The Adenovirus E4orf4 Protein Induces G2/M Arrest and Cell Death by Blocking Protein Phosphatase 2A Activity Regulated by the B55 Subunit

Human adenovirus E4orf4 protein is toxic in human tumor cells. Its interaction with the Bα subunit of protein phosphatase 2A (PP2A) is critical for cell killing; however, the effect of E4orf4 binding is not known. Bα is one of several mammalian B-type regulatory subunits that form PP2A holoenzymes with A and C subunits. Here we show that E4orf4 protein interacts uniquely with B55 family subunits and that cell killing increases with the level of E4orf4 expression. Evidence suggesting that Bα-specific PP2A activity, measured in vitro against phosphoprotein substrates, is reduced by E4orf4 binding was obtained, and two potential B55-specific PP2A substrates, 4E-BP1 and p70^S6K, were seen to be hypophosphorylated in vivo following expression of E4orf4. Furthermore, treatment of cells with low levels of the phosphatase inhibitor okadaic acid or coexpression of the PP2A inhibitor I₁^PP2A enhanced E4orf4-induced cell killing and G₂/M arrest significantly. These results suggested that E4orf4 toxicity results from the inhibition of B55-specific PP2A holoenzymes, an idea that was strengthened by an observed growth arrest resulting from treatment of H1299 cells with Bα-specific RNA interference. We believe that E4orf4 induces growth arrest resulting in cell death by reducing the global level of B55-specific PP2A activity, thus preventing the dephosphorylation of B55-specific PP2A substrates, including those involved in cell cycle progression.Our research group and others have shown that the 114-residue product of early region E4 of human adenoviruses, termed E4orf4, induces p53-independent cell death in human tumor cells (24, 25, 34-36, 55) and in Saccharomyces cerevisiae (23, 53). E4orf4 protein, which shares no obvious homology with other viral or cellular products, kills a wide range of human cancer cells but is believed to have reduced activity against normal human primary cells (6, 55, 56). Although in some cases E4orf4-expressing cells exhibit characteristics typical of apoptosis, including the presence of irregularly shaped and shrunken nuclei, cytoplasmic vacuolization, and membrane blebbing (24, 25, 50, 55), cell death may more typically be independent of caspase activation (24, 25, 30, 32, 50). With H1299 human non-small-cell lung carcinoma cells, death is characterized by rapid cell rounding, enlargement, release from the surface of culture plates, cell cycle arrest in G₂/M and possibly G₁, and eventually, after an extended period, loss of membrane integrity (30). Both cytoplasmic and nuclear pathways have been observed, the former involving interactions with c-Src family kinases, activation of calpain, and remodeling of the actin cytoskeleton (7, 24, 50, 51, 58). Little is known about the nuclear pathway, which may represent the predominant death-inducing process. Our current evidence suggests that H1299 cells die following prolonged irreversible cell cycle arrest leading to mitotic catastrophe and death by a necrosis-like process (30).E4orf4 is known to associate with the Bα regulatory subunit of protein phosphatase 2A (PP2A) (22, 34), and this interaction appears to be necessary for the majority of E4orf4 toxicity in both yeast (23, 53) and human tumor cells (34, 56). PP2A is an abundant serine-threonine phosphatase involved in regulation of metabolism, splicing, translation, morphogenesis, development, and cell cycle progression (15, 19, 27, 43, 59). PP2A holoenzymes exist as multiple heterotrimeric complexes composed of a catalytic C subunit, an A subunit that functions as a scaffold, and a B-type regulatory subunit. Two forms each of the A and C subunits exist in mammalian cells; however, more than 20 B-type subunits have been identified in three unique classes (B/B55, B′/B56, B″/PR72), plus striatin/SG2NA (sometimes called B‴) (10, 19, 26). Although one group has suggested that E4orf4 protein interacts with one or more members of the B′/B56 class (57), it is generally accepted that interaction with the Bα/B55 subunit (Cdc55 in yeast) is important for induction of cell death in both human tumor cells and yeast (53, 57). Interestingly, a recent report has also suggested that in yeast, growth suppression induced by E4orf4 is mediated only in part by the catalytic C subunit of PP2A (31).In the present report, we show that E4orf4 protein interacts uniquely with members of the B55 class of PP2A B-type subunits, and at sufficient concentrations, it appears to become toxic by reducing dephosphorylation of substrates of B55-containing PP2A holoenzymes. As cell death is preceded by cell cycle arrest, we believe that key substrates may include proteins required for cell cycle progression.     

In Vitro and In Vivo Biology of Recombinant Adenovirus Vectors with E1, E1/E2A,or E1/E4 Deleted

Isogenic, E3-deleted adenovirus vectors defective in E1, E1 and E2A, or E1 and E4 were generated in complementation cell lines expressing E1, E1 and E2A, or E1 and E4 and characterized in vitro and in vivo. In the absence of complementation, deletion of both E1 and E2A completely abolished expression of early and late viral genes, while deletion of E1 and E4 impaired expression of viral genes, although at a lower level than the E1/E2A deletion. The in vivo persistence of these three types of vectors was monitored in selected strains of mice with viral genomes devoid of transgenes to exclude any interference by immunogenic transgene-encoded products. Our studies showed no significant differences among the vectors in the short-term maintenance and long-term (4-month) persistence of viral DNA in liver and lung cells of immunocompetent and immunodeficient mice. Furthermore, all vectors induced similar antibody responses and comparable levels of adenovirus-specific cytotoxic T lymphocytes. These results suggest that in the absence of transgenes, the progressive deletion of the adenovirus genome does not extend the in vivo persistence of the transduced cells and does not reduce the antivirus immune response. In addition, our data confirm that, in the absence of transgene expression, mouse cellular immunity to viral antigens plays a minor role in the progressive elimination of the virus genome.Replication-deficient human adenoviruses (Ad) have been widely investigated as ex vivo and in vivo gene delivery systems for human gene therapy. The ability of these vectors to mediate the efficient expression of candidate therapeutic or vaccine genes in a variety of cell types, including postmitotic cells, is considered an advantage over other gene transfer vectors (3, 28, 49). However, the successful application of currently available E1-defective Ad vectors in human gene therapy has been hampered by the fact that transgene expression is only transient in vivo (2, 15, 16, 33, 36, 46). This short-lived in vivo expression of the transgene has been explained, at least in part, by the induction in vivo of cytotoxic immune responses to cells infected with the Ad vector. Studies with rodent systems have suggested that cytotoxic T lymphocytes (CTLs) directed against virus antigens synthesized de novo in the transduced tissues play a major role in eliminating cells containing the E1-deleted viral genome (56–58, 61). Consistent with the concept of cellular antiviral immunity, expression of transgenes is significantly extended in experimental rodent systems that are deficient in various components of the cellular immune system or that have been rendered immunocompromised by administration of pharmacological agents (2, 33, 37, 48, 60, 64).Based on the assumption that further reduction of viral antigen expression may lower the immune response and thus extend persistence of transgene expression, previous studies have investigated the consequences of deleting both E1 and an additional viral regulatory region, such as E2A or E4. The E2A region encodes a DNA binding protein (DBP) with specific affinity for single-stranded Ad DNA. The DNA binding function is essential for the initiation and elongation of viral DNA synthesis during the early phase of Ad infection. During the late phase of infection, DBP plays a central role in the activation of the major late promoter (MLP) (for a recent review, see reference 44). The E4 region, located at the right end of the viral genome, encodes several regulatory proteins with pleiotropic functions which are involved in the accumulation, splicing, and transport of early and late viral mRNAs, in DNA replication, and in virus particle assembly (reviewed in reference 44). The simultaneous deletion of E1 and E2A or of E1 and E4 should therefore further reduce the replication of the virus genome and the expression of early and late viral genes. Such multidefective vectors have been generated and tested in vitro and in vivo (9, 12, 17, 19–21, 23, 24, 26, 34, 40, 52, 53, 59, 62, 63). Recombinant vectors with E1 deleted and carrying an E2A temperature-sensitive mutation (E2Ats) have been shown in vitro to express much smaller amounts of virus proteins, leading to extended transgene expression in cotton rats and mice (19, 20, 24, 59). To eliminate the risks of reversion of the E2Ats point mutation to a wild-type phenotype, improved vectors with both E1 and E2A deleted were subsequently generated in complementation cell lines coexpressing E1 and E2A genes (26, 40, 63). In vitro analysis of human cells infected by these viruses demonstrated that the double deletion completely abolished viral DNA replication and late protein synthesis (26). Similarly, E1/E4-deleted vectors have been generated in various in vitro complementation systems and tested in vitro and in vivo (9, 17, 23, 45, 52, 53, 62). These studies showed that deletion of both E1 and E4 did indeed reduce significantly the expression of early and late virus proteins (17, 23), leading to a decreased anti-Ad host immune response (23), reduced hepatotoxicity (17, 23, 52), and improved in vivo persistence of the transduced liver cells (17, 23, 52).Interpretation of these results is difficult, however, since all tested E1- and E1/E4-deleted vectors encoded the bacterial β-galactosidase (βgal) marker, whose strong immunogenicity is known to influence the in vivo persistence of Ad-transduced cells (32, 37). Moreover, the results described above are not consistent with the conclusions from other studies showing, in various immunocompetent mouse models, that cellular immunity to Ad antigens has no detectable impact on the persistence of the transduced cells (37, 40, 50, 51). Furthermore, in contrast to results of earlier studies (19, 20, 59), Fang et al. (21) demonstrated that injection of E1-deleted/E2Ats vectors into immunocompetent mice and hemophilia B dogs did not lead to an improvement of the persistence of transgene expression compared to that with isogenic E1-deleted vectors. Similarly, Morral et al. (40) did not observe any difference in persistence of transgene expression in mice injected with either vectors deleted in E1 only or vectors deleted in both E1 and E2A. Finally, the demonstration that some E4-encoded products can modulate transgene expression (1, 17, 36a) makes the evaluation of E1- and E1/E4-deleted vectors even more complex when persistence of transgene expression is used for direct comparison of the in vivo persistence of cells transduced by the two types of vectors.The precise influence of the host immune response to viral antigens on the in vivo persistence of the transduced cells, and hence the impact of further deletions in the virus genome, therefore still remains unclear. To investigate these questions, we generated a set of isogenic vectors with single deletions (AdE1°) and double deletions (AdE1°E2A° and AdE1°E4°) and their corresponding complementation cell lines and compared the biologies and immunogenicities of these vectors in vitro and in vivo. To eliminate any possible influence of transgene-encoded products on the interpretation of the in vivo results, we used E1-, E1/E2A-, and E1/E4-deleted vectors with no transgenes.     

A Proteomic Approach To Identify Candidate Substrates of Human Adenovirus E4orf6-E1B55K and Other Viral Cullin-Based E3 Ubiquitin Ligases

It has been known for some time that the human adenovirus serotype 5 (Ad5) E4orf6 and E1B55K proteins work in concert to degrade p53 and to regulate selective export of late viral mRNAs during productive infection. Both of these functions rely on the formation by the Ad5 E4orf6 protein of a cullin 5-based E3 ubiquitin ligase complex containing elongins B and C. E1B55K is believed to function as the substrate recognition module for the complex and, in addition to p53, Mre11 and DNA ligase IV have also been identified as substrates. To discover additional substrates we have taken a proteomic approach by using two-dimensional difference gel electrophoresis to detect cellular proteins that decrease significantly in amount in p53-null H1299 human lung carcinoma cells after expression of E1B55K and E4orf6 using adenovirus vectors. Several species were detected and identified by mass spectroscopy, and for one of these, integrin α3, we went on in a parallel study to confirm it as a bone fide substrate of the complex (F. Dallaire et al., J. Virol. 83:5329-5338, 2009). Although the system has some limitations, it may still be of some general use in identifying candidate substrates of any viral cullin-based E3 ubiquitin ligase complex, and we suggest a series of criteria for substrate validation.During the past decade protein degradation has become increasingly recognized as a critical mechanism by which cells regulate a number of fundamental processes (reviewed in references 37, 57, and 59). Degradation frequently involves one of a variety of E3 ubiquitin ligase complexes in which a substrate recognition component introduces the target protein for ubiquitination and subsequent degradation by proteasomes (reviewed in reference 59). Several types of these complexes involve a member of the cullin family (reviewed in reference 59), and a considerable amount of information is known about those containing Cul2 or Cul5. In these cases the substrate recognition module is linked via elongins B and C to a subcomplex containing Cul2 or Cul5 and the RING protein Rbx1 (34, 58). This complex interacts with an E2 conjugating enzyme, often either Cdc34 or Ubc5, to conjugate ubiquitin chains to the substrate (44). With both Cul2- and Cul5-based complexes interaction with elongins B and C occurs via a single BC box sequence (42). The presence of either Cul2 or Cul5 is generally determined through the presence in the substrate recognition protein of specific Cul2- or Cul5-box sequences (35).Many viruses have evolved to encode products that inhibit cellular E3 ligases to protect important viral or cellular species or, in some cases, that highjack these cellular complexes to target key substrates for degradation, including components of cellular host defenses, to facilitate the infectious cycle (reviewed in reference 4). These strategies are quite common among the small DNA tumor viruses (7), and one of the most studied examples is the complex formed by the human adenovirus E4orf6 and E1B55K proteins. These proteins have been known for some time to interact (69) and to reduce the levels of the p53 tumor suppressor in infected cells (14, 47, 48, 62, 72, 73). In addition, they were shown to function in concert to block nuclear export of cellular mRNAs late in infection (2, 6, 29, 60) and to enhance the selective export of late viral mRNAs (2, 26, 29, 60, 78). Our group showed that the human adenovirus serotype 5 (Ad5) E4orf6 product interacts with several proteins (13), including components of what was at the time a unique Cul5-based E3 ubiquitin ligase containing elongins B and C and Rbx1 that degrades p53 (61). Curiously, Ad5 E4orf6 contains three BC boxes that we believe make it highly efficient in highjacking cellular elongin B/C complexes (8, 17, 41). The mechanism of selective recruitment of Cul5 by the Ad5 complex remains unknown as E4orf6 lacks a Cul5-box (17, 41). E1B55K seems to function as the substrate recognition module and, of considerable interest, both its association with E4orf6 and induction of selective late viral mRNA transport was found to depend on formation of the E3 ubiquitin ligase complex, suggesting that additional degradation substrates must exist (8, 9). This idea is not surprising since viruses, especially the small DNA tumor viruses, often evolve gene products that target multiple critical cellular pathways (32). In fact two additional E1B55K-binding substrates have now been identified, Mre11 from the MRN DNA repair complex (8, 75), and DNA ligase IV (3), the degradation of which prevent formation of viral genome concatemers, thus enhancing packaging of progeny DNA. Degradation of p53 has been suggested to promote enhanced progeny virus production by preventing the early apoptotic death of infected cells due to the stabilization of p53 by the viral E1A products (reviewed in reference 66). Nevertheless, degradation of these substrates seems unlikely to explain the observed effects on mRNA transport, suggesting that still more substrates remain to be identified. Although the studies described in the present report were in part launched to identify such substrates, as will become clear below, these targets remain to be identified.In an attempt to identify new substrates of the Ad5 E4orf6/E1B55K E3 ubiquitin ligase complex, a proteomics-based approach was initiated involving two-dimensional difference gel electrophoresis (2D-DIGE) analysis and subsequent mass spectrometry. As is well known, this technique has the advantage of improved sensitivity and accuracy provided by its ability to separate samples under two different conditions on a single gel together with a reference sample, thus reducing significantly the analytical coefficient of variation. It allows the quantification of differentially abundant proteins in complex biological samples, providing a tool to detect decreases in the levels of proteins in the cell due to targeted proteolytic degradation. We report here our attempts to identify substrates of the Ad5 E4orf6/E1B55K complex by comparing the proteomes of human non-small cell lung carcinoma H1299 cells expressing, by means of adenovirus vectors, both E1B55K and E4orf6 proteins or E4orf6 protein alone. Ten candidate proteins were identified, most having functions seemingly unrelated to our current understanding of the roles of the E4orf6/E1B55K complex. At least three showed promising features characteristic of substrates, and one has now been confirmed in a parallel study to be a bone fide E4orf6/E1B55K substrate (20). We suggest that this approach could be utilized to identify candidate substrates, among relatively high abundance proteins, that are degraded by other viral cullin-based E3 ubiquitin ligase complexes.