Abstract
Coronaviruses encode multiple interferon antagonists that modulate the host response to virus replication. Here, we evaluated pathogenesis and host transcription in response to infection with murine coronaviruses encoding independent mutations in two different viral antagonists: the deubiquitinase (DUB) within nonstructural protein 3 and the endoribonuclease (EndoU) within nonstructural protein 15. The virus with reduced ability to deubiquitinate proteins, herein termed the DUBmut virus, was engineered via X-ray structure-guided mutagenesis and activates an earlier interferon response than the wild type virus. However, the replication kinetics of DUBmut in cultured cells are similar to wild type virus and pathogenesis in mice is also similar to what was observed during infection with wild type virus. On the other hand, we previously reported that an EndoUmut virus containing an inactivated endoribonuclease activity elicited rapid and robust activation of type I interferon, which limited virus replication and pathogenesis. Here, using a transcriptomics approach, we compared the scope and kinetics of the host response to the wild type, DUBmut, and EndoUmut viruses in infected macrophages. We found that the EndoUmut virus activates a focused response, predominantly involving type I interferons and a subset of interferon-responsive genes, within 12 hours after infection. In contrast, the wild type and DUBmut viruses stimulate upregulation of over 2,800 genes, including activation of unfolded protein response (UPR) pathways and a proinflammatory profile associated with viral pathogenesis. This study highlights the role of viral interferon antagonists in shaping the kinetics and magnitude of the host response during virus infection and demonstrates that inactivation of a dominant viral antagonist, the coronavirus endoribonuclease, dramatically alters the host response in macrophages and the disease process.
Author Summary Macrophages are an important cell type during coronavirus infections because they “notice” the infection and respond by activating type I interferons, which then act to establish antiviral defenses and limit virus replication. In turn, coronaviruses encode proteins that mitigate the cell’s ability to detect virus replication or amplify the interferon response. Here, we evaluated the host macrophage response to two independent mutant coronaviruses: one with a reduced deubiquitinating activity (DUBmut), and the other containing an inactivated endoribonuclease (EndoUmut). We observed a rapid, robust, and focused response to the EndoUmut virus, which was characterized by enhanced expression of interferon and interferon-stimulated genes. These results indicate that coronaviruses utilize EndoU activity for preventing early activation of interferon in macrophages, thereby allowing for viral replication. In contrast, DUBmut elicited a transient interferon response and ultimately activated over 2,800 genes, including many well-known players in pro-inflammatory pathways and the unfolded protein response. These DUBmut-induced pathways are associated with development and progression of significant disease, similar to what is observed during wild type virus infection. This study demonstrates the distinct consequences of mutating different viral interferon antagonists and reveals that intact coronaviral EndoU activity substantially contributes to the ability of coronaviruses to replicate in macrophages.
Introduction
The ability to evade or delay activation of host innate and adaptive immune responses is now recognized as a characteristic of many pathogenic viruses. Viruses from diverse families, including filoviruses [1], poxviruses [2], influenza viruses [3], flaviviruses [4], and coronaviruses [5, 6], encode proteins that are not required for viral replication per se, but act as modulators of initial, innate host responses or of later, adaptive responses that aim to limit virus replication. By understanding how these viral antagonists regulate the immune response, we can fine-tune the rational design of therapeutics and vaccines to control existing and emerging viral pathogens.
Coronaviruses (CoVs) are members of the order Nidovirales, which includes enveloped viruses with large (∼30kb), positive-sense single-stranded RNA genomes that yield a characteristic nested set of subgenomic mRNAs during replication in the cytoplasm of infected cells [6, 7]. The genome organization for coronaviruses is highly conserved, with the 5’-most two-thirds of the genome encoding the replicase polyprotein, followed by sequences encoding the canonical structural proteins: spike, envelope, membrane, and nucleocapsid. Many CoVs contain accessory genes, which are interspersed among the genes for the structural proteins. Although these accessory genes are not necessarily required for virus replication and are, in general, not highly conserved within the virus family, many encode proteins that regulate the host response [8]. Interestingly, coronavirus replicase proteins, which are highly conserved, can also act as antagonists to block or delay the innate immune response [9–12]. That a slew of coronavirus-encoded accessory and non-accessory proteins have been shown to shape the host antiviral response suggests that viral-mediated subversion of host defenses is an important element of infection. Here, we focus on two highly conserved domains within the replicase polyprotein—the viral protease/deubiquitinase (DUB) and the endoribonuclease (EndoU)—with the goal of assessing the relative role(s) of each of these enzymatic activities in shaping the pathogenesis of mouse hepatitis virus (MHV), a model murine coronavirus.
Coronavirus protease activities are critical for processing the replicase polyprotein. Murine coronavirus MHV encodes three proteases: two papain-like proteases (PLP1 and PLP2) and one chymotrypsin-like protease (3CLpro, also termed Mpro). MHV PLP2 is similar to the single papain-like protease (termed PLpro) of Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV). Bioinformatic analyses and enzymatic studies revealed that the PLpros of SARS-CoV and MERS-CoV are multifunctional proteins that contain protease, deubiquitinase (DUB), and de-ISGylating activities [13–17]. However, it has been challenging to study the effects of a mutated DUB on virus replication and pathogenesis because the protease and DUB activities share the same catalytic site, disruption of which is lethal for virus replication. Therefore, it was necessary for us to identify a residue in the MHV protease/DUB-ubiquitin binding domain that can be mutated to result in reduced DUB activity while maintaining sufficient protease activity to allow for proteolytic processing of the replicase polyprotein. In this report, we describe the X-ray structure-guided generation of such a DUB mutant and our subsequent evaluation of its replication and of the host transcriptional response that it elicits during infection. We also sought to expand our understanding of the role of DUB interferon antagonism by comparing it with another recently identified antagonist, the coronavirus endoribonuclease (EndoU).
EndoU, a highly conserved enzyme within the coronavirus family, was initially thought to play a role in coronavirus RNA synthesis [18, 19]. Recent studies revealed that EndoU acts as an interferon antagonist by preventing activation of host sensors of viral double-stranded RNA (dsRNA) [10, 11]. Viruses encoding a catalytically inactive EndoU undergo initial RNA replication but accumulation of viral dsRNA intermediates during replication is detected by host sensors MDA5, PKR, and OAS. Activation of these dsRNA sensors initiates the innate immune response, including activation of type I interferons, interferon-responsive genes, and apoptosis-promoting caspases, all of which collectively limit virus replication. We previously reported that MHV-EndoU mutant viruses are profoundly attenuated in mice and elicit a protective immune response [10]. In the current study, we use transcriptome profiling to evaluate the kinetics of host gene expression in macrophages upon infection with MHV wild type (WT), DUB-mutant, and EndoU-mutant viruses. Our analyses of the respective transcriptional response to each virus reveal significant differences in the kinetics, magnitude, and breadth of host gene expression during infection and provide new information on the extent to which viral interferon antagonists manipulate the overall host response to viral infection.
Results
Mutation of murine coronavirus papain-like protease 2 ubiquitin-binding domain modulates deubiquitinating activity
To investigate the role of viral deubiquitinating activity in coronavirus replication and pathogenesis, we first needed to identify amino acid residues that, when mutated, result in reduced DUB activity while preserving the enzyme’s deISGylating and protease activities, the latter being necessary for viral replication. X-ray structural studies of SARS- and MERS-CoV papain-like protease/DUBs co-crystalized with ubiquitin modified at the c-terminus with a covalent warhead (aldehyde, 3-bromo-propylamine, or propargylamine) allowed for identification of residues that are important for direct interaction with ubiquitin [16,20,21]. Here, we took a slightly different approach and mutated the catalytic cysteine (C1716) of MHV PLP2 to a serine residue and co-crystallized it with free ubiquitin. The X-ray structure of MHV PLP2 (C1716S) in complex with ubiquitin was determined to a resolution of 1.85 Å and an overall Rfree = 19.6% and Rwork = 15.8% (PDB: 5WFI). The overall structure of the MHV PLP2-ubiquitin complex is similar to other PLP2/PLpro ubiquitin-bound structures (Fig 1A). Ubiquitin binds within the palm region and is gripped by the zinc-fingers motif while the c-terminus extends into the active site.
(A) Overall structure of MHV PLP2-CI94S-Ub con1plex. Do111ains are color coded: Ub in yellow, Ubl2 do111ain ofPLP2 in purple, thun1b domain in orange, pah11 do111ain in cyan, and fingers don1ain in green. Residue D1772 of PLP2 is located outside of the active site, which is circled in black. (B) Hydrogen bond interactions between MHV PLP2 R1803 and the backbone of ubiquitin A46. (C) Binding interactions between MHV PLP2 01772 and two arginine residues (R42 and R72). (D) Hydrophobic interactions between F1812 of MHV PLP2 and ubiquitin residues 144 and V70. The 2Fo-Fc 111aps (blue) surrounding the residue5 are contoured at la in each panel. The PDB coordinates for the MHV PLP2-Cl716S-Ub complex have been deposited under PDB Code 5WFI. (E) Sequence alignment of coronavirus papain-like protease/deubiquitinating domains from MHV (1606-1911 aa, accession# AAX23975), SARS (1541-1854 aa, accession# ACZ72209), and MERS (1480-1803 aa, accession# AHY21467). Numbering is based on MHV sequence. A111ino acids mutated in this study are indicated with an asterisk and those that bind ubiquitin are boxed in green. The sequence aligm11ent was created using ESPript.
Next, we aligned the primary amino acid sequence of the MHV PLP2 domain with the SARS-CoV and MERS-CoV papain-like protease domains (Fig 1E). The sequence alignment and X-ray structure of the MHV PLP2–ubiquitin complex were then analyzed in conjunction with the previous structural and mutagenesis studies on SARS- and MERS-CoV to identify candidate residues that could be mutated to render a loss of DUB activity in vitro. From this analysis, we identified three residues (R1803, D1772, F1812) in MHV PLP2 that form direct interactions with ubiquitin (Fig 2B-D). Two of the side chain guanidinium nitrogens of R1803 form direct hydrogen bonds with the backbone carbonyl oxygen of A46 in ubiquitin (Fig 1B). The two side chain carboxylate oxygens of D1772 in MHV PLP2 interact with ubiquitin by forming direct bonds with each of the guanidinium nitrogens of R42 and with one of the guanidinium nitrogens of R72 (Fig 1C). Finally, F1812 forms Van der Waals contacts with the side chains of I44 and V70 and the delta-carbon of R42 of ubiquitin (Fig 1D).
(A) Relative kinetic activities of three mutant MHY PLP2 enzymes toward three substrates: z RLRGG-AMC (green), Ub-AMC (blue), and ISGl5-AMC (yellow) compared to the wild type enzyme. (B) Steady state kinetic parameters for wild type and D 1772A mutant enzymes. (C) Sequence alignn1ent of Ub and ISG15 fro1n hun1an and 1nouse generated byClustal Omega. The two arginine residues of Ub (R42 and R72) that interact with D1772 are indicated by arrows. R72 is conserved between Ub and ISG15, whereas R42 (shaded in yellow) is only present in Ub. Accession nu1nbers: Human_Ub, Iubq; Mouse_Ub, P6299 I; Hu1nan TSG15, AAH09507; Mouse TSG 15, AAT09347. The sequence align1nent was created using ESPript.
To identify a mutant MHV PLP2 enzyme that retains protease activity but exhibits reduced DUB and/or deISGylating activity, we performed site-directed mutagenesis on each of the residues (R1803, D1772, F1812) by changing each to an alanine to disrupt interactions with ubiquitin. Each mutant enzyme was expressed, purified, and tested for its ability to hydrolyze three substrates: z-RLRGG-AMC, Ub-AMC, and ISG15-AMC. The activity of each mutant enzyme toward each substrate relative to the wild type enzyme is shown in Fig 2A. All three mutants retained their ability to hydrolyze the peptide substrate but each mutant had altered specificity toward Ub-AMC and ISG15-AMC substrates. Mutation of F1812 resulted in a substantial decrease in hydrolysis of both Ub-AMC and ISG15-AMC (Class I), whereas mutation of R1803 resulted in loss of activity only toward ISG15-AMC (Class II), and mutation of D1772 resulted in loss of activity only toward Ub-AMC (Class III).
Since one of the primary goals of this study is to understand the contribution of DUB activity to viral replication and pathogenesis, we next focused on quantitating further the effects of the D1772A mutant on the steady-state kinetic parameters of MHV PLP2 toward the three different substrates (Fig 2B). The RLRGG-AMC peptide substrate is often used as surrogate of the viral polyprotein substrate and the kinetic data in Fig 2B show that this substrate is still well-recognized and cleaved by the D1772A mutant. In fact, we observed a small rate enhancement in the catalytic efficiency (i.e., kcat/Km) compared to the wild type enzyme.
The Ub-AMC substrate, on the other hand, is poorly recognized and cleaved by the D1772A compared to the wild type enzyme. The wild type enzyme normally interacts strongly with Ub-AMC with a Km value of 0.67 µM. However, mutation of D1772 to an alanine significantly disrupts the interaction with ubiquitin, making it impossible to saturate MHV PLP2 under normal experimental conditions (Fig 2B). The net result is a significant reduction in the catalytic efficiency (kcat/Km) compared to the wild type enzyme, which was the goal of these trials.
The kinetic response of MHV PLP2 toward another substrate, ISG15-AMC, was also determined. ISG15 is an important ubiquitin-like modifier that is upregulated and used to ISGylate host proteins during viral infection. A number of viruses, including coronaviruses, engender ISGylation during infection but the function(s) and importance of this activity are not clear. For MHV, neither the wild type nor the D1772A mutant PLP2 enzyme can be saturated with ISG15-AMC, suggesting weak binding with this ubiquitin-like modifier (Fig 2B). Moreover, the R1772A mutation does not disrupt the interaction with ISG15 but in fact enhances it to some degree. A potential explanation for the observed selective disruption of ubiquitin binding stems from our analysis of a primary sequence alignment of ubiquitin and ISG15 and the residues that interact with D1772 (Fig 2C). The interaction between MHV PLP2 D1772 and the R42 residue in human and mouse ubiquitin is absent in human and mouse ISG15 since this residue is a tryptophan in human and mouse ISG15. Therefore, in line with our observations, D1772A mutation would not be expected to alter ISG15 binding. In contrast to R42, residue R72 is conserved in both ubiquitin and ISG15 and its interaction with MHV PLP2 for ubiquitin is likely weaker than that with ISG15.
The in vitro biochemical studies presented here support the notion that we are able to use a structure-guided approach to uncouple the DUB enzymatic activity from MHV PLP2 while preserving the peptide hydrolysis and deISGylating activities of PLP2. Next, we focused on comparing the activity of the mutant enzyme to its wild type counterpart for the ability to remove Flag-tagged-ubiquitin conjugated to host proteins in cultured cells (Fig 3A). We found that in cells, wild type PLP2 exhibits robust DUB activity and removes ubiquitin modifications from multiple cellular proteins. On the other hand, the PLP2-D1772A mutant exhibits reduced DUB activity, similar to that of the previously documented catalytic cysteine to alanine mutant, PLP2-CA [22]. To determine if this impaired DUB activity altered the ability of PLP2/DUB to act as an interferon antagonist, we transfected cells with RIG-I, an interferon-luciferase reporter, and either wild type or mutant PLP2 and measured luciferase activity at 18 hours post-transfection. In agreement with previous reports, we find that wild type PLP2/DUB acts as an interferon antagonist, reducing reporter activity by 50-80%. In contrast, PLP2-D1772A is unable to significantly reduce interferon activation in this assay despite similar expression levels of the wild type and mutant versions of the protein (Fig 3B). We also evaluated the protease activity of the enzymes in cells using two independent trans-cleavage assays and found that the wild type and DUB-mutant enzymes produce similar levels of cleaved products. These results indicate that the D1772A substitution did not alter protease activity (Fig 3C and D), in agreement with the in vitro kinetic results described above (Fig 2). Together, these studies reveal that aspartic acid residue 1772 of MHV-PLP2 is important for DUB-mediated interferon antagonism, but not for protease activity.
(A) Western blot assessing the DUB activity of PLP2. (B) IFN antagonism of PLP2 was determined using an IFN-luciferase reporter stimu lated by N-RIG-1 expression. Protease activity was evaluated using (C) a trans-cleavage assay that detects the cleaved products by western blot and (D) a pGlo biosensor assay which is activated by PLP2-mediated cleav age of the substrate. Data are representative of at least two independent experiments. Data in (B) and (D) are presented as means ± SD.
Since the D1772A substitution did not impact protease activity, we reasoned that we should be able to recover infectious virus containing this substitution, thereby allowing us to determine if the mutation has any effect on viral replication kinetics and interferon antagonism in the context of the live virus. We generated the mutant virus via reverse genetics [23], performed full genome sequencing to verify the genotype (2 nucleotide changes at positions 5525 and 5526, resulting in D1772A substitution in the replicase polyprotein), and designated the virus as DUBmut. Upon evaluating virus replication of the DUBmut virus by performing a growth kinetics experiment in parallel with wild type virus, we found that DUBmut replicates with essentially identical kinetics as the wild type in a murine astrocytoma cell line (DBT cells) (Fig 4A). These results are consistent with previous studies of coronavirus interferon antagonists, which showed in many cell lines that viral-mediated interferon antagonism is not essential for virus replication [10, 11].
(A) Replication kinetics of WT and DUBmut virus in DBT cells. (B) IFNcxl I mRNA levels in WT and DUBmut-infected BMDMs were assessed at indicated time points by qRT-PCR. (C) IFNcx protein levels in the supernatants of infected BMDMs were evaluated at the times indicated. (D) Comparison of IFNcx11 levels in B6 versus MDAS-/- BMDMs at 12 hours post-infection (HPI). (E) Assessing levels of viral nucleocapsid (N) mRNA by qRT-PCR. (F) Replication kinetics of WT and DUBmut virus in BMDM cells. Data are representativeof at least two independent experiments and are presented as means ± SD. Data in (B)and (C) were statistically analyzed using unpaired t-tests. *, p < 0.05; **, p < 0.01.
To determine if the impaired DUB activity of the DUBmut virus had an effect on interferon antagonism, we infected primary bone marrow-derived macrophages (BMDMs) and evaluated viral replication kinetics and levels of interferon mRNA and protein. We observed significant activation of interferon alpha (IFNα) mRNA expression (Fig 4A) that is coupled with release of IFNα protein into the supernatant, as detected by ELISA (Fig 4B). We show that this activation of IFNα is dependent on expression of pattern recognition receptor MDA5 (Fig 4D), in agreement with previous reports [10, 11]. To our surprise, we found that replication of DUBmut is not impaired relative to wild type in BMDMs, as measured by level of nucleocapsid RNA (Fig 4E) and evaluation of infectious virus particle production over time in the kinetics assay (Fig 4F). These results demonstrate that an elevated interferon response is generated during replication of the DUBmut virus, but that this interferon profile is not associated with reduced production of infectious particles in either DBT cells or BMDMs.
Disrupting DUB activity is mildly attenuating to coronavirus pathogenesis in mice
To complement our in vitro studies, we next sought to determine whether loss of DUB activity and the observed activation of interferon during DUBmut infection in macrophages is associated with an attenuated phenotype in mice. To test this, we inoculated mice intra-peritoneally with 6×104 plaque-forming units (pfu) of the designated virus and measured viral titer in the liver and spleen at the indicated days post-infection (dpi). We detected similar levels of infectious particles at 3 dpi, but reduced levels of infectious particles in the livers and spleens of DUBmut-infected mice at 5 dpi (Fig 5A). Similar pathology in liver sections was observed at 3 and 5 dpi (Fig 5B). These results suggest that the DUBmut virus is mildly attenuated compared to the wild type virus when evaluated using the intraperitoneal route of infection. We also used the more sensitive intracranial infection model to evaluate pathogenesis and found similar lethality among WT and DUBmut-infected mice (Fig 5C). We were surprised at the similarities between wild type and DUBmut infection in vivo since our previous study involving inactivation of a different viral interferon antagonist, EndoU, revealed profound attenuation and rapid clearance of an EndoU-mutant virus from infected mice [10]. Having shown that the DUBmut and the EndoUmut viruses each activate an innate immune response, we expected that infection of mice with either mutant virus would lead to similar outcomes. Therefore, as our expectations for the DUBmut virus were in stark contrast to what we observed, we next performed transcriptome profiling of infected macrophages at multiple time points in order to build a more complete understanding of the host response to the wild type virus, DUBmut virus, and the highly attenuated EndoUmut virus.
(A) Viral burden in livers and spleens isolated from WT- or DUBmut virus-infected mice was determined by plaque assay. The number of mice in each group is shown in parentheses. Data were statistically analyzed using unpaired t-tests and are presented as means ± SEM. (8) H&E staining of liver sections from infected mice at 3 and 5 days post-infection (DPI). Represen tative MHV-associate d liver lesions are indicated by arrows. (C) Weight loss (left panel) and survival curve (right panel) of mice after intracranial inoculation with I00 pfu of WT or DUBmut virus. The number of mice in each group is shown in parentheses. Data were statistically analyzed using log-rank tests and are presented as means ± SEM.
Transcriptome profiling reveals differences in kinetics and magnitude of host responses to mutant viruses
BMDMs were infected with the designated virus and total RNA was harvested for RNA-seq transcriptome profiling at 3, 6, 9, and 12 hours post-infection (hpi). Using the RNA-seq data, we identified differentially expressed genes by applying a cutoff of at least 4-fold elevated expression in wild type-infected cells at 12 hpi over mock (q < 0.05). We then utilized Cluster 3.0 software to visualize overall patterns of gene expression between all groups across all time points (Fig 6). These analyses reveal significant differences in overall patterns of gene expression between the EndoUmut-infected cells and wild type virus-infected cells, whereas the gene expression profile induced by the DUBmut virus is remarkably similar to wild type virus infection (Fig 6). More specifically, 2,879 genes are significantly transcriptionally activated (>4-fold, q < 0.05) in wild type- and DUBmut-infected macrophages by 12 hpi when compared to mock-infected macrophages. Interestingly, our analyses also identified a subset of 231 genes, indicated by the bracket in Figure 6, that were most highly upregulated in EndoUmut-infected cells. It is important to note that these genes were also induced during WT and DUBmut infections, but not to the same magnitude as detected in the EndoUmut-infected cells. We next sought to functionally cluster these 231 genes as a means of investigating the respective host transcriptional response to each mutant virus.
Total RNA was extracted from mock- and virus-infected BMDMs at 3, 6, 9, and 12 hours post-infec tion (HPI). Samples were subjected to SR50 Illumina sequencing. Raw reads were processed to generate differ ential expression data and normalized counts for each gene in each sample. The expression data from WT MHV-infected BMDMs at 12 HPI was compared with mock-infected cells; a statistical significance cutoff of q < 0.05 and a fold-change cutoff of > 4 were applied to select the most highly statistically significant differen tially expressed genes. The normalized counts for the 2,879 genes that met these cutoffs were averaged across replicates and converted to log2-transfonned z-scores. These expression values were then imported into Clus ter 3.0 software to mathematically arrange the gene list based on similarities in expression patterns between samples via the Pearson correlation (uncentered) and centroid linkage metrics. Java TreeView software was used to visualize the output from Cluster 3.0. Plotted are row z-score-standardized log2-transfonn ed values for each gene across all samples and timepoints. The color bar indicates the approximate row z-score that is associ ated with each color, with wanner colors corresponding to higher relative expression values within each row and cooler colors corresponding to lower relative expression values. Bracket indicates the genes selected for subsequent analysis in Figure 7.
EndoUmut infection activates genes associated with an antiviral response
Using an online tool called Database for Annotation, Visualization and Integrated Discovery (DAVID) to cluster the 231 genes bracketed in Figure 6 based on functional similarities [24, 25], we find that the protein products of these genes are predominantly involved in antiviral and signaling pathways. Notably, the DAVID analyses reveal a subset of 30 unique genes, including several interferon isoforms, which are clustered together into statistically significant pathways associated with the immune response and signaling cascades (Fig 7A). Interestingly, the heat map of these genes reveals a similar temporal trend in upregulation of IFN gene expression during infection with each of the three viruses; however, this expression profile is by far the most pronounced in the EndoUmut-infected cells. In this heatmap view, as in Figure 6, it is difficult to ascertain any difference in expression of IFN genes between wild type- and DUBmut-infected cells; therefore, to obtain more information on the kinetics of transcriptional activation between groups, we quantitated the reads associated with each gene and graphically display the normalized read counts in Figure 7B (also see Supplemental Fig 1). Using this method of plotting the read counts over time, we detected statistically significant transcriptional upregulation of IFN gene expression in the DUBmut-infected cells, in agreement with the data presented in Figure 4. However, the magnitude of DUBmut-induced IFN gene expression was dwarfed by IFN activation in EndoUmut-infected cells. EndoUmut-infected macrophages exhibited markedly elevated levels of IFN transcripts as early as 9 hpi. By 12 hpi, expression of the genes bracketed in Fig 6 upon EndoUmut infection was significantly higher than during WT or DUBmut infection, with our assay detecting 1,000-10,000 more reads per gene in the EndoUmut-infected cells. This focused response to EndoUmut is successful in limiting virus replication in cultured macrophages, as demonstrated in our previous report [10]. Notably, we evaluated the transcriptional response during the first 12 hpi, prior to the onset of apoptosis, which we and others have shown to occur in BMDMs upon EndoUmut infection [10, 11].
After applying DAVID functional clustering to the data shown in Figure 6, the expression profiles of30 genes within the cluster brack eted in Fig 6 were plotted as a heatmap and as line graphs to visualize the patterns of expression. (A) Plotted are row z-score-standardized log2-transformed means of replicates for all samples. (B) Normalized read count values for the most highly differentially expressed genes are plotted as line graphs (also see supplemental figures). Plotted are the average normalized counts for each gene in EndoUmut-(red), DUBmut-(green), WT MHV-(blue), and mock-infected (black) BMDMs infection groups over all four timepoints. The normalized counts from each infection group at 12 hours post-infection (HPI) were subjected to statistical testing using two-tailed Student’s t-tests. *, p < 0.05; **, p < 0.01; ***, p< 0.001; ****, p <0.0001; ns, not significant. Data are presented as means ± SD.
Wild type and DUBmut coronavirus replication activate the unfolded protein response and host response to cell damage
In contrast to the robust, but focused, antiviral gene expression response detected in EndoUmut-infected macrophages, we uncovered a complex response involving 2,879 genes in both WT- and DUBmut-infected macrophages by 12 hpi. The overall patterns of expression of these genes in DUBmut- and WT-infected BMDMs are strikingly similar (Figs 6 and 7). We therefore propose that the comparable outcomes of infection observed in DUBmut- and WT-infected mice are a consequence of transcriptionally similar host responses to each virus. Using DAVID-based functional clustering, we identified multiple statistically significant functional clusters within this set of 2,879 genes that are implicated in signaling cascades, the general immune response, and transcriptional regulation. These analyses reveal substantial upregulation in expression of many proinflammatory genes, including well-known players TNF, IL-6, IL-1b, CXCL-10, CXCL-11, and CCL3 (Fig 8). We noted significant differences in the activation of these genes associated with a proinflammatory response in BMDMs infected with either the wild type virus or DUBmut compared to those infected with EndoUmut. Potent activation of these genes by 12 hpi may stimulate recruitment of inflammatory mediators directed at tissue repair [26]. In addition to this pro-inflammatory signature, wild type- and DUBmut-infected cells also exhibit hallmarks of activation of the host unfolded protein response (UPR).
Normalized co unt values for these genes are plotted as line graphs as described in Figure 7. Plotted are the average normalized counts for each genein EndoUmut-(red), DUBmut (gree n), WT MHV-(blue), and mock-infected (black) BMDMs infection groups over all four timepoints. The normalized counts from each infection group at 12 hours post-infection (HPI) were subjected to statistical analysis using two-tailed Student’s t-tests. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. Data are presented as means ± SD.
UPR pathways are activated when unfolded proteins accumulate in the endoplasmic reticulum (ER), at which point the host cell initiates measures to resolve this overload [26]. In the context of virus infection, viral glycoproteins accumulate in the ER, triggering the release of BiP from three sensor proteins: IRE1, ATF6, and PERK. Activation of these sensors triggers signaling cascades resulting in transcriptional activation of genes encoding ER chaperones and proteins involved in lipid synthesis and amino acid transport (Fig 9A). Based on our analyses of differential gene expression, we report significant transcriptional upregulation of multiple genes associated with activation of ER sensors IRE1, ATF6, and PERK, such as Edem1, Hspa5 (encoding BiP protein), and Ddit3 (encoding CHOP), in response to virus replication, with the most robust response detected in cells infected with the wild type or DUBmut virus (Fig 9 B, C, and D). Overall, these differential gene expression analyses in macrophages reveal similar host responses to the wild type and DUBmut viruses that include activation of UPR pathways and proinflammatory genes, whereas a distinct transcriptional profile during infection with the EndoUmut virus is predominately defined by a focused, robust antiviral response.
(A) Schematic diagram of unfolded protein response. To evaluate the transcriptional response during viral infection, normalized count values for nine UPR-ind uced genes at 12 HP! are plotted. Genes are grouped into three panels according to whether their expressionis triggeredby (B) PERK, (C) ATF6, or (D) !REI pathway signaling. Data were subjected to statistical analysis using two-tailed Student’s t-tests. *, p < 0.05; **, p < 0.01; ***, p< 0.001; ****, p < 0.0001; ns, not significant. Data are presented as means ± SD.
Discussion
Here, we report that inactivation of a coronaviral interferon antagonist, EndoU, profoundly alters the host response to viral replication in macrophages. We find that the EndoUmut virus elicits a rapid, robust, and specific antiviral response that was effective in limiting virus replication. In contrast, our data show that the WT and DUBmut viruses ultimately elicit very similar host responses that are both characteristic of an unfolded protein response and consistent with a proinflammatory profile that is associated with viral pathogenesis [27]. The results from the DUBmut-infected macrophages indicate that the mere induction of type I IFN is not a sufficient marker for attenuation of the virus. Instead, these results suggest that the timing and the magnitude of the host antiviral response are critical for determining the outcome of infection in macrophages. Our observation that the EndoUmut virus induces an earlier and more profoundly elevated type I interferon response than even the DUBmut virus implies that there is a threshold of IFN expression that must be breached before the cell mounts an effective antiviral response.
The structure-guided approach used to generate the DUBmut virus allowed for characterization of three different classes of mutant enzymes: Class I, deficient in both DUB and de-ISGylating activity; Class II, deficient in de-ISGylating activity only; and Class III, deficient in DUB activity but competent in protease and de-ISGylating activity. We utilized three unique biochemical substrates, each with a conjugated fluorescent AMC reporter, to evoke the multi-functional activities of PLP2. Activity against the z-RLRGG-AMC peptide substrate represents the polyprotein processing activity of MHV PLP2, while UB-AMC and ISG15-AMC stimulate the deubiquitinating and deISGylating activities of the enzyme, respectively. The kinetic data for the D1772A DUBmut hydrolysis of the z-RLRGG-peptide provided in Figure 2B indicate that the polyprotein processing ability of this mutant is likely not affected by the D1772A substitution. The deISGylating ability of the enzyme is also not affected. In contrast, the mutant enzyme’s deubiquitinating activity is significantly reduced relative to wild type, which is most likely due to a lowered binding affinity for ubiquitin as the enzyme could not be saturated with Ub-AMC as a substrate.
We were able to reproduce the enzymatic profile of the purified PLP2-D1772A mutant protein when we expressed it in cell culture (Fig 3). Therefore, our finding that the DUBmut virus containing the PLP2-D1772A substitution activates an elevated antiviral response in macrophages compared to the wild type virus, but that this antiviral state results in only mild attenuation of disease in mice relative to WT infection, was unexpected. Previous studies demonstrated that ubiquitin has important roles in both the activation and the attenuation of innate antiviral pathways [28]; therefore, we anticipated a more remarkable phenotype for a DUB-mutant virus. We can imagine several possible explanations for our findings. First, it is possible that viral DUB activity has a relatively minor role in shaping pathogenesis in this system. In fact, our recent studies using SARS-CoV and SARS-related CoVs found that the papain-like protease domain/DUB is a virulence trait that varies among members of the SARS-coronavirus species [9]. In that study, we found that replacing the SARS PLP2/DUB domain with a SARS-related PLP2/DUB domain reduced the ability of that virus to antagonize the innate immune response. Together, this previous report in conjunction with the current study support the concept that different PLP2/DUB domains may have distinct effects on antagonism of the innate immune response depending on the virus and the host cell type. Another possibility is that the DUB-mutant virus we generated may not have been sufficiently debilitated in its DUB activity to result in altered pathogenesis. We found that it was difficult to recover viable DUB-mutant viruses; indeed, this D1772A mutant was the sole viable DUB-mutant representative of our many attempts. Because an elevated interferon response was elicited from DUBmut-infected cells, this mutant virus fulfilled our criteria demonstrating the inactivation of an interferon antagonist. However, we speculate that if we are able to recover mutants that exhibit a range of DUB activity, we may be able to more fully assess the role of DUB activity as a contributor to coronaviral pathogenesis. Despite these caveats, the MHV DUB-mutant generated in this study did exhibit a reproducible phenotype of eliciting an elevated interferon response in infected macrophages that was associated with mild attenuation of pathogenesis with reduced titers in the livers and spleens of mice at day 5 post-infection. Overall, we conclude that DUB activity is indeed a virulence trait, but not a major driver of virulence for MHV. Our results support the concept that multiple viral factors likely work in concert to shape and ultimately limit the innate immune response to coronavirus replication.
The other remarkable finding from this study was the distinct transcriptional profile elicited by the EndoUmut virus during infection of bone marrow-derived macrophages compared to the profiles of wild type- and DUBmut-infected cells (Fig 6). We detected elevated levels of interferon and interferon-responsive genes as early as 6-9 hpi, with EndoUmut-infected cells exhibiting by far the highest expression levels of these genes. We and others found that an antiviral response to EndoUmut infection results in activation of apoptosis, which subsequently limits virus replication in cell culture and in infected mice [10, 11]. The current study indicates that screening MHV mutants in interferon-responsive cells may be an effective approach to identifying strains that stimulate a robust innate immune response, which may then restrict virus replication in animals. Previous studies of coronavirus-encoded interferon antagonists focused on evaluating the host transcriptional response to infection at later time points (such as 24 and 48 hpi) and in a variety of cell types or the tissues of infected animals. For example, a study comparing infection of mice with wild type SARS-CoV versus a virus containing a mutation in the interferon antagonist nsp16, termed the dNSP16 CoV, revealed that the transcriptional profile in the lungs of the dNSP16-infected mice mirrored the response to wild type virus [12]. However, combining the dNSP16 mutation with a mutation in another interferon antagonist, ExoN, was shown to reduce disease in mice and elicit protective immunity. It would be interesting to determine if the transcriptional profile elicited by the double mutant SARS-CoV is altered compared to wild type virus, particularly at early times after infection. A study of MERS-CoV comparing the host response to wild type virus with the response to a mutant virus containing deleted accessory open reading frames (MERS-CoV dORF3-5) provides evidence of significant differences in the early transcriptional responses to infection in Calu3 2B4 cells [29]. This study revealed that MERS-CoV dORF3-5 infection prompted earlier (7, 12, and 24 hpi) and more robust type I and type III interferon responses, and that the mutant virus was more sensitive than wild type to pre-treatment with interferon. The MERS-CoV interferon antagonist mutant virus was attenuated in mice and elicited a protective response against subsequent lethal challenge. For coronaviruses, one difficulty that arises when evaluating the role(s) of viral-mediated modulation of the host response is that multiple viral proteins, including both structural and nonstructural proteins, have been shown to antagonize the innate immune response in vitro and/or in vivo. These include: nsp16-2’O MTase, nsp14-ExoN, nsp1, nsp7, E protein, N protein, M protein, SARS-CoV-ORF6, MERS-CoV-ORF3-5, MERS-CoV-4a, and MERS-CoV-4b [12,29–33]. Each of these interferon antagonists may play either a cell- or tissue-type specific role, or act in concert with other factors during viral replication to mitigate the innate immune response. Further studies are needed to fully understand if all or only a subset of these antagonists must be silenced to generate an effective, live-attenuated vaccine. The results presented here and from the study of the MERS-CoV dORF3-5 mutant virus indicate that inactivating interferon antagonists and screening for an early and robust antiviral transcriptional profile may represent an efficient and informative approach to evaluating live-attenuated vaccine candidate strains for existing and emerging coronaviruses.
Our results demonstrating upregulation of the unfolded protein response (UPR) in response to wild type and DUBmut coronavirus replication confirm and extend the work of earlier studies that documented activation of the ER sensors PERK, IRE1, and ATF-6 during coronavirus infection [32,34,35]. Heavy utilization of the endoplasmic reticulum for generating coronavirus replication complexes and of the ER-Golgi intermediate compartment for assembling virus particles places a substantial load on the host translational machinery during infection. Host sensors IRE1, ATF-6, and PERK are situated in the ER to sense and respond to such overload by prompting upregulated expression of genes encoding ER chaperones, amino acid transporters, and activators of lipid biosynthesis. Ironically, many of these proteins ultimately facilitate virus replication and assembly. Notably, it has been demonstrated that UPR pathways that promote apoptosis are blocked during coronavirus replication [32, 34]. The ability of viruses to modulate the UPR has important implications for the innate immune response to such viruses because the UPR has been shown to attenuate antiviral defenses by way of degrading the type I interferon receptor [36]. To our knowledge, the results presented here provide the first transcriptomic evidence of UPR activation in coronavirus-infected macrophages, underlining an important role for UPR pathways in the coronavirus life cycle. Our observation that EndoUmut-infected macrophages exhibit significantly lower expression of several genes involved in UPR pathways compared with wild type- and DUBmut-infected cells is consistent with the reduced levels of virus replication detected in EndoUmut-infected macrophages.
The notion of inactivating viral interferon antagonists as a means of generating live-attenuated vaccines is supported by recent reports of screening for inactivation of influenza A virus-encoded interferon antagonists [3], as well as studies that revealed that the classic vaccine strain of yellow fever virus encodes an interferon antagonist in the NS5 protein [37]. For coronaviruses, it is not yet clear if disabling a single interferon antagonist, such as the highly conserved EndoU, will be sufficient to attenuate viruses that infect different cell types in different species. Promisingly, our studies of a coronavirus that causes lethal disease in piglets, porcine epidemic diarrhea virus (PEDV), revealed that inactivation of EndoU activity is associated with attenuated disease [38]. However, additional work is needed to evaluate potential reversion of EndoU mutant viruses and to determine if inactivating multiple interferon antagonists is an effective approach for generating safe and effective live-attenuated coronavirus vaccines.
Materials and Methods
Ethics Statement
The mouse experiment in this study was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The experimental protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at Loyola University Chicago (IACUC#: 2016-029). C57BL/6 female mice were purchased from The Jackson Laboratory and maintained in the Comparative Medicine Facility of Loyola University Chicago. Mice were consistently monitored for signs of distress over the course of the experiments to be removed from the experiment and euthanized using carbon dioxide inhalation to prevent unnecessary suffering.
Cells
Human embryonic kidney (HEK) 293T cells were purchased the from American Type Culture Collection (ATCC, # CRL-11268) and maintained in DMEM (#10-017-CV, Corning) containing 10% fetal calf serum (FCS) and supplemented with 1% nonessential amino acids, 1% HEPES, 2% L-glutamine, 1% sodium pyruvate, and 1% penicillin/streptomycin. DBT cells were cultured in MEM (#61100-061, ThermoFisher) supplemented with 5% FCS, 2% L-glutamine, and 10% Tryptose Phosphate Broth (TPB). The 17Cl-1 cell line was maintained in DMEM containing 5% FCS. Baby hamster kidney cells expressing the MHV receptor (BHK-R) were kindly provided by Mark Denison (Vanderbilt University Medical Center) and maintained in DMEM supplemented with 10% FCS, 2% L-glutamine, and 0.8 µg/mL G418. Bone marrow-derived macrophages (BMDMs) were prepared and cultured as described previously [10].
Plasmids and mutagenesis
The sequence of the PLP2 domain (amino acids 1525-1911 of MHV pp1ab) in frame with a V5 epitope tag was codon-optimized, synthesized by Genscript (Piscataway, NJ) (sequence available upon request), and cloned into pCAGGS vector. For mutagenesis, an overlapping PCR strategy was used with synthetic primers (sequences available upon request). The introduced mutations were verified by sequencing. The RIG-I and nsp2/3-GFP expression plasmid was kindly provided Ralph Baric (University of North Carolina). The IFNβ-Luc reporter plasmid was a gift of John Hiscott (Jewish General Hospital, Montreal, Canada). The Flag-Ub plasmid was kindly provided by Adriano Marchese (University of Wisconsin-Milwaukee).
MHV PLP2 wild type and mutant purification, kinetics, and X-ray structure
The wild type, C1716S, D1772A, R1803A, and F1812A mutant enzymes were expressed and purified similar to our previously published methods except that the MHV PLP2 construct used here (amino acids N1609 to N1909) was absent the DPUP domain [39]. Crystallization and X-ray structure determination details will be published elsewhere. Steady-state kinetic studies on the wild type, D1772A, R1803A, and F1812A mutant enzymes with substrates Z-RLRGG-AMC (where Z is a carboxybenzyl protecting group), Ubiquitin-AMC, and ISG15-AMC were performed as described previously [39]. X-ray structural coordinates have been deposited under PDB code 5WFI. Structure figures were generated with the software program UCSF Chimera [40].
Protease and deubiquitinating activity assays
To determine the protease activity of PLP2, HEK293T cells grown to 70% confluency in 24-well plates (Corning) were transfected using TransIT-LT1 (MIR2300, Mirus) according to the manufacturer’s protocol. For the protease activity assay, HEK293T cells were transfected with 25 ng nsp2/3-GFP plasmid and 200 ng pCAGGS-PLP2-V5 expression plasmids (wild type and mutant). To determine deubiquitinating activity of the proteins, cells were transfected with 200 ng Flag-Ub plasmid and pCAGGS-PLP2-V5 expression plasmids (wild type and mutant). Cells were lysed 24 h post-transfection with 100 µL of lysis buffer (comprising 20 mM Tris [pH 7.5], 150 mM NaCl, 1mM EGTA, 1mM EDTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1mM β-glycerophosphate, 1mM sodium orthovanadate, 1 µg/mL leupeptin, and 1mM phenylmethylsulfonyl fluoride). Whole cell lysates were separated by SDS-PAGE and transferred to PVDF membrane in transfer buffer (0.025M Tris, 0.192M glycine, 20% methanol) for 1 hour at 60 Volts at 4°C. Following this, the membrane was blocked using 5% dried skim milk in TBST buffer (0.9% NaCl, 10mM Tris-HCl, pH7.5, 0.1% Tween 20) overnight at 4°C. The membrane was incubated with either polyclonal rabbit anti-GFP antibody (A11122, Life Technologies) for the protease assay, or mouse anti-flag (F3165, Sigma) for the DUB assay. The membrane was then washed three times for 15 minutes in TBST buffer followed by incubation with either secondary donkey anti-rabbit-HRP antibody (711-035-152, Jackson ImmunoResearch) or goat anti-mouse-HRP antibody (1010-05, SouthernBiotech). Then the membrane was washed three more times for 15 minutes in TBST buffer. Detection was performed using Western Lighting Chemiluminescence Reagent Plus (PerkinElmer) and visualized using a FluoroChemE Imager (Protein Simple). The expression of PLP2, β-actin, and calnexin were probed with mouse anti-V5 antibody (R960, ThermoFisher), mouse anti–β-actin (A00702, Genscript), or mouse anti-calnexin antibody (2433S, Cell Signaling), respectively.
Biosensor live cell assay
The protease activity of PLP2 was also assessed using a biosensor live cell assay as described previously [41]. Briefly, HEK293T cells in a 96 black-wall plate were transfected with 37.5 ng pGlo-RLKGG construct and 50 ng PLP2 expression plasmids. GloSensor (E1290, Promega) reagent diluted in DMEM+10% FCS was added at 18 hpi. Plates were read using a luminometer (Veritas) every hour over a course of 5 hours.
Generating DUB-mutant MHV
We used a previously described reverse genetics system of MHV-A59 [23] to generate the DUB-mutant virus. Briefly, the nucleotides coding for the Asp-1772 of the PLP2 domain were mutated via site-directed mutagenesis. Viral genomic RNA from in vitro transcription of ligated cDNA fragments using a mMESSAGE mMACHINE T7 Transcription Kit (AM1344, Thermal Fisher) was electroporated into BHK-R cells. Cell supernatants were collected as viral stock following observation of cytopathic effects. Rescued virus was plaque-purified, propagated on BHK-R cells, and titrated on 17Cl-1 cells. The stock virus was subjected to full-genome sequencing and the sequences were aligned to the parental strain, with the intended engineered mutation detected and no additional mutations detected (Kansas State University Diagnostic Laboratory).
Growth kinetics
DBT cells or BMDMs were infected with wild type icMHV-A59 or DUB-mutant virus at a multiplicity of infection (MOI) of 1 in serum-free medium. After a one-hour incubation, the inoculum was replaced with fresh, complete medium. Cell culture supernatants were collected at indicated time points and titrated by plaque assay on 17Cl-1 cells. Titers were obtained from three independent assays for each sample. Graphs of virus kinetics were generated using Prism software (GraphPad Software).
Quantification of IFNα production by RT-qPCR and ELISA
BMDMs in a 12-well plate were mock-infected or infected with MHV at a MOI of 1. At indicated time points, monolayer cells were collected for RNA extraction and cell culture supernatants were harvested for ELISA analysis. To determine IFN-α11, β-actin, or MHV-A59 N gene mRNA levels, total RNA was extracted from collected cells using a RNeasy Mini Kit (74104, Qiagen). The first strand cDNA was synthesized from an equal amount of RNA using Rt2 HT First Strand Kit (330401, Qiagen). qPCR was then performed with specific primers for mouse IFN-α11 (PPM03050B-200, Qiagen), mouse β-actin (PPM02945B-200, Qiagen), or MHV-A59 N gene using RT2 SYBR Green qPCR Mastermix (330502, Qiagen) in the Bio-Rad CFX96 system. The thermocycler was set as follows: one step at 95°C (10 min), 40 cycles of 95°C (15 s), 60°C (1 min) and plate read, one step at 95 °C (10 s), and a melt curve from 65°C to 95°C at increments of 0.5°C/0.05s. Samples were evaluated in triplicate and data are representative of three independent experiments. The levels of mRNA were reported relative to β-actin mRNA and expressed as 2−ΔCT [ΔCT = CT(gene of interest) − CT(β-actin)]. The secreted amount of IFN-α in culture supernatants was assayed using a mouse IFN-α ELISA kit (BMS6027, eBioscience) according to the manufacturer’s instructions.
Mouse experiments
Evaluation of MHV pathogenesis in laboratory mouse was previously described [10, 42]. Briefly, for intracranial infections, six-week-old C57BL/6 female mice were inoculated with 600 pfu of virus in 20 μL PBS. Infected mice were monitored for body weight daily and euthanized when weight loss surpassed 25%. Statistical analyses of survival rate were performed using the log-rank test. For intraperitoneal infection, each mouse was injected with 6×104 pfu of virus in 100 μL PBS. Organs were collected at indicated time points and evaluated for viral burden. Liver pathology was evaluated using H&E staining by the Tissue Processing Core Facility at Loyola University Chicago.
RNA-seq data analysis pipeline
Raw RNA-seq reads were subjected to analysis using Galaxy’s online platform to generate differential gene expression data between infection groups [43]. Reads were clipped to remove any residual unique barcode sequences that were added during preparation of each sample for sequencing. Clipped reads were concatenated to combine multiple files per sample into a single file per sample. These files were groomed to ensure that all reads were in Sanger FASTQ format. FASTQ reads were aligned to the GRCm38 Ensembl build of the C57BL/6J mouse genome using HISAT2 aligner, which locates the region of the genome to which each read corresponds, resulting in an output BAM file [44]. All reads that did not align to the mouse genome (i.e., reads that originated from viral RNA) were discarded. The BAM files contained the alignment information for each read in that sample and were used as inputs into featureCounts, which quantifies the number of reads in each sample that corresponds to each gene in the mouse genome [45]. Finally, the output count data from featureCounts was used as the input for DESeq2 to calculate differential expression for each gene across all samples and treatment groups. DESeq2 was used to generate normalized count values for each gene in all samples [46]. These normalized counts are intended to correct for size differences between samples that might otherwise skew differential expression calculations if some samples contained substantially different numbers of total reads. The normalized count values were plotted and visualized in heat maps, generated using R, and line graphs, generated using Prism software. Submission of all RNA seq data is in progress with NCBI GEO, and an accession number will be provided prior to publication of this manuscript.
Identifying differentially expressed genes (DEGs)
To identify and analyze differentially expressed genes (DEGs) between infection groups, we used as a starting point the list of DEGs between mock- and WT MHV-infected BMDMs at 12 hpi that was generated as an output by DESeq2. This list of genes was filtered based on the statistical significance associated with the fold-change differential expression value. A q-value (aka adjusted p-value, calculated by DESeq2 for each gene in each comparison using the Benjamini-Hochberg procedure) of < 0.05 was chosen as the cutoff for statistical significance; genes whose differential upregulation values did not meet this cutoff in WT-MHV-infected BMDMs at 12 hpi compared with mock-infected cells were removed from the list [46]. Next, a differential expression magnitude cutoff of > 4 was applied to the remaining genes. Genes that were not more highly expressed by at least 4-fold in WT-MHV-infected cells at 12 hpi compared to mock were removed from the list. After applying these cutoffs, 2,879 genes remained and were arranged in order of most- to least-highly upregulated in WT-MHV-infected cells at 12 hpi compared to mock. Cluster 3.0 software was then used to apply mathematical clustering to the z-score-standardized log2-normalized mean normalized count values associated with each gene in each time point and infection group [47]. Specifically, the default settings—the similarity metric “Pearson correlation (uncentered)” and the clustering method “centroid linkage”—were applied to the list of 2,879 genes and the corresponding expression values for each gene across all samples to produce a hierarchically clustered gene list based on how similar or different the expression patterns were between groups of genes across all samples. This new list of clustered genes and their associated expression values were then visualized as a heatmap using Java TreeView software.
Author Contributions
Conceptualization: XD, AV, ADM and SCB. Investigation: XD, AV, YC, KRK, MH, RCM, AO. Formal Analysis: XD, AV, YC, KRK, MH, RCM, AO, ADM and SCB. Writing – Original Draft Preparation: XD, AV, MH, ADM and SCB. Writing – Review & Editing: with comments from XD, AV, MH, RCM, AO, ADM and SCB. Visualization: XD, AV, YC, KRK, MH, AO, ADM and SCB. Funding Acquisition and Supervision: ADM and SCB.
Supporting Information Legends
Figure S1. Plots of normalized read counts of IFN and IFN-stimulated genes identified by functional clustering as shown in Figure 7. (A) Normalized read count values for the most highly expressed genes are plotted as line graphs. (B) Normalized read count values for genes with less than 100 reads at 12 hours post-infection. Plotted are the average normalized counts for each gene in EndoUmut-(red), DUBmut-(green), WT-MHV-(blue), and mock-infected (black) BMDMs infection groups over all 4 time points.
Acknowledgments
We thank Dr. Qunfeng Dong for assistance with the bioinformatic analyses. This work was supported by the National Institutes of Health (NIH) grant R01 AI085089 (to SCB and ADM). MH and RCM were supported by NIH T32 Training Grant for Experimental Immunology (#AI007508) and RCM was supported by the Arthur J. Schmitt Dissertation Fellowship in Leadership and Service (Arthur J. Schmitt Foundation). Crystallization and DNA sequencing were partially supported by the Purdue Center for Cancer Research Macromolecular Crystallography and DNA Sequencing Shared Resources, which are supported by NIH grant P30 CA023168.