFLIP), which have two forms, a short form structurally related to vFLIP, and a long form, cFLIPL, which contains an additional caspase-like domain but lacks proteolytic activity.42 Both cFLIP and vFLIP regulate NF-κB activation in response to TNF receptors.43

TNF Receptor Mimicry

A large proportion of the genes encoded by poxviruses, herpesviruses and others are not required for virus replication in laboratory settings, implying that these genes are critical for successful virus transmission in the environment. Many of these ‘nonessential’ genes are critical for immune evasion and encode mimics of the TNFR that bind to TNF, such as the cytokine response modifier proteins CrmB, -C, -D and -E of cowpox virus, the M-T2 protein of myxoma virus and S-T2 of Swope fibroma virus.29, 44, 45 Alternative strategies to accomplish similar effects were adopted by other viruses that encode vTNF-BP proteins (e.g. the L2 and related proteins encoded by Tanapox virus, Yaba monkey tumor virus, swinepox and Mule deer poxvirus), which share significant amino-acid sequence similarity to MHC class I molecules but lack transmembrane domains.46 The heterotrimeric receptor internalization and degradation (RID) complex composed of two adenovirus transmembrane proteins, E3-10.4K and E3-14.5K, also inhibits TNRF1 signaling. RID components bind TNFR1 leading to internalization and degradation. RID also blocks the association of RIP kinase with TNFR1, and inhibits NF-κB activation by inhibiting assembly of the requisite IKK (IkB kinase) complex.47, 48 RID can also target other death receptors such as Fas and TRAIL-R1.49

Still other viruses encode proteins that act directly or indirectly to dampen TNF and other death receptor signaling pathways by reducing TNFR expression and/or assembly of TNFR1 complexes and downstream effectors. Immediate-early transcription factors of Epstein–Barr virus (EBV) alter expression levels of TNFR1 and its negative regulators.50 The unrelated bovine leukemia retroviruses, human papillomaviruses (HPV), poliovirus, human T-cell leukemia virus (HTLV-1 tax protein), rotavirus (VP4 protein), herpes simplex virus glycoprotein gD and many others also modulate NF-κB activation, which presumably promotes cell survival and inhibits cell death, but the details are likely to be more complicated.29

The E6 protein of HPV16, the etiologic agent of cervical and other cancers, binds the cytoplasmic tail of TNFR1 to inhibit TRADD association and subsequent cell death.51 The 3A protein of poliovirus increases turnover of TNFR1 from the plasma membrane.52 The NS5A protein promotes persistence of hepatitis C virus (HCV) by binding to TRADD, thereby interfering with FADD binding to death receptors, whereas another HCV protein, the 22 kDa core/capsid precursor, was reported to either promote or inhibit TNF-induced cell death.29 However, detailed molecular mechanisms are still in progress for most of these viral mechanisms.

Molecular Target of Hepatitis C Virus NS3-4A Protease

Headway towards a molecular understanding of virus-regulated signaling comes from the recent identification of intracellular sensors of viruses and how viruses sabotage these sensor-mediated signals to subvert host–defense strategies. In addition to Toll-like receptor 3 (TLR3), which recognizes dsRNA intermediates generated during virus replication, two CARD-containing cytosolic RNA helicases, RIG-I/Ddx58 (retinoic acid-inducible gene I) and Mda5/Ifih1/Helicard1, also bind to dsRNA to trigger activation of NF-κB and IRF3 (interferon regulatory factor 3).53 An exciting recent development was the identification by several different groups of a new CARD-containing adaptor protein, Cardif/MAVS/IPS-1/VISA, that activates NF-κB in response to virus infection, here referred to as MCARD for the lack of a better term.54, 55 MCARD binds RIG-I to promote activation of NF-κB and IRF3. But HCV, an important cause of chronic liver disease, specifically targets MCARD. The HCV serine protease NS3-4A, which is required to cleave viral polyproteins during virus replication, also cleaves after Cys508 of MCARD, causing a potent inhibition of signaling.54 Therefore, the anti-viral compound BILN2061, which inhibits the NS3-4A protease, will presumably also impair the ability of HCV to overcome host–defense mechanisms.

Alternative Strategies of Interference

Delineation of this mechanism of MCARD has stimulated investigations of other viruses. As expected, however, there will be many alternative strategies to regulate the pathways downstream of the cellular sensors that are triggered upon infection with a virus. Kaposi's sarcoma-associated herpesvirus (KSHV), which causes skin lesions of immunocompromised HIV patients, encodes its own vIRF3 to inhibit TNF-induced NF-κB activation, whereas adenovirus E1A sensitizes cells to TNF-mediated cell death by inhibiting cFLIP expression.56, 57 Rotavirus VP4 contains a TRAF-binding motif impair caspase-8 activation and promote NF-κB activation, African swine fever virus (ASFV) encodes a mimic of IκB to inhibit NF-κB activation, and vaccinia virus encodes at least two inhibitors of NF-κB activation N1L and K1L.29

Regardless of where in the pathway these viral factors are eventually found to function, questions remain as to what are their ultimate purposes. An obvious goal of those viral modulators considered above is for avoidance of host–defense responses that trigger cell death to guard against unscheduled proliferation and the spread of invading viruses. These same functions of viral proteins can also promote prosurvival and proliferative signals needed to provide the cellular machinery required for completion of virus replication. It remains possible, however, that these factors also have additional functions and that cell survival and cell death are only the final outcomes. For example, cellular energetics is closely linked to cell proliferation and programmed cell death. Thus, it is conceivable that many viruses encode factors that specifically alter the bioenergetic status of host cells. Afterall all of the processes described above require energy. The vaccinia virus N1L protein was suggested recently to enhance ATP utilization, but the possible mechanisms are unknown.58 Other candidates include a host of different viral factors that localize to mitochondria.

Mitochondrial Factors Encoded by Viruses

Death factors with day-jobs

Cellular Bcl-2 family proteins are key regulators of permeability of the outer mitochondrial membrane, inhibiting the release of prodeath factors such as cytochrome c, SMAC/Diablo, AIF, EndoG and HtrA/OMI from the intermembrane space.59, 60 Yet, the biochemical mechanisms by which Bcl-2 family members achieve these functions are still a mystery. Cellular Bcl-2 proteins are generally divided into three functional subgroups, antiapoptotic (Bcl-2, Bcl-xL, Mcl-1, Bcl-w), proapoptotic Bcl-2 family members (Bax and Bak) and proapoptotic BH3-only proteins (Bad, Bid and others). Although these classifications generally apply and were confirmed using knockout mice, there are clear examples where the endogenous functions of these proteins are reversed. For example, Bax, Bak and Bad can be antiapoptotic,15, 61, 62 whereas Bcl-2 and Bcl-xL can promote cell death when their N-terminus is cleaved off.63, 64, 65 An extensive analysis of Bak knockout mice revealed that Bak function is dependent on the cell type, the developmental stage and the specific death stimulus.15 However, until the biochemical functions of Bcl-2 family proteins are fully delineated, the apparently reversible functions of Bcl-2 family proteins remain an enigma. Nevertheless, these observations led us to consider that perhaps all Bcl-2 family proteins have prosurvival ‘day-jobs’, but that under specific circumstances can be converted into prodeath factors. In this manner, Bcl-2 family proteins are analogous to the dual roles of cytochrome c in energetics versus caspase-9 activation, and to caspase-8 in promoting cell proliferation versus apoptosis. However, few investigators study the functions of Bcl-2 family proteins in the absence of a death stimulus.

Viral Bcl-2 family proteins

Given the importance of cellular Bcl-2 family proteins, it is no surprise that many different viruses encode sequence homologs and other factors that modulate Bcl-2 protein function. Viral Bcl-2 homologs are encoded by γ-herpesviruses such as EBV and KSHV (HHV8), and several different poxviruses including the myxoma virus M11L protein and fowlpox virus-Bcl-2.46 Even though viral homologs share relatively low amino-acid sequence identity with cellular Bcl-2 proteins, the NMR structures of BHRF1 from EBV and of KSBcl-2 from KSHV reveal similar protein folds compared to cellular Bcl-xL and Bax, but can be less capable of binding prodeath BH3 peptides.66, 67, 68, 69 Another mechanism by which viral Bcl-2 homologs escape cellular regulatory mechanisms is that they cannot be converted into prodeath factors by proteolysis like their cellular counterparts.70 Nevertheless, these viral factors still prevent the loss of mitochondrial membrane potential and permeabilization of the outer mitochondrial membrane like their cellular counterparts. For example, the myxoma poxvirus protein M11L, which is required to prevent apoptosis during virus infection, is targeted to mitochondria through a C-terminal hydrophobic domain flanked by positively charged residues.46 M11L inhibits mitochondrial permeability transition (loss of ΔΨm) that occurs following a death stimulus, apparently by binding to the mitochondrial PBR and/or Bak.

Viral Proteins that Target Cellular Bcl-2 Proteins

Adenovirus E1B-19K, vaccinia virus F1L and cytomegalovirus vMIA (viral mitochondrial inhibitor of apoptosis) and vICA (viral inhibitor of caspase-8 activation) proteins also function at the same step as Bcl-2 as they bind and regulate cellular Bax and/or Bak.46 vMIA is the product of immediate-early gene UL37 exon 1 and localizes to mitochondria.71 vMIA surprisingly causes mitochondrial fragmentation/division that is usually only observed during cell death, yet vMIA potently suppresses apoptosis triggered by diverse stimuli. A recent study identified GADD45 (growth arrest and DNA damage 45) as a binding partner that binds to the domain of vMIA required for antideath activity.72 Furthermore, depletion of three GADD45 isoforms abrogated the antideath activity of vMIA, suggesting a possible link between DNA damage responses and vMIA.

Bcl-2 Family Proteins in Mitochondrial Biogenesis, Energetics and Autophagy

Cellular Bcl-2 family proteins were recently implicated in regulating mitochondrial fission and fusion. The Bcl-2 homolog encoded by Caenorhabditis elegans, CED-9, was recently found to interact with Mfn2 (mitofusin 2; homolog of yeast fuzzy onion 1/Fzo1), a regulator of mitochondrial fusion.73 Genetic studies in C. elegans demonstrated that worm Bcl-2, normally an antideath factor, can promote cell death by activating Drp1, a dynamin-like GTPase that was first identified in yeast as a factor that is required for mitochondrial fission.74, 75 The human and yeast homologs of Drp1/Dnm1 also promote programmed cell death in mammalian and yeast cells, respectively.76, 77, 78 Like Bax, Drp1 relocalizes to mitochondria during cell death and colocalizes with Bax at mitochondrial constriction sites.79, 80 The connection between mitochondrial fission and cell death is still unclear but may be involved in the release of cytochrome c from mitochondria or in mitochondrial degradation.81, 82 On the other hand, mitochondrial fission is important for mitochondrial maintenance and biogenesis. Interestingly, Drp1 is also required to insert a mitochondrion into neuronal structures rich in synapses.83 This finding fits with the idea that Drp1 mediates fission to send newly produced mitochondria to ATP-requiring locations. The observations that Bcl-xL, Bax and Bak can regulate synaptic activity, and the studies that connect human Drp1 and worm Bcl-2, suggest that Bcl-2 family proteins may regulate synaptic activity via a functional interaction with Drp1 or another mitochondrial factor.15, 84, 85, 86 Perhaps, these scenarios reveal aspects of the ‘day-jobs’ of Drp1, Bax and Bak that are otherwise known better as prodeath factors. These findings raise the possibility that viral Bcl-2 proteins also regulate mitochondria by affecting membrane fission and fusion.

Bcl-2 family proteins are also implicated in regulating cellular bioenergetics, perhaps by regulating availability of mitochondrially produced ATP.28 Although the results of microarray screens have suggested that viruses have a strong influence on the expression of cellular proteins that control lipid metabolism, amino-acid metabolism and oxidative phosphorylation, the mechanisms involved are not known (Figure 1c).87, 88, 89, 90 Cellular Bcl-2 proteins have also been shown to interact functionally with mitochondrial membrane proteins VDAC and ANT that regulate the exchange of ATP and ADP between the mitochondrial matrix and the cytosol.91, 92 It would not be surprising if viral factors specifically modulate cellular energetics to benefit the virus, but this area is understudied.

Viral and cellular proteins are reported to regulate autophagy, a recycling process first delineated in yeast that promotes cell survival when nutrients are limiting.93, 94, 95 In a yeast two-hybrid screen, cellular Bcl-2 pulled out Beclin, the mammalian homolog of yeast ATG6 that regulates autophagy.96 Beclin also protects mice from a fatal Sindbis virus infection.96 We also identified Beclin in a yeast two-hybrid screen using the viral Bcl-2 homolog encoded by mouse gamma herpesvirus 68 as bait (B Burns and JM Hardwick, unpublished). Another herpesvirus Bcl-2 homolog, KSBcl-2, like its cellular homolog Bcl-2, inhibits autophagy at least in part by interfering with the interaction of Beclin with IP3K/Vsp34.97 However, the molecular details of these processes are still being uncovered.

Considering the Complexities and Caveats

Part of the difficulty in understanding the complexities of virus–host cell interactions is that the various model systems used to study these processes each have different hard-wired cellular programs that dictate outcomes. Thus, it is challenging to correlate even the most obvious experimental observations in cultured cells/extracts with the relevant balance of forces that determine the outcome of virus infections in animals. This difficulty is further compounded by temporal events, such as those described above where caspases may promote cell proliferation before promoting cell death.98 Thoughtful considerations are also required to distinguish between direct and indirect effects of these viral factors, especially when pathways are circular, or when outcomes are dependent on amplification of signals to reach detectable levels, such as caspase activity. Artificially elevated levels of caspase activity resulting from experimental conditions may not normally occur, or only occur late in the process when their prodeath functions are opposite to their day jobs. For example, low caspase-8 activity levels (indistinguishable from background) could promote, or be required for, cell proliferation, while enhanced caspase-8 activation causes apoptosis (Figure 1b).

Although many are reluctant to consider that caspases play key roles in cell proliferation, a number of compelling studies with knockout animals and cells derived from patients support this idea. This role for at least some caspases would also explain the exceptional difficulty that we and others have had in establishing stable cell lines expressing viral caspase inhibitors, which bind and directly inhibit caspase active sites such as cowpox virus CrmA and baculovirus P35 (unpublished data).

These observations evoke another cautionary message; stably expressed viral (or cellular) regulators of cell death that impede cell growth and/or survival will automatically provide strong selection pressures for compensatory mutations in the same or alternate pathways. This principle of compensatory mutations is an established phenomenon, but its prevalence may be severely underappreciated. This was made obvious to us by studying a knockout mutant of the yeast Saccharomyces cerevisiae that lacks an inhibitor of programmed cell death. This mutant has genetically separable mutations not present in parental strains that result in altered growth phenotypes (W-C Cheng and JM Hardwick, unpublished). This observation caused us to reconsider other observations in our lab, such as newly established human Bcl-2- and Bcl-xL-expressing cell lines that exhibit growth-inhibited phenotypes in the first week of selection compared to matched controls, but lose this slow-growth phenotype thereafter (DG Kirsch and JM Hardwick, unpublished).6, 63, 99 Newly established cell lines expressing protease-resistant murine Bad, a BH3-only protein, readily reveals the antideath function of Bad, but only for the first several passages. This antideath function of Bad was gradually lost relative to matched controls over the next few weeks, but was repeatedly recovered by generating new cell lines (SY Seo and JM Hardwick, unpublished).62 Similarly, cIAP1 function must be analyzed de novo (T-T Sheu and JM Hardwick, unpublished). However, by considering the nature of these compensatory mutations that permit survival and growth in the absence of antideath factors, we will likely learn about the underlying functions of viral antideath factors.