Transcriptomic analysis of the dialogue between Pseudorabies virus and porcine epithelial cells during infection

Background Transcriptomic approaches are relevant for studying virus-host cell dialogues to better understand the physiopathology of infection and the immune response at the cellular level. Pseudorabies virus (PrV), a porcine Alphaherpesvirus, is a good model for such studies in pig. Since PrV displays a strong tropism for mucous epithelial cells, we developed a kinetics study of PrV infection in the porcine PK15 epithelial cell line. To identify as completely as possible, viral and cellular genes regulated during infection, we simultaneously analyzed PrV and cellular transcriptome modifications using two microarrays i.e. a laboratory-made combined SLA/PrV microarray, consisting of probes for all PrV genes and for porcine genes contained in the Swine Leukocyte Antigen (SLA) complex, and the porcine generic Qiagen-NRSP8 oligonucleotide microarray. We confirmed the differential expression of a selected set of genes by qRT-PCR and flow cytometry. Results An increase in the number of differentially expressed cellular genes and PrV genes especially from 4 h post-infection (pi) was observed concomitantly with the onset of viral progeny while no early global cellular shutoff was recorded. Many cellular genes were down-regulated from 4 h pi and their number increased until 12 h pi. UL41 transcripts encoding the virion host shutoff protein were first detected as differentially expressed at 8 h pi. The viral gene UL49.5 encoding a TAP inhibitor protein was differentially expressed as soon as 2 h pi, indicating that viral evasion via TAP inhibition may start earlier than the cellular gene shutoff. We found that many biological processes are altered during PrV infection. Indeed, several genes involved in the SLA class I antigenic presentation pathway (SLA-Ia, TAP1, TAP2, PSMB8 and PSMB9), were down-regulated, thus contributing to viral immune escape from this pathway and other genes involved in apoptosis, nucleic acid metabolism, cytoskeleton signaling as well as interferon-mediated antiviral response were also modulated during PrV infection. Conclusion Our results show that the gene expression of both PrV and porcine cells can be analyzed simultaneously with microarrays, providing a chronology of PrV gene transcription, which has never been described before, and a global picture of transcription with a direct temporal link between viral and host gene expression.


Background
In vitro analyses of host cell/pathogen interactions are essential to unravel the mechanisms of infection and to investigate the host response to infection. Pseudorabies virus (PrV) belongs to the Alphaherpesvirinae subfamily as for example the human herpes simplex virus 1 (HSV-1) and is a well-known pig pathogen responsible for Aujeszky's disease, causing considerable economical losses worldwide in this species [1]. Although some countries have succeeded in eradicating Aujeszky's disease through vaccination and health policies, the disease prevalence still remains variable in other countries. Young piglets are more severely affected by PrV infection often resulting in fatal encephalitis, than older infected pigs, which can remain asymptomatic or develop mild to severe respiratory disease symptoms associated with a limited mortality. Indeed, PrV displays a strong tropism for epithelial cells of the oronasal respiratory tract, which are the first cells targeted by virions [1,2]. Abortions, stillbirths or weak piglets that die within 48 h of birth are also observed when pregnant sows are infected [1]. Moreover, PrV can infect a broad range of vertebrates resulting in a uniform lethality but it is generally considered as a nonpathogenic agent for man [3]. Because PrV is easy to propagate in cells of several mammalian species including rodents and is not harmful to laboratory workers, PrV is a highly relevant model to study the biology of alphaherpesviruses and their interactions with host cells in vitro [1]. In addition, its genome has been reconstructed from sequences of six different strains (Kaplan, Becker, Rice, Indiana-Funkhauser, NIA-3, and TNL) and 70 genes encoding structural and non structural proteins have been annotated [4].
Viruses have evolved strategies to evade the host immune response. In particular, herpesviruses interfere with the Major Histocompatibility Complex (MHC) class I antigen presentation pathway to avoid the Cytotoxic T Lymphocyte (CTL) response [5]. MHC class I molecules are expressed on almost all nucleated cells and present peptides, including peptides derived from viral antigens, to CTL, which play a critical role in the defense mechanisms against viral infection. It has been reported that PrV infection decreases the expression of MHC class I molecules on the cell surface [6]. This down-regulation is partly explained by the inhibition of the ABC transporter TAP activity, due to interactions with the viral gN protein encoded by the UL49.5 gene [7,8]. This inhibition is independent of the non-specific mRNA cellular shut-off produced by the virion host shut-off (vhs) protein encoded by the UL41 gene [9]. However, mechanisms other than TAP inhibition may be involved in avoiding the MHC class I presentation pathway.
A precise and more complete identification of cellular and viral genes, which are up-or down-regulated during the time course of infection, is essential to better understand the physiopathology of infection and to identify the molecules involved in host resistance/susceptibility mechanisms. During recent years, DNA microarray technology has proven to be a very efficient high-throughput tool to study the gene expression profiles of infected host cells or pathogens [10,11]. To date, three transcriptomic analyses focused on cellular gene expression have been carried out in non-porcine PrV infected cells [12][13][14]. Ray and Enquist have compared the cellular pathways regulated by PrV and HSV-1 during infection of rat embryonic fibroblast cells using a rat microarray [12]. In a similar system, Brukman and Enquist have explored how PrV evades the IFNmediated immune response [13]. Finally, Blanchard et al have used a human microarray to characterize the impact of PrV infection in human embryonic kidney cells (HEK-293) [14]. These studies have identified many biological processes and host cell genes regulated during infection. The next step in a transcriptomic approach would be the simultaneous analysis of viral and cellular modifications of transcription during PrV infection [10,11] using porcine genomic tools. Since the pig whole genome assembly is not yet achieved, no complete pan-genomic array exists and only partial generic microarrays are commercially available [15]. However, the pig MHC region referred to as the SLA (Swine Leukocyte Antigen) complex, located on chromosome 7, is the first region of the pig genome that is entirely sequenced and annotated [16].
In this context, our aim was to study the dialogue between PrV and the PK15 porcine epithelial cell line, which mimics the first porcine target cells. Using two different porcine microarrays, we followed both the viral and cellular transcriptome kinetics during infection. These microarrays were the Qiagen-NRSP8 commercial array [17] and a microarray we constructed, referred to as SLA/PrV, which combines probe sets specific to genes localized in the SLA complex, genes encoding other important immunological molecules [18] and all the PrV genes. Here, we present a large-scale analysis of the porcine physiological pathways regulated during viral infection with a special focus on genes in the SLA complex together with the modifications of the PrV transcriptome.
The Qiagen-NRSP8 oligonucleotide microarray contains 13297 probes, which match 8541 unique human or mouse RefSeq or pig annotated gene NCBI accession numbers and 1.5% of these encode immune proteins [17]. Only 48 of the 420 probes from the extended SLA probe set and 41 of the 73 from the immune probe set are present on both microarrays.

Expression of PrV genes during the time course of infection
The six time points, which were studied in this experiment -i.e. 0, 1, 2, 4, 8 and 12 hours (h) post-infection (pi), were chosen according to viral growth kinetics observed in PK15 cells in our experimental conditions ( Figure 2). The expression of viral genes was detected between 2 and 12 h pi and increased during time and most of the genes were expressed at 8 and 12 h pi. The hierarchical clustering (HCL) of viral gene expression levels according to all conditions (time and infection status) allowed us to distinguish two main groups: i) mock-infection at all time points and infection until 2 h pi ii) infection from 4 until 12 h (data not shown). With the k-means method, we identified three transcript clusters with similar expression profiles ( Figure 3). The average expression levels for the first cluster (29 probes) showed little variation and only from 8 h pi. The second cluster contained 30 probes corresponding to genes, the expression level of which increased from 4 h pi. The last group (21 probes) displayed a higher increase of expression level from 2 to 8 h pi.
Using a linear model and student t tests, the analysis of viral genes, which were differentially expressed between infected and mock-infected cells at each time pi (FDR = Map of the PrV transcriptome Figure 1 Map of the PrV transcriptome. The relative location of transcripts and designed amplicons are shown. UL and US regions of the PrV genome are represented in black while IR and TR regions are in pink. Transcripts corresponding to each gene are represented with arrows (coding region in orange and non-coding regions or introns in green). Amplicons are represented in blue on the genome. The name of each amplicon is written above the corresponding genome location. Size is measured in kb. 0.05) indicated that three differentially expressed probes were observed as soon as 1 h pi and that this number increased drastically between 2 and 8 h (39 additional probes) and reached a plateau after 8 h pi (Table 1). The important increase between 4 and 8 h confirmed the HCL and k-means results and correlated with the PrV growth curve ( Figure 2). Nineteen probes were not differentially expressed between infected and mock-infected cells at any time and 16 of them were found in the less variable cluster by the k-means method ( Figure 3). Most of the differentially expressed probes belonged to the two most variable clusters. Among the 58 differentially expressed probes, 26 recognized two to four transcripts and 32 were specific to a single transcript (Table 2). When examining these latter 32 genes, the first differentially expressed genes observed 1 h pi were US1 and UL29 encoding two non structural proteins RSP40 and ICP8, respectively, and UL49.5 encoding the virion envelope glycoprotein gN. However, IE180 encoding the immediate early protein IEP, a transactivator of early gene expression was found differentially expressed at 4 h pi. Moreover, the four genes, which were differentially expressed between 8 and 12 h, specify non structural (UL9), capsid (UL28), tegument (UL36) and envelope (US8) transcripts. The differential expression of UL41 encoding vhs protein was first detected at 8 h pi. Among the probes recognizing two transcripts, PKNIA3 specific of US3min and US3maj involved in the inhibition of apoptosis was also differentially expressed from 8 h pi and UL37 specific of UL37min and UL37maj encoding a tegument protein was differentially expressed at 12 h pi.

Global PK15 differential gene expression during the time course of infection
The number of differentially expressed cellular probes increased with time in parallel to viral gene expression (Table 1). Between 0 and 2 h pi most of the SLA/immune probes showed no change and few differentially expressed genes were detected from 1 h pi with the Qiagen-NRPS8 microarray ( Table 1). As shown in Table 1 (Table 1). With the k-means method, the expression levels for each condition (time and infection status) of the SLA/PrV differentially expressed probes set were clustered in three groups ( Figure   Clusters of PrV gene expression levels identified by the k-means method Figure 3 Clusters of PrV gene expression levels identified by the k-means method. The average of normalized intensities for all mock-infected conditions and for each infected condition, centered (median) by genes were analyzed with the k-means method (three clusters). For each cluster, one graph and one clustering picture are represented. The graph shows the mean of the expression levels of all genes (x = Time; y = mean of levels of expression). The clustering picture depicts the mean of each gene expression level for all mock-infected time points and for each infected time point (x = Time; y = level of expression). The list of PrV amplicons belonging to each cluster, the list of the corresponding viral genes and the location of the corresponding proteins in virion structure (for amplicons that hybridize to a single transcript) are represented on the right. The names of the probes that are differentially expressed are represented in blue. To distinguish immediate early, early and late genes, early genes are underlined. All other genes are late genes except IE180 which is the only immediate early gene of PrV.

PrV infection alters multiple biological processes and cellular functions
For each time point, the differentially expressed genes from the Qiagen-NRSP8 microarray were classified into biological processes using GO terms when available ( Table 3). The biological processes that contained more than 5% of the differentially expressed genes during the period between 2 and 8 h pi included: protein metabolism and modification (BP0060), nucleoside, nucleotide and nucleic acid metabolism (BP00031), developmental process (BP00193), signal transduction (BP00102), transport (BP00141), cell cycle (BP00203), immunity and defense (BP00148), intracellular protein traffic (BP00125) and cell structure and motility (BP00285). Several biological processes were predominantly regulated 1 h pi such as developmental processes (BP00193) and signal transduction (BP00102). Other biological processes were regulated later such as cell adhesion (BP00124) or apoptosis (BP00179) from 2 h pi and homeostasis (BP00267) from 4 h pi.
The Ingenuity Pathway Analysis (IPA) of the differentially expressed probes from the Qiagen-NRSP8 microarray identified 82 different top functions associated with significant networks (Table 4)

PrV infection modifies the expression of genes involved in MHC antigenic presentation pathways
The expression of many genes belonging to the SLA class I antigenic presentation pathway was modulated during PrV infection according to the results of both microarrays (Table 5). SLA Ia genes were down-regulated from 4 h pi with the SLA/PrV microarray and from 8 h pi with the Qiagen-NRSP8 microarray. TAP1 and TAP2 genes, encoding molecules involved in peptide transport from the cytosol to the endoplasmic reticulum, were also down-regulated 8 h pi according to the results of the Qiagen-NRSP8 microarray. Surprisingly, TAP1 was up-regulated 8 h pi with the SLA/PrV microarray. PSMB8 (alias LMP7), one of the genes encoding immunoproteasome molecules was up-regulated from 4 h pi on the Qiagen-NRSP8 microarray.
Unexpectedly, our results show that transcript levels of genes belonging to the MHC class II antigenic presentation pathway were also modulated during PrV infection. Expression of SLA-DOB and SLA-DMB decreased at 4 h pi according to the results from the SLA/PrV microarray (Table 5). SLA-DMB was also down-regulated with the Qiagen-NRSP8 microarray 8 h pi while SLA-DOA, SLA-DQA1 and CIITA were up-regulated at this time point (Table 5).

Immune response, apoptosis, nucleic acid binding and actin cytoskeleton pathways are modulated during PrV infection
Among all the biological processes (Table 3) and top functions (Table 4), shown to be regulated during PrV infection, we examined in greater detail genes differentially expressed in four pathways i.e. immune response, apoptosis, nucleic acid binding and actin cytoskeleton (Table 5).
For genes involved in immune response, we observed that CD4 and CD69 were up-and down-regulated from 4 h pi, respectively (Table 5) and that several chemokine ligand and interleukin genes such as IL12A, IL12B and IL17 were down-regulated at 4 and 8 h pi. These observations have a poor biological significance. Since the relevant gene products are known to be specific of immune cells, it is probable that these transcript expressions are not correlated UL14 UL10 ----6.6 6.1 gM; inhibits glycoprotein-mediated membrane fusion; type III membrane protein  US2 ---6.2 6.9 7.1 28K, tegument protein; membrane associated protein tegument *gene function and structural role [1] are only specified for amplicons that hybridize to a single transcript.
-gene not found differentially expressed between PrV infected and mock-infected cells with significant protein synthesis in the present epithelial cell context. Among the genes involved in interferonmediated immunity, many were modulated during PrV infection ( Clusters of PK15 gene expression levels identified by the k-means method   T8  MI T0 T1  T4  T12  T2   I   T8  MI T0 T1  T4  T12  T2   I   Cluster 1   T8  MI  T0  T1  T4  T12  T2   I   1   0   -1   Cluster 2   T8  MI  T0  T1  T4  T12  T2   I   1   0   -1   Cluster 3   T8  MI  T0  T1  T4  T12  T2   I For the apoptosis pathway, genes belonging to the BCL-2 molecule family, FAIM2, CASP1 and CASP3 were downregulated whereas CASP7 and NF-KB2 were up-regulated. Transcript levels of JAK1, XBP1, ATF4 and HSPA5 were decreased at 8 h pi. HSPA1A, HSPA1B, HSPA2, HSPA4, HSPA4K and HSPA8 were up-regulated at 8 h pi and EIF2A from 4 h pi. HSPA6 was first down-regulated at 2 h pi and then up-regulated from 4 h pi. Several differentially expressed genes, which belong to the apoptosis pathway, were also involved in the stress response. Among the differentially expressed genes that play a role in nucleoside, nucleotide and nucleic acid binding, the expression of histone genes HIST1H2AL, HIST1H4J, HIST1H2BK was decreased from 4 h pi. Expression of several histone deacetylases was also regulated during infection: HDAC2 and HDAC10 expression levels increased, while HDAC3, HDAC6, and HDAC9 expression decreased. HDAC2 and HDAC9 were regulated early from 2 h pi. Several genes encoding signal transducers and activators of transcription (STAT1, STAT3, STAT5B and STAT6) were down-regulated during PrV infection.
Within the actin cytoskeleton pathway, ACTG1 was upregulated very early i.e. as soon as 1 h pi. Other genes such as ACTC1, ACTRT2, ACTA4, MYO1D and MYO5A were all down-regulated from 4 or 8 h pi and ACTL6A was up-regulated at 8 h pi.

Validation of microarray results by quantitative real-time PCR (qRT-PCR)
Five genes involved in the presentation antigen class I pathway were studied by qRT-PCR: SLA Ia, TAP1, TAP2, PSMB8 and PSMB9. PPIA, down-regulated during infec-  tion, and TNF, even if not detected as differentially expressed in our transcriptome experiment, were also chosen for validation (Table 6). qRT-PCR were performed for a subset of conditions at 0, 2, 4, 8, and 12 h pi (see materials and methods). We confirmed that SLA Ia genes were down-regulated during infection from 8 h pi. We also observed a clear down-regulation of TAP1 and TAP2 from 8 and 4 h pi, respectively. An early down-regulation of PSMB8 and PSMB9 was detected before 2 h pi. TNF was strongly up-regulated from 4 h pi and PPIA was down-regulated from 2 h pi.   cell transcriptome modifications can be examined with a unique microarray combining viral and host cell probe sets. Indeed, a majority of transcriptomic studies have focused either on host or on pathogen gene expression profiling [11,22,23] and only a few studies report the simultaneous detection of pathogen and mammalian host transcriptomes i.e. Plasmodium berghei ANKA and mouse [24] and EBV-NK/T cell lymphoma and man [25]. Our work confirms the feasibility and the relevance of this kind of approach to establish a direct link between pathogen and cellular gene expression. In order to explore porcine cellular gene expression with even more detail, we chose to supplement the SLA/PrV microarray with the Qiagen-NRSP8 microarray. The sensitivity of each microarray differed according to the nature of the probes (70mer oligonucleotides versus DNA/cDNA) as shown by comparative studies [26]. Seventy-mer oligonucleotides give better results in terms of specificity and sensitivity compared to cDNA microarrays and this could explain some discrepancies observed between both microarrays in particular for the TAP1 gene [27] as confirmed in our study. With this integrated approach, a parallel increase in the number of differentially expressed PrV and cellular genes was detected illustrating the viral and cellular transcript modifications during infection.

A picture of PrV gene transcription during PK15 cell line infection
In our experimental conditions, we obtain a picture of the global PrV gene transcription during the lytic cycle. PrV transcription was monitored in single cycle conditions using a high MOI that guarantees that more that 90% cells are infected. Despite the presence of nested transcription units in PrV preventing the design of probes specific of unique transcripts for some genes, we were, however, able to confidently report viral gene expression for probes specific of unique viral transcripts and draw a general picture of PrV transcription during the time course of infection.  with viral growth and thus coincides with the beginning of the release of extracellular progeny (see Figure 2, Figure  3, Table 1 and Table 2). It has already been reported that the beginning of viral progeny usually occurs between 4 and 5 h pi but without any description of the global viral transcription [1]. We observe a continuous increase in transcript levels and in the number of differentially expressed viral probes between 4 and 8 h pi followed by a stabilization when virion production is maximum. This suggests that the transcriptional machinery is fully active at 8 h pi thus permitting a massive virion production. All the different classes of viral transcripts are represented SLA I cell surface decrease on PrV infected PK15 cells   [1]. After binding of the viral particle and fusion of the virion envelope with the cell membrane, the release of capsid and tegument proteins into the cell and the takeover of host cell protein synthesis machinery, the IE180 protein encoded by the immediate early gene is expressed in the cytoplasm and translocated to the nucleus. This protein can further transactivate the RNA polymerase II mediated transcription of early genes including transactivators of transcription (EP0, US1, UL54) and proteins essential for viral replication (UL5, UL8, UL9, UL29, UL30, UL42, UL52, UL28, UL39, UL40, UL23, UL50, UL12, UL2) [1]. The expression of IE180 has been reported to begin between 40 min and 1 h pi and last until 3 h pi [1]. In our experiment, the IE180 probe (IEP2) was differentially expressed only at 4 h pi, when the transcript level probably reaches its peak value. This suggests that a low level of IE180 transcripts is sufficient to induce the transcription of early genes. In this experiment, the differential expression of US1 and UL29 was detected as early as 1 h pi but other early genes appeared differentially expressed later. Interestingly, the UL49.5 probe corresponding to the gN protein, responsible for TAP inhibition, was differentially expressed at 1 h pi, even if this gene is not described as an early gene. The synthesis of late proteins, such as capsid (UL6, UL18, UL19, UL25, UL35 and UL38), tegument (UL11, UL13, UL16, UL21, UL36, UL37, UL37, UL41, UL46, UL47, UL48, UL49, UL51, US3, US2) and envelope proteins (gE, gI, gD, gM, gH, gC, gB, gN, gK) are reported to occur during the PrV replication cycle [1]. In our study, the two late transcripts UL6 and UL22 encoding the gH protein were differentially expressed as early as 2 h pi. The four latest differentially expressed genes mostly encoded envelope or tegument proteins, except UL9.
We described for the first time a global analysis of PrV gene transcription using a microarray. The results of our analysis is consistent with what is known about PrV viral cycle and with the kinetic classification of individual transcripts [1]. Similar approaches have been developed for other alphaherpesviruses such as HSV-1 [28,29] and Varicella-Zoster virus (VZV) [28,29]. It is difficult to compare our results with those obtained in the VZV study because this viral system does not allow cell infection under single cycle synchronized conditions (one limitation of this viral system), which is required to establish reliable kinetics of viral gene expression. However our results are consistent with the transcriptomic study reported for HSV-1 [28,29]. It is clear from figure 3 that PrV early homologues of HSV-1 immediate early genes (EP0, UL54, US1) and early genes (UL23 encoding thymidine kinase, UL30 encoding a DNA polymerase subunit, UL39 encoding the large subunit of ribonucleotide reductase) are expressed at early times before most of the late genes encoding structural proteins. A clear distinction between immediate early, early and late genes for PrV will require transcriptomic analysis in the presence of the translation inhibitor cycloheximide (to identify immediate early genes) or the viral DNA replication inhibitor phosphonoacetic acid (to distinguish early and late genes) as was done for HSV-1 [28,29].

PrV and cellular shutoff
A cellular shutoff during infection has been described for herpesviruses including PrV [1]. In our experiment, a shutoff of PK15 genes is observed during infection since many cellular genes are down-regulated between 4 and 12 h pi.
In contrast, at the 4 h time point, 42.5 % of the viral genes are up-regulated. Our transcriptomic analyses reveal that the shutoff occurs in porcine cells earlier than that previously reported in other transcriptome studies i.e. between 8 and 12 h pi in rat embryonic fibroblasts and from 6 h pi in human embryonic kidney cells [12,14]. It is assumed that the virion host shutoff protein (vhs/UL41 transcript) causes cellular shutoff. The vhs protein is an RNAse located in the viral tegument, which degrades host and viral RNA just after infection for HSV-1 [30]. Unlike HSV-1, it has been suggested that for cellular shutoff, PrV requires a fresh round of viral protein synthesis explaining the observed delayed shutoff [9]. In our experiment, UL41 transcripts appear to be differentially expressed only at 8 h pi suggesting that the vhs activity can be attributed to the newly synthesized proteins and not to the vhs proteins present in the virion tegument at the moment of infection and that the vhs protein should be active at low level. The activity of HSV-1 vhs is modulated by the UL48 product (VP16), which can bind to vhs to allow viral mRNA accumulation [31]. However, our study does not show any differential expression of the UL48 transcript.

PrV infection and immune evasion strategies
To evade host response PrV develops several strategies that probably disturb different biological pathways including the MHC class I presentation pathway. We observed a decrease of SLA-Ia and TAP2 transcript levels in PK15 cells infected with the PrV NIA3 strain as previously reported in infected PK15 and bovine kidney cells respectively (IND-F PrV strain) [9]. A down-regulation of TAP1 and TAP2 genes encoding immunoproteasome catalytic subunits, PSMB8 and PSMB9, involved in the MHC class I antigenic presentation pathway was also detected in our experiment. Moreover, we checked that at 8 h pi the PK15 cells expressed 50% less MHC class I proteins than mockinfected cells in our culture conditions. These results confirm previous reports describing the reduced capacity of infected cells to present viral peptides to CTL [6,7]. The viral gene UL49.5, encoding the gN protein, is one of the earliest differentially expressed genes in our study (from 1 h pi). This viral protein has been shown to inhibit TAP activity and induce degradation of TAP molecules by the proteasome [5,8,32]. Our results strongly suggest a very early production of gN protein and agree with the detection of TAP inhibition from 2 h pi [7]. This TAP inhibition has been shown to be independent of vhs activity [9] and we demonstrate here that UL41 encoding vhs is differentially expressed later than UL49.5, indicating two successive steps i.e. TAP inhibition followed by cellular shutoff.
Since the level of several transcripts involved in the MHC class I presentation pathway (MHC class Ia, TAP1, TAP2, PSMB8 and PSMB9) decreased, it is possible that PrV has developed complementary strategies to evade this pathway i.e. turning off the peptide pump with inhibition of TAP activity and transcription alteration of key players [5].
Other viruses, such as the human cytomegalovirus, downregulate the transcription of key players of the MHC class I antigen presentation pathway [5]. Unexpectedly, some MHC class II genes were also regulated during PrV infection in PK15 cells. In particular, a down-regulation of class II-like chaperones SLA-DOB and SLA-DMB was observed. MHC class II molecules that are constitutively expressed on professional antigen presenting cells (APCs), present peptides derived from exogenous antigens to CD4+ T-helper cells playing an important role in the induction and maintenance of CTL immunity. Epithelial cells can also constitutively express MHC II molecules but at a lower level than professional APCs [33]. In our experiment, we detected a small subpopulation of uninfected PK15 cells constitutively expressing MHC class II molecules. However, we did not detect modifications of the MHC class II expression at the cell surface. Since a downregulation of constitutive and IFNγ induced HLA class II expression has been observed in cells infected by other herpesviruses [33], our preliminary data suggest that it would be highly relevant to explore how PrV may interfere with the MHC class II presentation pathway in professional APCs.
In addition to genes belonging to MHC antigen presentation pathways, several other genes, playing a role in antiviral response are regulated during PrV infection such as genes belonging to the IFN signaling pathway. Indeed, it has been reported that in primary rat fibroblasts, PrV infection could suppress the establishment of the IFNβinduced viral state [13]. IRF3, which is a transcriptional factor involved in IFNβ production by epithelial cells, is down-regulated together with a set of other IRF. Constitutively expressed in the cytosol, IRF3 is phosphorylated during herpesvirus infection and translocated into the nucleus to target the IFNβ promoter. Many viruses interfere with IRF activities [34]. A decrease of IRF1 mRNA and protein levels has also been detected in cells infected with hepatitis C virus, resulting in the transcriptional repression of several IFN-stimulated genes [35]. In addition, TNF-alpha, which is a multifunctional cytokine with potent antiviral activities and which mediates protection against HSV-1 in the mouse [36] was analyzed by qRT-PCR. A strong up-regulation of TNF was detected from the beginning of PrV infection and until 12 h pi. These results suggest that the TNF transcription increase that is usually expected during an infection is not suppressed by PrV infection and that the cellular shutoff does not target TNF.

Other cellular pathways modulated during PrV infection
Our transcriptome analysis confirms that many other biological processes and functions are modulated during PrV infection in porcine PK15 cells as previously observed in rat embryonic fibroblast and human embryonic kidney cells [12,14]. We have focused our study on a limited number of pathways and genes. Interestingly, we observe both by transcriptome analysis and qRT-PCR that PPIA gene expression is clearly down-regulated during PrV infection. PPIA encodes cyclophilin A, a peptidyl-prolyl isomerase, which catalyzes the isomerization of peptide bonds from the trans to cis form at proline residues and facilitates protein folding [37] and which acts as a cytosolic molecular chaperone. Cyclophilins have been discovered because of their high affinity for cyclosporine, an immuno-suppressive drug, which prevents allograft rejection. This immunosuppressive effect is due to the calcineurin inhibition by a cyclosporin-cyclophilin complex. Calcineurin is required for transcriptional activation of many cytokines in stimulated T cells. Cyclophilin A can also interact with HIV-1 Gag polyprotein and is involved in HIV-1 replication kinetics and modifies the infectivity of HIV-1 virions in Jurkat T cells [38]. Indeed, virions produced by PPIA -/cells are less infectious than virions produced by PPIA +/+ cells. Since we observed a downregulation of PPIA before the global cellular shutoff, cyclophilin A may be a target for PrV and play a role in infection via an unknown mechanism.
Several cellular genes involved in apoptosis were regulated during PrV infection such as BCl-2 molecules and caspases. Viral infection of mammalian cells tends to generate proapoptotic signals to limit viral replication but viruses and, in particular, herpesviruses produce molecules acting as modulators of apoptosis [1]. Thus, US3 products from PrV play an anti-apoptotic role [39]. The PIKNIA3 probe specific to US3 long and short isoform transcripts was up-regulated from 8 h pi in our experiment, suggesting a possible late antiapoptotic role of the US3 products. In addition, PrV genes homologous to other HSV-1 antiapoptotic genes may also possess an antiapoptotic role such as UL54 or US1 [1]. UL54 was not differentially expressed in our experiment in contrast to US1, which was up-regulated very early as soon as 1 h pi.
Genes belonging to nucleic acid metabolism were differentially expressed from very early time points. A repression of many genes encoding histones and nucleosome components occurred in PK15 cells during infection. These results are concordant with a former study, which has shown a gradual inhibition of histone synthesis in RK13 rabbit cells during PrV infection [40]. We also observed a modulation of many histone deacetylases (HDAC). Acetylation of newly synthesized histones is required for their assembly into nucleosomes by histone chaperones and regulates the formation of heterochromatin that is critical for cellular gene transcription. US3, which was up-regulated from 8 h pi in our experiment, can suppress histone acetylation during HSV1 infection [41]. Indeed, PrV US3 could also inhibit histone acetylation during infection.
PrV infection regulates the expression of several genes involved in actin cytoskeleton signaling. A probe specific to ACTG1 was strongly up-regulated as soon as 1 h pi and reached a peak at 4 h pi. Cytoskeleton actin is involved in PrV assembly and in virus movement within the host cell.
In particular, viral capsids can travel along nuclear actin filaments using myosin-directed transport in neurons but also in PK15 cells [42]. Moreover, actins present in the nucleus participate in transcription [43]. Among PrV proteins, the US3 protein kinase contributes to cellular cytoskeleton modifications via the formation of actinand microtubule-containing cell projections, a phenomenon associated with an increase of PrV intercellular spread [44].

Conclusion
The originality of our approach lies in the simultaneous investigation of transcript levels of both host and pathogen genomes using a partial generic microarray and a dedicated microarray (SLA/PrV) combining all the PrV genes and probes from the SLA complex. It is now necessary to extend our analysis of the interactions between PrV and porcine cells to other target cells, such as immature dendritic cells (iDC) that are the first immune cells interacting with the virus. This kind of approach should also be efficient (i) to study viral and cellular gene expression using mutant viruses in order to better understand the role of each viral gene and (ii) to help identify species or strain specific transcriptomic signatures in host cells.  [47] and porcine cDNA clones were selected when they were available in the AGENAE library [19] or in two American libraries MARC1PIG and MARC2PIG constructed by the United States Department of Agriculture (USDA). When no porcine cDNA clone was found, exons were identified by comparison between ESTs and genomic DNA using the ICCARE comparison mapping tool [48,49]. Primers were designed with Primer 3
To design the PrV amplicons, we first established a complete composite sequence of the PrV genome by merging genomic sequences available in Genbank from seven strains of PrV. Positions of the 5' and 3' ends of viral transcripts (true or putative) as well as ORFs were established according to published data or Genbank annotations (F. Lefèvre, unpublished). The location of the 80 amplicons, at least one per transcript, was chosen according to this map, generally close to the 3' end of the transcripts ( Figure  1). In the case of nested transcription units, we designed as many amplicons as the number of nested transcripts so that each amplicon was located between the 5' ends of two consecutive transcripts. We used Primer 3 to design primers for amplification ( The SLA/PrV DNA/cDNA microarray platform has been submitted to the Gene Expression Omnibus (GEO) repository [52]. The accession number is GPL5622.

The generic porcine array
The second microarray is a commercial generic microarray spotted on slides (Qiagen-NRSP8) and contains 13297 oligonucleotides (70-mers) specific of 8541 porcine genes [17]. The microarrays used in this study were provided by Dr Max Rothschild (Department of Animal Science, Iowa State University). We used the annotation of Qiagen-NRSP8 slides given by Zhao and collaborators [17].  [20].

RNA labeling, hybridization scheme, microarray scanning and signal quantification
The SLA/PrV and the Qiagen-NRSP8 microarray data have been submitted to the GEO and received accession numbers GSE8676 and GSE9259, respectively.

Statistical data analysis
The normalization and statistical analysis steps were performed with scripts written with R software. Functions contained in stats, anapuce and varmixt packages were used. Raw data were log2 transformed, normalized by lowess (f = 0.3) and the median of each block spotted on the slide was subtracted. Normalized data were analyzed with a linear model. For one gene and one target, the following linear model was used: Hybridization scheme of the SLA/PrV and the Qiagen-NRSP8 microarrays Figure 6 Hybridization scheme of the SLA/PrV and the Qiagen-NRSP8 microarrays. A Hybridization scheme of the SLA/PrV microarrays. Forty-eight hybridizations corresponding to 24 PrV/SLA slides were hybridized (two arrays per slide, see Methods). B Hybridization scheme of the Qiagen-NRSP8 microarrays. Thirty-two Qiagen-NRSP8 slides were hybridized. Each labeled target is represented by its dye Cy3 or Cy5. Each arrow linking Cy3 to Cy5 represents one hybridization on one array. T0  T1  T2  T4  T8  T12   I   MI   I  with Y cfnds = normalized intensity, μ c = condition effect (1 ≤ c ≤ o), α f = fluorochrome effect (1 ≤ f ≤ 2), β n = network effect (1 ≤ n ≤ q), δ d = effect of experiment day (1 ≤ d ≤ 2), ε s = sample effect (1 ≤ s ≤ r), E cfnds = random error, o = number of conditions, q = number of arrays, r = number of samples. Since the effect of samples was not significant (Student t test, p > 0.05), we did not keep it in the final linear model. Selected contrasts (defined by a linear combination of 2 or more factor level means with coefficients that sum to zero) were tested with the Student t test: infected condition against mock-infected condition at each time point. The differences observed for one time point were tested taking into account all the data for all the time points leading to a powerful statistical test. The false Discovery Rate (FDR) was calculated after the Student t test (R software, varmixt package). We selected differentially expressed genes with an FDR = 0.05.

PK15 cells
Hierarchical clustering analysis (HCL, Eisen) was performed to analyze viral genes and cellular genes that were differentially expressed during infection (euclidian distance, average linkage) using TMeV software [53]. Averages of normalized intensities for all the mock-infected conditions and for each infected condition were calculated and these values were centered (median) by genes. The "k-means" method was used to identify groups of genes with similar expression kinetics (TMeV software, [53]). The optimal number of groups was determined by the " Figure of Merit" method (FOM).

Functional analysis/Network and pathway analysis
The "Gene Ontology" annotations of differentially expressed cellular genes were obtained with the Panther software based on human orthologous genes [54]. The biological process sub-ontology was used to classify genes in different functional groups. The Ingenuity Pathways Analysis (IPA) program permitted the determination of the significant networks and the top functions associated with the differentially expressed genes at each time point (IPA 5.0, Ingenuity Systems Inc., USA; [55]). The IPA program searches the Ingenuity Pathway Knowledge Base (IPKB) for interactions (known from the literature) between the uploaded genes and all other genes contained in IPKB and generates a series of networks. The genes selected by the IPA network analysis are called focus genes. The Fisher exact test was used to assign statistical significance, and each network's score is displayed as thelog (P-value). Significant networks with a score greater than 3 were selected (p < 0.001) and top functions associated with these networks are presented (Table 4).

Quantitative real time RT-PCR (qRT-PCR)
RNAs corresponding to mock-infected T0 and T1 and infected T0, T2, T4, T8, T12 samples from three replicates were used for real-time PCR analysis. Primers were chosen with the Primer express software or manually, one of each primer couple located on the exon-exon boundary ( Table  7). Five differentially expressed porcine genes and two other porcine genes were chosen for real-time PCR testing. Each amplicon was sequenced and its sequence was aligned with the gene reference sequence. Two and a half μg of cleaned total RNA were reverse transcribed using Superscript II enzyme (Invitrogen, France) with Oligo(dT) primers (Invitrogen, France) and cDNAs were quantified with the 2100 Bioanalyzer (Agilent Technologies, France) and diluted to obtain a 30 ng/μl final concentration for all samples. Triplicate reactions were set up in a 20μl volume using 10 ng cDNA, PCR primers (300 nM final concentration) and SYBR Green PCR Master Mix (Applied Biosystem, USA). An ABI PRISM 7900 HT sequence detection system was used for monitoring the level of SYBR green fluorescence. A dissociation curve was produced at the end of the cycling phase to ensure a single PCR product and no primer dimer. Amplification efficiencies were determined for all genes by serial dilution of cDNA using E = 10 [-1/slope] . We checked that a slope between -3.2 and -3.5 was obtained for each primer couple to calculate the relative expression levels of the selected genes by using the 2 -ΔΔCt method. The gene encoding Ribosomal Protein L32 (RPL32) was chosen as the internal reference for each sample. The mock-infected sample at T0 of each replicate was used as a calibrator for all samples of one replicate. The mean and the standard deviations for each time of each replicate were calculated.

Flow cytometry
The anti-porcine MHC class I monoclonal antibody (mAb) PT85A and the anti-porcine MHC class II mAb MSA3 were purchased from VMRD (USA). The mAb HOPC-1 (IgG2a, Beckman Coulter, USA) was used as control Ab. PE-conjugated goat Abs to mouse IgG2a were purchased from Southern Biotech (USA).