Major Human Cytomegalovirus Structural Protein pp65 (ppUL83) Prevents Interferon Response Factor 3 Activation in the Interferon Response

Viruses and cells.

Primary human foreskin fibroblasts were grown and maintained as described (16). CMV strain AD169varDE (wild type [wt]) and RVAd65 (pp65 mutant virus derived from AD169varDE, a German variant of this widely used AD169 strain) (59), as well as strains AD169varATCC (American Type Culture Collection) (64), TownevarRIT3 (64), and Toledo (passage 10) were propagated, purified, and plaque assayed in HFs in complete medium as described (16). We employed a high MOI (4 PFU/cell) to give uniform infection levels of cells. RVAd65 and AD169varDE were confirmed to replicate to equivalent levels, as described when the mutant was first reported (59). Infections were carried out as previously described (16) with a parallel set of cultures evaluated as infection controls. RVAd65 and AD169varDE infected >98% of cells at an MOI of 4. Virus-free infected cell supernatants (wt or mutant-infected) collected at 4 or 6 h postinfection (hpi) after exposure to a high MOI inoculum failed to induce cellular gene expression or to interfere with replication of vesicular stomatitis virus (data not shown), consistent with previous observations (55). IFN-β is induced between 8 and 16 hpi at low MOIs with CMV (9, 55) but is not induced under high MOI conditions. UV inactivation, reducing plaque formation by >99.9999%, was carried out with 160 mJ/cm² in a UV 1800 Stratalinker (Stratagene, La Jolla, Calif.) for 10 min with virus at a concentration of 4 × 10⁷ PFU/ml in 4 ml. Inactivated virus was held for 1 h and diluted 10-fold in complete medium prior to use. CMV strain Towne pp65-expressing HFs transduced with MSCVpp65 or control cells transduced with empty LNCX vector were prepared and selected as described previously (66). pp65 expression was monitored by pp65 immunofluorescence staining, and the experiments shown in the present study were performed when pp65 was expressed in >95% of G418-selected HFs. PBMC from a CMV-seronegative adult donor were prepared by using Lymphoprep (Axis-Shield, Oslo, Norway) and suspended in complete medium prior to infection.

cDNA microarrays.

Polyadenylated RNA from 10⁷ confluent HFs was purified by using Oligotex (QIAGEN, Valencia, Calif.) from Trizol (Invitrogen)-extracted total RNA. This RNA was prepared, reverse transcribed, labeled with Cy3-dUTP or Cy5-dUTP (Amersham, Little Chalfont, Buckinghamshire, United Kingdom) by random primed synthesis with DNA pol I Klenow (Amersham Life Science, Inc., Cleveland, Ohio), and hybridized to spotted, human cDNA microarrays as previously described (19). Sequence-verified human cDNA microarrays (HE and HG series, 31,000 spots; HD51 series, 17,000 spots) were produced at Stanford. Images were collected by using a GenePix 4000B microarray scanner, manually flagged to eliminate poor spots and analyzed by GenePix Pro 2.0 (Axon, Union City, Calif.) in combination with established methods (18, 25, 62). The data are available from the Stanford Microarray Database (SMD) (http://genome-www5.stanford.edu/MicroArray/SMD) Data were filtered for intensity (>150 pixels) and for regression correlation (>0.6) to remove dim spots. Cluster analysis was performed on a randomized seed of data (20). Background variability (normalized ratio of ± 1.4) was determined by hybridizing Cy3- and Cy5-cDNA from the same source to arrays (HE series). Significance analysis of microarrays (SAM) software was used as described in a one-class analysis (71). Expressed sequence tags (ESTs) and unnamed genes were evaluated by using BLAST (National Center for Biotechnology Information).

RNA blot analysis.

Total RNA (5 μg) was isolated and resolved by electrophoresis through 1% agarose, blotted, and hybridized by using biotinylated probes (interleukin-6 [IL-6], WARS, GBP-1, TAP-1, Mx1 [p78], ISG20, RANTES, Mip-1α, cig5, β-actin) derived from HF cell cDNA as previously described (16), with β-actin as a loading control with NorthernMax (Ambion). When used, actinomycin D (Sigma, St. Louis, Mo.) was added at a concentration of 5 μg/ml for 2 h at 4 hpi.

Antibodies, immunofluorescence microscopy, and immunoblot analysis.

Cells on 13-mm glass coverslips were fixed in ice-cold methanol for 20 min, rinsed, and blocked with phosphate-buffered saline containing 10% bovine serum albumin (Sigma). Mouse monoclonal antibodies to IRF-3 (SL12.1; BD Pharmingen, San Diego, Calif.) and pp65 (28-19 from William Britt, University of Alabama) and rabbit polyclonal antibodies to NF-κB (p50 subunit H-119), IRF-7 (H-246), STAT-1 (E-23), STAT-3 (H-190) (all from Santa Cruz Biotechnology, Santa Cruz, Calif.) were employed. Rabbit polyclonal antisera to IRF-3 from Santa Cruz Biotechnology used for localization in other reports (9, 12) exhibited only nonspecific immunofluorescence in our hands. Primary and Texas Red or fluorescein isothiocyanate-conjugated secondary antibodies (Vector Laboratories, Burlingame, Calif.) were in 2% bovine serum albumin. Nuclei were counterstained with Hoechst 44432 (Molecular Probes, Eugene, Oreg.). Localization was evaluated by epifluorescent microscopy. For immunoblot analysis, cells or isolated nuclei (4 × 10⁵ cells or nuclei) were electrophoretically separated in 7.5 or 10% denaturing polyacrylamide gels, transferred, and probed with SL12.1 as described (67). Horseradish peroxidase-conjugated anti-mouse antibody (DakoCytomation Denmark A/S, Glostrup, Denmark) and an ECL Western blotting kit (Amersham) were employed. Nuclei were isolated following cell lysis in 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1% NP-40, and 0.5% deoxycholate. The IRF-3 phosphorylation state was determined at 4 hpi in buffer after collecting cells in the presence of 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10 mM EDTA, 1%NP-40, 1 mM sodium orthovanadate, 30 mM NaF, 0.1 mg of leupeptin/mg, and 1 mM phenylmethylsulfonyl fluoride.

IRF-3 nuclear translocation assay.

Cells in 12-well culture dishes were exposed to Superfect (QIAGEN)-loaded plasmid pcDNA3-EYFP (Clontech, Palo Alto, Calif.) for 2 h, according to the manufacturer's method. After treatment, the complete medium was added, and data were collected at 4 h (49).

pp65 modulation of host cell transcriptome.

Spotted human cDNA microarrays were used to assess the impact of CMV infection in the presence or absence of the major virion protein, pp65, on host cell transcript levels. We focused on very early times (1, 2, 3, and 4 hpi) following initial viral binding and penetration and employed a pair of viruses whose virions differ only in the presence of pp65 (UL83). Our experimental design employed two complementary microarray analyses to unveil the specific impact of pp65: (i) direct (or type I) analysis in which cDNA generated from the pp65 mutant was compared to cDNA from wt virus-infected cells, and (ii) indirect (or type II) analysis in which either mutant or wt virus-infected cell cDNA was compared to mock-infected cell cDNA collected at the same time point (19). These microarray approaches utilized replicates and cDNA synthesized directly from cellular mRNA template without any amplification. Data were subjected to a twofold change cutoff with clustering of data (20) as well as the stringent statistical criteria established by SAM (71).

Initially, we investigated the global impact of CMV strain AD169varDE infection on permissive HFs. Host transcript abundance in AD169varDE-infected cells was compared to mock-infected reference samples collected at the same time points and hybridized to 31,000-spot cDNA arrays. We observed changes in the pattern of gene expression at early times after infection that were consistent with previous reports (13, 63, 80, 81). When evaluated by using a twofold cutoff, 652 cDNAs were differentially expressed, with most being induced (584 cDNAs; 90% of spots) by virus infection at the 3 and/or 4 hpi time points. SAM analysis scored over 10,000 cDNAs as significantly altered by viral infection over this 4-h time course. Again, most (80%) were more abundant in virus-infected cells (http://genome-www5.stanford.edu/). Importantly, 565 (97%) virus-induced cDNAs identified by using a twofold cutoff criterion were included in this set. When genes represented by the cDNAs were grouped based on characteristics of regulation, an overwhelming majority (70%) of the 584 cDNAs represented IFN response-regulated genes, consistent with previous reports studying smaller data sets (13, 63, 80, 81). The extended lists of candidate cellular genes that respond to CMV infection can be found in the supplemental data (http://genome-www5.stanford.edu/MicroArray/SMD/).

To investigate the influence of pp65 on CMV-mediated induction of cellular gene expression, we performed a direct microarray comparison of host transcript abundance in pp65-mutant and wt CMV-infected cells over a 4-h time course. When evaluated with a twofold cutoff, 220 cDNAs were differentially expressed at one time point or more over this time course. Cluster analysis placed these changes into two distinct categories that did not overlap (Fig. 1). A total of 101 cDNAs were induced more strongly by pp65 mutant virus (Fig. 1A, yellow), and 119 cDNAs were induced more strongly by wt virus (Fig. 1B, blue). The most dramatic differences in magnitude and numbers were observed at either 3 or 4 hpi. For example, at 4 hpi, 78 of the 101 cDNAs scored as more highly induced in pp65 mutant-infected cells, while 59 of the 119 cDNAs scored as more highly induced in wt virus-infected cells. Sixty-one annotated genes plus an additional 31 ESTs were represented in the set of 101 cDNAs that were more highly induced during mutant virus infection. Eighty annotated genes plus 36 ESTs were represented within the set of 119 cDNAs that were more strongly induced by wt virus infection with three cDNAs (CREM, PPIF, and ANLN) spotted in duplicate and giving similar patterns. When we sought common characteristics of host genes whose expression was coordinately regulated, 35 of the 61 genes (57%) induced more strongly by the mutant virus were identified as IFN-response genes (17, 26, 70). We could not find any recognizable pattern for the 80 genes induced more dramatically by wt virus infection, and so nothing more was done with this set. When the data from the time course were subjected to SAM analysis, which evaluates statistical consistency independent of the factor of change, 1,646 cDNAs were judged to be more strongly induced by mutant virus, and 1,604 cDNAs were induced more strongly by wt virus (http://genome-www5.stanford.edu/). The percentage of IFN-regulated genes increased as the stringency of cutoff was raised such that 92% of the genes meeting both the SAM significance and 2.0-fold cutoff criteria were in the IFN response group. Thus, our evaluation made use of several independent experimental strategies to show that pp65 mutant virus induced a response similar to that induced by a matched wt virus; however, the IFN-like component of this response was considerably stronger. When additional selection criteria were applied to this data set (twofold cutoff in two arrays), 19 cDNAs corresponding to 18 genes emerged as most strongly induced in pp65 mutant-infected cells (ISG20 was duplicated) (Fig. 2). All of these genes had been characterized as IFN-response genes (17, 70). We found similar results on smaller microarrays comparing mutant- and wt-infected cell cDNAs and also confirmed that mutant virus induced global IFN-like changes that were stronger than wt virus compared to mock infection at each time point (http://genome-www5.stanford.edu/). To illustrate how well this indirect analysis agreed with the direct analysis described above, Fig. 2 shows the primary data from both types of assay for the 18 genes that were most strongly induced by mutant virus. Thus, both direct and indirect microarray comparisons proved very powerful in showing the stronger stimulatory impact of the pp65 mutant over that of a matched pp65-expressing virus.

We confirmed the microarray data by RNA blot analysis of nine genes that were induced more dramatically by mutant than by wt virus (Fig. 3). To determine whether the difference in gene expression patterns reflected the delivery of virion pp65, we followed RNA levels of three genes (WARS, IL-6, and GBP-1) after exposure to UV-inactivated viruses. Both replication-competent and inactivated viruses induced similar RNA levels at 4 hpi (Fig. 3B) as expected (11, 13, 63, 76, 77, 80, 81), with the response to the pp65 mutant much stronger. These data implicated input virion pp65 as a modulator of cellular gene expression immediately following virus binding and penetration. We also found that newly synthesized pp65 made late during CMV infection (48 hpi) altered the expression of IFN-regulated genes GBP-1 and IL-6, mirroring the observations at the earlier time points (data not shown). To determine whether levels of pp65 in virions influenced the response, we investigated the behavior of two independently generated viral ie2 mutants, IE2 86ΔSX-EGFP, an AD169varATCC-based deletion mutant (56) and RC2933, a TownevarRIT3-based virus (J. Xu, D. Formankova, and E. S. Mocarski, unpublished data), that both fail to express late IE2 gene products (reference 56 and data not shown). Infection with either of these ie2 mutants resulted in a significantly reduced level of pp65 made late during the virus replication cycle (56), and this corresponded to a dramatic reduction in the incorporation of pp65 into virus particles (data not shown). Both of these mutants induced IL-6 and GBP-1 transcripts to higher levels than did the IE2 86ΔSX-EGFP revertant (Fig. 3C) or other control viruses (data not shown), which is consistent with the hypothesis that the amount of pp65 incorporated into virions influenced the IFN-like response to virus infection.

pp65 control of the cytoplasmic localization and hypophosphorylation state of IRF-3.

Binding sites for transcription factors (NF-κB, STATs, and IRFs) known to play crucial roles in the induction of IFN-response (54, 68) were contained in the promoter regions of the genes that responded differentially to wt and mutant virus infection. Upon activation, all of these factors translocate to and accumulate in the cell nucleus. We investigated the localization of IRF-3, NF-κB, STAT-1, STAT-3, and IRF-7 at 4 hpi under conditions leading to a uniform infection of >98% of cells (MOI of 4). Under these conditions an overwhelming majority of wt virus-infected HFs were pp65 antigen positive shortly after virus adsorption (data not shown), as expected from published reports (58). There was a dramatic difference in the localization of IRF-3 in wt (Fig. 4A and D) and mutant (Fig. 4B and E) virus-infected cells. IRF-3 remained localized to the cytoplasm following infection with wt virus, a pattern similar to mock-infected HFs (Fig. 4C and F), but localized to the nucleus in pp65 mutant virus-infected cells. Similar IRF-3 patterns were observed at 8 hpi in HFs (Fig. 4G to L) as well as in human PBMC at 4 hpi (Fig. 4M to P). In contrast to the differential impact we have seen on IRF-3 localization patterns, NF-κB translocated to the nucleus by 4 hpi and remained nuclear through 8 hpi, irrespective of whether mutant or wt virus was used (Fig. 5), results that are at variance with a recent report (12). In our hands, IRF-7 expression was not detectable by immunofluorescence analysis in either uninfected or infected cells (data not shown).

In contrast to the activation pattern of NF-κB, both STAT-1 and STAT-3 remained cytoplasmic at 4 hpi with either wt or mutant virus (data not shown). The failure of STAT-1 to localize to the nucleus is consistent with earlier work that showed that STAT-1 remains inactive at early and late times during infection (41, 47) but is at variance with a recent report (12) showing that this transcription factor localizes to the nucleus shortly after exposure to virus.

Overall, IRF-3 levels detected by immunoblotting were similar in wt virus-, mutant virus-, and mock-infected cells (Fig. 4Q). Consistent with the immunofluorescence results, nuclei fractionated from mutant virus-infected HFs contained higher levels of IRF-3 than nuclei from either wt- or mock-infected cells (Fig. 4R). Further analysis in the presence of phosphatase inhibitors revealed a slower migrating form of IRF-3 (Fig. 4S), suggesting a difference in phosphorylation states (72). Thus, pp65 appeared to counteract the hyperphosphorylation of IRF-3 that is associated with nuclear accumulation (54, 68, 73). Consistent with the IRF-3 results shown for AD169varDE, other pp65-expressing strains (AD169varATCC, TownevarRIT3, and Toledo) showed a cytoplasmic IRF-3 localization at 4 h after virus infection at high MOIs (Fig. 6). This result suggests that our observations were not dependent upon a single strain or strain variant. Given the behavior of CMV strain Toledo, which is fully virulent and expresses a complete set of CMV gene products (43), we believe the prevention of IRF-3 activation is likely to represent a normal activity of natural CMV strains.

To determine whether RNA degradation played any role in the IRF-3-dependent activation of gene expression, we inhibited transcription with actinomycin D and compared RNA levels in pp65 mutant- and wt virus-infected cells. Both GBP-1 and IL-6 mRNAs were stable during a 2-h treatment at 4 hpi with no significant or differential effect on overall RNA stability (Fig. 4T; data not shown). Thus, it does not appear that CMV pp65 acts in a manner analogous to the virion host shut-off function (vhs/UL41) of herpes simplex virus (36).

To determine whether pp65 was itself sufficient to prevent IRF-3 translocation independent of other virion structural proteins, IRF-3 localization was evaluated in retrovirus-transduced HFs that stably expressed pp65 (Fig. 7). We chose retrovirus vector-transduced cells for this study because this strategy does not in itself activate an IFN-like response that might interfere with interpretation of any experiments. We employed treatment with plasmid pcDNA3-EYFP-loaded Superfect liposomes to induce IRF-3 nuclear translocation in HFs independent of CMV infection, a method that activates IRF-3 (49), and compared empty LNCX vector-transduced HFs to pp65-transduced HFs. While nontransduced HFs (Fig. 7A and D) and HFs stably transduced with an empty LNCX vector (data not shown) supported nuclear accumulation of IRF-3, pp65-expressing HFs (Fig. 7G) exhibited cytoplasmic localization of IRF-3 following exposure to DNA-loaded liposomes (Fig. 7B and E). As expected, IRF-3 also remained predominantly cytoplasmic in pp65 HFs infected with pp65 mutant virus (Fig. 7C and F). Thus, pp65 alone was sufficient to prevent IRF-3 activation by a nonviral inducer.