Release of Severe Acute Respiratory Syndrome Coronavirus Nuclear Import Block Enhances Host Transcription in Human Lung Cells

Raw microarray data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus database (30) and are accessible through the GEO series under accession number GSE33267 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE33267). Raw proteomics data corresponding to peptide identifications used to populate the AMT tag database are available at the PRoteomics IDEntification (PRIDE) database (http://www.ebi.ac.uk/pride/) under the project name “A Systems Biology Approach to Emerging Respiratory Viral Diseases” in the PRIDE Public Projects folder and corresponding to PRIDE accession numbers 19877 to 19890. The raw quantitative proteomics data can be accessed at the Pacific Northwest National Laboratory (PNNL) Biological Mass Spectrometry (MS) Data and Software Distribution Center (http://omics.pnl.gov/) in the Systems Virology Contract Data folder within the Browse Available Data folder. All data sets and associated metadata have been submitted to the Virus Pathogen Resource (ViPR; http://www.viprbrc.org). Additional details from this study and similar studies can be accessed through the Systems Virology website (http://www.systemsvirology.org).

Infections for microarray and proteomics analyses and validation studies were performed in a clonal population of Calu3 cells (human lung adenocarcinoma cells) sorted for high levels of expression of the SARS-CoV cellular receptor angiotensin-converting enzyme 2 (ACE2), referred to as Calu3 2B4 cells (31). Calu3 2B4 cells were grown in minimal essential media (MEM; Invitrogen-Gibco) containing 20% fetal bovine serum (HyClone) and 1% antibiotic-antimycotic mixture (Invitrogen-Gibco). Viral titration assays were performed in Vero E6 cells. Vero E6 cells were maintained in MEM (Invitrogen-Gibco) containing 10% fetal clone II (HyClone) and 1% antibiotic-antimycotic (Invitrogen-Gibco).

Human tracheobronchial epithelial cells were obtained from airway specimens resected from patients undergoing surgery under University of North Carolina Institutional Review Board-approved protocols by the Cystic Fibrosis Center Tissue Culture Core. Primary cells were expanded to generate passage 1 cells, and passage 2 cells were plated at a density of 2.5 × 10⁵ cells per well on permeable Transwell-COL (12-mm-diameter) supports. Human airway epithelium (HAE) cultures were generated by provision of an air-liquid interface for 6 to 8 weeks to form well-differentiated, polarized cultures that resembled in vivo pseudostratified mucociliary epithelium (32).

Wild-type infectious clone-derived SARS-CoV (icSARS-CoV) and icSARS-CoV ΔORF6 and their corresponding MA15 mouse-adapted variants were derived from the Baric laboratory's infectious clone constructs as previously described (see Section S1 and Fig. S1 in the supplemental material) (9, 33, 34). Briefly, genome fragments were amplified in Escherichia coli, excised from plasmids by restriction digestion, ligated, and purified prior to in vitro transcription reactions to synthesize full-length genomic RNA, which was transfected into Vero E6 cells. The media from transfected cells were harvested and served as seed stocks for subsequent experiments. Viral genomes were confirmed by sequence analyses prior to use in any experiments. All work was performed in a biosafety level 3 facility supported by redundant fans.

Relative quantities of viral genomic or subgenomic mRNA were determined by quantitative real-time PCR (qRT-PCR). First-strand cDNA synthesis was performed using 500 ng of total RNA and ThermoScript reverse transcriptase (Invitrogen) according to the manufacturer's protocol. The qPCR assay was performed using a SYBR green kit (Applied Biosystems, Carlsbad, CA) with specific primers for the different RNA species, according to the manufacturer's standard protocol. Relative RNA quantities were determined using the comparative threshold cycle (C_T) method, with human RPL14R2 serving as the endogenous reference and mock-infected samples serving as the calibrators. Primer sequences are available upon request.

To determine the pattern of differential gene expression or protein expression for icSARS-CoV-infected, icSARS-CoV ΔORF6-infected, and mock-infected cells, Calu3 2B4 cells were plated in triplicate under each condition at each time point, washed prior to infection, infected at a multiplicity of infection of 5 (MOI 5), and incubated at 37°C for 40 min. The inoculum was removed, cells were washed 3 times with 1× phosphate-buffered saline (PBS), and then fresh medium was added prior to time zero. For both microarray and proteomic analyses, at 0, 3, 7, 12, 24, 30, 36, 48, 54, 60, and 72 h postinfection, medium was collected to determine viral titers at each time point for each well and cells were either washed in 1× PBS and then harvested in TRIzol (Invitrogen) and stored at −80°C (for RNA) or washed 3 times in cold 150 mM ammonium bicarbonate buffer, lysed for 15 min in 8 M urea, and stored at −80°C (for protein).

Infection of HAE cultures with icSARS-CoV and icSARS-CoV ΔORF6 was performed as previously described by our group (35–37). Briefly, triplicate cultures were washed with 1× PBS, and 200 μl of mock, icSARS-CoV, and icSARS-CoV ΔORF6 inocula was added to the apical surface. Cultures were incubated at 37°C for 2 h; the inoculum was removed, and unbound viruses were removed by washing three times with 1× PBS. Apical wash samples were harvested to analyze viral growth kinetics at 2, 24, 48, and 72 h postinfection and were assayed by plaque assay in Vero E6 cells (36).

RNA isolation from Calu3 2B4 cells and the subsequent microarray processing and identification of differentially expressed transcripts were performed as described previously (38). All probes were required to pass Agilent quality control (QC) flags for all replicates of at least one infection time point (33,631 probes passed). Differential expression was determined by comparing icSARS-CoV- and icSARS-CoV ΔORF6-infected replicates to mock-infected replicates for each time point, based on a linear model fit for each transcript. Criteria for differential expression were an absolute log₂-fold change of 1.5 and a false discovery rate (FDR)-adjusted P value of <0.05 for a given time point. Differential expression was also calculated directly between icSARS-CoV and icSARS-CoV ΔORF6 for each time point by using the criteria of a 2-fold change and an FDR-adjusted P value of <0.05. Significant transcript values were transformed for clustering and network analysis to the fold change (log₂) of icSARS-CoV- or icSARS-CoV ΔORF6-infected samples compared to time-matched mock-infected samples.

To process primary HAE samples, both the apical and basolateral surfaces of the cultures were washed 3 times in 1× PBS and stored at −80°C in RNAlater (Ambion/Invitrogen). To isolate total RNA, each membrane was washed twice with 500 μl of 4 M guanidinium thiocynate, 25 mM sodium citrate, 0.5% Sarkosyl, 0.1 M 2-mercaptoethanol. The washes were combined and, following shearing of the DNA, total RNA was isolated through phenol-chloroform extraction. The RNA was further purified using Qiagen RNeasy minicolumns per the manufacturer's instructions (38). Equivalent amounts of RNA from three biological replicates from each condition were pooled. Microarray analysis was performed as previously described (39) using Agilent 4x44K whole human gene expression microarrays.

Functional enrichment statistics and network analysis were determined using DAVID (40, 41) and Metacore (GeneGo, St. Joseph, MI) to identify the most significant processes affected by infection. The DAVID functional annotation tool utilizes the Fisher exact test to measure gene enrichment in Gene Ontology (GO) biological process category terms for significant genes compared to background, which included all genes on the Agilent platform that passed the QC criteria. To identify major transcriptional regulators whose nuclear import is controlled by karyopherins, the statistical Interactome tool in MetaCore was used to measure the interconnectedness of genes in the experimental data set relative to all known interactions in the background data set. Statistical significance of overconnected interactions was calculated using a hypergeometric distribution, where the P value represented the probability of a particular mapping arising by chance for experimental data compared to the background (42). In order to determine the consequence of removal of the nuclear import block in SARS-ΔORF6 infection, significantly overconnected transcription factors were filtered for those whose transport is regulated by karyopherins in the cell. Networks were constructed in MetaCore for experimental data by using an algorithm that identified the shortest path to directly connect nodes in the data set to transcription factors. Network visualizations were created in MetaCore or Cytoscape (43).

The detailed proteomic methodology, including sample preparation, processing, and analysis methods, are provided in Section S2 of the supplemental material. Cell lysates were trypsin digested and fractionated by strong cation exchange as previously described (44, 45). A novel accurate mass and time (AMT) tag database (46) for virus-infected Calu3 2B4 cells was generated by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis (44, 47), using combined aliquots of the icSARS-CoV-, icSARS-CoV ΔORF6-, and mock-infected samples. Following AMT tag database generation, LC-MS analyses were performed on all icSARS-CoV-, icSARS-CoV ΔORF6-, and mock-infected samples to generate quantitative data, using identical chromatographic and electrospray conditions as for LC-MS/MS analyses. The LC system was interfaced to an Exactive mass spectrometer (ThermoScientific), and the temperature of the heated capillary and the electrospray ionization voltage were 250°C and 2.2 kV, respectively. Data were collected over the mass range 400 to 2,000 m/z. Quantitative LC-MS data sets were processed using the PRISM data analysis system (48), which is a series of software tools developed in-house (e.g., Decon2LS [49] and VIPER [50], which is freely available at http://ncrr.pnl.gov/software/). Individual steps in this data processing approach are reviewed in reference 46. The peak intensity values (i.e., abundances) for the final peptide identifications were processed in a series of steps using MatLab R2010b, including quality control (51, 52), normalization (53), and quantification to protein level (54). Comparative statistical analyses of time-matched mock-infected with icSARS-CoV- and icSARS-CoV ΔORF6-infected samples was performed using a Dunnett adjusted t test to assess differences in protein average abundance, and a G-test was used to assess associations among factors due to the presence/absence of a response (51).

Chromatin immunoprecipitation (ChIP) analysis was performed by using the ChIP assay kit (Millipore). Briefly, Calu3 2B4 cells were infected with icSARS-CoV or icSARS-CoV ΔORF6 or mock infected, as described above, and harvested at 0, 24, and 48 h postinfection. Sonication conditions were chosen to produce the desired size distribution of chromatin, between 200 and 1,000 bp. Samples were then immunoprecipitated with anti-CREB antibody (clone aa5-24; Millipore), anti-VDR antibody (clone ab3508; Abcam), or an anti-mouse IgG (Jackson Laboratory) as a control. To verify the presence of a particular promoter fragment following ChIP, qRT-PCR was performed. Response-specific promoter regions of MMP19, CDKN1A, and MCL-1 (identified as target genes downstream of CREB or VDR) were chosen and amplified by using the following primers: MMP19_f, TCT CCC ACC AAT ACC AGC AGT TCA; MMP19_r, GGA TAC TCG GGA GGG TGG ACG TAG; CDKN_f, TCT TGG ATT GAG GAA CAG GCA ATG; CDKN_r, TCC CAA CAA ACA AGG GGT GGT T; MCL1_f, AGC CTG TTT GGT GGT GTC TTC ACA; MCL1_r, GAG ATG GGG TTT TCA CGA TGT TGG. To determine the levels (relative fold enrichment) of immunoprecipitated chromatin (specific promoter region), the C_T values were analyzed by the standard curve method, and each sample was normalized to the appropriate IgG sample and to the corresponding time-matched mock sample (55). Data presented are the means ± standard errors of means for triplicate samples.

Female C57BL/6J mice (B6; 20-week-old mice from Jackson Laboratories) were anesthetized with a ketamine (1.3 mg/mouse)–xylazine (0.38 mg/mouse) mixture administered intraperitoneally in a 50-μl volume. Each mouse was intranasally inoculated with 10⁵ PFU wild-type icSARS-CoV mouse-adapted virus or icSARS-CoV ΔORF6 mouse-adapted backbone diluted in PBS in a volume of 50 μl. Mice were weighed daily, and at 1, 2, 4, and 7 days postinfection, 5 animals from each infection group were euthanized and the lungs were removed to determine viral titers. The large, lower lobe of the right lung was homogenized in 1 ml of sterile PBS with glass beads by using a Magnalyser (Roche) at 6,000 rpm for 60 s. Aliquots of 200 μl of lung homogenate were plated on Vero E6 cells in serial 10-fold dilutions to determine virus titers. All mice were housed using individually ventilated Sealsafe cages and the SlimLine system (Tecniplast, Exton, PA) under biosafety level 3 conditions. Experimental protocols were reviewed and approved by the Institutional Animal Care and Use Committee of the University of North Carolina, Chapel Hill.

To determine if deletion of the SARS-CoV ORF6 gene caused a specific or more general effect on host transcription following the release of the block on karyopherin-mediated nuclear import, Calu3 2B4 cells were infected with icSARS-CoV or icSARS-CoV ΔORF6 at a high MOI (MOI of 5) to minimize paracrine signaling effects in uninfected cells. Medium, total RNA, and protein were harvested at 0, 3, 7, 12, 24, 36, 48, 54, 60, and 72 h postinfection. Serial dilutions of the medium samples were generated to determine viral titers, and the results for six replicate infections, graphed as the PFU per ml, are shown in Fig. 1B. For both icSARS-CoV and icSARS-CoV ΔORF6, viral titers increased by 4 logs by 30 h postinfection, with peak titers reaching ∼10⁸ at 36 h postinfection (Fig. 1B). No significant difference in titer between icSARS-CoV and icSARS-CoV ΔORF6 was detected at any time postinfection, and a similar percentage of cells was infected (data not shown). At the high MOI, icSARS-CoV and icSARS-CoV ΔORF6 genome and subgenomic RNA transcript levels were detected by 6 h postinfection and remained at nearly identical levels at early and late times postinfection (Fig. 1C). In wild-type- but not icSARS-CoV ΔORF6-infected cells, ORF6 protein expression was detected after 24 h and increased through 48 to 54 h postinfection (Fig. 1D). Under identical conditions, membrane glycoprotein expression was detected at 24 h and peaked between 54 and 60 h during both virus infections (Fig. 1E). These results extend previous studies that demonstrated that the SARS-CoV accessory ORF6 interferon antagonist is dispensable at an MOI of >1 (9, 17).

To determine the effect of the removal of ORF6 protein on mRNA synthesis levels and species in SARS-CoV-infected Calu3 2B4 cells, total RNA was harvested in triplicate at the time points described above, and global mRNA expression was analyzed. Little host gene differential expression was detected during the first ∼24 h of icSARS-CoV or icSARS-CoV ΔORF6 infection (Fig. 2A to C), supporting previous studies that indicated that coronaviruses enter the host cell “quietly,” perhaps by sequestering double-stranded RNA away from or thwarting recognition by the host cell sensing machinery early after infection (56, 57). Overall, a total of 6,947 genes were differentially expressed (P < 0.05; 2-fold change) when icSARS-CoV ΔORF6 and icSARS-CoV samples were directly compared across each time point. Figure 2A shows patterns of gene expression after hierarchical clustering with expression values. Most genes were more highly up- or downregulated in icSARS-CoV ΔORF6-infected than in icSARS-CoV-infected cells compared to the mock-infected samples, such that the primary difference between infections was the magnitude of response. This response differential between the two viruses was also apparent when we compared the total number of differentially expressed genes over time for each virus (Fig. 2B). Differential gene expression changes peaked between 48 and 72 h postinfection for both icSARS-CoV and icSARS-CoV ΔORF6. During this time period, significantly more differentially expressed genes were transcribed in icSARS-CoV ΔORF6-infected Calu3 2B4 cells than in icSARS-CoV-infected cells (P < 0.0001, chi-square test), likely due to one of the functions described for the ORF6 protein as a transcriptional block, mediated by the prevention of karyopherin nuclear translocation (Fig. 1A and 2B). After 48 h, there was 50 to 80% overlap in differentially expressed genes between time points, suggesting that the pool of differentially expressed genes is relatively consistent later in the time course of infection (Fig. 2C). This pattern of expression, which is also reflected in Fig. 2A, is likely a consequence of the differences in upstream gene regulation between icSARS-CoV ΔORF6 and icSARS-CoV and indicates that early transcriptional regulation by the ORF6 protein results in dramatic changes in host gene expression that are maintained throughout the 72-h time course.

In addition, we noted a trend for delay in the host response to icSARS-CoV ΔORF6 infection, such that icSARS-CoV had more differentially expressed genes early in the time course between 0 and 36 h postinfection (Fig. 2B). Many of the 202 differentially expressed genes between icSARS-CoV ΔORF6 and icSARS-CoV at 0 to 24 h (Fig. 2C) were expressed at similar levels at later time points, indicating that the host response eventually “catches up” for icSARS-CoV ΔORF6. However, some of the most highly differentially expressed genes between the viruses were detected as early as 24 h postinfection, including matrix metalloproteinase 19 (Mmp19), calcitonin α (Calcα), and calcitonin beta (Calcβ). This early response period included genes enriched for interferon signaling and innate immune response pathways, specifically the Jak-STAT, Th17, and interleukin-4 signaling pathways, suggesting that the presence or absence of the ORF6 gene may either mediate early differences in the kinetics of nuclear import or promote an early replication-enhancing phenotype (87). As the latter phenotype was not evident at a high multiplicity of infection (Fig. 1B and C), the former possibility may be more likely.

Although previous studies have indicated that the ORF6 protein is an interferon antagonist, it is less clear whether the block in nuclear import specifically targets interferon signaling or represents an outcome associated with a more global block in the import of nuclear cargo, including transcription factors (17). A summary of our overall modeling approach with details of the individual steps is described in Section S3 of the supplemental material and is outlined in Fig. S2 of the supplemental material. To determine the significant biological processes associated with icSARS-CoV ΔORF6 differential gene expression, the data set was first reduced to six clusters by K-means based on common patterns of expression across genes (Fig. 2A), and then significant enrichment (P < 0.05) of biological process Gene Ontology categories was calculated for each cluster individually (Fig. 2A). Processes related to transcription, nuclear signaling, cell proliferation and death, and host antiviral and the immune response were upregulated in icSARS-CoV ΔORF6 compared to icSARS-CoV (clusters C4, C5, and C6), while cell cycle and select metabolic processes, like DNA and lipid metabolism, were downregulated (clusters C1, C2, and C3) (Fig. 2A). These data suggest that the ORF6 protein antagonizes nuclear import processes and host transcription.

Viral antagonists of nuclear import have been well documented (10–14); however, it is not clear whether these antagonists selectively and/or differentially regulate nuclear cargo importation, resulting in hierarchical and cell-type-specific patterns of antagonism of host gene expression. The SARS-CoV model is particularly appropriate to address this question, as the ORF6 protein binds karyopherin α2, suggesting a selective targeting of cargo. However, the process is further complicated by the ORF6 protein-karyopherin α2 complex's capacity to also sequester karyopherin β1, which is essential for all nuclear import via karyopherins (Fig. 1A). The pattern of expression in gene cluster C4, which included 1,674 genes exclusively upregulated in icSARS-CoV ΔORF6 but unchanged in icSARS-CoV, was enriched in biological processes for chromosome organization and regulation of gene expression and nucleosome assembly. This was particularly interesting, given that one of the ORF6 protein mechanisms of action is prevention of nuclear translocation of karyopherin-dependent cellular factors. Therefore, we focused on the genes in cluster C4 to identify potential downstream targets of karyopherins whose transcription is blocked in icSARS-CoV infection. In particular, we evaluated whether transcription factors whose transport is regulated by karyopherins were overconnected (transcription factors with a significant number of downstream target genes that were differentially expressed, in this case either up- or downregulated [P < 0.05]) to downstream gene expression networks (or to downstream genes whose expression is regulated by the same transcription factor) by hypergeometric distribution within the C4 cluster. When the number of connections between target genes in the data set and upstream regulatory transcription factors are greater than would be expected by chance (based on the number of known connections), this indicates that these transcription factors are enriched and are important regulators of the host response by the ORF6 protein. All of the transcription factor hubs (schematically represented by large circles in Fig. 3) associated with gene cluster C4 that were identified as being regulated by karyopherins in MetaCore are listed in Table 1. Six of these transcription factors (VDR, CREB1, Oct3/4, HIFα2/Epas1, p53, and SMAD4) were significantly (P < 0.05) overconnected to the data set (Table 1), indicating that there was differential signaling through these transcription factor hubs when comparing icSARS-CoV ΔORF6 to icSARS-CoV. Over 350 gene nodes in the C4 cluster directly interact with these six transcription factors (Fig. 3), based on information in the Metacore knowledge base, suggesting their transcription is specifically blocked in the presence of functional ORF6 protein. As a summary of our network analysis, we graphed the fold change over mock for all of the genes regulated by each of the six transcription factors identified by our modeling approaches (Fig. 4A). In each graph, the differentially expressed networks in the icSARS-CoV ΔORF6-infected cells are significantly upregulated (7- to 22-fold increase over mock, which was significant from 48 to 60 h postinfection for all of the transcription factors) compared to steady levels in the icSARS-CoV-infected cells, suggesting that the release of the ORF6 protein-mediated nuclear importation block does affect cellular transcription and is not limited to immune-responsive transcription factors. In contrast, two transcription factors (beta-catenin and androgen receptor) that do not require karyopherin to translocate to the nucleus are shown in Fig. 4B and demonstrate similar differential expression patterns between icSARS-CoV ΔΟRF6 and wild-type virus (only a 1.5-fold difference above mock, and the changes were significant only at 72 h postinfection for beta-catenin) through 60 h.

The initial targeted approach that identified the six transcription factors was restricted to the subset of genes in cluster C4 (Fig. 2A), which were upregulated exclusively in icSARS-ΔORF6-infected cells. However, we wondered if additional karyopherin-regulated transcription factors were also present in networks generated from the original 6,947 genes (in all 6 clusters) identified (Fig. 2A), which would suggest that karyopherins import transcription factors that both positively and negatively regulate downstream target genes during the course of SARS-CoV infection. To answer this question, we expanded our transcription factor analysis to include all significant genes in the data set, resulting in 27 overconnected transcription factors whose nuclear importation was mediated by karyopherins (Table 2). This analysis provided a more global understanding of the potential impact of the nuclear importation block during SARS-CoV infection compared to the more targeted approach illustrated in Fig. 3 and 4. For example, STAT1 was identified as an important karyopherin-mediated regulator of gene expression during infection. Our laboratory and others have previously observed nuclear translocation of STAT1 in the absence of the ORF6 protein and the upregulation of common targets (i.e., cyclooxygenase 1 and 2/prostaglandin G/H synthase 1 and 2) in icSARS-ΔORF6-infected but not in wild-type-infected cells (see Fig. S3A and B in the supplemental material) (17, 58). For Calu3 2B4 cells, our data suggest that one of the functions of the ORF6 protein is to mediate complex hierarchical antagonism phenotypes of the nuclear import machinery, antagonizing different cellular responses during virus infection.

To independently confirm the transcription factors identified from network-based modeling analysis of the microarray data with Calu3 2B4 cells, similar modeling approaches were performed to identify transcription factors from microarray data for icSARS-CoV- and icSARS-CoV ΔORF6-infected primary HAE cell cultures. Total RNA was harvested and analyzed by using an Agilent microarray from wild-type- and icSARS-CoV ΔORF6-infected HAE cultures at multiple times postinfection, and the data were compared to Calu3 2B4 transcriptomic data sets to determine if similar targets (genes downstream of the transcription factors identified from just the Calu3 2B4 microarray data) could be identified in both primary and traditional cell isolates. We first examined the viral growth kinetics in HAE cultures and determined that icSARS-CoV titers peaked at 48 h postinfection (Fig. 5A). In contrast, replication titers for icSARS-CoV ΔORF6 increased over the entire 72-h time course (Fig. 5A). At a slightly reduced MOI (Calu3 2B4 cells at an MOI of 5 versus HAE at an MOI of 2), icSARS-CoV ΔORF6 demonstrated slower growth kinetics until later in infection in primary cells, consistent with a role in early replication (Fig. 5A) and (59). Transcription factor analysis of the HAE data set resulted in 7 enriched transcription factor hubs regulated by karyopherins (P < 0.05), including RelA, C-jun, CREB1, Hif1α, C-fos, VDR, and SMAD3. Two of the enriched transcription factor hubs in HAE cultures, CREB1 and VDR, were also identified as important in the targeted Calu3 2B4 analysis of exclusively upregulated differentially expressed genes (cluster C4), whose transcription factors enter the nucleus in a karyopherin-mediated process (Table 1; Fig. 3 and 4), while the other 4 transcription factors overlapped the hubs identified from the global Calu3 2B4 gene microarray data set (Table 2), independently confirming the overlap of target genes regulated by the VDR and CREB1 transcription factors between HAE and Calu3 2B4 cells. A comparison of the CREB1 and VDR transcription factor networks, which were significantly enriched in both HAE and Calu3 2B4 cells, is shown in Fig. 5B (see also Fig. S4 in the supplemental material) and illustrates the genes that are either uniquely regulated by the overconnected transcription factor gene targets in each cell type or are genes that are common to both cell types. The mRNA expression levels (from the microarray data) for representative individual genes that were differentially expressed in both the HAE and Calu3 2B4 microarray data sets are graphed in Fig. 5C for comparison, including the genes B cell translocation gene 2 (Btg2), forkhead box 03a (Foxo3a), hypermethylated in cancer 2 (Hic2), human p-thromboglobulin gene (Ptg), thiamine transporter gene (Scl19a2), glucose transporter gene (Scl2a6), transforming growth factor β3 (TGFβ3), and POK family transcription factor (Zbtb5). Transcription factor analysis in HAE cultures resulted in strong overlap with the transcription factor hubs identified in Calu3 2B4 cells and further supported the importance of karyopherin-mediated nuclear importation during SARS-CoV infection.

To independently validate our transcriptomic results, we performed shotgun proteomics using the AMT tag approach and ChIP-PCR, with parallel sets of Calu3 2B4 cells infected with icSARS-CoV and icSARS-CoV ΔORF6. Following proteomic analysis, a total of 864 proteins were significantly (P < 0.05) differentially expressed between icSARS-CoV- and icSARS-CoV ΔORF6-infected Calu3 2B4 cells across all time points. To most directly determine how the proteomics contributes to our understanding of the role(s) for the ORF6 protein in karyopherin-mediated nuclear translocation and host gene expression, we first integrated the transcriptomic and proteomic data sets. From the integrated data, we obtained a more comprehensive view of the changes mediated by the ORF6 protein in Calu3 2B4 cells at both the gene and protein levels, allowing us to determine whether karyopherin-mediated transcriptional hubs are further enriched (i.e., more significant) with the addition of the proteomic data. For example, if the proteomic data support a role for the ORF6 protein in karyopherin-mediated nuclear transport, then we would expect an increase in the enrichment scores (increased significance) of these hubs after addition of the proteomics; otherwise, the values would decrease. From a biological perspective, the enrichment scores (P values) represent how connected the transcription factors are to downstream targets (genes or proteins) in the data set compared to the number of targets expected by chance. If a transcriptional hub is significantly (P < 0.05) more connected to the data set, this suggests that signaling through this hub is uniquely regulated, based on the experimental conditions: in this case, through deletion of ORF6. We can use the integrated gene and protein data to identify significant hubs whose downstream targets may primarily be detected at the protein level and whose enrichment scores are greatly elevated by inclusion of the proteomics data. In our study, transcription factor analysis of the combined Calu3 2B4 proteomics and transcriptomics data resulted in an increase in enrichment scores, ranging from 1.5 × 10¹²- to 2 × 10¹²-fold, for 22 out of 27 of the transcription factor hubs requiring karyopherin for nuclear importation, including 5 out of the 6 transcription factor hubs uniquely upregulated during icSARS-CoV ΔORF6 infection (Table 2). This suggests that many protein nodes in the proteomic data set share common upstream transcription factor regulators with gene nodes in the transcriptional data, including those that require karyopherin for nuclear import. For four transcription factor hubs in particular, C-myc, Rela, specificity protein 1 (Sp1), and STAT1, values from the combined transcriptomic/proteomic analysis were >1,000-fold more significant than for the transcriptomics alone, demonstrating the added value of including targets from both data types (Table 2). In support of the transcription factor analysis, expression levels of individual downstream targets were confirmed between Calu3 2B4 transcriptional and proteomics analyses. Enhancer of mRNA-decapping enzyme 3 (EDC3) and Golgi apparatus adapter-related complex of proteins mu 1 subunit (AP3M1), two target gene nodes predicted to be a part of CREB1 networks, were measured at both the transcript and protein levels (Fig. 6). Individual RNA expression values derived from microarray analysis demonstrated increased RNA expression trends in icSARS-CoV ΔORF6 versus icSARS-CoV in both EDC3 and AP3M1 (Fig. 6A and C). Coordinately, protein abundance also increased at late times during icSARS-CoV ΔORF6 infection; in contrast, icSARS-CoV infection protein levels failed to rise above mock infection values (Fig. 6B and D). When directly compared, both gene and protein analyses demonstrated augmentation of these gene nodes in icSARS-CoV ΔORF6 infection at late times compared to icSARS-CoV infection. Together, these data confirm increased expression of targeted gene nodes downstream of identified transcriptional factors in icSARS-CoV ΔORF6 infection, increases that are absent in icSARS-CoV infection.

Finally, to demonstrate that the transcription factors we identified were actively engaged in transcription during icSARS-CoV ΔORF6 infection, we performed ChIP followed by real-time PCR (ChIP–RT-PCR) of target gene transcripts. Calu3 2B4 cells were infected with icSARS-CoV or icSARS-CoV ΔORF6 or mock infected, protein and DNA were cross-linked, genomic DNA was sheared, and VDR- and CREB1-bound DNA promoter regions were precipitated with antisera directed against each of these transcription factors. Next, gene-specific primer pairs were used to amplify DNA regions of interest by real-time PCR. The results for transcription factor-specific antisera were compared to control IgG precipitations and time-matched mock-infected controls. We chose to amplify three representative downstream target genes: the matrix metalloproteinase 19 (MMP19) and cyclin-dependent kinase inhibitor 1A (CDKN1A) CREB1-regulated genes, as well as the myeloid leukemia cell differentiation gene (Mcl-1), a downstream target of VDR. Figure 7 shows the significant increase in the amount of chromatin precipitated relative to the fold enrichment in icSARS-CoV ΔORF6-infected samples for MMP19, CDKN1A, and Mcl-1 at both 24 and 48 h postinfection compared to wild-type icSARS-CoV. These results confirm that at least two transcription factors identified by modeling are significantly enriched on the promoter elements of their target genes following icSARS-CoV ΔORF6 infection, supporting the microarray data and validating our modeling approaches.

Airway epithelial cells are major targets for SARS-CoV infection in humans and in mice (36, 60–62). Functional characterization of the gene sets differentially induced following icSARS-CoV ΔORF6 and icSARS-CoV infection in either traditional human lung cell monolayers or primary HAE cultures revealed significant induction of a number of key antiviral response genes (Fig. 8A) that may play critical roles in the host response or in promoting efficient virus replication. These data are consistent with the possibility that icSARS-CoV ΔORF6 might be less capable of maintaining efficient viral replication under more natural conditions (i.e., at a lower MOI) in vitro. For example, low numbers of virus-infected cells early in infection would afford robust antiviral host gene expression and paracrine signaling, potentially limiting secondary rounds of viral replication. To test this hypothesis in vitro, we infected Calu3 2B4 cells at a low MOI (MOI of 0.01) with icSARS-CoV or icSARS-CoV ΔORF6 and assessed viral titers by plaque assays in Vero E6 cells. Peak titers for wild-type virus were detected at 72 h postinfection (∼10⁷ PFU/ml), while in contrast, peak titers for icSARS-CoV ΔORF6 (∼10⁶ PFU/ml) were detected at 96 h postinfection and titers were reduced by 1 to 2 logs at all time points examined (Fig. 8B). These data are in contrast to previously published reports (9, 59) in which titers in Vero E6 cells were compared, and the differences can likely be attributed to the intact innate immune and other antiviral signaling pathways present in the human conducting airway Calu3 2B4 cell line. Subsequent studies using another recombinant SARS-CoV that does not express the ORF6 protein have also demonstrated that at a low MOI (MOI of 0.01) in nonhuman primate kidney cells, 2- to 5-fold reductions in viral titers were detected between 6 and 24 h postinfection (77).

To test the hypothesis that the release of the nuclear importation block mediated by the ORF6 protein might attenuate pathogenesis, 20-week-old B6 mice (n = 5 mice/time point) were infected with 10⁵ PFU of icSARS-CoV mouse-adapted virus or icSARS-CoV ΔORF6 mouse-adapted virus. Weight loss was measured each day, and lungs were harvested to assess viral titers at days 1, 2, 4, and 7 postinfection. Mice infected with icSARS-CoV mouse-adapted virus steadily lost weight over the 7 days of infection, while in contrast, mice infected with icSARS-CoV ΔORF6 mouse-adapted virus lost weight through 4 days postinfection and then began to recover and gain weight (Fig. 8C). Weight loss in animals infected with icSARS-CoV ΔORF6 mouse-adapted virus was significantly different than in mice infected with icSARS-CoV mouse-adapted virus at days 2 to 7 postinfection (P < 0.01 by Student's t test). Mock-infected mice lost no weight over the course of the infection. Lung titers for both viruses were determined by plaque assay, and no significant differences were measured at days 1, 2, or 7 postinfection. However, at 4 days postinfection, titers in the mouse-adapted icSARS-CoV ΔORF6-infected mice were significantly higher than mouse-adapted icSARS-CoV-infected mice, despite their weight gain, suggesting that viral growth kinetics were uncoupled from pathogenic outcome (Fig. 8D) following the removal of ORF6 protein expression. Future studies will examine the transcriptional and proteomic profiles of the infected mouse lungs, including those of targeted knockout mice, but these studies are beyond the scope of the current work.