INTRODUCTION
Novel research strategies are needed to elucidate the complex virus-host interaction networks that regulate viral pathogenesis and to provide rapid response strategies for control of newly emerging viral pathogens. Prior to 2003, human coronaviruses were categorized as mildly virulent upper respiratory pathogens; however, severe acute respiratory syndrome coronavirus (SARS-CoV) infection results in high mortality rates (∼10%) (
1,
2). SARS-CoV emerged suddenly from zoonotic reservoirs and rapidly circumnavigated the globe in 2003 (
3–8). The SARS-CoV positive-stranded RNA genome encodes a variety of novel genes that do not exist in other human coronaviruses, which likely contributes to the alteration of virulence and disease severity (
9). In our study, a systems biological approach was used to examine the consequences of antagonizing karyopherin-dependent nuclear importation during SARS-CoV infection.
Signal-mediated macromolecular transport between the cytoplasm and nucleus is an integral part of cellular processes, including gene expression, signal transduction, development of antiviral states, and cell cycle progression. Highly pathogenic RNA viruses, including SARS-CoV, enteroviruses, Ebola virus, human immunodeficiency virus, cardioviruses, and Nipah viruses, encode proteins that antagonize nuclear importation processes, suggesting a common modality for regulating viral pathogenesis and disease outcomes across disparate virus families (
10–14). Interestingly, the consequences of viral antagonism on host nuclear import and mRNA expression have not been carefully evaluated using genomic-based strategies. Rather, elegant reductionist approaches have demonstrated targeted antagonism of innate immune signaling pathways, typically involving interferon (IFN) regulatory factor 7 (IRF-7) and STAT transcription factors (
15–20). Reflecting this approach, the interferon antagonist activity of the SARS-CoV ORF6 protein has been demonstrated to mediate its function by binding to the nuclear importation chaperone protein karyopherin. ORF6 protein binds specifically to karyopherin α2, trapping the import factor on intracellular membranes, where the complex then sequesters karyopherin β1 (
17), preventing nuclear transportation of cargo into the host cell nucleus (see
Fig. 1A). As karyopherin β1 is essential for all nuclear import by karyopherin α proteins, depletion of this factor may dramatically reduce or alter the transport of other cargo by karyopherin-based transport mechanisms (
21). We hypothesized that the ORF6 protein may actively manipulate the translocation of multiple transcription factors, coordinately modulating the levels of host transcription during infection.
Using SARS-CoV as a model system, we investigated whether the ORF6 accessory protein mediated a specific or more general block of karyopherin-mediated nuclear translocation and host gene expression. We identified a cluster of genes that are uniquely upregulated during infection with a mutant SARS-CoV strain that does not express ORF6 protein. Our data showed that ORF6 protein expression effectively ablates the activity of numerous karyopherin-dependent host transcription factors that are critical for establishing antiviral responses and regulating other key host responses during virus infection. The transcription factors identified in these studies (vitamin D receptor [VDR], cyclic AMP receptor binding protein 1 [CREB1], Oct3/4, hypoxia-inducible factor α2 [HIFα2]-Epas, p53, and SMAD4) play important roles in the regulation of a variety of cellular processes, including transforming growth factor beta induction, maintenance of normal lung cell functions, prevention of lung disease phenotypes, and proper immune cell functions (
21–29). Finally, we verified our hypothesis that the upregulation of nuclear translocation attenuates viral pathogenesis with both
in vitro and
in vivo studies. Together, these data suggest that the ORF6 protein mediates the establishment of an intracellular environment that enhances SARS-CoV replication later in infection by suppressing host antiviral and innate immune expression cascades. In addition, these studies suggest that other viral nuclear import antagonists will also mediate pleiotropic alterations in host gene expression during infection, potentially leading to broad-based strategies for intervention and control of viral pathogenesis
in vivo.
DISCUSSION
Viral antagonism of host cellular processes is well recognized as a major mechanism for regulating viral pathogenesis and virulence. SARS-CoV encodes several interferon antagonists that delay host cell recognition of infection, innate immune sensing, and signaling pathways, as well as interferon-stimulated gene expression; one antagonist, ORF6 protein, does so by blocking nuclear import. Many other highly pathogenic RNA viruses encode proteins that specifically antagonize nuclear import to prevent host innate immune and other critical cellular macromolecular processes to enhance virus replication and transmission between hosts. For example, Ebola virus VP24 binds karyopherin α1 and blocks STAT1 nuclear import (
10). The Nipah virus W protein is localized to the nucleus, where it inhibits both virus- and Toll-like receptor 3-triggered signaling in the infected cell by preventing the phosphorylation and activation of STAT1 and subsequent downstream interferon-stimulated gene induction (
11). Some cardiovirus L proteins interact directly with Ran-GTPases, which are required for the export of new nuclear mRNA (
12). In these systems, observations were conducted using candidate gene markers, limiting full recognition of the impact of viral antagonism gene function on nuclear translocation during infection. To redress this limitation, we used systems biology approaches that integrated transcriptomic and proteomic data sets in primary and traditional human lung epithelial cells to identify the impact of the ORF6 protein on host macromolecular processes. An advantage of the SARS-CoV ORF6 protein model is that recombinant viruses lacking ORF6 expression are viable and replicate efficiently
in vitro and
in vivo (
9,
59). Our data support the hypothesis that antagonists of nuclear translocation differentially target host signaling pathways, perhaps in a tissue- and cell-specific manner, to prevent antiviral defenses and other subcellular responses that may limit virus replication. The data reported herein suggest that other viral nuclear import antagonists likely antagonize multiple host transcription factors and cellular processes to allow efficient virus replication, transmission, and spread. Although speculative, these antagonists likely block paracrine signaling of cytoplasmic cargo (e.g., cytokines, steroids, hormones), blocking cellular antiviral responses during virus infection.
Viral antagonists that target nuclear translocation provide a novel model system to study the regulated inhibition of host response networks in the context of virus infection. One of the current ORF6 protein functional paradigms is that it blocks nuclear import by binding karyopherin α2 on internal membranes, sequestering karyopherin β1 and preventing karyopherin-regulated nuclear import of key antiviral transcription factors, like STAT1. However, STAT1 typically uses karyopherin α1 and β1 for import, suggesting that the primary targets for the ORF6 protein may actually be other key host response transcription factors and expression networks during infection. Given the unique biochemical targeting of the ORF6 protein and the fact that SARS-CoV also encodes other proteins that antagonize innate immune signaling (
63,
64), it is not surprising that STAT1 was not positioned more prominently as one of the key transcription factors targeted by ORF6 protein antagonism of nuclear import in either Calu3 2B4 and HAE cultures. The global genomics-based technologies used in this study provide an alternative approach to evaluate the role of specific viral genes on host transcription and proteomic regulatory networks during infection. We recognize that karyopherin and cargo concentrations are heavily cell type dependent, and thus the ORF6 protein's effect on nuclear import in other permissive cell types might result in dramatically different hierarchical activities, leading to variations in host gene expression networks and/or the activation of alternate transcription factors (
17,
31,
65,
66). Although our initial studies focused on a recombinant virus that lacked the entire ORF6 gene, future studies using recombinant viruses lacking either the N- or C-terminal portion of the ORF6 protein in early virus replication and nuclear import may provide further resolution to the precise domains in the ORF6 protein that mediate these phenotypes.
In support of earlier reports (
9,
17,
59), deletion of ORF6 had minimal effects on the efficiency of virus replication in Calu3 2B4 and HAE cultures; however, these growth comparisons were performed at a high MOI, which is often a less powerful barometer of virus growth than studies with a low MOI in culture. The Calu3 2B4 and HAE culture systems represent continuous and primary models of the human airway epithelium, a major target for early SARS-CoV infection and replication
in vivo in many species (
36,
59,
67). Thus, host response outcomes in these cells likely model
in vivo responses that inform downstream innate and adaptive immune responses during infection. At lower MOIs, more subtle effects of the ORF6 protein have been reported on virus replication rates, especially early, but not late, in infection (
59,
68). While our data in HAE and Calu3 2B4 cultures support these earlier findings, we designed the current experiment to include a high MOI to reduce the impact of paracrine signaling between infected and uninfected cells, allowing us to specifically focus on host responses during infection. Importantly, in the presence or absence of the ORF6 protein, similar RNA expression kinetics and levels of virus replication were noted in both Calu3 2B4 cells and HAE cultures; thus, the virus effects on differential gene expression likely reflected the targeted activities of the ORF6 protein directly on virus-host regulated interactions, like nuclear translocation. However, there were dramatic differences in the levels and rates of differential gene expression detected between the two viruses despite similar replication kinetics. icSARS-CoV differential gene expression was detectable as early as 24 h postinfection and was higher than the levels detected for icSARS-CoV ΔORF6 through 36 h postinfection. At 48 h postinfection, a drastic change occurred with the number of differentially expressed genes still increasing for icSARS-CoV, but the number of genes detected for icSARS-CoV ΔORF6 jumped to twice the level of the wild type (
Fig. 2B, ΔORF6 36 h 1600 versus 48 h 4750), a trend which continued through 72 h postinfection. These data tracked with levels of detectable ORF6 protein expression (
Fig. 1D) and suggest that the changes mediated by the ORF6 protein during infection can likely be uncoupled from replication kinetics but still modulate pathogenesis, as seen with the weight loss recovery in the mouse model (
Fig. 8C and
D).
Several of the identified transcription factors play critical roles in lung cancer but can also mediate many acute and chronic lung disease phenotypes. During normal lung development, CREB1 plays an important role in the differentiation of the mucin- or mucus-producing cells (
69). In contrast, studies have also demonstrated that CREB1 and its associated pathways contribute to pathological lung disease progression via inflammatory response-mediated lung remodeling postinfection/post-lung injury (
70,
71). VDR transcription factor levels are reduced in patients with chronic obstructive pulmonary disease (COPD), and vitamin D has been shown to be important in the onset of COPD (
72,
73). In VDR knockout mice, increased lung inflammation and emphysema were noted, suggesting an important role for VDR signaling in normal lung function and lung disease prevention or exacerbation (
74). The transcription factor EpasI/HIFα2 is critical for responding to reduced oxygen levels in the intracellular environment (
75). EpasI/HIFα2 is part of the hypoxia response and was one of the genes detected in the microarray analysis of wild-type SARS-CoV-infected nonhuman primates, whereas in the presence of the ORF6 protein its expression was downregulated (
76,
77). The SMAD family of transcription factors are potent inducers of TGF-β, which can activate apoptosis signaling pathways during influenza virus infection (
78). SMAD proteins as well as p53 have also been implicated as being underexpressed in a wide variety of cancers, including lung carcinomas (
23,
79). Many viruses have designed strategies to counter p53 signaling (
80–83). Importantly, p53 expression enhances innate immunity by enhancing IFN-dependent antiviral activity via IRF-9 activation, independent of its functions as a proapoptotic and tumor suppressor gene (
80). In some cell lines, influenza virus infection can increase p53 expression, where it is essential for the induction of cell death phenotypes, and loss of p53 expression enhances virus growth (
81). SARS-CoV targets and infects Oct 4-expressing pluripotent lung stem cells, which differentiate into type 1 and type 2 pneumocytes, and is essential for repair and function of the alveoli (
84). Lung stem cells are reported targets for SARS-CoV infection, and so blocking Oct 4 expression may well slow the rate of recovery from lung injury by preventing differentiation of stem cells that are essential for normal lung function and wound repair (
85). All of the transcription factor hubs identified by modeling of systems biological approaches play critical roles in the regulation and maintenance of lung cell homeostasis and would not have been collectively identified in more-targeted studies.
Enrichment of karyopherin-mediated transcription factors in icSARS-CoV ΔORF6 infection compared to infection with the wild type was independently confirmed and validated from analysis of different data sets, including proteomics and ChIP-PCR data from infected Calu3 2B4 cells and microarray data from infected HAE cultures. It is reassuring that bioinformatic approaches independently validated earlier biochemical studies, demonstrating targeted antagonism of nuclear import by using a candidate gene approach. ChIP-PCR analysis demonstrated that some transcription factors identified by modeling approaches were actively bound to the promoters of specific downstream target genes (identified by microarray analysis), lending further credence to the modeling analysis (
55,
86). The strength of the current approach is that global integration of data in the context of biological pathways or functional processes, such as transcriptional factor regulation, provide a complementary interpretation of genomic and proteomic results that is necessary for systems-level comparisons (
87). Direct comparison or integration of heterogeneous data sets is complicated by inherent differences among cell types, platforms, and technologies. However, the identification of common regulatory components that induce gene or protein signatures across model platforms can provide important information about underlying biological processes during infection (
88). The overlap in regulatory events across systems, including comparison of infection of traditional cell lines to infection of primary human cells, can be considered independent confirmation of the mechanism of action for the ORF6 protein during SARS-CoV infection. In this study, a range of approaches were applied to provide a comprehensive view of karyopherin-dependent nuclear importation mediated by the ORF6 protein during SARS-CoV infection, and these approaches were not only validated across systems but also could be interpreted within the context of
in vitro and
in vivo viral endpoints.
Removal of the ORF6 protein-mediated nuclear importation block attenuated SARS-CoV pathogenesis in a 20-week-old B6 mouse model. There was no difference in viral growth kinetics between icSARS-CoV and icSARS-ΔORF6 in human lung cells (Calu3 2B4) at high MOIs, and in the 20-week-old mouse model replication patterns were similar between the two viruses, with significant titer differences at only a single time point. However, despite similar virus loads in the lung, weight loss was significantly reduced in mice infected with icSARS-CoV ΔORF6 mouse-adapted virus compared to those infected with wild-type mouse-adapted virus, suggesting that viral titers do not necessarily correlate with disease outcomes. Similar findings have been noted with recombinant SARS-CoV bearing zoonotic Spike glycoproteins, like HC/SZ/61/03 (
77,
89). The category of genes most affected by release of ORF6 protein nuclear importation block
in vitro-regulated overall gene expression, cellular rearrangement for division, and factors required for differentiation, suggesting that the intracellular environment may play a substantial role in determining levels of viral pathogenesis and controlling the intracellular antiviral state.
Antagonists of nuclear import likely contribute to the virulence of many highly pathogenic viruses. Importantly, some of the transcription factors identified in cell culture models, like STAT1, which is significantly regulated by ORF6 protein function, have also been demonstrated to alter
in vivo pathogenesis (
60). In STAT1 knockout mice, but not IFNR knockout mice, icSARS-CoV mouse-adapted infection resulted in dysregulation of T cell and macrophage differentiation, leading to a Th2-biased immune response and the development of alternatively activated macrophages that mediated a lethal, profibrotic environment within the lung (
90). Future studies will evaluate icSARS-CoV and icSARS-CoV ΔORF6 pathogenesis in p53, CREB1, EpasI, and VDR knockout mice. In addition to antagonizing innate immune signaling, our data argue that antagonists of nuclear import likely target alternative and perhaps unappreciated host signaling networks to maximize a favorable environment for virus replication and pathogenesis, both
in vitro and
in vivo.