Kaposi’s sarcoma-associated herpesvirus (KSHV/HHV-8) is etiologically linked to the endothelial tumor Kaposi’s sarcoma and with two lymphoproliferatve disorders, primary effusion lymphoma and multicentric Castleman’s disease. HHV-8 infects a variety of target cells both in vivo and in vitro, binds to the in vitro target cells via cell surface heparan sulfate, and uses the α3β1 integrin as one of the entry receptors. Within minutes of infection, HHV-8 induced the integrin-mediated signaling pathways and morphological changes in the target cells (S. M. Akula et al., Cell, 108: 407–419, 2002; P. P. Naranatt et al., J. Virol., 77: 1524–1539, 2003). As an initial step toward understanding the role of host genes in HHV-8 infection and pathogenesis, modulation of host cell gene expression immediately after infection was examined. To reflect HHV-8’s broad cellular tropism, mRNAs collected at 2 and 4 h after infection of primary human endothelial [human adult dermal microvascular endothelial cells (HMVECd)] and foreskin fibroblast [human foreskin fibroblast (HFF)] cells and human B cell line (BJAB) were analyzed by oligonucleotide array with ∼22,000 human transcripts. With a criteria of >2-fold gene induction as significant, ∼1.72% of the genes were differentially expressed, of which, 154 genes were shared by at least two cells and 33 genes shared by all three cells. HHV-8-induced transcriptional profiles in the endothelial and fibroblast cells were closely similar, with substantial differences in the B cells. In contrast to the antiapoptotic regulators induced in HMVECd and HFF cells, proapoptotic regulators were induced in the B cells. A robust increase in the expression of IFN-induced genes suggestive of innate immune response induction was observed in HMVECd and HFF cells, whereas there was a total lack of immunity related protein inductions in B cells. These striking cell type-specific behaviors suggest that HHV-8-induced host cell gene modulation events in B cells may be different compared with the adherent endothelial and fibroblast target cells. Functional clustering of modulated genes identified several host molecules hitherto unknown to HHV-8 infection. These results indicate that early during infection, HHV-8 reprograms the host transcriptional machinery regulating a variety of cellular processes including apoptosis, transcription, cell cycle regulation, signaling, inflammatory response, and angiogenesis, all of which may play important roles in the biology and pathogenesis of HHV-8.
Kaposi’s sarcoma (KS) is the leading vascular tumor of HIV-infected patients (1) . The development of KS in the four epidemiologically distinct forms (classic, endemic, posttransplant, and AIDS-KS) is associated with the KS-associated herpesvirus or human herpesvirus 8 (KSHV/HHV-8; Ref. 2 ). HHV-8 is also etiologically linked with two lymphoproliferatve disorders, body cavity-based B-cell lymphoma or primary effusion lymphoma, and a subset of multicentric Castleman’s disease (2 , 3) . Although studies have suggested a decline in the incidence of KS in United States coinciding with the introduction of highly active antiretroviral therapy, mounting noncompliance failure rates in highly active antiretroviral therapy patients suggest that KS will represent a major health problem for years to come (4) . In Africa, widespread infection with HIV has resulted in an alarming prevalence of HHV-8 infection and the associated diseases (5) . Moreover, there is an increasing concern regarding posttransplant KS developing in solid-organ transplant patients, either caused by HHV-8 reactivation or primary infection transmitted from the donor (6) .
HHV-8 is in a latent state in the KS endothelial cells and expresses the genes encoding the latency-associated nuclear antigen (LANA) and a subpopulation of KS lesion inflammatory and spindle cells displays lytic HHV-8 replication (7) . In a marked contrast to the well-established linkage between B-cell lymphomas and latent infection by the related γ1-EBV, studies have indicated the important roles for both latent and lytic infection in KS pathogenesis (4) . KS tumorigenesis appears to require an ongoing lytic infection because interruption of lytic replication by drugs such as ganciclovir appears to prevent KS development (8) . Detection of lytic replication in only a small percentage of KS cells coupled with the signaling properties of several lytic cycle HHV-8 proteins suggest that products of lytic infection may act in a paracrine fashion to promote KS tumorigenesis (9) . KSHV lytic proteins have been shown to induce growth deregulation and angiogenic factors via activation of multiple host cell signaling cascades, including extracellular signal-regulated kinase (ERK), c-Jun-NH2-terminal kinase, and p38 pathways.
In vitro, HHV-8 has been shown to infect human B, epithelial, endothelial, foreskin fibroblast cells, and keratinocytes (10, 11, 12) . The EBV infection of primary B cells results in latent infection, immortalization of B cells, and the maintenance of latent viral episomes replicating along with host cell division. Unlike EBV, infection of primary B cells by HHV-8 does not result in a sustained latent infection and immortalization. Our studies have shown that via its envelope glycoproteins gpK8.1A and glycoprotein B (gB), HHV-8 binds to the ubiquitous cell surface heparan sulfate (HS) molecules (13, 14, 15) , binds subsequently to the α3β1 integrin via its gB (possessing the integrin binding Arginine-Glycine-Aspartic Acid motif), and uses α3β1 integrin as one of the cellular receptors for its entry into the target cells (16) . Our studies have also shown that within minutes of target cell infection, in a gB-integrin α3β1-dependent manner, HHV-8 activated the phosphatidylinositol 3′-kinase, protein kinase C-ζ, mitogen-activated protein/ERK kinase signaling cascade and cytoskeletal rearrangements (17) . Pretreatment of cells with inhibitors specific against members of this cascade blocked HHV-8 infectivity significantly (17) , suggesting that by orchestrating the signal cascade, HHV-8 may create an appropriate intracellular environment to facilitate the infection.
In addition to the induction of preexisting host cell signal pathways, similar to other viruses, HHV-8’s interactions with host cell surface molecules may be triggering cellular transcriptional and other responses. The binding of herpes simplex virus type-1 to cell surfaces induces cellular genes and an antiviral state, which are countered by virion entry and expression of newly synthesized viral protein(s) (18) . The binding of human cytomegalovirus viral gB to an unknown cell surface receptor induced the IFN-responsive RNAs in the absence of viral and cellular protein synthesis (19) . Integrin interactions with extracellular matrix proteins induce robust host cell gene expression and mediate a variety of cell functions such as regulation of gene expression, cell survival, cell cycle progression, cell growth, apoptosis, and differentiation (20) . Because integrin-mediated signaling was shown to actively contribute to the HHV-8 infectious process (16, 17) , it can be speculated that target cell gene responses induced during the initial stages of infection may have pivotal roles in HHV-8 infection. The identification of such cellular signatures might eventually lead to the development of novel strategies to control HHV-8 infection.
As an initial step toward understanding the role of host genes potentially involved in dictating the outcome of HHV-8 infection and pathogenesis, we undertook this study to analyze the cellular transcriptional responses in three different target cells in vitro at 2 and 4 h after HHV-8 infection. Our analysis reveals that HHV-8 triggers a robust modulation of cellular gene expression with cell type-specific and common responses, which may potentially play important roles in the biology of HHV-8 and pathogenesis.
MATERIALS AND METHODS
Human adult dermal microvascular endothelial cells (HMVECd), human foreskin fibroblast (HFF), body cavity-based B-cell lymphoma-1 cells (HHV-8-positive and EBV-negative human B cells), recombinant green fluorescent HHV-8 (GFP-HHV-8) carrying body cavity-based B-cell lymphoma-1 cells (12) , and BJAB cells (HHV-8-negative B cells) were grown as per procedures described before (13) .
HHV-8 for array analysis was purified from GFP-body cavity-based B-cell lymphoma-1 cells (12) after cells were stimulated with 20 ng/ml 12-O-tetradecanoylphorbol-13-acetate (Sigma, St. Louis, MO) for 6 days. GFP-HHV-8 in the spent culture medium was concentrated and gradient purified by using Nycodenz (Sigma) as described previously (14) . Purity of the virus was judged using general guidelines established in our laboratory (17) .
Human genome HG-U133A and B (Affymetrix, Santa Clara, CA) are oligonucleotide probe-based gene arrays containing >100,000 unique oligos representing ∼39,000 transcript variants, which in turn represents >33,000 well-substantiated human genes. The HG-U133B chip represents the majority of the expressed sequence tags, which are not well annotated. Because the HG-U133A chip represents ∼22,283 most well-characterized genes, we used this chip for analysis.
Gene Array Expression Analysis.
HMVEC-d, HFF, and BJAB cells were washed twice in serum-free DMEM or RPMI and incubated in the serum-free medium for 6–8 h to reduce the effect of the serum or other growth factors in the analysis. Serum-starved cells were infected with GFP-HHV-8 (at five viral genome copies/cell), incubated at 37°C for 2 or 4 h. Total RNA from uninfected (UI) controls and infected cells were isolated using RNeasy mini columns (Qiagen, Valencia, CA). First strand synthesis was performed using 10–15 μg of total RNA, a T7-(dT)24 oligomer and the Superscript Choice System (Invitrogen, Carlsbad, CA). The T7 promoter introduced during the first strand cDNA synthesis was then used to direct the synthesis of cRNA by using T7 RNA polymerase (Enzo Diagnostics, Farmingdale, NY) and biotinylated deoxynucleotide triphosphates. The biotin-labeled cRNA was fragmented to a mean size of 200 bp before hybridization. Total RNA and biotin-labeled cRNA were tested for the integrity and size by resolving on Agilent RNA 6000 Nano LabChips (Agilent, Palo Alto, CA). Hybridization was performed at 45°C for 16 h (0.1 m MES [4-morpholinepropanesulfonic acid (pH 6.6), 1 m NaCl, 0.02 m EDTA, and 0.01% Tween 20] and washed under both nonstringent (1 m NaCl, 25°C) and stringent (1 m NaCl, 50°C) conditions. Chips were stained with phycoerythrin-streptavidin (10 μg/ml), scanned, and analyzed with Microarray Suite 5.0 software (MAS 5.0; Affymetrix). Each infection was repeated twice, and each RNA sample was hybridized to two chips.
Data Analysis and Statistics.
The primary data captured using MAS 5.0 software resulted in a single raw value for each probe set based on the mean of the differences between the hybridization intensity for the perfect match features and the mismatch features for a particular transcript (data analysis fundamentals: GeneChip expression analysis can be found online). 4 Three types of normalizations were applied to this data before additional sorting and analysis. The hybridization intensities across the treatments of a particular cell type were first normalized by applying the global normalization algorithm that trims the mean signal intensity of the experiment to the trimmed mean signal of the baseline or control treatment. To facilitate comparison between samples and experiments, the globally normalized data were further subjected to (a) a per chip normalization to account for chip-wide variations in intensity by dividing each intensity value by the 50th percentile of all values on the chip, and (b) a per gene normalization where each gene is divided by the intensity of that gene in the control sample. To perform the global normalizations with high confidence, a regression analysis (Pearson correlation coefficients) was done using the raw signal intensities of various treatments.
These files from expression analyses were then exported via MicroDB and Data Mining Tool (Affymetrix) for additional filtering and analysis. In these analyses, genes designated as significantly changed were those (a) that possessed a reliably detectable signal (absolute call ≠ “Absent” or “Marginal” in the case of repressions, whereas, in the case of inductions, differentially induced genes were taken into analysis), (b) have detection P ≤ 0.05, and (c) as determined by the statistical algorithm to be changed ≥2-fold (change call ≠ “no change” or “marginal”). To additionally increase the significance of expression changes, we interrogated our data sets for an increase in average difference (intensity) of at least 2-fold at both time points of HHV-8 infection (2 and 4 h) or 3-fold or more at a given time point. The primary oligonucleotide hybridization data as well as all of the tables created after filtering can be obtained by contacting the authors via e-mail ().
Clustering and Gene Ontology (GO) Consortium Linking.
To classify gene expression profiles into groups according to their behavioral patterns, cluster analysis was conducted. Two separate unsupervised clustering algorithms, a hierarchical clustering and a nonhierarchical K-means clustering, were performed (GeneSpring version 5.0; Silicon Genetics, Redwood City, CA). Hierarchical clustering was done using average linkage method and standard correlation as a similarity measure. K-means clustering was performed after subjecting the data sets first for bioscript analysis (GeneSpring) that performs 3, 5, 8, and 10 K-means clusters and returns the classification with the highest explained variability. Data were subsequently analyzed for their functional affiliations by GO linking (GeneSpring).
Reverse Transcription-PCR (RT-PCR).
Total RNAs isolated from the UI or HHV-8-infected cells using RNeasy (Qiagen) were DNase treated (Invitrogen) and subjected to cDNA synthesis using Thermoscript reverse transcriptase and random oligonucleotides (Invitrogen). Primer pairs designed using Oligo 4.0 (Molecular Biology Insights, Cascade, CO) were used to amplify specific genes from 250 ng of cDNA using HotStar TaqDNA polymerase (Qiagen). Amplifications were carried out in parallel for both infected and UI samples. For manual quantitation, successive samples were removed every three cycles, beginning with cycle 14 and continuing through cycle 44. Progressive PCR samples were resolved on agarose gels, visualized after ethidium bromide staining, and quantitated using AlphaImager 2000 (Alpha Innotech, San Leandro, CA). For all genes, integrated density values corresponding to the sum of pixel intensities after background corrections were recorded for both the UI and infected samples at linear points on the amplification curve and fold changes were created after normalizing to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Expression changes of 18 cellular genes were analyzed by RT-PCR using primer sequences that are summarized in the supplementary data. 5
Western Blot Analysis.
Activation of p21CIP that was shown to be up-regulated in all of the 3 cell types by the array was confirmed by Western blot analysis. Target cells were grown to 70–80% confluence or to logarithmic phase by feeding the day before infection, serum-starved, and infected with GFP-HHV-8 at 37°C. At different time points, cells were rinsed with PBS and lysates were prepared (17) . Ten μg of lysates were resolved on SDS-12%-PAGE and immunoblotted with antibody detecting p21CIP (Santa Cruz Biotechnology, Santa Cruz, CA) and antimouse-IgG-horseradish peroxidase (KPL, Inc., Gaithersburg, MD). The blots were stripped and reprobed with anti-β-actin antibodies as a loading control (Sigma).
Northern Blot Analysis.
The activation of ICAM-1, DUSP5, and integrin α6 genes was confirmed by northern analysis. Ten μg of total RNA from HHV-8-infected or UI target cells were separated on a 1% agarose-5% formaldehyde gel in 4-morpholinepropanesulfonic acid buffer [20 mm 4-morpholinepropanesulfonic acid (pH 7.0), 5 mm sodium acetate, and 1 mm EDTA] and transferred to positively charged nylon membranes (Sigma) using 20× SSC (3 m NaCl and 0.3 m sodium citrate). The membranes were hybridized with a DNA probe specific for ICAM-1, DUSP5, and integrin α6 and prepared as detailed below. Total RNA isolated from HHV-8-infected BJAB cells (4 h) was converted to cDNA and subjected to PCR using primers specific for ICAM-1, DUSP5. or integrin α6. The PCR products were resolved on agarose, purified from the gel, and labeled with [α-32P]dATP using Klenow fragment (Promega, Madison, WI). The membranes were hybridized with 1 × 106 to 2 × 106cpm of probes/ml at 65°C overnight in Church buffer [1% BSA, 1 mm EDTA, 0.5 m sodium phosphate (pH 7.2), and 7% SDS]. Internal control hybridization was performed for GAPDH in parallel. Blots were subjected to two low stringency washes (2× SSC, 0.1% SDS, 15 min at room temperature) and one high stringency wash (0.5× SSC, 0.1% SDS, 15 min at 65°C) and exposed to a phosphorimager screen for quantitation and subsequently to XAR film at −70°C.
Induction of Host Gene Expression by HHV-8 and the Validity of Gene Array.
We have previously demonstrated that within minutes of infection, HHV-8 entered the target cells and induced the preexisting host cell signal pathways (16 , 17) . Target cells underwent morphological changes, accumulated actin stress fibers, and induced the formation of filopodia and lamellipodia (17) . These observations indicate that virus greatly influences the cellular functions early during infection. To further assess the impact of HHV-8 and its components on the target cells early during infection, comparisons of global gene expression profiles between UI and HHV-8-infected cells were examined. HMVECd (primary human endothelial cells) and BJAB (human B cell line) represent two major in vivo targets of HHV-8 infection. HFF (primary human foreskin fibroblast) cells were included because activation of a variety of signaling cascades in response to HHV-8 infection was observed in these cells (16 , 17) . An initial time course experiment revealed the induction of a large number of host cell gene expression after 90 min of infection (data not shown). To provide the best snapshot of HHV-8-induced genome-wide changes early during infection, we examined the gene expression profiles of cells at 2 and 4 h after infection.
Each RNA samples isolated from different treatments were hybridized with two HG-U133A chips representing ∼22,283 annotated transcripts. Transcription of ∼50% of the total 22,283 transcripts (∼9000) was detected (a present call) at each time point of treatment, which is probably well representative of the total gene expression in the human cells. The gene expression profiles of HHV-8-infected cells over the respective UI controls are depicted in the scatter plot shown in Fig. 1 ⇓ . Expression of only a small number of transcripts changed in consequence to HHV-8 infection at both time points. Expression ratios of most of the genes were close to 1 or changed only ≤2-fold cutoff line, suggesting that they were not significantly changed in consequence to HHV-8 infection (Fig. 1) ⇓ . These few changes indicate that changes in expression unrelated to HHV-8 infection did not present a problem in our assays. Comparison of the signal intensities between the arrays from one cell type revealed a low interarray variation (data not shown). Distributions of signal intensities among treatments were additionally analyzed by comparing mean signal intensities. This revealed a higher Pearson correlation coefficient between 2 and 4 h postinfection samples of one cell type than samples from different cell types (data not shown). This further increases the reliability of the data and validates our gene array results.
HHV-8 Induces Cell Type-Specific as well as Conserved Host Cell Gene Expression.
Our gene array design included the analysis of HHV-8 interaction with three target cells to reflect the broad cellular tropism shown by this virus (10, 11, 12) . To obtain a measure of similarity between the three target cells in their capacity to respond to HHV-8 infection and to compare the behavior among the cells, average fold change after 2- and 4 h postinfection for each cell type was determined, and a correlation matrix with Pearson correlation coefficient was then calculated between the three cell types. Although the correlation was relatively low as expected, correlation between HMVECd and HFF cells was 10 times higher compared with the Pearson correlation coefficient of BJAB versus HMVECd and about three times higher than BJAB versus HFF cells (data not shown).
Results of UI controls at two time points revealed high correlation values, with most of these signals changing <1.5-fold, thus justifying the 2-fold threshold filter used for calculating the significant modulation (data not shown). To increase the significance further, only those genes with >2-fold change in their expression at both time points and/or >3-fold at one time point were considered. Such filtering revealed that a total of 324, 374, and 175 transcripts changed in HMVECd, HFF, and BJAB cells, respectively. A further breakdown of these results revealed the up-regulation of 215, 243, and 170 transcripts, whereas 109, 131, and 5 transcripts were down-regulated in HMVECd, HFF, and BJAB cells, respectively (data not shown). For delineating the behavioral patterns, this data were then subjected to hierarchical clustering, and the results obtained for individual cells are shown Fig. 2A ⇓ . Overall, more genes were up- or down-regulated in the HMVECd (Fig. 2A ⇓ , top panel) and HFF cells (Fig. 2A ⇓ , middle panel) than in the BJAB (Fig. 2A ⇓ , bottom panel) cells. These differential or cell type-specific expression changes were more prominent for the down-regulated genes (represented in shades of blue) because very few genes were repressed in BJAB cells compared with HMVECd or HFF cells (Fig. 2A) ⇓ .
Comparison of transcriptional profiling between cells was subsequently done, which yielded fewer common genes. This is probably because as the number of arrays in each data set increases, it becomes less likely for a particular gene to pass such strict filtering criteria. Nevertheless, genes emerge from such comparisons will be of great interest for studying their role in the unique biology of HHV-8. Several features were apparent from these analyses: (a) many of the genes with altered expression were detected in more than one sample, arguing against spurious changes; (b) list generated for dynamically expressed genes is much shorter for BJAB cells; and (c) there was virtually nonoverlapping list of genes in all of the three cell types analyzed. The Venn diagram shown in Fig. 2B ⇓ represents the cell type-specific distribution of HHV-8 regulated genes and their overlap in expression. Of the total 686 genes, 33 were regulated across all of the cell types, 121 genes were regulated by two of the three cells, whereas the rest did not share the transcriptional responses (Fig. 2B) ⇓ .
Semiquantitative RT-PCR of Selected Genes Confirms the Gene Array Results.
To confirm the oligonucleotide gene array results by an independent method, semiquantitative RT-PCR analysis on UI and HHV-8-infected samples were carried out for 18 selected genes. These included 7 genes in which the differential regulation was seen in all three target cells, 7 in two target cells, and 4 genes that were uniquely regulated in a single cell type (Table 1) ⇓ . As an example, RT-PCR confirmation of DUSP5 gene that was up-regulated in all of the three target cells is shown in Fig. 3 ⇓ . DUSP5 RT-PCR product from the infected RNA was readily detectable by sample 8 (cycle 35; Fig. 3A ⇓ , Lane 9), whereas the RT-PCR product from the UI RNA sample was only detectable by sample 11 (cycle 44; Fig. 3A ⇓ , Lane 12). In contrast, the amplification of constitutively expressed GAPDH gene reached the linear phase equally in both UI and HHV-8-infected RNA samples starting from sample 3 (cycle 17; Fig. 3B ⇓ ). Comparison of the integrated density value scores of the DUSP5 gene from the infected and UI control RNA is shown in Fig. 3C ⇓ . These results suggested that the DUSP5 gene in the infected RNA underwent linear amplification nine PCR cycles earlier than did the parallel UI cell RNA.
We carried out the above described comparison for 18 selected genes, normalized them to the GAPDH mRNA levels and calculated the fold induction. For the 7 genes whose expression changed in all of the three target cells, a positive correlation was obtained for all of the 7 at both time points in all of the three different target cells. For the rest of the genes, RT-PCR was done using RNA from all three target cells collected 4 h after infection where greater changes was observed (Tables 1 ⇓ and 2) ⇓ . Larger changes were generally observed by RT-PCR, with the exception of MAP3K8, where lower fold activation was observed in the RT-PCR reactions (20.11- versus 2.8-fold, Table 1 ⇓ ). Inhibitor of apoptosis homologue C was up-regulated only in the HMVECd and BJAB by the array; however, by RT-PCR, up-regulation was detected in all three infected target cells (Table 1) ⇓ . Similarly, Coxsackie and adenovirus receptor was identified as a down-regulated gene by the array only in the HMVECd cells. In HFF and BJAB cells, although the results (raw data) showed a down-regulation, the detection Ps were higher (>0.99), and hence, the Coxsackie and adenovirus receptor was considered an outlier and was filtered out in the analysis. Results of RT-PCR analysis showed an up-regulation of Coxsackie and adenovirus receptor in HMVECd-infected cells while no change was seen in the BJAB and HFF cells (Table 1) ⇓ . The observed differences between the array and RT-PCR assays and the differences in the magnitude of induction detected could be attributable to the increased sensitivity of RT-PCR and/or the potential for chip saturation. Nevertheless, importantly, RT-PCR confirmed the induction or repression for the majority of the genes tested in the RNA samples used for array analysis, as well as in the independently derived set of RNA samples derived from target cells infected with wild-type HHV-8 (data not shown).
Northern and Western Blot Assays for Selected Genes Verify the Gene Array Data.
To obtain additional confirmation of array results (Table 2) ⇓ by another independent method, Northern blot analysis was performed using RNA preparations tested by gene array. Total RNA from HHV-8-infected cells at 4 h after infection and controls were fractionated by gel electrophoresis, blotted, and probed with 32P-labeled probes. The Northern blots (Fig. 4A) ⇓ confirmed the gene array results (Table 2) ⇓ for ICAM-1, DUSP5, and integrin α6 genes. A low level of constitutive expression of ICAM-1 and DUSP5 was detected in the UI cells (Fig. 4A ⇓ , top two panels, Lanes 1, 3, and 5). Upon HHV-8 infection, ICAM-1 was strongly up-regulated by 22-, 20-, and 18-fold in BJAB, HFF, and HMVECd cells, respectively (Fig. 4A ⇓ , top panel, Lanes 2, 4, and 6). A similar pattern was observed for the DUSP5 gene (Fig. 4A ⇓ , panel 2, Lanes 2, 4, and 6). Unlike ICAM-1 and DUSP5 genes, integrin α6 gene expression was detected only in the infected BJAB cells by the gene array (data not shown). Northern blot analysis confirmed the array data and a 10-fold induction was observed (Fig. 4A ⇓ , panel 3, Lanes 1 and 2). Neither an endogenous expression (Fig. 4 ⇓ A, panel 3, Lanes 3 and 5) nor an induction (Fig. 4A ⇓ , panel 3, Lanes 5 and 6) was observed for integrin α6 in HMVEC-d and HFF cells. Equal amount of GAPDH was detected in all these samples demonstrating the equal loading of RNA in these samples (Fig. 4A ⇓ , bottom panel).
To correlate the gene array detected transcription changes with protein levels, we analyzed the expression levels of p21CIP protein by immunoblot. Compared with UI controls, after HHV-8 infection, gene array detected the consistent induction of p21CIP gene in all three target cells at both time points (Table 2) ⇓ . Western blot analysis at 2, 3, 4, 6, and 8 h postinfection of all of the target cells showed that p21CIP was up-regulated with a sustained activation up to 6 h postinfection (Fig. 4B ⇓ , Lanes 2–6). Immunoblot for β-actin detected similar expression in all these samples, demonstrating the equal protein loading in these experiments (Fig. 4B ⇓ , bottom panel). In summary, our data revealed a good correlation between array and other means of target verification, thereby validating our methodology and analysis.
HHV-8-Induced Host Cell Genes Cluster Based on Their Kinetics of Activation.
As our experimental data contains only two conditions (infection at two time points versus UI), a simple hierarchical clustering could provide sufficient visual clustering of the data set. Because a visual inspection might fail to detect any subtle variation patterns hidden in the data set, a more sophisticated K-means clustering was performed. The initial bioscript analysis on 154 most informative genes suggested that K-means clustering into five groups would represent the highest variability. Thus, in the first stage, the algorithm was iterated to create five clusters with an additional five clusters option for tightening the clustering process. In the second fine-adjustment phase, the algorithm was iterated for additional clusters. However, the results with six or more clusters returned with duplication of kinetic patterns.
An example of the K-means clusters for BJAB cells is shown in Fig. 5 ⇓ . Each of the five clusters represented different kinetics of gene activation that included clusters I and II showing repression kinetics and clusters III–V exhibiting up-regulation (Fig. 5) ⇓ . Cluster I contained 29 transcripts (18.83%) that were down-regulated during the infection. Cluster II genes (15 genes, 9.74%) also were down-regulated and is the only cluster that demonstrated a generalized repression pattern that was maintained from 2–4 h after infection. Cluster III contained 18 transcripts (11.68%), including genes in which induction peaked at 2 h after infection and began to return to the basal level between 2 and 4 h after infection. Cluster IV, which represents the maximum number of genes (54 genes, 35.06%), showed a sustained up-regulation pattern at 2 and 4 h postinfection. Cluster V represented by 36 genes showed up-regulation that was maintained at both 2 and 4 h after infection. Although these clusters included genes involved in a variety of functions, some general patterns were prominent. Cluster I represented the majority of cytokines such as interleukin (IL)-1β, IL-6, chemokine (CXC) ligand 2,5, CXCL10 (IP10), cytokine Gro-α, β, and γ, and the IFN-stimulated genes such as 2′5′OAS, MxA, Mr 15,000 protein, T-cell α chemoattractant precursor, and guanylate binding protein 1. The tables listing members of different clusters I–V are presented in the supplementary data. 5
Although similar kinetic patterns were observed for the differentially expressed HFF and HMVECd genes, the following differences were prominent: (a) variation in the number of genes in each cluster; and (b) variation in the functional groups to which genes in each cluster belonged. For example, a majority of the IFN-stimulated genes showed the repression pattern and belonged to gene cluster I in the BJAB cells. In contrast, in HFF and HMVECd cells, IFN-responsive genes and other cytokine genes belonged to cluster IV with a sustained up-regulation at 2 and 4 h postinfection (data not shown), a pattern similar to clusters IV and V of the BJAB cells. Another clear difference was the observation that majority of the up-regulated angiogenic factors such as (vascular endothelial growth factor-A (VEGF-A), angiopoietin-related protein-4, stanniocalcin, and endothelial differentiation gene-1 in the HFF and HMVECd cells were down-regulated in the BJAB cells. The lack of functional correlations in these cells was additionally supportive of our initial observation of low correlation between the BJAB and other two primary adherent cells. These results suggested that HHV-8-induced modulation of gene expression in B cells was different compared with the adherent cells.
Cell Type-Specific Gene Activations Early during HHV-8 Infection.
To further investigate the HHV-8 induced cell type-specific gene modulations, the uniquely regulated genes in BJAB (n = 102), HFF (n = 239), and HMVECd (n = 191) cells that were not shared with other cells (Fig. 2B) ⇓ were analyzed for their biological functions. The GO building primarily identified GO terms involved in various biological processes, cell components, and molecular functions. Nodes for molecular functions were more elaborate and appropriate, and hence, we focused more on the GO molecular functions. Notable GO molecular functions that associated with all of the cells include major processes such as apoptosis, cell cycle, cancer, microtubular dynamics, structural proteins, transport, and signal transduction (Fig. 6) ⇓ . The top-ranked biological process also reflected the same molecular function, i.e., cell signaling and cell-cell communication accounted for ∼90% of the responses (data not shown). Such mapping of biological functions again showed the following behavioral difference in BJAB cells compared with the two primary adherent target cells: (a) despite the smallest number of uniquely regulated genes in BJAB, it included a large number of apoptosis regulators; (b) majority of apoptotic regulators in B cells were proapoptotic unlike the antiapoptotic mediators induced in the HFF and HMVECd cells; and (c) there was a total lack of immunity-related proteins in BJAB cells. The striking cell type-specific behaviors additionally supported the notion that at least in the initial stages of infection, HHV-8-induced host cell gene modulation events in B cells were different compared with the adherent endothelial and fibroblast target cells.
Biological Implications of Host Cell Genes Modulated during the Early Phases of HHV-8 Infection.
To explore the relevance of HHV-8-modulated genes in its biology, an interactive exercise involving several databases such as Unigene, LocusLink, Gene Cards, PubMed, and extensive review of literature was conducted. For this analysis, only the list of 154 genes shared by at least two cell types was used. Of these dynamically regulated 154 transcripts, functions of 13 sequences were not known. The remaining gene products were classified according to known functions in signal induction, apoptosis, transcription, host defense, inflammatory responses, angiogenesis, tumorigenesis, cell and tissue structural dynamics, metabolic pathways, and various other functions (Table 2) ⇓ . These analyses revealed the modulation of several genes not previously implicated in KS or linked to HHV-8 biology. With a specific focus on the most informative genes, here, we discuss some of these gene products with respect to their role in the biology of HHV-8 and potential involvement in KS.
Up-Regulation of Genes Regulating the Signaling Networks during the Early Phases of HHV-8 Infection.
HHV-8 activates a variety of cellular signaling molecules early during infection (16 , 17) . Inducible signaling agonists and antagonists play a vital role in regulating the strength, duration, and range of action of cellular signals. Our array results demonstrated the modulation of a number of molecules involved in the regulation of signaling cascades. This included IER3/IEX-1, which is a new type of ERK substrate (does not interact with c-Jun NH2-terminal kinase or p38). IEX-1 has a dual role in ERK signaling by acting both as an ERK downstream effecter mediating survival and as a regulator of ERK (21) . HHV-8 also activated other regulatory molecules of cell signaling like sprouty homologues and DUSP5 genes. Human sprouty 1–4 are orthologs of Drosophila sprouty, which is a general intracellular tyrosine kinase signaling inhibitor (22) . DUSPs inactivate kinases by dephosphorylating both the phospho-serine/threonine and phospho-tyrosine residues. They negatively regulate the members of the mitogen-activated protein kinase superfamily and DUSP5 has maximal activity toward ERK (23) . HHV-8 induced significant activation of ERK in HMVECd and HFF cells early during infection, and such activations were sustained for the first 30 min, after which, the activation lowered to the background level (17) . Activation of Sprouty 2 and 4 and DUSP 1, 5, and 6 observed during HHV-8 infection (Table 2) ⇓ may constitute a significant feedback inhibitory mechanism for deactivating ERK1/2, thereby restoring these signaling pathways to their virus sensitive preinfection state.
Up-Regulation of Genes Encoding Antiapoptotic Proteins during the Early Phases of HHV-8 Infection.
HHV-8 encodes and expresses several antiapoptotic proteins during its latency (v-FLIP) and lytic cycle (v-BCl-2, K7, and K15; Ref. 4 ). However, HHV-8 also must have developed additional mechanisms early during infection to block apoptosis that is probably triggered by virus binding and entry into the target cells (Table 2) ⇓ . These factors mediate both transient and sustained protection against various apoptotic stimuli. Our results show that HHV-8 induces the inhibitor of apoptosis in HMVECd cells by ∼14- and 4-fold at 2 and 4 h after infection, respectively, and by ∼2- (2 h) and 8 (4 h)-fold in BJAB cells (Table 2) ⇓ . The inhibitor of apoptosis family of proteins prevents cell death by binding to and inhibiting active caspases 3 and 9 (24) . HHV-8 induced the bcl-2-related protein A1 in HMVECd and BJAB cells (Table 2) ⇓ . Bcl-2-related protein A1 does not block proapoptotic caspases but is believed to be a temporary protection mechanism against apoptotic stimuli (25) . Myeloid cell leukemia-1 was another antiapoptotic Bcl-2 family member that was up-regulated by HHV-8 (Table 2) ⇓ . Similar to other Bcl-2 family members, myeloid cell leukemia-1 localizes in the mitochondria and can associate with other proapoptotic family members (25) . The only virus infection that is known to induce bcl-2-related protein A1 and myeloid cell leukemia-1 is EBV mediated by its latent membrane protein-1 (26) . Neither of the proteins was shown to be induced in gene array experiments with cells latently infected with HHV-8 (27 , 28) . The antiapoptotic proteins induced by HHV-8 such as bcl-2-related protein A1 can promote viability on a rapid short-term basis early during infection and thus allowing the recruitment of other Bcl-2 family members for further cell fate decisions.
Up-Regulation of Genes Encoding Angiogenic Signatures during the Early Phases of HHV-8 Infection.
KS lesions are characterized by extensive neoangiogenesis (29) . A striking finding of our array results is the induction of several genes involved in the control of vascular remodeling and angiogenesis. HHV-8 induced VEGF, thrombomodulin, urokinase-type plasminogen activator receptor, matrix metalloproteinase (MMP-1), tissue inhibitor of matrix metalloproteinase-1, and angiopoietin-related protein 4 (Table 2) ⇓ . During angiogenesis, local coagulation and fibrinolysis must be modulated in a controlled fashion. The urokinase-type plasminogen activator system is one of the most efficient proteolytic systems active in the extracellular environment (30) . Our finding of significant induction of urokinase-type plasminogen activator receptor by HHV-8 is exciting and consistent with a role of urokinase-type plasminogen activator receptor for the metastatic phenotype (31) because 51.6% of the KS tissues were positive for urokinase type plasminogen activator receptor immunostaining (31) . MMPs are believed to be pivotal enzymes in invasion and angiogenesis, whereas tissue inhibitors of metalloproteinase-1 are antagonists to a number of MMPs and reduce the neovascularization process (32) . HHV-8 induced a robust activation of MMP-1 and tissue inhibitors of metalloproteinase-1, and by 4 h, the up-regulation of MMP-1 was stronger than tissue inhibitors of metalloproteinase-1 (2–6-fold above, Table 2 ⇓ ), suggesting the induction of a powerful angiogenic signal early during infection.
Among the several positive regulators of angiogenesis, two families of growth factors are largely specific for vascular endothelium by virtue of having receptors that are mostly restricted to endothelial cells, namely VEGF and the more recently discovered angiopoietins (33) . These two families seem to work in complementary and coordinated fashion during vascular development. We observed a strong activation of VEGF-A, C, and angiopoietin like-4 genes by HHV-8 (Table 2) ⇓ . VEGFs have been shown to be up-regulated by several HHV-8 proteins such as v-IL6, vGCR, vMIP-I, and vMIP-II (4) . Though angiopoietins display strong angiogenic activity independent of VEGF, their role in KS is not yet defined.
KS lesions are composed of a complex mixture of different cell types with a prominent infiltrate of extravasated erythrocytes and lymphocytes. During tissue inflammation, normal endothelial cells can be induced to become adhesive for circulating blood cells and to support their transmigration into inflamed tissue. Galea et al. (34) have shown that the lymphocyte function-associated antigen-1-ICAM-1 interaction is the primary one involved in the adhesion of peripheral blood lymphocytes to KSY1 cells, a KS cell line. HHV-8-encoded ORF74 up-regulates the expression of VCAM-1, ICAM-1, and E-selectin, whereas HHV-8-K5 is known to down-regulate it (35) . The strong induction of ICAM-1 gene during the early stages of infection in the HMVECd and HFF cells (Table 2) ⇓ is suggestive of its potentially important role in KS pathogenesis and in the activation of inflammatory responses.
Differential Regulation of Host Cell Defense Genes during the Early Phases of HHV-8 Infection.
Although no direct activation of IFN-α, β, or γ or its receptors was observed in the gene array, up-regulation of IFN-regulated genes was observed (Table 2) ⇓ . This differential regulation of IFN-regulated genes was not observed in the B-cell line, BJAB. Among IFN-regulated genes activated by HHV-8, OAS2, MxA, and guanylate-binding protein-1 have been shown to restrict the growth of certain viruses (36) . HHV-8-induced IFN-stimulated gene 15 is a cytokine responsible for augmenting and amplifying the immunomodulatory effects of IFN-α or IFN-β. HHV-8 infection also revealed the activation of IFN regulatory factors 1 and 7, the actions of which are known to be down-regulated by HHV-8-encoded proteins vIRF1 and ORF 45, respectively (4) .
Modulation of Cytokine Genes during the Early Phases of HHV-8 Infection.
KS is a multifocal angiogenic tumor consisting of characteristic spindle cells and infiltrating leukocytes. Unlike most cancers, KS does not appear to be caused by clonal expansion of a transformed cell. Instead, it appears to be a hyperplastic disorder caused, in part, by local production of inflammatory cytokines such as IL-1, IL-6, IFN-γ, and tumor necrosis factor α, and growth factors such as basic fibroblast growth factor and VEGF. This is supported by the fact that infiltration of inflammatory cells in KS lesions, including CD8+ T cells, monocytes, macrophages, and dendritic cells, precedes the proliferation of the spindle-shaped endothelial cells. Infiltrating cells systematically produce inflammatory cytokines that are likely responsible for the activation of vessels and endothelial cells, increase of adhesiveness with extravasation, and recruitment of lymphocytes and monocytes. In total agreement with this, our experiments have shown that HHV-8 induces a variety of cytokines, including IL-8 and Gro1, Gro2, and Gro3 at early time points of infection.
Modulation of Stress Response Genes during the Early Phases of HHV-8 Infection.
Many inhibitors of stress responses are known to inhibit virus infection (37) . Notable among the HHV-8-induced stress response genes are manganese superoxide distmutase and cyclooxygenase-2 (Cox-2). Cox-2 is a proinflammatory stress compound whose expression was the most strongly up-regulated gene (by 84.65-fold at 2 h and 44.79-fold at 4 h) in HHV-8-infected HMVECd cells (Table 2) ⇓ . Cox-2 is believed to promote viral infection by inhibiting the target cell nitric acid synthesis that function to induce antiviral status (37) . Human cytomegalovirus induces Cox-2, and inhibition of Cox-2 by specific inhibitors significantly reduced the yield of human cytomegalovirus (37) . Identification and abrogation of such stress responses offer clues to gene expression that could be a rate-limiting step in the efficient establishment of virus infection. Studies using specific Cox-2 inhibitors to determine the role of Cox-2 in HHV-8 infection are in progress.
Our studies describe a comprehensive picture of global gene changes soon after HHV-8 infection of three susceptible cell types. Our experimental design of analyzing the host cell transcriptional changes immediately after infection is quite distinct from the previous gene induction studies that were carried out at later time points of virus infection. Data presented here provide the framework and starting point for the further detailed analysis of the induced factors’ roles in the biology of HHV-8 early during infection.
As with other viruses, several events occur during the early stages of target cell infection by HHV-8 that must be playing active roles in deciding the outcome of infection. For better conceptual purposes, we have divided these early events into six overlapping dynamic phases. Phase I involves the binding of virus to cell surface via its interactions with HS (13, 14, 15) , integrins (16) , and possibly to other yet to be identified molecule(s). This is followed by virus entry into the target cells (Phase II; Refs. 16 , 17 ), probably overlapping with the induction of host cell signal pathways during Phase I (17) that facilitate the entry. In Phase III, the viral capsid/tegument moves in the cytoplasm facilitated by the induced signal pathways and probably overlaps with Phase IV host cell gene transcription and expression. In Phase V, viral DNA enters into the nucleus followed by viral gene expression (latent and/or lytic), which is greatly influenced by the HHV-8-induced signal pathways and expressed host cell genes. Phase VI involves the overlapping viral gene-induced host cell gene expression, which may exert an influence on success of viral infection. As part of understanding the early events of HHV-8 infection, this study was designed to analyze the Phase IV host cell gene expression immediately after infection.
Several evidences presented in this article such as the changes in only a small number of transcripts at both time points, gene expression differences in the three cell types, close similarity of gene expression in the adherent cells, RT-PCR and Northern blot confirmation of array data with RNA samples derived from different set of experiments, as well as the Western blot assays, asserted that the observed host gene modulation was caused by HHV-8 infection. The impact of HHV-8 binding and entry into the target cells (Phases I–III) upon the host cell gene expression was not analyzed here because of the following reasons. The binding and entry processes of herpesviruses are complex events involving multiple cell surface receptors. Although HHV-8 binding to cell surface HS and α3β1 integrin has been demonstrated, it must be also interacting with other yet to be unidentified molecule(s). We did not use virus pretreated with heparin in our studies to study the impact of binding on host cell gene expression because soluble heparin though lowered the binding of radiolabeled HHV-8 to the target cells significantly, certain percentage of virus still bind and entered the cells resulting in low level of infection (13) . Similarly, infectivity neutralizing HHV-8 anti-gB, gpK8.1A, gH, and gL antibodies, Arginine-Glycine-Aspartic Acid peptides, anti-integrin antibodies, and soluble integrins did not prevent the binding of HHV-8 to the target cells (14 , 16) . Hence, we could not use these treatments in the array analyses. Moreover, although these treatments blocked infection at a post-HS binding stage of infection, whether virus still enters the target cells is not known at present and is under investigation. Similarly, UV-inactivated herpesviruses have been shown to enter the target cells. Our recent studies show that HHV-8 uses multiple integrin molecules during the binding and entry process (F. Z. Wang, P. P. Naranatt, N. S. Walia, H. H. Krishnan, L. Zeng, and B. Chandran, unpublished observations), making the choice of integrin molecule harder to block HHV-8 entry. Currently, there are no characterized HHV-8 mutants that cannot bind or enter the target cells. The complex dynamics of HHV-8-receptor(s) interaction is at the very early stage of understanding, and it is not known whether HS and integrin interaction occur simultaneously or sequentially. Hence, in the present set of experiments, we set out to study the transcriptional reprogramming early during infectivity rather than delineating the responses that are HHV-8 binding/entry specific (Phases I–III).
Previous studies of analyzing host cell gene modulation by viruses have been limited to the expression of one or limited numbers of genes after infection. However, gene array technology makes it possible to analyze the induction of multiple target genes at a genome-wide scale at a given time point. This technology has been used to analyze the effects of a number of human virus infections on cell physiology beginning with human cytomegalovirus (38) . A common finding in many of these investigations is the up-regulation of genes involved in the inflammatory response (39) . Beyond a common innate immune response to infection, different cells may have characteristic signatures to different pathogens, often as a result of the highly specific activities of a particular pathogen’s virulence determinants. Thus, a number of molecules with potential usefulness in controlling virus infections have emerged from such analyses (27 , 37) .
Our array data revealed a number of unique observations. For example, HHV-8 infection has a major impact on the expression pattern of cellular genes that exhibited cell-type specificity. The differentially expressed genes belonged to a variety of cellular pathways. The striking cell type-specific behaviors suggest that at least in the initial stages of infection, HHV-8-induced host cell gene modulation events in B cells may be different compared with those in the adherent endothelial and fibroblast cells. The differences in B cells are very interesting and may potentially reflect the biology and differences in the outcome of HHV-8 infection in vivo. Human B cells are a reservoir of latently infected HHV-8. HHV-8-associated primary effusion lymphoma and multicentric Castleman’s disease differ from KS in many respects, most notably in the expressed viral genes. In addition to the latency-associated ORF 73, ORF 72, K13, and K12 genes that are expressed in the KS endothelial cells, HHV-8 also expresses the ORF 10.5 (LANA2/vIRF3) and K2 (vIL-6) in primary effusion lymphoma cells, both in vivo and in vitro, and additional lytic cycle genes in the multicentric Castleman’s disease cells (2) . Although the BJAB cells that we analyzed have been used by others extensively as representative of B cells (10 , 40) , it may be important to confirm these observations using other B-cell lines and most importantly using primary B cells as characterization of HHV-8 infection in these cells becomes established. Nevertheless, this is an interesting observation, and additional work is needed to correlate the relevance of B-cell-specific changes induced by HHV-8 with its B-cell pathogenesis.
Why should HHV-8 need to modulate the host cell transcriptional machinery during the initial phases of infection? Early during infection of target cells, HHV-8 has to overcome several host-mediated obstacles. For example, HHV-8 has to (a) block the apoptosis of host cells triggered by the virus binding and entry processes, (b) modulate host cell transcription to overcome the restriction on virus gene transcription, (c) block the activation and effects of innate immune responses, and (d) evade the elimination by the constant surveillance pressure from the host immune system such as IFNs, tumor necrosis factor α, complement, antibodies, ADCC, natural killer cells, CTL, and phagocytic cells. To establish a successful infection, HHV-8 must have developed many ways to manipulate and overcome these obstacles, using both viral and host proteins. Our observation of modulation of host genes that govern vital cellular processes such as apoptosis, transcription, host defense and inflammation, extracellular matrix remodeling, angiogenesis, and protein processing during the early course of infection is exciting because these cellular transcriptional reprogramming probably be serving vital roles in overcoming the above-mentioned obstacles and establish a successful infection.
Poole et al. (28) have used a different primary effusion lymphoma cell line (JSC1) to produce the virus that was used to infect primary endothelial cells for a period of 3–5 weeks, and gene array experiments were done when almost all of the cells changed from typical cobblestone to spindle-shaped morphology and were positive for LANA. Moses et al. (27) infected endothelial cells previously immortalized by retroviral expression of human papillomavirus E6 and E7 genes, and arrays were done with after ∼4 weeks of HHV-8 infection when >90% of the cells were LANA positive and showed the spindle cell phenotype. Mikovits et al. (41) reported the expression changes after bone marrow-derived primary CD34+ cells were infected with HHV-8 and maintained for 30 days before the analysis. These three studies were done after at least several days after infection, where latency had been established and with cells appearing to resemble a transformed phenotype. In contrast, our objective was to analyze the host gene expression changes at very early time points during the primary HHV-8 infection of target cells. In the examination of host genes modulated early during infection in the present study, a strong up-regulation of IFN-responsive genes such as IRF 7, Mx1, IFN-inducible transmembrane protein 1, OAS, IFN-stimulated protein 15, IRF-1, and IFN-inducible Mr 67,000 guanylate binding protein 1 was observed. Studies by Poole et al. (28) showed the up-regulation of other IFN-responsive genes such as IFN-induced transmembrane protein 3, Mx R2, IFN-α-inducible protein (IFI-6–16), and IFN-inducible protein 56. Thus, although both of these studies showed similar induction of IFN-responsive genes, the effect on specific genes seemed to vary. This is probably expected because the Poole et al. (28) array was done after the virus had established latency, during which time the effect on some of the IFN-responsive genes may have been reduced. Moreover, it is known that several viruses like herpes simplex virus type-1 (18) and vesicular stomatitis virus (42) are known to first induce an antiviral status and disarm this at later time points of infection to favor the virus infection. Other common genes that were shared between our early time points of infection and the previous latently infected later time point analyses included SSI-3, vEts transcription factor, tissue plasminogen activator, IL-8, BCl-3, nucleoside phosphorylase, and tissue inhibitors of metalloproteinase-1. Thus, 10 of the genes that were up-regulated in our array analysis were also up-regulated in those Poole et al. (28) . Comparison of transcriptional profiles between Moses et al. (27) and Pool et al. (28) showed that only 7 of 124 induced and 3 of 60 repressed genes were common (27) . Authors argued that such low correlation was because 117 of 124 of the induced genes were not spotted in both arrays. Our comparison between the elements used for HG-U133A and Human UniGem V2.0 microarrays, used by Poole et al. (28) , revealed that nearly one-third of the most informative genes detected by us were absent in the cDNA arrays used by Poole et al. (28) . This, together with the differences in the time points of analysis, might have profoundly influenced the transcriptional changes observed between our study and previous studies (27 , 28 , 41) .
Another important observation in our study is the repression of mRNA for a subset of genes in the infected cells. Repression does not appear to be a general degradation phenomenon because we did not observe a great reduction in the number of transcripts. Because HHV-8 infection can affect both the synthesis and the stability of cellular mRNAs, the data obtained here probably reflect the interference of HHV-8 at multiple steps in host gene expression. Among the genes that were down-regulated at least in two cells, dickkopf homologue 1 was of particular interest because this molecule is known to be one of the secreted inhibitors of Wnt signaling. Wnt signaling is a highly conserved developmental pathway in which β-catenin mediates changes in gene expression (43) . HHV-8-encoded LANA has been shown to stabilize β-catenin by binding to its negative regulator glycogen synthase kinase-3β and inducing its nuclear accumulation (44) . Our results indicated the down-regulation of one of the negative regulators of Wnt signaling, thereby suggesting that additional mechanisms may be operative as positive feedback of Wnt signaling. Among other genes that were down-regulated, the majority belong to the molecular chaperons (BCL2-associated athanogene family member-2, nuclear receptor subfamily, and Zinc finger proteins 133 and 238) and tumor suppressors (absent in melanoma 1 and Thioredoxin-interacting protein). Although the biological significance of down-regulating chaperons is not known, deregulations of tumor suppressors may be beneficial to HHV-8 and associated oncogeneic process.
HHV-8’s exploitation of cell cycle regulatory α3β1 integrin molecule for entry into the target cells and the induction of integrin-mediated mitogenic signaling pathways may have important implications in the unique biology of KS lesions and HHV-8’s role in KS pathogenesis. Besides the delivery of viral DNA into the cells, HHV-8 interactions with the host cell receptor(s) induced signal pathways, and the observed host cell gene expression such as the induction of genes encoding angiogenic signatures, antiapoptosis, cytokines, and stress response genes may bring about important biological consequences and play a significant role in KS pathogenesis. In addition to identifying specific host response genes within each functional group that are already known to be important for HHV-8 infection, our findings identify novel candidate genes that are not yet known to HHV-8 biology. For example, up-regulated stress gene Cox-2 and signaling regulatory Heparin binding-epidermal growth factor and repressed tumor suppressor absent in melanoma 1 and thioredoxin-interacting protein. The roles of these molecules in HHV-8 infection and KS development remain to be studied. The close similarity of RT-PCR and Northern and Western blot assays with the array data in terms of similar magnitudes and direction of changes in the genes identified suggests that the array data may serve as a guide post for evaluating interesting genes.
Similar to HHV-8 interactions with UI endothelial cells, interactions with latently infected cells may also stimulate the production of cytokines/growth factors, which in turn may stimulate the endothelial growth. A synergism may exist between latently infected endothelial cells and incoming infection that may control endothelial cell growth by an autocrine-parocrine loop. Additional studies are needed to examine the consequences of HHV-8-induced signaling pathways and transcriptional reprogramming in the regulation of virus gene expression during a primary infection of endothelial and B cells and on cells that are already programmed by the cell growth-modulating HHV-8 latency-associated proteins and whether such interactions play essential roles in the establishment and/or maintenance of latent infection and/or cellular proliferation of latently infected endothelial cells and/or B cells. Additional studies with virus strains, mutants, viral proteins, and host cell types such as primary B cells will be required to obtain a more complete picture of host cell gene expression and biological effects after HHV-8 infection. A greater understanding of host cell gene reprogramming induced by HHV-8 may eventually lead to the development of novel therapies to control KS lesions.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Grant support: USPHS Grants CA 75911 and 82056 (B. C.), a University of Kansas Medical Center Biomedical Research Training Program Postdoctoral Fellowship (P. P. N.), and USPHS Grant P20 RR16475 (S. R. S.).
Requests for reprints: Bala Chandran, Department of Microbiology, Molecular Genetics and Immunology, Mail Stop 3029, The University of Kansas Medical Center, 3901 Rainbow Boulevard, Kansas City, KS 66160. Phone: (913) 588-7043; Fax: (913) 588-7295; E-mail:
↵4 Internet address: http://www.affymetrix.com/support/technical/manual/expression_manual.affx.
↵5 Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org).
- Received September 3, 2003.
- Revision received October 14, 2003.
- Accepted October 20, 2003.
- ©2004 American Association for Cancer Research.