This Article Abstract Full Text (PDF) Supplemental material Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hertel, L. Articles by Mocarski, E. S. Search for Related Content PubMed PubMed Citation Articles by Hertel, L. Articles by Mocarski, E. S.

Global Analysis of Host Cell Gene Expression Late during Cytomegalovirus Infection Reveals Extensive Dysregulation of Cell Cycle Gene Expression and Induction of Pseudomitosis Independent of US28 Function

Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, California

Received 26 March 2004/ Accepted 24 May 2004

	ABSTRACT

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Replication of human cytomegalovirus (CMV) depends on host cell gene products working in conjunction with viral functions and leads to a dramatic dysregulation of cell cycle gene expression. Comprehensive transcriptional profiling was used to identify pathways most dramatically modulated by CMV at late times during infection and to determine the extent to which expression of the viral chemokine receptor US28 contributed to modulating cellular gene expression. Cells infected with the AD169 strain of virus or a fully replication competent US28-deficient derivative (RV101) were profiled throughout the late phase of infection (50, 72, and 98 h postinfection). Although sensitive statistical analysis showed striking global changes in transcript levels in infected cells compared to uninfected cells, the expression of US28 did not contribute to these alterations. CMV infection resulted in lower levels of transcripts encoding cytoskeletal, extracellular matrix, and adhesion proteins, together with small GTPases and apoptosis regulators, and in higher levels of transcripts encoding cell cycle, DNA replication, energy production, and inflammation-related gene products. Surprisingly, a large number of cellular transcripts encoding mitosis-related proteins were upmodulated at late times in infection, and these were associated with the formation of abnormal mitotic spindles and the appearance of pseudomitotic cells. These data extend our understanding of how broadly CMV alters the regulation of host cell cycle gene products and highlight the establishment of a mitosis-like environment in the absence of cellular DNA replication as important for viral replication and maturation.

	INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Human cytomegalovirus (CMV) infection has a dramatic impact on the cell that starts immediately after infection (4) and continues through late times (34, 67). The replication cycle follows a temporal cascade of events that depends upon both viral and host cell functions. Viral DNA replication begins between 14 and 24 h postinfection (hpi), and release of progeny starts between 36 and 48 hpi, reaching maximal levels between 72 and 96 hpi (67). This process causes profound changes in host cell shape, metabolism, and gene transcription, components of which are suspected to be critical for efficient replication. Previous studies in primary fibroblasts have revealed the global impact of viral infection on signaling and transcriptional changes that start as early as 15 min and last as long as 48 hpi (4, 16, 50, 51, 85, 112, 113). These studies have largely focused on the immediate impact of the virus on cells and have revealed a dramatic upmodulation of cellular inflammatory and immune gene expression due to virus binding and penetration. Based on this work, selected cellular signaling events (51, 103) and cellular proteins (13, 89, 114) have been implicated as important regulators of infection. There has been a less concerted effort to understand the global impact of CMV infection at late times during infection (16), despite the fact that maximal modulation would be expected at late times. Also, remarkably little information has been presented on the contribution of virus-encoded signaling proteins that are expressed at late times, such as the CMV chemokine receptor US28.

CMV encodes at least four apparent seven transmembrane-spanning proteins (21, 42), pUL33, pUL78, pUS27, and pUS28, one of which (pUS28) is a G-protein-coupled receptor that binds a wide range of CC chemokines (9, 38, 56, 72, 100), as well as fractalkine/CX₃CL1 (55). pUS28-mediated signaling is both chemokine ligand dependent (38) and constitutive (19, 101) and stimulates mitogen-activated protein kinase extracellular-signal-regulated kinase 2, focal adhesion kinase 1, and Src (9, 19, 91, 101). pUS28-mediated signaling occurs in CMV-infected cells (9, 98), as well as in US28-transfected cells. US28 transcripts are readily detected by 24 hpi and continue to rise through late times when pUS28 scavenges chemokines from the infected cell culture fluid (12, 98). Although many of the US28 mutant viruses exhibit modest replication defects, US28 itself is completely dispensable for replication in cultured fibroblasts (12, 98), as shown through studies on one viral mutant in particular, RV101 (12). The phenotypic consequences of signaling through this receptor during infection in fibroblasts have not been investigated.

CMV infection of resting primary fibroblasts dysregulates expression of several genes encoding cell cycle regulators (34), such as the G₂/M cyclin B (26); cell cycle checkpoint proteins, such as p53 and pRb (49); and DNA replication effectors, including components and regulators of the prereplication complex (10, 107).Through the induction of these changes, CMV has long appeared to stimulate the generation of an intracellular environment similar to S phase based initially on cellular activation (4) while simultaneously inhibiting cellular DNA synthesis in a manner analogous to a G₁/S block (34, 67). The execution of these strategies presumably allows the virus to maximize the availability of cellular functions required for successful replication of the viral genome, to eliminate the competition from the cellular genome, and to avoid the onset of apoptosis.

We used cDNA microarrays to evaluate the impact of CMV infection, as well as the contribution of US28 expression, on cellular gene transcription in fibroblasts at late times postinfection. We used a replication-competent US28 mutant virus, RV101 (12), to avoid contributions from adventitious mutations that are a common occurrence in CMV mutants (67). The levels of a large number of transcripts encoding functions involved in a wide range of cellular pathways were altered by infection, but none of these changes could be ascribed to US28 expression. The most remarkable findings revealed a substantial increase in the expression levels of mitochondrial genes, a considerable alteration in expression levels of multiple GTPase family members, and an intense dysregulation of cell cycle pathways that normally control mitosis. These latter changes were linked to the appearance of pathological mitosis in culture and were likely part of a virus-induced strategy to enhance viral genome replication and virion maturation.

(These results were presented at the 9th International Cytomegalovirus Workshop and the First International Betaherpesvirus Workshop in Maastricht, The Netherlands, on 20 to 25 May 2003.)

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Cells, viruses, and microarrays. Primary human foreskin fibroblasts (HFs) were cultured as described previously (23) and used between passage 15 and 17 postisolation. The human CMV strain AD169 obtained from the American Type Culture Collection (ATCC) has been denoted AD169varATCC (AD) to distinguish it from other variants of this strain in current use (41, 87). The US27/US28 deletion mutant virus RV101 (RV) was made by using AD169varATCC by T. R. Jones (Wyeth-Ayerst Research, Pearl River, N.Y.) and exhibits growth characteristics indistinguishable from parental virus (12) that we independently confirmed for the present study (data not shown). Microarrays were purchased from the Stanford Functional Genomics Facility (http://www.microarray.org/sfgf/jsp/home.jsp). About 85% of the spots printed on the polylysine-coated glass microscope slides were from the I.M.A.G.E. consortium clones from the Research Genetics Sequence Verified clone set, 10% were from the Cancer Genome Anatomy Project clone set (http://www.ncbi.nlm.nih.gov/ncicgap), and 5% were from a set of Methanococcus jannaschii controls. The batch of arrays (SHDB) used in the present study contained 41,792 total spots (39,781 unique and 2,011 repeated spots) representing 29,593 unique genes. After purchase, the microarrays were postprocessed as described previously (http://derisilab.ucsf.edu/).

Cell infections. CMV was propagated at a multiplicity of infection (MOI) of 0.01 in confluent HFs and purified as described previously (23). For microarray analysis, HF monolayers plated at a density of 3 x 10⁴ cells/cm² in 850-cm² roller bottles were infected at confluency (day 5 postseeding) with purified virions (23) at an MOI of 10. After virus adsorption for 1 h at 37°C in 5% CO₂, the inoculum was removed, and the cells were washed twice with culture medium prior to the addition of fresh medium. At 50, 72, and 98 hpi, infected monolayers were harvested, snap-frozen, and stored at –80°C until the time course was completed. Uninfected confluent HF cultures were used as a reference. For all other analyses described here, HF monolayers were plated at a density of 5 x 10⁴ cells/cm² on coverslips in 24-well plates 3 days before infection with purified AD virions at an MOI of 10. After virus adsorption for 1 h, cells were washed twice prior to the addition of fresh medium. For drug treatment, medium plus 300 µg of phosphonoformic acid (PFA; Sigma, St. Louis, Mo.)/ml or medium plus 1 mM or 10 mM hydroxyurea (HU; Sigma) was added to the washed cells. Virus titers in the supernatant of wells treated with each of the drugs were determined by plaque assay and were consistent with previous reports (6, 99).

RNA isolation and microarray hybridization. mRNA was extracted by using the FastTrack 2.0 mRNA Isolation Kit (Invitrogen, Carlsbad, Calif.). Only RNA preparations reaching optical density at 260/280 nm (OD_260/280) ratio values of 2.0 were used. Labeled cDNA populations were synthesized from 2 µg of mRNA per sample by reverse transcription and simultaneous incorporation of Cy3-dUTP (reference sample, green) or Cy5-dUTP (infected samples, red) (Amersham Biosciences, Piscataway, N.J.) with an oligo-dT₂₀ primer and the Superscript II Reverse Transcriptase (Invitrogen). Reference and sample cDNA populations were mixed, and unincorporated nucleotides were removed by using a CyScribe GFX purification kit (Amersham Biosciences). Then, 5 µg of human Cot-1 DNA (Invitrogen), 20 µg of poly(A) RNA (Sigma), 20 µg of yeast tRNA (Invitrogen), 3.4x SSC (1x SSC is 0.15 M NaCl plus 0.015 M sodium citrate), and 0.3% sodium dodecyl sulfate were added to the probe prior to denaturation by heating at 100°C for 2 min and application on the arrays. Hybridization of two arrays per virus per time point (total of 12 arrays) were performed at 65°C for 16 to 18 h in a custom slide chamber with humidity maintained by a small reservoir of 3x SSC. The arrays were then washed as described previously (29) and scanned with a Gene Pix 4000A Scanner (Axon Instruments, Foster City, Calif.).

Microarray data analysis. Array images were analyzed with the Gene Pix Pro 4.0 software and unreliable spots identified by visual inspection were marked with a "flag." After submission to the Stanford Microarray Database (SMD; http://genome-www.stanford.edu/microarray/), infected/uninfected ratio values were log₂ transformed, normalized by using the default normalization protocol of the SMD, and retrieved by spot number (so that no averaging of the ratio values from duplicate spots was performed). To eliminate spots of poor quality from subsequent analyses, spot data were filtered by applying three criteria: no flag, brightness (channel 1 [Cy-3] net mean and channel 2 [Cy-5] normalized net mean intensities) of 150 fluorescence units, and even distribution of color brightness across the whole spot area (regression correlation values of 0.6). Statistical analysis was performed by applying the one- or two-class (unpaired) response type of Significance Analysis of Microarrays (SAM), version 1.20 (http://www-stat.stanford.edu/tibs/SAM/) (96), to three sets of four arrays each (at 50, 72, or 98 hpi) or one set of 10 arrays (50, 72, and 98 [50-72-98] hpi). The number of permutations was set to 24 (4-array sets) or to 300 (10-array set), and the random number seed was set to 94332998 (one class) or to 91797601 (two classes, unpaired). The lists of SAM significant spots generated by the one-class analysis of the 50-, 72-, and 98-hpi array sets were intersected with the list generated from the 50-72-98-hpi array set by using the I.M.A.G.E. number as spot identifier. The identity of SAM significant spots was obtained through the SOURCE database (http://source.stanford.edu; time of the analysis, April 2003) (25), and gene grouping in functional categories was performed according to information retrieved from Online Mendelian Inheritance in Man (OMIM) and Medline databases. Genes belonging to the same cellular subcompartments or functional pathways were assigned to a specific class within each category. This classification process brought genes having similar roles into close proximity within each category's Excel spreadsheet (see information file S6 in the supplemental material), allowing us to immediately appreciate if they were part of extensive cellular pathways. Functional maps were generated by either modifying existing maps on the Gene Microarray Pathway Profiler (GenMAPP) website (http://www.genmapp.org/) or by creating new maps by using GenMAPP software (24) in combination with the Kyoto Encyclopedia of Genes and Genomes database (KEGG, http://www.genome.ad.jp/kegg/) and literature searches. Microarray images and raw data are available on the SMD website, SAM analysis Excel files and functional grouping Excel files are included as six information files in the supplemental material.

Immunofluorescence analysis. For tubulin 1 (tub1) and immediate-early protein 1 and 2 (IE1/IE2) double staining, cells were fixed in cold methanol for 5 min at –20°C, permeabilized with 0.1% Triton X-100 in phosphate-buffered saline (PBS) for 20 min on ice, incubated in blocking buffer (1% fetal calf serum in PBS) for 10 min at room temperature, and stained with an anti-tub1 monoclonal antibody (MAb; 1:10,000; clone GTU-88, Sigma), followed by a Texas red-conjugated goat anti-mouse antibody (1:100; Vector Laboratories, San Bruno, Calif.). After three washes with PBS, cells were incubated with normal mouse immunoglobulin G (1:10; Caltag, Burlingame, Calif.) for 30 min at room temperature before being stained with a fluorescein isothiocyanate (FITC)-conjugated anti-IE1/IE2 antibody (1:800; MAB810F; Chemicon, Temecula, Calif.). Cellular DNA was highlighted with Hoechst 44432 (Molecular Probes, Eugene, Oreg.). For ppUL44, pp150, and pUL43 staining, infected cells were fixed with 1% paraformaldehyde in PBS for 15 min at room temperature, permeabilized, and blocked as described above and then stained with MAbs to ppUL44 (1:500; Goodwin Institute, Plantation, Fla.), pp150 (MAb 36-14), or pUL43 (MAb 20392), followed by an FITC-conjugated goat anti-mouse antibody (1:100; Vector Laboratories). Cellular DNA was highlighted with propidium iodide (Molecular Probes). For tub1 and bromodeoxyuridine (BrdU) double staining, cells were labeled with 50 µM BrdU (Sigma) for 1 h at 37°C with 5% CO₂, washed with PBS, fixed with 1% paraformaldehyde for 30 min at room temperature, and incubated with 2 M HCl for 1 h at 37°C. Cells were then washed twice with borate buffer (pH 8.5) and three times with PBS, permeabilized, blocked, and stained with anti-tub1 and Texas red anti-mouse antibodies as before. After being washed, the cells were labeled with a FITC-conjugated anti-BrdU antibody (1:100; Molecular Probes) for 1 h at room temperature. Coverslips were mounted with FluoroGuard Antifade Reagent (Bio-Rad, Hercules, Calif.) before being analyzed on an Olympus BX60 epifluorescence microscope equipped with x60 or x100 phase-contrast objective lenses. Images were collected with a Hamamatsu ORCA-100 digital camera and Image Pro Plus 4.0 software (MediaCybernetics, Silver Spring, Md.). Digitized images were stored, electronically colorized, and overlaid for evaluation of two- and three-color experiments.

Transmission electron microscopy. Cells were fixed in 2% glutaraldehyde in PBS, postfixed with 1% osmium tetroxide for 1 h, blocked with 1% aqueous uranyl acetate, dehydrated through graded ethanol steps, and embedded in Poly/Bed (Polysciences). Thin sections (70 nm) were then prepared, stained with 1% uranyl acetate and 1% lead citrate, and viewed on a Phillips CM12 microscope.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Trends in cellular gene expression at late times during CMV infection and impact of viral chemokine receptor expression. To identify cellular genes modulated by CMV at late times postinfection, we compared the transcriptional profiles of HFs infected with either wild-type CMV strain AD169varATCC (AD) or RV101 (RV), a mutant virus derived from AD169varATCC exhibiting full replication potential despite the complete absence of US27 and US28 expression (12). We used this particular mutant to avoid potential contribution of adventitious mutations that appear to be carried in US28 mutants that exhibit growth defects. Transcriptional profiling was performed according to the scheme outlined in Fig. 1A. Cy5-labeled cDNAs from HFs infected with either AD or RV for 50, 72, or 98 h were combined with Cy3-labeled cDNA from confluent uninfected HFs used as a reference (29) and hybridized to 12 replicate 42,000 spot human cDNA arrays. After normalization of the spot ratio values, usable data were selected according to an established set of filtering criteria (84). The percentage of qualified spots identified on each array ranged from 44 to 83%, with a mean of 65% ± 11% (Table 1). The data from replicate arrays were organized into four groups to facilitate analysis, three with the 4 arrays at 50, 72, or 98 hpi and one with the 12 arrays from the combined time points (50-72-98 hpi). Each group contained a common set of spots that passed filtering criteria on all arrays contained in that group. These sets consisted of 20,123 spots from the 50-hpi time point (48% of total spots), 14,856 (35%) spots from the 72-hpi time point, 21,940 (52%) spots from the 98 hpi-time point, and 11,942 (29%) spots from the combined 50-72-98-hpi time points. In order to maximize the usable data in the analysis, two arrays containing <50% qualified spots (AD 72 hpi 1 and RV 72 hpi 1) were excluded from the combined time point group. In this way, 10 replicates composed the 50-72-98-hpi group, and this group had 14,126 (34% of total spots) qualified spots.

View larger version (37K): [in this window] [in a new window]   FIG. 1. (A) Flow chart of microarray data generation and analysis. Cy5-labeled cDNA from infected HFs and Cy3-labeled cDNA from uninfected HFs were mixed and hybridized to replicate 42,000 spot human cDNA arrays to generate data that was normalized and filtered by the SMD and subjected to statistical analysis by SAM software. (B) SAM one-class analysis (left) and SAM two-class (unpaired) analysis (right) of the 14,126 data set from the combined 50-72-98-hpi time points. For each spot, the observed SAM score is plotted against the expected SAM score. The middle line intersecting the origin in each graph represents observed equals expected score values, whereas the additional lines define the upper and lower significance threshold specified by the lowest allowed FDR parameter (96). Significantly regulated spots are shaded gray. Gray spots in the upper right quadrant are upmodulated, and gray spots in the lower left panel are downmodulated. (C) Functional categories (right column) represented in the 1,983 SAM-significant upmodulated (white bars) and 3,675 SAM-significant downmodulated (gray bars) spots from the 50-72-98-hpi data set. The series of bars indicates the distribution of total spots (All spots), unnamed (All unknown spots), and named (All known genes), as well as the distribution of the named genes into functional groups based on their principle role according to the literature. Functional groups include soluble factors and cell surface receptors (Factors Receptor), cytoskeleton-extracellular matrix-adhesion (Cytosk-ECM-Adh), nucleic acid metabolism (Nucl Acid metab), protein metabolism (Protein metab), vesicles and intracellular transport (Vesicles transport), enzymes (Enzyme), transcription factors (Transcr factor), intracellular signaling (Signaling), cell cycle (Cell cycle), other (Other), GTPases (GTPase), lipid metabolism (Lipid metab), immune system (Imm system), and apoptosis (Apoptosis). The categories are ranked according to the total number of spots or genes in each (left column).

Expression levels of US28, a signaling G-protein-coupled receptor, have been shown to be elevated at late times of infection (12). To evaluate the contribution of US28 signaling to the host cell response during the time of maximal expression, we compared the transcriptional profiles of AD and RV virus-infected cells by using SAM two-class (unpaired) analysis. Given the consistency of changes at all late time points in the combined data set, this method was initially applied to the 14,126 spot data set, comparing the results of the five replicate AD virus- and five replicate RV virus-infected cell arrays where >50% of spots were qualified. Figure 1B (right) shows the plot generated by this analysis. The distribution of the data fell along a line representing observed equals expected score values, with few deviations. This analysis showed that genes were not expressed differently in AD and RV virus-infected cells. No FDR values of <43 were available for this analysis, indicating that even spots appearing slightly outside the significance threshold had a relatively high (43%) likelihood of being false (see information file S5, FDR and delta values, in the supplemental material). Thus, we concluded that these spots were unlikely to represent true positives and that cellular gene expression was not perceptively influenced by US27 or US28. A comparison of the transcriptional profiles generated by each virus at each time point, conducted by using two arrays per virus per time point, also failed to reveal any differences (data not shown). It is important to consider that the viruses we compared exhibited equivalent replication in HFs (12) and that previous studies (12, 90, 98) have used some US28 mutant viruses that exhibit a variety of growth defects that appear to be unrelated to the disruption of the US28 gene and US28 signaling (12). Growth defects may have been more important than US28 expression in some of the reported differences ascribed to US28. No matter how our data were analyzed, either with replicates from each time point or total combined AD and RV virus-infected cell arrays, a US28 mutant virus that retained normal growth properties did not impact the cellular transcriptional profile any differently than control virus.

Functional analysis of cellular gene expression at late times of infection. Genes whose expression was altered by viral infection in the combined 50-72-98-hpi data set relative to uninfected HFs were categorized according to functional activity, cellular localization site, or metabolic pathway. According to the SOURCE database (http://source.stanford.edu; see Materials and Methods), 3,210 spots were found to represent 2,227 functionally characterized genes (983 spots were represented more than once on each array), with 2,448 spots representing either expressed sequence tags (ESTs) or genes that had not been annotated. Based on known properties listed in SOURCE, as well as in literature searches, each gene was assigned into one functional category shown in Fig. 1C (see information file S6 in the supplemental material). The five categories with the greatest numbers of genes, soluble factors and cell surface receptors (14%), cytoskeleton-extracellular matrix-adhesion (13%), nucleic acid metabolism (10%), protein metabolism (10%), and vesicle and intracellular transport (9.6%) accounted for more than half of the genes whose expression was altered by infection. Of the 2,227 functionally characterized genes, a smaller percentage (927; 42%) was upmodulated, and a greater percentage (1,300; 58%) was downmodulated. Genes belonging to most of the functional categories showed similar distributions; however, the cytoskeleton-extracellular matrix-adhesion, vesicles and intracellular transport, GTPase, and apoptosis categories contained even greater percentages (68%) of downmodulated genes, and the nucleic acid metabolism, enzyme, cell cycle, and immune system categories contained greater percentages of upmodulated genes (Fig. 1C). Our finding that 2,448 ESTs and 2,227 named genes out of a survey of ca. 14,000 host genes were modulated by CMV infection contrasts with a previous published analysis (16). Although past work focused on early times of infection comparing infectious virus and UV-inactivated virus particles, a dramatic drop in numbers of modulated genes from 900 to 650 was reported to occur between 24 and 48 hpi. Our preliminary analysis had revealed no such drop (S. Watanabe, M. B. Eisen, P. O. Brown, and E. S. Mocarski, Abstr. 24th International Herpesvirus Workshop, Cambridge, abstr. 15.019, 1999). Comparison of previously published 48 hpi data (16) with the analysis reported here is difficult due to differences in the application of statistical analysis.

Relationships between the genes in a category were queried by using OMIM and literature searches. Genes within the same category showing similar transcriptional changes were often found to encode proteins that were functionally related and could be organized into classes within each category (see information file S6, column class). Certain categories (cytoskeleton-extracellular matrix-adhesion, GTPase, enzyme, cell cycle, and nucleic acid metabolism) contained a large proportion of genes in functional relationships that were diagrammed by using GenMAPP software in combination with the KEGG database. Although most of the maps were assembled by using the SAM-significant genes from the combined 50-72-98-hpi set, we added several SAM-significant genes from single time points when they were excluded from the combined set because spots representing the gene did not qualify at other time points. We found that the pathways emerging from these analyses were more valuable than a list of genes to illustrate the impact of the virus on the host cell transcriptome during the late phase of infection. Gene maps are shown in Fig. 2 to 5. Being microarray-derived, all changes described in the present study refer exclusively to transcript levels. In addition to transcriptional changes, other cellular mechanisms affecting translation, posttranslational modifications, protein stability, and activity affect the amount and functionality of the products of the genes displayed on the maps.

View larger version (35K): [in this window] [in a new window]   FIG. 2. GenMAPP of virus-induced gene expression changes affecting integrin subunits, integrin-to-actin signaling pathways, and actin cytoskeleton. This map was generated by using the "integrin-mediated cell adhesion" KEGG pathway as a template. Genes (abbreviated names based on SwissProt accession number) are shown in boxes (see Materials and Methods for website and other information files). The map also includes two genes from the signaling category (the phosphatidylinositol-3-kinase subunits p110 [PI3K p110] and p55 [PI3K p55]) and eight genes from the GTPase category (rhoA, rac1, cdc42, rho-associated coiled-coil containing protein kinase 1 [ROCK1], LIM motif-containing protein kinase 1 [LIMK1], and p21-activated kinases 2 [PAK 2], 3 [PAK 3], and 6 [PAK 6]). Genes represented by more than one spot in the SAM-generated list are shown in a dashed box. Genes consistently upmodulated are bright orange, genes consistently downmodulated are green, genes not found in any of the data sets are uncolored, and one gene, which scored as SAM significantly upmodulated at one time (integrin ß4 was upmodulated in the 98 hpi data set), is light orange. The mean fold change (calculated by SAM) is noted to the right of each box, with positive values indicating upmodulation and negative values indicating downmodulation. The two dashed parallel lines indicate the plasma membrane. Solid arrows denote proximal events, and dashed arrows denote distal processes in the pathways shown. MLC, myosin light chain.

View larger version (62K): [in this window] [in a new window]   FIG. 5. GenMAPP of virus-induced gene expression changes affecting cell cycle and DNA synthesis. (A) Cell cycle phases G1, S, G2, and M are depicted by blue arrows, with genes modulated by infection listed underneath. (B) DNA synthesis functions in the prereplication complex (left) and replication fork machinery (right) regulated by infection. The same scheme as shown in Fig. 2 is used with the addition of inhibitory relationships denoted by a . Ink 4 a/c contains ink a, b, and c; HDAC 2/10 contains the histone deacetylases 1, 2, 3, 4, 5, 6, 7A, 8, 9, and 10; E2F 2/6 contains E2F 2, 3, 4, 5, and 6; and Apc 1/11 contains apc 1, 2, 4, 5, 6, 7, 10, and 11.

These findings demonstrated that the genes encoding proteins involved in the assembly and disassembly of actin fibers were downmodulated by CMV at late times. Such changes, along with a similar impact on all three major rho family GTPases and their downstream effectors, are likely to disrupt the organization of the cytoplasmic microfilaments and to strongly impair the formation of stress fibers, filopodia, and lamellipodia, eventually leading to the rounded, enlarged cell cytopathic effect observed at late times of infection. The same alterations, combined with the reduction in cellular adhesion transcript levels, would also predictably affect the potential for infected cells to migrate, weakening GTPase-mediated mechanisms on one hand and diminishing adhesiveness on the other hand.

GTPases and related genes. Transcripts encoding GTPase category gene products involved in vesicle transport were also strongly downmodulated and extended beyond those functions involved in integrin-to-actin signaling (Fig. 3). The 112 genes in this category were assigned to one of six classes, with 32 genes (29%) encoding rho-related proteins, 26 genes (23%) encoding rab-related proteins, 17 genes (15%) encoding ras-related proteins, 16 genes (14%) encoding G-protein-related proteins, 14 genes (12.5%) encoding arf-related proteins, and seven genes (6%) encoding ran-related proteins. Forty-nine of these were assembled in the map, which depicts the main members of the six major GTPase families. Expression levels of 11 main G-protein family members were altered by infection, with eight subunits, including two G subunits (Gi2 and -12), three Gß subunits (Gß1, -4, and -2-like), and three G subunits (G10, -11, and -12), consistently downmodulated. Three G subunits (Golf, G13, and G14) showed upmodulation. Heterotrimeric G proteins play an important role in transducing signals derived from the activation of cell surface receptors by various hormones, neurotransmitters, chemokines, and sensory stimuli. Based on the observed downmodulation of most G and all Gß and G transcripts that scored in the analysis, we would predict that the sensitivity of infected cells to signals conveyed via G-protein-coupled receptors would be reduced. This downmodulation may contribute to the failure of pUS28 to have a detectable impact on host gene expression at late times of infection.

View larger version (38K): [in this window] [in a new window]   FIG. 3. GenMAPP of virus-induced gene expression changes affecting functions involved in vesicular traffic, including members of the G protein, ras, rho, rab, arf, and ran families of GTPases. Family names are typed in red, and the principal subcellular system controlled by each family is listed underneath. Members belonging to each family are grouped by braces (}). The same scheme shown in Fig. 2 is used with the addition of one gene, which scored as SAM significantly downmodulated at one time (rhoI was downmodulated in the 98 hpi data set), is light green.

Three of the six consistently upmodulated rabs transcripts encode proteins involved in melanosome transport and morphogenesis (rab8a, rab38, and rab27a). Although rab38 is necessary to target the melanin biosynthesis enzyme tyrp1 to end-stage melanosomes (60), rab8 plays a role in melanosome trafficking (20) and rab27a is required for the recruitment of myosin Va to melanosomes, which then stimulates the microtubule-mediated transport of melanosomes to the tips of dendrites in melanocytes and their transfer to adjacent keratinocytes (48). Since fibroblasts do not possess melanosomes, it is tempting to speculate that the observed transcriptional upmodulation of these genes might be part of a virus-controlled strategy to promote its maturation and release through the specific vesicle traffic mechanisms active in melanocytes.

Lastly, the downmodulation of rab13, a structural and functional regulator of tight junctions (65), was in optimal agreement with our results pointing to cell rounding, whereas the upmodulation of rab32, which participates in synchronizing the mitochondrial fission process (5), was particularly interesting in connection with the observed disruption of the reticular mitochondrial network in CMV-infected HFs (66). Altogether, these data are the first to highlight GTPases and related genes as major targets of transcriptional modulation at late times during CMV infection.

Nuclear DNA-encoded mitochondrial genes. The enzyme category had the most broadly upmodulated transcript levels (Fig. 1 and 4). The 146 genes in this category were assigned to one of five classes, with 55 genes (38%) encoding mitochondrial functions, 39 genes (27%) associated with carbohydrate metabolism, 11 genes (7%) involved in detoxification, 8 genes (5%) participating in the synthesis and/or utilization of glutathione, and 33 genes (23%) representing miscellaneous enzymes. Mitochondrial functions stood out as being the largest class in this category and had a remarkably high proportion of upmodulated genes (91%). Fifty-one genes from this class and two genes from the carbohydrate metabolism class were assembled in a map depicting the mitochondrial oxidative phosphorylation (oxphos) and the related biochemical pathways involved in oxidation of fatty acids (ß-oxidation cycle), catabolism of amino acids, and transport of NADH, acetyl coenzyme A (CoA), and ATP (Fig. 4). Strikingly, all 24 genes encoding components of the electron transport chain complexes I to IV, representing one-third of the known electron transport chain subunits, were upmodulated by infection. Half of the 18 genes that are part of the ATP synthase complex and 8 genes involved in the assembly, biosynthesis, or activity regulation of specific components of the oxphos (coenzyme Q7 [COQ7], cytochrome c synthase, ATPase inhibitory factor 1, chaperone ABC1, F1 complex assembly factor 1 and 2, and cytochrome c oxidase 15 [COX15] and COX17) were consistently upmodulated during infection. Altogether, these data revealed a very strong positive effect of CMV infection on the expression levels of multiple genes related to ATP synthesis, which has not been observed previously. This impact likely results in enhanced energy production that benefits infection. Transcripts for four (short and medium chain acyl-CoA dehydrogenases, 3-ketoacyl-CoA thiolase, and 3-hydroxyacyl-CoA dehydrogenase) of the six standard enzymes in the ß-oxidation cycle and one (enoyl-CoA isomerase) of the three auxiliary enzymes involved in the oxidation of unsaturated fatty acids were also upmodulated. Interestingly, the only downmodulated enzyme in the ß-oxidation cycle was the short chain enoyl-CoA hydratase, which reverses the direction of the cycle by catalyzing the opposite reaction of the long chain enoyl-CoA hydratase. Both the electron transfer flavoprotein (ETFP), which transfers the electrons generated by the ß-oxidation cycle to complex II of the oxphos system, and the glutaryl-CoA dehydrogenase, which transfers electrons generated from the degradation of the amino acids lysine, hydroxylysine, and tryptophan to the ETFP, were upmodulated. These findings demonstrated that the entire mitochondrial system involved in respiration, as well as the degradation of fatty acids and the utilization of amino acid catabolism products, was consistently upmodulated at late times of infection, eventually leading to an increased exploitation of the fatty acid deposits of the cell in order to produce ATP. Further support for this hypothesis was provided by the consistent downmodulation of the key mediators of the fatty acid neosynthetic pathway (citrate synthase, citrate transporter, and ATP citrate lyase), which use the acetyl-CoA produced by the ß-oxidation cycle to store energy for later use instead of synthesizing ATP for immediate consumption. Expression levels of nine genes belonging to four transport systems, one importing NADH from the cytosol and three exporting ATP from the inner mitochondrial matrix, were also modified by infection. The cytoplasmic and mitochondrial isoforms of glutamate oxaloacetate (GOT1 and GOT2, respectively) and the mitochondrial malate dehydrogenase (part of the NADH transport system), the fibroblast isoform of adenine nucleotide translocator (ANT2), and the ubiquitous isoform of creatine kinase (part of the ATP transport system) were upmodulated, whereas liver ANT3 and sarcomeric creatine kinase were downmodulated. VDAC1, which forms voltage-gated pores in the outer mitochondrial membrane to allow diffusion of small hydrophilic molecules and adenine nucleotides in normal situations and of cytochrome c during apoptosis was also downmodulated, whereas the low-abundance VDAC2 isoform that inhibits mitochondrion-mediated apoptosis when overexpressed (22) was upmodulated. Taken together, these data show that the vast majority of the genes composing the two main mitochondrial energy production systems and their associated substrate import and product export mechanisms were consistently upmodulated. In contrast, the genes mediating the first three steps of the pathway leading to the reconstitution of the fatty acids pools were downmodulated. These findings indicate that increased mitochondrial energy production is important at late times during CMV infection and suggest that fatty acid catabolism may dominate over neosynthesis, leading to the depletion of cellular fat stores during infection.

View larger version (59K): [in this window] [in a new window]   FIG. 4. GenMAPP of virus-induced gene expression changes affecting the mitochondrial oxphos, fatty acid ß-oxidation, malate-aspartate transport, ATP transport, and citrate transport systems. The oxphos system is depicted in five large boxes showing subunits (sub; small boxes) of complexes I to IV and ATP synthase. The dashed parallel lines to the left and right of the large boxes represent the mitochondrial inner membrane. The map also includes a gene from the amino acid metabolism category (glutaryl-CoA dehydrogenase), three genes from the vesicle and intracellular transport category (citrate transporter and the voltage-dependent anion channel 1 [VDAC1] and 2 [VDAC2]) and seven genes from the lipid metabolism category (short and medium chain acyl-CoA dehydrogenase, enoyl-CoA isomerase, short chain enoyl-CoA hydratase, 3-hydroxyacyl-CoA dehydrogenase, 3-ketoacyl-CoA thiolase, and ATP citrate lyase). The same scheme as shown in Fig. 2 is used. TCA, tricarboxylic acid cycle; catab, catabolism; synth, synthesis. In complex I, Chain 1//6 contains chains 1, 2, 3, 4, 4L, 5, and 6; B14.5A//18 Sub contains the B14.5A, B14.5B, B15, B17, B17.2, and B18 subunits; 9//24 kDa Sub contains the 9, 9.6, 13A, 13B, 15, 18, 19, 20, 23, and 24 kDa subunits; and 14//75 Sub contains the 14, 51, and 75 kDa subunits. In Complex IV, COX1//4 contains COX1, -2, -3, and -4.

Cell cycle- and cellular DNA replication-related genes. Consistent with data in the literature (10, 32, 34, 52, 83, 107), both cell cycle and nucleic acid metabolism categories were strongly and positively affected by infection (Fig. 1 and 5). The functional genomics approach used here allowed us not only to expand the number of known cell cycle- and cellular DNA replication-related genes upmodulated by CMV but also to gain a better understanding of how these functions may be linked and have a concerted impact on the cell. The 113 genes in the cell cycle category were assigned to one of five classes, with 42 genes (37%) associated with the G₂/M phase, 27 genes (24%) involved in general growth control, 19 genes (17%) associated with the G₁ or G₁/S phases, 13 genes (11%) associated with the S phase, and 12 genes (11%) related to p53 or pRb activities. The 229 genes in the nucleic acid metabolism category were divided into three classes: RNA metabolism (103 genes, 45%), DNA metabolism (101 genes, 44%), and nucleotide metabolism (25 genes, 11%). In the cell cycle category, the S-phase class contained the highest percentage of upmodulated genes (85%), followed by the p53-pRb-related class (75%) and the G₂/M phase class (64%). In the nucleic acid metabolism category, the DNA metabolism class stood out as having the largest percentage of upmodulated genes (66%), followed by the mostly downmodulated nucleotide metabolism (42% upregulated) and RNA metabolism (38% upregulated) classes. These data revealed a very strong positive effect of CMV infection on the expression of multiple S-phase, G₂/M-phase, and DNA activity regulators, probably all of which are required for efficient viral replication. Interestingly, DNA and RNA metabolism gene transcript levels were affected in opposite ways by viral infection, suggesting that genes involved in DNA metabolism might be preferentially transcribed during viral infection. A total of 112 genes from both categories were organized into the maps shown in Fig. 5A and B, which depict cell cycle control and DNA synthesis functions, respectively. A total of 78% of the genes present on the maps and altered by infection were upmodulated. Expression levels of several G₁/S- and S-phase regulators were modified by infection: while both cyclin D1 and the associated cyclin-dependent kinase 6 (CDK6) were strongly downmodulated, the G₁ and S cyclin-CDK complex inhibitors ink4d, kip1, kip2, and cip2 were consistently upmodulated. Transition through the G₁/S boundary is normally promoted by the activity of the cyclin D1-CDK4 or cyclin D1-CDK6 complexes and is normally inhibited by members of the ink, kip, and cip families of CDK inhibitors. The observed expression pattern of these key regulators, therefore, was indicative of a virus-induced cell cycle arrest in G₁, as proposed elsewhere (14, 26, 62). Thedownmodulation of the CDK2, -3, -4, and -6 inhibitor cip1 and of the G₀/S transition inhibitor growth arrest-specific 1 (GAS1), together with the upmodulation of cyclin A, cyclin G₂, the cyclin E- and cyclin A-CDK2 complex activator cdc25A, the transcription factor E2F1, and the E2F1 dimerization partner DP2, however, suggested a transition to an S-phase-like environment, a possibility consistent with earlier studies (49, 81, 86). The induction of cyclin A transcript levels at late times of infection, in particular, was reported previously (49), although this comparison must be made to nonstimulated confluent HFs and not to serum-stimulated cells (83). The upmodulation of nine genes encoding core components of the prereplication complex (all six minichromosome maintenance genes [MCM2 to -7] and three of the six origin recognition complex genes [ORC1L, -3L, and -4L]) and of 10 genes encoding core components of the DNA replication system (DNA polymerase [POLA], primase 1 [PRIM1], the processivity factor proliferating cell nuclear antigen [PCNA], two of the three subunits of replication factor A [RPA2 and RPA3], and four of the five subunits of replication factor C [RFC1 to -4]) further supported this hypothesis. Finally, the expression of several functions that interact with the DNA synthesis complex was altered by infection: the prereplication complex loading inhibitor geminin, the origin activation stimulator cdc7, the POLA-recruiting factor cdc45L, the DNA repair polymerases ß (POLB) and (POLE), and the E1A-binding protein p300 (EP300), which facilitates the DNA repair functions of PCNA, were all upmodulated, whereas the inhibitor of the initiation step in DNA replication deleted in oral cancer 1 (DOC1) and four additional DNA polymerases (polymerase 2 [POLA2], the regulatory subunit of polymerase [POLD2], and the two proofreading polymerases [POLZ] and [POLK]) were downmodulated. In agreement with other reports (10, 107), these data indicate a broad enhancement in the expression of cellular DNA synthesis machinery and related components, possibly reflecting a need for their presence to support viral DNA replication. In this regard, the twofold upmodulation of DNA topoisomerase II (TOP2A), an enzyme that catalyzes the relaxation of supercoiled DNA and the catenation and decatenation of circular DNA and is essential for CMV replication (8), may contribute to the proposed directional rolling-circle model of CMV genome replication (67). Altogether, these results further extend and deepen our understanding of the well-established dysregulation of multiple G₁, S, and DNA replication genes (34, 67), unveiling a complex system of virus-induced events potentially aimed at optimizing viral replication rates.

Multiple components of two fundamental cell cycle control mechanisms, the pRb and p53 systems, were also affected by infection. In addition to E2F1 and DP2, eight pRb interacting partners were upmodulated: five pRb-binding proteins (RBBP2, -4, -5, -6, and -8), two histone deacetylases (HDAC1 and -11) and prohibitin, which is a protein that represses E2F-mediated transcription (102) and, interestingly, also impacts the assembly of mitochondrial respiratory chain complexes (73). Two other genes encoding pRb-binding proteins, the Abelson murine leukemia viral oncogene homolog 1 (ABL1) and the E2F1 partner DP1, were downmodulated. Expression levels of six p53-related genes were modified; four were upmodulated (p53 itself, tumor protein p53-binding protein 2 [TP53BP2], apoptosis-stimulating protein of p53 1 [ASPP1], and protein phosphatase 2C [PP2C]), and two were downmodulated (mouse double minute homolog 2 [mdm2] and ataxia telangiectasia mutated [ATM]). p53-mediated transcription of proapoptotic genes is activated by ATM as a result of DNA damage, is enhanced by TP53BP2 and ASPP1, and is inhibited by mdm2 and PP2C. Thus, the persistent upmodulation of p53 and the activators TP53BP2 and ASPP1, coupled with the downmodulation of the inhibitor mdm2, might result in the induction of apoptosis were it not for the concurrent downmodulation of the stimulator ATM and the upmodulation of the inhibitor PP2C. These results complement and extend data in previous reports (49), revealing a multifaceted effect of CMV infection on the expression of several pRb- and p53-related genes and emphasizing the ability of this virus to simultaneously induce an S-phase-like environment in cells, achieve replication of the viral genome, and avoid the onset of apoptosis triggered by the checkpoint effectors.

Quite surprisingly, the majority (50 genes, 61%) of the 82 cell cycle genes whose expression was modified by infection (Fig. 5A) were related to the G₂/M phase, highlighting the likely importance of functions associated with this part of the cycle for CMV infection. Expression levels of the G₂/M transition regulators wee1, myt1, cyclin B2, CDK1, cdc28 kinase regulatory subunit 2 (CKS2), polo-like kinase 1 (plk1), peptidyl-prolyl cis-trans isomerase NIMA-interacting 1 (pin1), protein phosphatase 1 (PP1A), and the PP1A regulatory subunit 7 (PP1R7) were upmodulated. Entry into mitosis relies upon the properly timed activation of the cyclin B-CDK1 complex. The activity of this complex is controlled by the antagonistic actions of wee1 and myt1 kinases, which suppress CDK1 activity, and the cdc25B and cdc25C phosphatases, which promote CDK1 activation (79). In turn, cdc25C activation in prophase is mediated by phosphorylation through the cyclin B-CDK1-mediated feed-forward loop, by the PP1A-mediated removal of an inhibitory phosphate on serine 216, by a pin1-catalyzed peptidyl-proline isomerization, and by the plk1-mediated phosphorylation of serine 198, which induces the translocation of cdc25C into the nucleus (94). The consistent upmodulation of all of these regulators revealed a very strong positive effect of CMV infection on G₂/M transition control mechanisms, possibly leading to the stimulation of infected cell progression toward a mitosis-like state. Additional evidence supporting this hypothesis was provided by the observation that a large proportion (41 genes, 82%) of the 50 altered G₂/M-related genes encoded proteins specifically involved in the implementation of M-phase events. More than half (22 genes, 54%) of these genes encode proteins implicated in the correct formation of the mitotic spindle and of stable spindle-chromosome attachments. These included six genes controlling the centromere and kinetochore creation process, 10 genes required for centrosome and mitotic spindle functions, and six components of the spindle assembly checkpoint (SAC) (Fig. 5A). Five of the six centromere and kinetochore genes were upmodulated and encoded proteins involved in marking a chromosomal region for centromere formation (centromere protein A [CENPA]), in the assembly of the kinetochore complex (suppressor of G₂ allele of SKP1 [SGT1]), in the attachment of the kinetochore to microtubules (cdcA1), and in general centromere activation (centromere protein F [CENPF] and ZW10 interactor [zwint]). The only downmodulated gene in this group, centromere protein C1 (CENPC1), encodes a protein required for the inactivation of one centromere and the creation of a functionally monocentric chromosome. Eight of the ten centrosome and mitotic spindle genes were upmodulated and encoded proteins important for centrosome integrity (NIMA-related kinase 2 [NEK2]), duplication (centrin 3), and maturation into a microtubule organizing center (centrosomal protein 1 [cep1]), for the nucleation of microtubule assembly at the spindle poles (tub1), for the regulation of the number of microtubule ends in the spindle (katanin p60 and p80), and for the positioning (kinesin family member 4A [kif4a]) and segregation (mitotic centromere-associated kinesin [MCAK]) of mitotic chromosomes. The two downmodulated genes encoded nuclear mitotic apparatus protein 1 (numa1), which is essential for the organization and stabilization of spindle poles and checkpoint with forkhead and ring finger domains (CHFR), which monitors centrosome separation and retards entry into mitosis in the absence of this event. The observed global upmodulation of numerous genes encoding proteins that regulate the assembly and functionality of the mitotic spindle was unexpected and prompted us to consider the possible progression of the infected cells into a metaphase-like state. The downmodulation of CENPC1, CHFR, and numa1 expression, however, suggested the presence of abnormalities in this process, as did the consistent upmodulation of six genes encoding the SAC components budding uninhibited by benzimidazoles 1 (bub1), mitotic arrest-deficient 2A and 2B (mad2A and mad2B), monopolar spindle 1-like (mps1L/TTK), aurora A, and CENPE. In cells containing disrupted or mispositioned spindles, mps1L/TTK recruits active CENPE at the kinetochores. This is followed by the association of mad1 and mad2 and by the mad2-mediated inhibition of the anaphase promoting complex/cyclosome (APC/C), which effectively blocks mitotic cells in metaphase (1). The upmodulation of these key genes hinted at the presence of an active SAC-like process in infected cells. This change, along with the upmodulation of aurora A, which is associated with centrosome amplification and multipolar spindles (111), and of NEK2, which induces splitting of centrosomes and dispersal of centrosomal material (36), suggested a damaging effect of infection on centrosome structure. The expression levels of three APC/C core components were modified in infected cells, with one (cdc27/apc3) downmodulated and two (apc7 and cdc23/apc8) upmodulated. Two APC/C regulatory factors, cdc20 and snk/plk-akin kinase (sak), were also upmodulated. The APC/C ubiquitin ligase complex is activated by cdc20 and controls exit from mitosis by targeting B-type cyclins and securin (also called pituitary tumor-transforming gene 1 [PPTG1]) for degradation. The destruction of securin stimulates the activity of separase (also called extra spindle poles-like 1 [espl1]), a caspase-like protease, which triggers sister chromatid separation by cleaving the scc/rad21 subunit of the cohesin complex (43). Both cohesin and condensin complexes ensure the correct segregation of replicated chromosomes by organizing chromosomes into highly compact mitotic structures (condensin) and by keeping sister chromatids together until they split at anaphase (cohesin). Expression levels of securin, separase, two of the four cohesin subunits (scc/rad21 and structural maintenance of chromosomes 3 [SMC3]), and one of the five condensin subunits (chromosome associated protein E [CAPE]) were upmodulated, whereas the expression of PTTG1-interacting protein (PTTG1-IP), which facilitates securin nuclear translocation, was downmodulated. These data indicate a substantial increase in transcript levels of chromosome management factors, suggesting their importance during the late phase of infection, possibly to properly organize the mass of newly replicated viral DNA, which can reach 4,000 copies/infected cell (69). Finally, four genes encoding NIMA-related kinases (NEK4, -6, -7, and -9) and four genes encoding septins (septin2/nedd5, septin4/pnutl2, septin7/cdc10, and septin9/msf) were transcriptionally altered in infected cells and all except septin4/pnutl2 were downmodulated. Elimination of NEK9 results in spindle abnormalities and chromosomal misalignments (80), whereas inhibition of NEK6 arrests cells in M phase and triggers apoptosis (110). The consistent downmodulation of NEK9 could therefore negatively impact the process of spindle assembly, in agreement with the observed alterations in the expression of spindle-morphogenesis-related genes. The downmodulation of NEK6 and of septin2/nedd5 and septin9/msf, whose depletion results in the accumulation of binucleated and cells arrested during cytokinesis (93), also suggested that exit from a mitosis-like environment might be impaired.

Altogether, these results show a global and consistent increase in the transcript levels of numerous cell cycle regulators and DNA replication machinery components, indicating that the ability to control cell cycle phases and functions is important for the successful completion of CMV replication.

Time-dependent induction of a mitosis-like state during CMV infection and association with CMV replication components. The persistent transcriptional upmodulation of a very large number of M-phase genes suggested that CMV-infected cells might proceed through a mitosis-like state even though cellular DNA replication is blocked. To investigate this process further, uninfected or AD-infected HFs (MOI of 10) were subjected to immunofluorescence analysis for tub1, a protein known to localize at centrosomes, mitotic spindle microtubules, and spindle poles (58) and whose expression levels were consistently upmodulated by infection. Cellular DNA was detected by Hoechst 44432 dye staining, and infected cells were highlighted by immunolabeling with an MAb to the viral nuclear antigens IE1/IE2. In uninfected interphase cells, the IE1/IE2 antigens were absent and tub1 was diffusely cytoplasmic with a concentration in one or two spots corresponding to the centrosomes (Fig. 6A). In infected interphase-like cells, the IE1/IE2 antigens were confined to nuclei bounded by an intact nuclear membrane (Fig. 6C), and tub1 was observed in both cytoplasm and at centrosomes (Fig. 6B). Unlike uninfected interphase cells, however, several of these cells contained more than two tub1-labeled centrosomes, usually positioned adjacent to the nucleus (Fig. 6B and D). These supernumerary centrosomes likely originated from aberrations in some or all of the mechanisms controlling centrosome duplication, maturation, and separation. Thus, CMV infection appears to be associated with centrosomal injuries, a finding consistent with early observational reports (18, 39, 71). The data obtained from functional genomic analysis, however, allows us to understand how the complex interplay between CMV and the cell cycle machinery may be affecting the functioning of the centrosome and to speculate which genes might be involved in the generation of the observed injuries. The upmodulation of NEK2, aurora A, and mps1L/TTK in infected cells may be directly related to this process, given that overexpression of these gene products in mammalian cells results in aberrant centrosome duplication (31, 36, 111). The product of the upmodulated gene PP1A is also a good candidate as an effector function, because this protein localizes at centrosomes during mitosis (7) and modulates NEK2 kinase activity (46). The microtubule organizing center, where centrosomes normally reside, is where final steps of viral particle production occur, likely as the ultimate site of virion tegumentation and envelopement (82). Moreover, this structure is visible as a cytoplasmic inclusion and functions as a gathering site for cellular organelles, dense bodies, virions, and other virus-like particles at late times postinfection (88). Thus, the increased expression levels of several genes whose products are involved in governing the functionality of the microtubule organizing center, such as centrin 3, cep1, plk1, and cdk1, might be part of a viral strategy to gain control of this structure.

View larger version (24K): [in this window] [in a new window]   FIG. 6. Abnormal centrosome numbers in nucleated CMV-infected cells. (A to D) Uninfected or CMV-infected HFs stained with an anti-tub1 MAb (red) and a FITC-conjugated MAb to the viral nuclear proteins IE1/IE2 (green). (A) tub1 localization in a representative uninfected interphase cell; (B to D) tub1 (B) and IE1/IE2 (C) localization in a representative infected nucleated cell and a B+C merged image (D). Arrows point at the two additional centrosomes in the infected cell. Magnification, x1,000.

View larger version (62K): [in this window] [in a new window]   FIG. 7. Abnormal mitotic figures in nonnucleated CMV-infected cells. (A to M) Uninfected or CMV-infected HFs stained with an anti-tub1 MAb (red), an FITC-conjugated MAb to the viral nuclear proteins IE1/IE2 (green) and Hoechst 44432 to detect cellular DNA (blue). (A to C) tub1 (A) and cellular DNA (B) localization in a representative uninfected mitotic cell and A+B merged image (C); (D to I) tub1 (D), cellular DNA (E), and IE1/IE2 (G) localization in a representative infected nonnucleated cell; phase-contrast image of the same cells (H), D+E merged image (F), and F+G+H merged image (I). (J and K) IE1/IE2 (J) and cellular DNA (K) localization in a representative infected nonnucleated cell; (L and M) IE1/IE2 (green) and tub1 (red) localization in representative infected nonnucleated cells. Magnification, x1,000.

View larger version (19K): [in this window] [in a new window]   FIG. 8. Dependence of pseudomitosis on viral and cellular DNA synthesis. (A) Percentage of IE1/IE2 antigen-positive pseudomitotic cells at four different times postinfection in three independent experiments. Each bar represents the mean percentage ± the standard deviation. (B) Percentage of IE1/IE2-positive pseudomitotic cells at 72 hpi in infected HF cultures maintained in the presence of medium (•), medium plus 1 mM HU to block cellular DNA synthesis (), medium plus 10 mM HU to block both viral and cellular DNA synthesis (), or medium plus PFA to block viral DNA synthesis (). Each symbol represents an independent experiment. Horizontal bars indicate mean values for all experiments.

To determine whether the viral replication cycle progressed normally in pseudomitotic cells, infected cultures were subjected to immunofluorescence analysis for the viral delayed early protein ppUL44 (68, 105), the early-late protein pp150 (82), and the true-late protein pUL43 (2). A