Gene Expression and Transcription Factor Profiling Reveal Inhibition of Transcription Factor cAMP-response Element-binding Protein by γ-Herpesvirus Replication and Transcription Activator*

Herpesvirus replication involves the expression of over 80 viral genes in a well ordered sequence, leading to the production of new virions. Viral genes expressed during the earliest phases of replication often regulate both viral and cellular genes. Therefore, they have the potential to bring about dramatic functional changes within the cell. Replication and transcription activator (RTA) is a potent immediate early transcription activator of the γ-herpesvirus family. This family includes Epstein-Barr virus and Kaposi sarcoma-associated herpesvirus, human pathogens associated with malignancy. Here we combine gene array technology with transcription factor profiling to identify the earliest DNA promoter and cellular transcription factor targets of RTA in the cellular genome. We find that expression of RTA leads to both activation and inhibition of distinct groups of cellular genes. The identity of the target genes suggests that RTA rapidly changes the cellular environment to counteract cell death pathways, support growth factor signaling, and also promote immune evasion of the infected cell. Transcription factor profiling of the target gene promoters highlighted distinct pathways involved in gene activation at specific time points. Most notable throughout was the high level of cAMP-response element-binding protein (CREB)-response elements in RTA target genes. We find that RTA can function as either an activator or an inhibitor of CREB-response genes, depending on the promoter context. The association with CREB also highlights a novel connection and coordination between viral and cellular “immediate early” responses.

Herpesvirus replication involves the expression of over 80 viral genes in a well ordered sequence, leading to the production of new virions. Viral genes expressed during the earliest phases of replication often regulate both viral and cellular genes. Therefore, they have the potential to bring about dramatic functional changes within the cell. Replication and transcription activator (RTA) is a potent immediate early transcription activator of the ␥-herpesvirus family. This family includes Epstein-Barr virus and Kaposi sarcoma-associated herpesvirus, human pathogens associated with malignancy. Here we combine gene array technology with transcription factor profiling to identify the earliest DNA promoter and cellular transcription factor targets of RTA in the cellular genome. We find that expression of RTA leads to both activation and inhibition of distinct groups of cellular genes. The identity of the target genes suggests that RTA rapidly changes the cellular environment to counteract cell death pathways, support growth factor signaling, and also promote immune evasion of the infected cell. Transcription factor profiling of the target gene promoters highlighted distinct pathways involved in gene activation at specific time points. Most notable throughout was the high level of cAMP-response element-binding protein (CREB)-response elements in RTA target genes. We find that RTA can function as either an activator or an inhibitor of CREB-response genes, depending on the promoter context. The association with CREB also highlights a novel connection and coordination between viral and cellular "immediate early" responses.
The cellular environment plays a significant role in the outcome of virus infection, illustrated most clearly in the case of herpesvirus infections. In permissive cells, these viruses undergo productive replication, releasing virus progeny and killing the host cell. In non-permissive cells, they persist in a latent form with minimal gene expression for the lifetime of the cell. However, activation of certain cellular signaling pathways within the latently infected cell leads to reactivation and resumption of productive infection.
How these processes are controlled by the cellular environment is of great interest, particularly in the case of human ␥-herpesviruses (i.e. Epstein-Barr virus, Kaposi sarcoma-associated herpesvirus (KSHV 3 /HHV-8)) where infection has been linked to the development of malignancies, including lymphoma, nasopharyngeal carcinoma, gastric carcinoma, and Kaposi sarcoma (1)(2)(3)(4)(5)(6). Although epithelial and endothelial cells are most permissive for replication, these viruses primarily infect B lymphocytes, where they establish latent infection and express only a small subset of their genes. Activation of the protein kinase A (PKA), RAS/MEK/ERK, and protein kinase C pathways (7)(8)(9)(10) or inhibition of NF-B and Akt (11,12) has been shown to reactivate the latent virus and restore lytic replication.
These cellular pathways are thought to regulate the balance between latency and lytic replication via expression of an immediate early viral gene product, replication and transcription activator (RTA). In KSHV, the expression of RTA is an essential prerequisite for productive replication and is also sufficient to reactivate the virus from latency (13)(14)(15). The RTA homologue in Epstein-Barr virus functions in a similar manner, although it requires cooperation with another viral gene product ZEBRA (reviewed in Ref. 16). The RTA protein is a potent transcription factor, with a highly conserved N-terminal DNA binding domain, a basic leucine zipper dimerization domain, and a C-terminal activation domain. Although there is little overall sequence similarity between the activation domains of RTA homologues, one 50-amino acid sequence close to the C terminus is well conserved (see Fig. 1A). The sequence lies within a region required for binding of RTA to essential components of the cellular transcription machinery, including Brg1 and TRAP230, in addition to co-activator cyclic AMP-response element-binding protein (CREB)-binding protein (CBP) (13,17,18). Deletion of this region severely compromises the transcription activation and virus reactivation functions of RTA (13).
Although RTA is capable of activating transcription independently via direct promoter binding (19 -22), it can also cooperate with a number of cellular transcription factors, such as Oct-1 (23,24), the CCAAT/enhancer-binding protein (C/EBP) family (25), Notch coactivator RBP-J (26), and STAT3 (27). The number of different interacting partners and associated mechanisms of transcription regulation illustrates how RTA has adapted to utilize diverse cellular pathways to support virus replication.
Microarray technology is a powerful technique to investigate the way in which ␥-herpesviruses interact with their host cell. However, this virus family presents significant obstacles to this approach. The first is the size of the viral genome (ϳ150 kb) and the associated large number of open reading frames expressed during infection. This produces very complex cellular responses and difficulty attributing a particular response to a specific viral protein. The second obstacle is the lack of cell lines permissive to some herpesviruses. This is a particular problem for the human ␥-herpesviruses, where latent, not lytic, infection predominates in most epithelial and endothelial cells in vitro. Nevertheless, the genome-wide analyses of herpesvirus infection have generated valuable information and identified virusinduced cellular changes, including activation of the MAK kinase and Src kinase pathways (28) and the interferon pathway (28 -30), up-regulation of metalloproteinases (29 -31), and expression of anti-apoptotic proteins (30). In addition, a cell line conditionally expressing KSHV RTA has been used to identify specific changes occurring in B cells during virus reactivation, 24 and 48 h after expression (32), highlighting the up-regulation of B cell activation markers CD21 and CD23a.
We chose an alternate approach, to focus specifically on the effects of an immediate early viral protein within a time frame more closely related to its temporal expression in the virus life cycle. To this end, we generated two cell lines, expressing KSHV wild type RTA and an RTA mutant severely compromised in its activation capacity (RTA⌬608 -651). Because RTA functions during the lytic cycle, we chose the most permissive cell line available as a host: 293 epithelial cells (33,34). Expression of the RTA proteins in these cell lines was tightly controlled and inducible by tetracycline, allowing us to minimize any long term toxicity of RTA expression. With this method, we were also able to uniformly and simultaneously express RTA in a cell population over a short time frame (4 -12 h) without experiencing the delay caused by viral entry and uncoating. A comparison between the gene changes recorded by wild type RTA and the transcriptionally attenuated mutant RTA allowed us to isolate specific targets of wild type RTA and exclude gene changes caused by foreign gene expression or experimental manipulation.
Using these conditions, we generated array data that we present in two ways: -fold change in gene expression (compared with time 0) and transcription factor binding site frequency within the gene promoters (transcription element listening system (TELiS) analysis). This broad approach to identifying changes occurring within a narrow time frame gave us a more comprehensive picture of the changes in the cellular environment initiated by RTA. We find a number of new targets for transcription activation and inhibition. Transcription factor binding sites within the target gene promoters reflect many of the known cooperating interactions between RTA and cellular transcription factors, at distinct time points. In addition, they also show a high frequency of CREB-response elements. We find that RTA regulates CREB-driven transcription in a promoter-specific manner and suggest that this allows RTA to divert a key component of the cell's immediate early response to virus infection.
Transient Transfections-10 5 293T cells/well were seeded onto a 24-well dish 16 h before transfection. 400 ng of DNA/ well was transfected using the calcium phosphate method. 10 ng of pLucϪ122, MJmulti, K2p(Ϫ827), or K2p(Ϫ414) were transfected with the indicated amounts of RTA expression vectors. Cell extracts were harvested 24 h post-transfection and assayed using the Dual Luciferase reporter assay system (Promega). T-RExRTA cell lines were transfected with 300 ng of the indicated reporter constructs using Lipofectamine Plus reagent (Invitrogen). 3 h after DNA addition, medium was added with and without 1 g/ml (wild type RTA) or 0.01 g/ml (RTA⌬) tetracycline. Whole cell extracts were assayed for luciferase activity 24 h later, using the Dual Luciferase system (Promega). Transfections using the PCSK1 promoter were performed using Lipofectamine 2000 reagent (Invitrogen). Cells co-transfected with 200 ng of CREB were stimulated with 300 M protein kinase A inducer dibutyryl cyclic AMP (Sigma) 3-4 h post-transfection. Cell extracts were harvested 24 h after transfection.
Cell Lines-293RTA and 293RTA⌬ tetracycline-inducible cell lines were generated using the T-Rex system (Invitrogen). FLAG-tagged RTA cDNAs were PCR-cloned from pFLAGcRTACMV2 into pCDNA5/TO (Invitrogen), sequenced, and transfected into the parent 293T-Rex cell line using Lipofectamine Plus reagent (Invitrogen). 24 h post-transfection, cells were trypsinized and reseeded at 1:5-1:20 dilutions in the presence of blasticidin (5 g/ml) and hygromycin (200 g/ml). Single clones were isolated, and whole cell extracts were screened by Western blot for the expression of FLAG-RTA after incubation with 1 g/ml tetracycline for 24 h.
Western Blot-Single clones isolated after transfection of the T-RExRTA expression plasmid and hygromycin selection were grown to the 24-well stage and induced with 1 g/ml or 0.01 g/ml tetracycline for the indicated times. Cell extracts were harvested in 50 l of 1ϫ SDS loading dye, boiled, and sonicated, and 10 l was analyzed by SDS-PAGE. Proteins were transferred to nitrocellulose and probed with primary antibodies recognizing the FLAG epitope, ␤-actin (both from Sigma), or polyclonal rabbit serum recognizing KSHV RTA, in 4% reconstituted nonfat dried milk in Tris-buffered saline with 0.05% Triton X-100. Final images were collected using a Storm Phos-phorImager (GE Healthcare).
Preparation of RNA-T-RExRTAwt and T-RExRTA⌬ cell lines were seeded onto 6-well dishes and induced with 1 g/ml (wt) or 0.01 g/ml (⌬) tetracycline, in duplicate. Cells were harvested at 0, 4, 8, and 12 h postinduction, and RNA was prepared using the Qiagen RNeasy kit according to the manufacturer's instructions (Qiagen, Valencia, CA). The integrity of the RNA was determined by agarose gel electrophoresis before cDNA synthesis.
Microarray Analysis-Microarray analysis was performed essentially as described (37). Briefly, RNA was first checked for integrity using an Agilent Bioanalyzer. 5 g of total RNA was then used to synthesize biotinylated cDNA using BioArray kits (Enzo Life Sciences, New York, NY). cRNA was fragmented and hybridized to Affymetrix U133A human genome high density oligonucleotide arrays. The arrays were then washed using a custom array washer and scanned with an Affymetrix G2500A scanner. Images were analyzed using the gcRMA algorithm (38), and data were normalized using global median scaling.
The condensed data files are available from ArrayExpress (accession number E-MEXP-1437).
Methods and Tools for Microarray Data Analysis-16 microarray data sets were condensed as described above and globally normalized so that the median intensity across genes and arrays is a value of 100. Duplicates were then averaged to generate data sets for both 293T-RExRTA and 293T-RExRTA⌬ cell lines of each time point, which were used in further analysis. By comparing with the 0 h time point, the -fold change in expression of each gene at each time point was obtained (the intensity in at least one of the two compared data sets was above 200, which was twice the median value, and set as a base-line threshold). The FOCUS program, a tool for hypothesis-driven analysis of microarray data, was used to identify genes that followed a specified pattern of change across the experiment (39). TELiS was used for transcription factor analysis (40). The analysis of potential CREB-regulating genes was done by using the CREB target gene database (41). The transcriptional start sites of genes of interest were obtained from the DBTSS (Database of Transcriptional Start Sites) (42).
Microarray procedure (Suppl. Fig. 1) shows tetracyclineinducible cell lines expressing either KSHV RTA or firefly luciferase that were generated using the RevTet system (Clontech, Mountain View, CA). Parallel cell lines (three KSHV RTA and two luciferase) were induced with 1 g/ml doxycycline, and RNA was collected at 24 h for array analysis. After global normalization of each array, results were calculated as mean -fold change using data from the luciferase-expressing cell line as a base line. Raw data are available through GEO (accession number GSE21987).

Generation of Inducible Cell Lines Expressing Wild Type and
Mutant RTA-Our aim was to generate a system where the immediate effects of RTA on cellular gene expression could be isolated and gene changes resulting from experimental manipulation could be minimized. To this end, we introduced a FLAG-RTAwt and also a transcriptionally attenuated RTA (RTA⌬608 -651), lacking a critical region of the activation domain ( Fig. 1A), into a 293 cell line, where expression could be controlled by tetracycline (T-REx system, Invitrogen).
We first confirmed that the wild type and mutant proteins were expressed at comparable levels in a transfection system (Fig. 1B). We then compared their transcriptional activity by co-transfecting each RTA expression plasmid with viral reporters highly responsive to RTA (pLucϪ122 (19) (Fig. 1C) and K2p(Ϫ827)Luc (35) (Fig. 1D)). We confirmed that the mutant RTA was severely defective, compared with wild type protein.
We then generated clonal cell lines inducibly expressing each protein by transfecting RTA expression vectors into a 293 cell line expressing the tetracycline repressor (T-REx system, Invitrogen; see "Experimental Procedures"). After selection with hygromycin, ϳ10 single clones were expanded for each RTA construct and tested for expression of FLAG-RTA by Western blot. Expression of RTA appeared to be tightly controlled and could be detected 12 h postinduction ( Fig. 2A) and in subsequent experiments, as early as 4 h (data not shown). We observed very different levels of expression in the wild type and mutant RTA-expressing clones, which was not the result of clonal variation, because all clones transduced with RTAwt expressed the same low level of protein compared with the consistently higher level of expression in all clones expressing RTA⌬608 -651 ( Fig. 2A) (data not shown). This may instead be the result of increased stability of the mutant protein caused by the removal of sequences known to negatively regulate stability (43). Nevertheless, in order to accurately compare the effects of equivalent amounts of protein it was necessary to establish conditions that produced more comparable levels of expression. Inducing the cells expressing RTAwt with higher concentrations of tetracycline did not increase expression any further. We therefore decreased the concentration of tetracycline used to induce expression of RTA⌬608 -651 to a level where it was comparable with the wild type protein (Fig. 2B). Transfection of these cell lines with the viral reporters shown in Fig. 1, B and C, followed by tetracycline induction (1 and 0.01 g/ml for wild type and mutant cell lines, respectively) confirmed that expression of only RTAwt resulted in robust transcription activation and that this activation was diminished by mutations or deletions in the RTA-response elements of each reporter (MJmulti and K2p(Ϫ414)) ( Fig. 2, C and D).
Cellular Response to Expression of RTA-The clonal cell lines expressing either wild type (T-RExRTAwt) or mutant (T-RExRTA⌬) RTA were then used to assay gene changes after 4, 8, and 12 h of expression (using Affymetrix arrays; see "Experimental Procedures"). Analysis of the raw data indicated good consistency between RNA derived from duplicate cell samples collected at each time point, with R 2 values of 0.988 or greater for each set of duplicates ( Fig. 3) (data not shown). Each set of data were normalized, and genes were recorded in terms of their -fold change, relative to time 0 (before tetracycline induction) for each cell line. (full data sets are available from ArrayExpress; see "Experimental Procedures").
Using a 2-fold increase as the threshold, we found that 56 transcripts were up-regulated in the T-RExRTAwt cell line compared with 17 in T-RExRTA⌬ cells after 4 h of induction. At 8 h, 97 transcripts were up-regulated in T-RExRTAwt cells but only three in T-RExRTA⌬ cells, with one transcript in common. By 12 h, 147 transcripts were up-regulated in the wild type cell line with only four up-regulated by mutant RTA, and only two of these were shared by both lists. Although there was minimal overlap between the transcripts up-regulated by wild type and mutant proteins, we excluded any in common from further analysis, in order to focus on the most responsive targets of RTA.
The identity of the genes up-regulated (49 genes) and downregulated (41 genes) at 4 h is shown in Table 1. As an illustration, the genes affected by expression of both wild type and mutant protein RTA (and therefore excluded from further analysis) are shaded gray. Notably, this did not exclude any of the genes up-or down-regulated to the greatest extent, indicating that the transcriptional effects of RTA predominated over any experimental manipulation.
A group of the most highly up-regulated genes (SERPIN B2, TNFRSF21, IFI16, STK3, CXCR4, and MICB) are consistent with a proinflammatory response. Up-regulation of CXCR4 and also intracellular adaptor kinase SHC1 are indicative of an increase in MAPK pathway activation, which may be further enhanced by the down-regulation of mitogen-and stress-activated kinase (MSK) inhibitor DUSP1. HES1, the inhibitor of Notch target gene transcription is also down-regulated, along with genes encoding a group of proapoptotic proteins (RHOB, CASP8AP2, and KLF10). There is therefore strong evidence that the gene changes initiated within hours of RTA expression promote cell survival and increase growth factor signaling via pathways known to support efficient virus replication (44) (see "Discussion"). In addition, we note the early up-regulation of cell surface proteins CD55 and CD59, two inhibitors of complement-mediated cell death (45)(46)(47). The enhanced expression of these genes may allow the infected cell to support viral gene expression while also avoiding immune recognition.
Using "Sustained Change" to Identify RTA-associated Cellular Signaling Pathways-From 4 h, we then extended our attention to the gene lists collected 8 and 12 h after RTA induction. These indicated that many of the gene changes observed at 4 h were sustained over the course of the experiment.
Although RTA is a potent transcription factor with direct promoter-binding activity, mounting evidence has shown that it also activates via interactions with cellular signaling proteins, including STAT3, Oct-1, and RBP-J (23,24,26,27). We reasoned that sustained changes in gene expression were more likely to be the result of combined effects of RTA and cellular transcription partners and/or secondary effects of RTA on cellular signaling pathways. We therefore sought to identify these relationships. The microarray data were first reanalyzed to identify genes with a sustained pattern of up-regulation (sustained change), after expression of wild type RTA but not mutant RTA, using the FOCUS program (39) (see "Experimental Procedures"). The intersection of the original -fold change lists with these "FOCUS lists" generated sustained change lists of 32 genes from the 8 h time point and 65 genes from the 12 h time point, shown in Table 2. The list shows sustained up-regulation of many cellular genes, up to 70-fold, by 12 h. In addition, it highlights new targets, with delayed kinetics, which were not sufficiently up-regulated at 4 h to pass the threshold for Table 1. An example is gp130 (TNFRSF21), encoding one of the IL-6 receptor components. This gene is particularly interesting because its expression is not sufficient for signaling of cellular IL-6 (cIL-6) but is sufficient for the cell to respond to the viral IL-6 homologue vIL-6. vIL-6 is expressed during  (19). K2p(Ϫ414) is a deletion of the vIL-6 reporter K2p(Ϫ827), which is less responsive to RTA (35). Error bars, S.D.

TABLE 1 Genes up-regulated (A) and down-regulated (B) after induction of T-RExRTAwt cells for 4 h
RNA samples from two separate experiments were assayed, and average results are expressed as -fold change over baseline (time 0; see "Experimental Procedures"). Genes are listed in descending order of -fold change. Genes shaded in gray were also up-regulated (A) or down-regulated (B) after expression of mutant RTA. An asterisk in the shaded genes in B indicates genes that were down-regulated by wild type RTA but up-regulated by mutant RTA.

TABLE 2 Short list of genes with consistent up-regulation ("sustained change") at earlier time points, generated from the interest of FOCUS analysis and -fold change lists
-Fold increases shown in italics were generated from raw data lower than the 200 minimum threshold. Ϫ indicates a decrease in gene expression compared with uninduced cells at this time point.
virus replication and is functionally equivalent to cIL-6, promoting cell proliferation. RTA also activates the vIL-6 promoter during lytic replication (35) (Fig. 1D). RTA may therefore be increasing signaling by vIL-6 in two ways: by increasing the expression of the cytokine and by increasing the sensitivity of the infected cell to vIL-6. Transcription Factor Analysis Reveals Overrepresentation of CREB Binding Sites in Early Targets of RTA-To determine which cell signaling pathways might cooperate in RTA-mediated gene activation, the 8 and 12 h sustained change lists were analyzed using the TELiS database, which identifies transcription factor binding sites overrepresented in a set of gene promoters, compared with the average of all promoters of genes present on a particular microarray (40). Because the 4 h gene list had only one prior time point (time 0), it could not be reliably filtered for sustained change using the FOCUS program. This list (with a 1.5-fold change threshold) was therefore analyzed through TELiS without prior filtering. Table 3 shows the transcription factor binding sites significantly overrepresented at each time point (p value Ͻ0.05). The V$ motifs listed in the left-hand column indicate specific sequences in the database, alongside the transcription factor(s) that binds those sequences. Because some transcription factors bind more than one motif, they may be present several times in the same list. At each time point, the binding motifs are listed in order of the number of gene promoters in which they occurred (percentage of all genes screened at each time point, or "% input").
In the lists compiled at 4, 8, and 12 h, we note that binding sites for a number of transcription factor families are overrepresented. Although the presence of a transcription factor binding site within a promoter does not conclusively prove that this factor is involved in regulation, several of the sites identified in this screen represent factors known to cooperate with RTA in transcription activation, including the Sp1, C/EBP, and STAT families and Oct-1 (23-25, 27, 48). Interestingly, at individual time points, specific transcription factor families stand out; the earliest targets of RTA show a high frequency of binding sites for Sp1. Interestingly, several reports have showed that RTA requires Sp1 binding sites to activate early viral genes, including the RTA promoter, ORF60, and the viral thymidine kinase gene (48 -50). Furthermore, both Sp1 and RTA are required for viral DNA replication via the Ori-Lyt sequence element on the viral genome. An increase in Sp1 activity is also closely associated with cell cycle progression and tumorigenesis. Therefore, the ability of RTA to rapidly target Sp1-driven cellular genes represents a potential point at which immediate early viral and cellular genes may be coordinately activated.
By 8 h, more than half of the promoters of the up-regulated genes contain binding sites for C/EBP␤. Similar to Sp1, this family of factors has been reported to cooperate with RTA to activate viral promoters during replication. However, existing reports limit this observation specifically to the C/EBP␣ family member. Both C/EBP␣ and C/EBP␤ can be activated by MAPK phosphorylation, and, because the gene changes also indicate that RTA increases MAPK activity, it is possible that C/EBP␤ is activated indirectly by RTA via phosphorylation. Interestingly, in an adipocyte model, MAPK activates C/EBP␤, which drives several rounds of cell division before activating expression of C/EBP␣, which then inhibits replication (51). We can envisage the same process occurring during virus replication, where early events require stimulation of cell synthesis pathways, followed by cell cycle arrest as resources are channeled into viral DNA and protein pathways.
Neither Sp1 nor C/EBP␤ appear to be overrepresented in the genes consistently up-regulated over a 12-h period. However, we note the appearance of binding sites for interferon regulatory factors. This is noteworthy, given that KSHV RTA has been found to have structural homology with interferon regulatory factor proteins and to be able to activate genes via binding to interferon-stimulated response elements (21).
The most unexpected motifs in the TELiS results, at all time points, are those binding the CREB family of transcription factors. CREB and RTA have many features in common, including a basic leucine zipper DNA binding/dimerization domain, the ability to utilize the CBP/p300 family of coactivator proteins, and phosphorylation and activation by PKA. These proteins are also both able to activate the viral RTA promoter, leading to the conclusion that activation of the CREB pathway can reactivate latent KSHV by driving the lytic cycle (10). However, to our knowledge, this is the first evidence that RTA preferentially targets CREB-responsive promoters on the cellular genome. We therefore further investigated the effects of RTA on CREBdriven transcription.
Dual Effects of RTA on a CREB-driven Target Promoter-The transcription factor binding site screen highlighted 40 RTAresponsive genes with binding sites for CREB family proteins. These proteins bind to both "full" (TGACGTCAA) and also "half" (TGACGT) CREB-response elements (CREs), either as homodimers or cooperatively with other bZIP family transcrip-

TABLE 3 Transcription factors with binding sites overrepresented in promoters of genes up-regulated by RTA at 4, 8, and 12 h
Genes up-regulated more than 1.5-fold at 4 h or up-regulated by more than 2-fold at 8 and 12 h with consistent up-regulation at earlier time points (see Table 2 and "Experimental Procedures") were analyzed using the TELiS database. Results with a p value of less than 0.05 are shown. % input refers to the number of gene promoters bearing each motif, compared with the total number screened at each time point. tion factors (52)(53)(54). CREB-1 can bind to promoters in a transcriptionally inactive form and requires phosphorylation at several amino acids, including the well characterized serine 133 residue, for full activity (55). For this reason, CREB-transfected cells were also treated with the PKA activator dibutyryl-cAMP.
To determine the relationship between RTA and CREB-mediated transcription activation, we chose the promoter of PCSK1 (proprotein convertase subtilisin/Kexin type 1) (up-regulated 12-fold by 12 h) as a model. The PCSK1 promoter has well characterized full and half CREs within 700 bp of the transcription start site (56,57). We first confirmed that a PCSK1 reporter plasmid bearing the CRE (PCSK1Ϫ769; Fig. 4A) was responsive to RTA after transfection into 293 cells (Fig. 4B) and also following transfection and tetracycline induction of the clonal cell lines (Fig. 4C). Because higher levels of tetracycline were used to induce the expression of wild type compared with mutant RTA, we also confirmed that the induction of PCSK1 in the cell lines was not due to tetracycline. Transfection of the PCSK1 reporter into the parent T-REx 293 cell line (i.e. in the absence of RTA) and treatment with 1 g/ml tetracycline did not result in any transcription activation (data not shown).
We then determined if RTA required CREB for activation, using a dominant negative CREB clone, "A-CREB," which specifically inhibits DNA binding by CREB-1 (58). As expected, the CREB-mediated activation of the PCSK1 promoter was highly sensitive to A-CREB (Fig. 4D), but RTA-mediated activation of the same promoter was unaffected (Fig. 4E). To confirm that RTA did not require CRE to activate PCSK1, we mapped the CREB-and RTA-responsive regions, using two PCSK1 deletion constructs illustrated in Fig. 4A. Transfection of either the fulllength reporter (PCSK1Ϫ769) or a deletion lacking the upstream half CREB site (PCSK1Ϫ710) resulted in full activation by both CREB and RTA (Fig. 4F). However, a deletion mutant bearing only the proximal full and half CRE sites (PCSK1Ϫ113) was responsive to CREB but lacked any response to RTA. This indicated that the RTA-responsive region lies between Ϫ710 and Ϫ113 and does not overlap with the CRE.
We then used the minimal PCSK1Ϫ113 reporter to investigate how RTA influences a CREB-driven gene, in the absence of any direct RTA transactivation. Surprisingly, we found that cotransfection of RTA with the minimal CREB-driven promoter (PCSK1Ϫ113) leads to inhibition of activation. When CREB was activated fully by dibutyryl-cAMP, the inhibitory effect of RTA was reduced but still dose-dependent (Fig. 4G). The inhibitory effect of RTA also required the intact activation domain of RTA because the mutant RTA⌬608 -651 was not able to inhibit CREB. We also confirmed that this effect was specific to CREB using a synthetic reporter pCRE-Luc, bearing only multimerized CREB-response elements. As for PCSK1Ϫ113, wild type but not mutant RTA inhibited pCRE-Luc in a dose-dependent manner (data not shown).
Inhibition of CREB as a Mechanism for RTA-mediated Down-regulation-Because RTA expression led to a number of genes being down-regulated in the array, and some of these genes are also CREB-responsive (e.g. HES1, FOXF2, and DUSP1), we asked whether the ability of RTA to inhibit CREB might be responsible for this effect. We chose the DUSP1 promoter to confirm the ability of RTA to inhibit promoter activity and to test whether this inhibition co-localizes with CREB-response elements. The DUSP1 promoter bears one full CRE and three half CREs, clustered within 500 bp of the transcription start site (Fig. 5A). We first compared the transcription activity of the full-length (ϳ3 kb) promoter with a deletion, encompassing all but one of the CREs (DUSP1Ϫ460). Transfection of these reporters into the RTA-expressing cell line confirmed that RTA expression inhibits promoter activity. Transfection of RTA-expression plasmids into 293 cells with these reporter constructs also confirmed dose-dependent inhibition of the DUSP1 promoter by wild type but not mutant RTA (Fig. 5C). In addition, consistent with our hypothesis, the inhibition mapped to the 460-bp fragment of the promoter, bearing the majority of CREB-response elements. Once again, the inhibitory effect of RTA could be reduced by the addition of CREB activator dibutyryl-cAMP, although the effect was still evident. We also confirmed that the inhibition of DUSP1 was not due to tetracycline in the cell line system because the same levels of tetracycline used to induce expression of wild type RTA did not affect DUSP1 promoter activity in the parent T-REx 293 cell line (data not shown).
RTA-specific Coactivators Partially Restore CREB-mediated Transactivation-Finally, we investigated the mechanism of RTA inhibition. We note the activation of CREB by dibutyryl-cAMP partially overcomes inhibition, suggesting some level of competition between CREB and RTA, although RTA has no inherent ability to transactivate CREB promoters. We therefore attempted to strengthen this conclusion by facilitating other points of the CREB transactivation and assaying the extent of inhibition. CREB can be activated by phosphorylation, which promotes its interaction with coactivator CBP. To determine if RTA was interfering with this CREB phosphorylation, we utilized a CREB mutant, Y134F, where the amino acid change mimics constitutive activation at the phosphorylation site (36). Fig. 6A shows that, on the pCRE-Luc reporter, this mutant has higher activity than wild type CREB but is equally sensitive to inhibition by RTA. This suggests that RTA is not interfering with the phosphorylation of CREB. Consistent with this, Western blot using an antibody specific for CREB phosphorylated at Ser 133 showed no effect of RTA on phosphorylation at this site (data not shown). This also showed that the inhibitory effect of RTA could not be explained by any degradation of CREB protein (59), which would also have been visible by Western blot. We therefore looked downstream at the recruitment of CREB coactivator CBP and several other mediators of transcription. We reasoned that wild type RTA, which binds CBP, and also coactivator Med 12 (TRAP230 (17)), may be sequestering these factors, preventing CREB from making functional connections with the core transcription machinery. In support of this, the C-terminal residues responsible for binding these coactivators are also essential for inhibition by RTA (13,17). Fig. 6B shows that overexpressing RTA co-activators CBP and Med12 but not adenovirus E1A and Elk-1-dependent coactivator Med23 partially restores activity to the pCRE-Luc reporter in the presence of RTA. All three of these coactivators also demonstrated some activation of the internal control reporter used to normalize the data, reducing the scale of the final result. Nevertheless, within the limits of this transfection system, the data suggest that the inhibition is due to a lack of productive contacts between CREB and its cofactors, which can be partially reversed by overexpressing these factors. The ability of RTA to bind to specific components of the cellular transcription machinery gives this viral protein the capacity to either activate or inhibit cellular transcription, depending on the transcription factor responsiveness of a given promoter.

DISCUSSION
Viral pathogens are adept at utilizing the cellular environment to support their replication. This involves both stimulating the pathways that support new protein and nucleic acid synthesis and avoiding the cell's stress-induced defense mechanisms (see Refs. 60 and 61 for recent reviews). Because some of these cellular defenses are initiated rapidly after infection, pathogens must also respond rapidly to gain control of the environment. In this respect, the proteins expressed earliest in infection have the greatest influence. Using the ␥-herpesvirus KSHV as a model, we sought to characterize these early effects on the cell, focusing on targets of the immediate early KSHV transactivator RTA. The resulting data included many new targets of RTA, which cannot all be discussed in detail here. However, we have compared the results of our array analysis with the results of an independent study, where the effects of RTA expression in 293 cells were analyzed at 24 h (supplemental Fig. 1). 4 We find that a number of the most highly up-regulated genes from our sustained change list (i.e. significantly upregulated over a period of 12 h) are also strongly represented in this separate study at 24 h. We conclude that these genes are particularly robust targets of RTA, independent of differences in experimental set up or analysis between the studies. For the remainder of the discussion, we will highlight several groups of genes whose expression changes we consider particularly interesting, before discussing the mechanistic insights that came from the TELiS analysis.
Only 4 h after induction of RTA expression, we found cellular genes both up-and down-regulated, by as much as 10 -20-fold. Although well characterized as a transcription activator, this finding demonstrated that RTA expression can also effectively down-regulate cellular gene expression. Furthermore, the identity of the down-regulated genes suggests broad downstream effects on cellular transcription, via the down-regulation of, for example, ID4, the tumor suppressor and inhibitor of basic helix-loop-helix transcription factors, and also HES1, the inhibitor of Notch signaling. It has recently been reported that RTA can inhibit the stability of cellular proteins by acting as an E3 ubiquitin ligase, targeting the proteins for degradation by the proteosome (59). The targets of RTA-mediated degradation also include a number of transcription inhibitors (62), one of which is HEY1, the binding partner of HES1 (63). The activity of the HEY1-HES1 transcription repressor complex is therefore diminished by RTA, via two distinct inhibitory mechanisms. In addition, at 4 h, we find that RTA decreases the expression of a number of proapoptotic proteins, including the RhoB member of the Ras family, transforming growth facto ␤-inducible early gene KLF10 (TIEG1), and caspase 8-associated protein (CASP8AP2). Our data and the data of others therefore support inhibition as an increasingly significant mechanism by which RTA regulates cellular gene expression.
Our attention was also drawn to the inhibition of DUSP1, the dual specificity phosphatase that specifically inhibits MSKs and thereby inhibits cell proliferation (64). Its expression was down-regulated 4-fold shortly after RTA expression, and this 4 -5-fold decrease was sustained over the 12-h period. MSKs stimulate gene expression in response to growth factor signaling and inflammation, by phosphorylating key cellular factors, including CREB, NF-B, and histone H3. A decrease in expression of the MSK inhibitor DUSP1 would therefore be expected to result in higher MSK target phosphorylation and increased activity of immediate early transcription factors that support cell proliferation. This fits well with published data illustrating 4 Su-Fang Lin, unpublished results. the importance of growth factor signaling pathways, specifically the MEK/ERK branch of the MAPK pathway, in KSHV infection. MAPK activity in the host cell is essential for infectivity and the expression of viral early genes after de novo infection (44,65) and reactivation from latency (9). In fact, the MAPK pathway is activated within a target cell at the earliest stages of KSHV infection, when viral glycoprotein B binds to integrin receptors at the cell surface (66). Our array data showed early up-regulation of chemokine receptor CXCR4 and, subsequently, cytokine receptor component gp130 (IL6ST), which both signal through MAPK. We also find an increase in the intracellular adaptor SHC1 (Src homology domain-containing transforming protein 1), which provides an essential link between membrane-bound growth factor receptors at the cell surface and activation of the MEK/ERK branch of the MAPK pathway (reviewed in Ref. 67). We conclude that RTA plays a significant role in maintaining activation of this growth factor pathway, using both positive and negative effects on gene transcription.
The array data also implicate RTA in supporting cell survival via immune evasion. We find that two cellular complement decay-accelerating factors, CD55 and CD59, are up-regulated. Expression of CD55 peaked at 4 h (2.6-fold; Table 1) and declined thereafter. CD59 expression continued to rise over 12 h, from 1.7-to 3-fold, (although its expression pattern did not pass the criteria for TELiS analysis). Increased expression of CD55 and CD59 on the membranes of infected cells allows them to evade complement-mediated lysis in two ways: by accelerating the decay of complement proteins at the cell surface, blocking cell lysis, and also by supplying these proteins for incorporation into newly formed, enveloped virions. Because virions themselves are targets for complement-mediated lysis, this can protect the new viral progeny and deliver CD55 and CD59 to the surface of newly infected cells (68). During lytic replication, KSHV expresses KCP (KSHV complement control protein), a functional homologue of the human RCA protein, which also inhibits complement-mediated lysis of infected cells (69). The related ␥-herpesvirus Herpesvirus saimiri encodes a viral homologue of cellular CD59 (vCD59). Complement evasion is therefore particularly important to this family of viruses, and multiple viral and cellular mechanisms are engaged to maximize the protective effect (70,71).
The second level of analysis, using the TELiS data base, provided a broad view of the cellular transcription factors involved in RTA-induced gene changes at 4, 8, and 12 h. Notably, several factors with documented ability to cooperate with RTA in transcription activation were highlighted in this screen, namely, Oct-1, STAT3, and C/EBP (23)(24)(25)27). The binding site sequence motif for RTA cofactor RBP-J is not present in the TELiS data base. Therefore, using the Transcriptional Regulatory Element Database (Cold Spring Harbor Laboratories) and CisGenome Software (Johns Hopkins School of Public Health), we independently searched for sites in the array data and found them within the promoters of the most rapidly up-regulated genes, including CXCR4, TRIM5, ANXA1, and SHC1. Based on this information, we conclude that the array data represents a broad view of RTA transcription partners and mechanisms.
The presence of CREB binding motifs in the TELiS analysis drew our attention, particularly because RTA does not have any documented ability to interact with CREB. We find that RTA and CREB activate the PCSK1 promoter via independent response elements. More surprisingly, in the absence of an RTA-response element, RTA is a potent inhibitor of CREB. RTA and CREB do share a number of similarities, including the recruitment of CBP as a coactivator (17,18,55). CBP is a large protein capable of interacting with a number of components of the transcription apparatus simultaneously. It is therefore a critical structural component of the transcription machinery, in addition to regulating chromatin remodeling via its inherent acetyl transferase activity. In this respect, it is not surprising that intercepting CBP is a common mechanism for viruses to dysregulate cellular transcription (reviewed in Ref. 72). Consistent with a role for transcription coactivators in the RTAmediated inhibition of CREB, we find that this inhibition can be partially reversed by overexpression of CBP and to a lesser extent RTA mediator Med12. This is supported by the inability of an RTA mutant lacking only the small coactivator binding domain to inhibit CREB. Levels of CBP inside the cell are reported to be limiting for transcription activation, priming the environment for competition between factors that require CBP for transcription (73). Consistent with this, we find that increasing the ability of CREB to bind to CBP, by mimicking phosphorylation of the CBP binding residue serine 133, without increasing intracellular levels of CBP, is insufficient to overcome inhibition (Fig. 6A). However, more general activation of CREB via the PKA pathway (i.e. the addition of dibutyryl-cAMP) does increase activation and diminish the effect of RTA. This may be the result of the broader effects of PKA activation, phosphorylating a number of transcription coactivators, all of which contribute to the elevated level of transcription. CREB itself is also phosphorylated at multiple residues other than serine 133 and binds several other coactivators, including the TORC family of factors (74). These cooperative interactions may also be increased after PKA activation and partially overcome the inhibitory effect of RTA. Taken together, these data suggest that RTA targets CREB-driven genes and that the result of RTA recruitment depends on the sequence motifs present in the promoter. RTA may activate transcription, via a direct response element. However, in the absence of an RTA-response element, the result is inhibition.
The TELiS analysis raises the broader question of how RTA specifically targets gene promoters bearing a number of specific transcription factor binding sites. In this respect, we notice that the factors most strongly overrepresented across all time points (Sp1, STAT, CREB, and C/EBP␤) all associate with CBP and are mediators of the immediate early response. We propose that these two observations are connected and that the ability to RTA to target APR-driven promoters early in the virus replication cycle is due to the defining mechanism of action of these transcription factors. Specifically, these factors are characterized by post-translational activation and rapid recruitment to promoters. In the case of Sp1 and CREB, the factors may even be "preloaded" onto certain promoters, so that phosphorylation events occur in situ and transcription rapidly initiates (75). In the case of STAT proteins and C/EBP␤, activation occurs in the cytoplasm, leading to translocation into the nucleus and transcription activation. We propose that the ability of RTA to bind to common cofactors with these APR factors allows it to colocalize onto critical gene promoters early after infection (Fig.  6C). The outcome of this recruitment then depends on the sequence elements present on the promoter. If the promoter also bears sequences enabling RTA to make contact with the DNA (either directly or via an intermediate, such as RBP-J) and thereby make productive contact with the preinitiation complex, the result is promoter activation (Fig. 6C, middle). However, if the promoter lacks an RTA-response element, the association with critical cofactors is non-productive, and the result is transcription inhibition (Fig. 6C, bottom). This mechanism therefore gives RTA the potential to elicit diverse effects on cellular gene expression, based on the specific cofactors associated with the transcription start site. It is also possible that other RTA cofactors, such as Med12, can facilitate this mechanism and generate some preference for genes regulated by Med12. Interestingly, the only other transcription factor reported to bind directly to Med12 is ␤-catenin (76). Because high levels of ␤-catenin are associated with cell proliferation during viral latency (77), it would be interesting to see if RTA reverses this effect during lytic replication, inhibiting ␤-catenin by interfering with its association with Med12.
KSHV is a large ␥-herpesvirus with many open reading frames expressed during productive replication. Dissecting the effect of each of these viral proteins on the cellular genome is therefore difficult in the context of whole virus infection. Nevertheless, it has been firmly established in recent years that RTA is the most critical factor for initiation of replication, and without RTA, infection is non-productive (13,14). We therefore isolated RTA from the rest of the genome and made a broadly based analysis of its effect on cellular transcription. This provided both new cellular targets of RTA and a potential point of coordination between the viral and cellular "immediate early" response. Several future lines of investigation are possible. Further genome-wide approaches, using transcription factor profiling, could be used to analyze the down-regulated genes and thereby characterize the role of RTA in transcription inhibition in more detail. Alternatively, investigations could focus on RTA-mediated protein-protein interactions at specific promoters, to elucidate how RTA competes effectively with cellular factors to facilitate gene changes that are favorable to virus replication.