|Home | About | Journals | Submit | Contact Us | Français|
The host response to a virus is determined by intracellular signaling pathways that are modified during infection. These pathways converge as networks and produce interdependent phenotypes, making it difficult to link virus-induced signals and responses at a systems level. Coxsackievirus B3 (CVB3) infection induces death of cardiomyocytes, causing tissue damage and virus dissemination, through incompletely characterized host cell signaling networks. We built a statistical model that quantitatively predicts cardiomyocyte responses from time-dependent measurements of phosphorylation events modified by CVB3. Model analysis revealed that CVB3-stimulated cytotoxicity involves tight coupling between the host ERK and p38 MAPK pathways, which are generally thought to control distinct cellular responses. The kinase ERK5 requires p38 kinase activity and inhibits apoptosis caused by CVB3 infection. By contrast, p38 indirectly promotes apoptosis via ERK1/2 inhibition but directly causes CVB3-induced necrosis. Thus, the cellular events governing pathogenesis are revealed when virus-host programs are monitored systematically and deconvolved mathematically.
Coxsackievirus B3 (CVB3) is among the most common causes of viral myocarditis-associated heart failure in infants and young children (Esfandiarei and McManus, 2008). A major component of CVB3 pathogenesis is cell death of infected cardiomyocytes, which leads to immediate tissue damage and the subsequent release of virulent CVB3 progeny that furthers disease progression (Kawai, 1999). Intervening at the early stages of CVB3 cytotoxicity could potentially reduce the severity of the disease and the need for heart transplantation in patients with viral myocarditis.
Throughout infection, CVB3 modulates various cell-signaling pathways that enable virus propagation (Esfandiarei and McManus, 2008; Garmaroudi et al., 2010). Inhibiting these pathways may provide a therapeutic opportunity to restrict CVB3 pathogenesis. But, an important hurdle is our fragmentary understanding of how the CVB3 infection cycle intersects with the host network. Viruses such as CVB3 have evolved to modulate cell-signaling networks in ways that allow them simultaneously to evade host defenses, promote cell entry, and undergo replication in a changing environment (Esfandiarei and McManus, 2008; Ribet and Cossart, 2010). Blocking individual signaling pathways in host cells often reduces CVB3 infectivity but does not prevent infection entirely (Garmaroudi et al., 2010). It remains unclear whether such “partly required” pathways converge upon a common set of host effectors or instead make independent contributions to pathogenesis (Miller-Jensen et al., 2007). The challenge is that CVB3 adaptively perturbs a collection of host pathways, which must be examined concurrently with time to understand how they interact and give rise to viral functions.
Here, we took a multi-pathway systems approach to connect signaling and host-cell responses in an in vitro model of CVB3 infection (Tan et al., 2007). The dynamics of eight signaling phospho-proteins were monitored together with six CVB3-induced host-cell readouts at five different dosings of CVB3. We then linked CVB3-induced signaling to host-cell readouts by building a predictive data-driven model using two time-dependent combinations of measured phospho-proteins. The results of this analysis revealed unexpected connections between the ERK1/2, ERK5, and p38 MAPK pathways related to their control of apoptotic caspases and overall cell death induced by CVB3. Combined perturbations of these pathways validated the predictions of the model and deconstructed the CVB3 response as a mixture of apoptosis (involving ERK5) and necrosis (involving p38). Our results illustrate how viruses such as CVB3 hijack multiple host signaling pathways simultaneously but use relatively straightforward strategies to manipulate host responses.
To determine whether known CVB3-induced signaling events were sufficient to predict viral propagation and host-cell toxicity, we sought to build a predictive mathematical model based entirely on quantitative experiments. Data-driven modeling identifies higher-order statistical covariations that can be used for prediction and analysis (Janes and Yaffe, 2006). Unlike other modeling formalisms (Aldridge et al., 2006), data-driven approaches can accommodate variegated datasets and make predictions without detailed knowledge of the underlying biochemical mechanisms.
To build the model, we systematically assembled a host-cell signaling and response dataset in virus-infected cardiomyocytes. At five different CVB3 multiplicities of infection (M.O.I.), we profiled eight signaling phospho-proteins by ELISA at six time points over 24 hr together with six CVB3-induced host-cell readouts at three time points over 24 hr (Figures 1A and 1B). Each phospho-protein and host-cell readout was selected based on previous studies suggesting that they were critical for CVB3 pathogenesis (Table S1). Analyzing the information contained in this mechanism-rich signature would then allow us to examine how host-cell pathways were coordinately perturbed during CVB3 infection.
We found that CVB3-induced host-cell responses showed time and dose dependencies that were expected for end-stage readouts (Figure 1B). Activation of the initiator caspases, caspase-8 and caspase-9, was accelerated with increasing M.O.I., corresponding to more-complete activation of the effector caspase for apoptosis, caspase-3 (Riedl and Shi, 2004). Interestingly, readouts of CVB3 propagation, such as expression of the VP1 capsid protein and the titers of released viral progeny (RVP), did not accelerate appreciably as they increased with CVB3 M.O.I.. This finding suggests intrinsic limits to the timing of the CVB3 replication cycle downstream of the M.O.I.-dependent rate processes of viral docking and internalization. The pattern of overall CVB3 cytotoxicity fell in between that of caspase and viral readouts, showing some acceleration as host-cell viability dropped with increasing CVB3 titers. Thus, CVB3 infection of cardiomyocytes elicits a collection of host-cell and viral phenotypes that are monotonic in time but differ in their kinetics and dose-dependent behaviors.
By comparison, we found that the dynamic patterns of protein phosphorylation stimulated by CVB3 were substantially more complex than the associated phenotypic readouts (Figures 1A and 1B). As before, we observed accelerated phosphorylation of some CVB3-induced pathways with increasing M.O.I., such as p38 and Hsp27, but not others, such as ERK (Figure 1A). In addition, biphasic activation patterns were common, and many individual activation peaks appeared or disappeared above a critical threshold of CVB3 M.O.I. (e.g., ATF2, CREB, and IκBα). Sham infection with 0 M.O.I. did not lead to any meaningful changes when compared to 0.5 M.O.I. (Figure S1), confirming that the measured signaling events were due to CVB3 infection. The internal consistency of our phospho (p)-ELISA measurements was also verified by the strong correlations between p-Akt and p-GSK3β (R = 0.6), a direct substrate of Akt (Cross et al., 1995), and between p-p38 and p-Hsp27 (R = 0.8), a direct substrate of the MK2 kinase that is a substrate of p38 (Rouse et al., 1994). The p-ELISA signatures thus provided a reliable starting point for connecting CVB-induced signaling to host-cell outcomes.
One way that the observed CVB3-induced pattern of readouts could be coordinated is if each phospho-protein contributed incrementally to the pattern based on its extent of phosphorylation. Host cells would then “integrate” the intracellular state established by the level of CVB3 infection and gauge their responses accordingly. To test the feasibility of this network mechanism, we used partial-least-squares modeling to link linear combinations of measured phospho-proteins to observed CVB3-induced readouts (Janes et al., 2005; Janes and Yaffe, 2006). In a partial-least-squares model, linear combinations take the form of principal components, which are latent dimensions in the underlying dataset that are derived to be optimally efficient at predicting response outcomes (Jensen and Janes, 2012).
To build the model, we first subdivided the phospho-protein time courses into early (0–8 hr) and late (8–24 hr) phases and then time-integrated each early and late phospho-protein measurement for every CVB3 M.O.I.. This subdivision allowed us to separate biphasic activation profiles into early and late peaks. Using the phospho-protein data as a set of predictor variables, we next sought a partial-least-squares model that could predict all of the CVB3-induced readouts accurately and simultaneously. We found that a model with the two leading principal components could capture all of the measured readouts to within 97% accuracy (Figure 1C). Importantly, this model also accurately predicted readouts for individual M.O.I. conditions that were left out of the model training during crossvalidation (Figure 1D). The model thus supported a network mechanism in which multiple intracellular pathways work together by independently contributing to CVB3-induced readouts.
Principal components can be further analyzed by plotting the weighted linear combinations of the original measurements that provided the basis for accurate model predictions (Figure 2A) (Janes et al., 2005; Janes and Yaffe, 2006; Jensen and Janes, 2012). In this mapping, early and late phospho-proteins are depicted together with CVB3-induced readouts. Clusters of phospho-proteins and readouts indicate measurements with close association in principal-component space and highlight correlations in the data that are most worthy of follow-up experiments (Janes et al., 2006; Jensen and Janes, 2012).
Inspection of this principal-component mapping revealed that all CVB3-induced readouts were densely clustered in one region (Figure 2A, contours), suggesting that they were tightly coupled. Within the cluster lay the transcription factor ATF2, which is critical for CVB3 pathogenesis in vivo (Reimold et al., 2001), and p38, a MAPK that we recently showed is the dominant ATF2 kinase during CVB3 infection (Garmaroudi et al., 2010). We also found p-Hsp27 in the cluster, which was expected because of its strong concordance with p-p38 (see above). Conversely, we were surprised to find p-ERK located in the cluster together with p-p38, because the ERK1/2 and p38 pathways are generally thought to be activated by distinct stimuli and often serve antagonistic functions (Xia et al., 1995). Nevertheless, their tight association in the model suggested that ERKs and p38 might be functionally interlinked during CVB3 infection.
An important consideration for this prediction was the high-throughput data upon which the model was founded (Janes and Yaffe, 2006). The commercial p-ELISA used to assemble the p-ERK dataset is marketed as specific for ERK1/2. However, ERK1 and ERK2 share ~50% identity with ERK5, a third MAPK whose regulation is distinct (Nishimoto and Nishida, 2006). All three ERKs have a Thr–Glu–Tyr motif that is bis-phosphorylated upon activation, and the sequence surrounding this motif is so similar that many p-ERK1/2 antibodies will cross-react with ERK5 (K.J.J. and K.A.J., unpublished observations). p-ERK5 cross-reactivity is readily distinguished from p-ERK1/2 during immunoblotting (ERK5 ~ 80–100 kDa vs. ERK1 ~ 44 kDa, ERK2 ~ 42 kDa), but the ELISA format cannot resolve proteins by molecular weight. Because ERK5 signaling is important for cardiovascular tissues (Regan et al., 2002), we decided to investigate the individual contributions of ERK1/2 and ERK5 by independent methods.
We first monitored the kinetics of ERK1/2 and ERK5 phosphorylation by blotting with antibodies that were specific for each pathway (Figure 2B). Both ERK1/2 and ERK5 were strongly phosphorylated shortly after CVB3 infection at 0.17 h p.i. and also after host cytotoxicity was evident at 24 h p.i. However, ERK5 showed a more-sustained phosphorylation up to 1 h p.i. and p-ERK1/2 exhibited a second peak at 8 h p.i., illustrating differences in their regulatory kinetics. The multiphase activation of ERK1/2 was further confirmed by measuring phosphorylation of RSK, a specific ERK1/2 substrate (Figure S2A) (Sturgill et al., 1988). Next, we used a pair of MEK inhibitors (PD184352 [PD] and U0126 [U0]) to separate the ERK1/2 and ERK5 contributions to the ERK p-ELISA. PD at low concentrations selectively blocks MEK1/2 and ERK1/2 phosphorylation, whereas U0 inhibits MEK1/2-ERK1/2 and MEK5-ERK5 equally (Davies et al., 2000) (Figures S2B and S2C). Thus, the contribution of ERK5 can be inferred from the difference between PD (ERK1/2 inhibition) and U0 (ERK1/2 + ERK5 inhibition). When cells were preincubated with U0 and treated with CVB3 for 10 min, we found that the measured p-ERK ELISA signal was reduced to background levels (Figure 2C). By contrast, pretreatment with PD reduced the ELISA signal by only ~30%, despite that ERK1/2 phosphorylation was completely inhibited (Figures 2C and S2C). This indicated that the p-ERK ELISA data was a convolution of ERK1/2 and ERK5 pathway activities and further implied that the predicted ERK–p38 associations (Figure 2A) could be between ERK1/2 and p38 or ERK5 and p38, or both.
We tested for crosstalk between p38 and ERKs by using SB203580 (SB), an ATP-competitive small-molecule inhibitor of p38 (Lee et al., 1994). We monitored p-ERK1/2, p-ERK5, and p-p38, as well as the major ERK1/2 and p38 effector kinases, RSK and MK2 (Rouse et al., 1994; Sturgill et al., 1988). We found that SB potently inhibited p38 activity in cardiomyocytes, as expected, blocking phosphorylation of MK2 at early and late times after CVB3 infection (Figures 2D and 2E). Acute SB treatment was also specific, because we did not observe any effect on early CVB3-induced ERK1/2 phosphorylation or activity (Figure 2D). Upon prolonged SB treatment, however, we observed a modest-but-reproducible increase in ERK1/2 phosphorylation (Figures 2E and 2F). We attribute this to secondary inhibition of PP1 and PP2A phosphatases, which are normally activated by p38 signaling and serve to dephosphorylate MEK1/2 upstream of ERK1/2 (Westermarck et al., 2001). Subsequent control experiments showed that SB-induced upregulation of ERK1/2 was independent of CVB3 treatment (Figure 2G). Thus, p38 signaling antagonizes late ERK1/2 signaling, prompting a reevaluation of earlier p38-inhibition experiments involving CVB3 (see below) (Si et al., 2005).
A second finding from these experiments was that SB treatment potently blocked both early- and late-phase phosphorylation of ERK5 (Figures 2H–J). The p38–ERK5 coupling was consistent with predictions of the model (Figure 2A), but such crosstalk had not previously been reported. To exclude the possibility that ERK5 inhibition was caused by off-target effects of SB, we repeated the experiments with BIRB796 (BIRB), an allosteric inhibitor of p38 (Pargellis et al., 2002) (Figure S2D). SB and BIRB have completely different mechanisms of inhibition and off-target signatures; therefore, a common phenotype with SB and BIRB strongly implicates p38 signaling (Bain et al., 2007). When cardiomyocytes were pretreated with BIRB, we observed the same blockade of CVB3-induced ERK5 phosphorylation as with SB (Figures S2E and S2F). Last, to examine the generality of the p38–ERK5 connection, we treated human embryonic kidney cells with sorbitol as an osmotic stress that activates both p38 and ERK5 (Abe et al., 1996; Raingeaud et al., 1995). SB and BIRB each blocked hyperosmolarity-induced ERK5 phosphorylation (Figure S2G), suggesting that p38 is generally required for proper activation of the MEK5–ERK5 pathway. Taken together, the molecular consequences of SB and BIRB indicate that p38 is functionally interconnected with both ERK1/2 and ERK5, as predicted by the model of CVB3 pathogenesis (Figure 2A).
An important category of host-cell responses in the starting dataset was the activity of apoptotic caspases (Figure 1B). ERKs and p38 mapped closely to these readouts in the model and could conceivably control CVB3-induced apoptosis directly (Figure 2A). Both ERK1/2 and p38 have been reported to be important for proper caspase activation (Luo et al., 2002; Si et al., 2005). However, these earlier studies used a dual MEK1/2–MEK5 inhibitor (U0) and were not aware of the antagonism between p38 and ERK1/2 (Figures 2E–G and S2C). We thus pursued follow-up studies using gain- and loss-of-function approaches for ERK–p38 together with direct measurements of caspase processing.
We began with ERK5, as it inhibits cardiac apoptosis in other contexts (Kimura et al., 2010; Yan et al., 2007) but had not been previously implicated in CVB3 infection. To block ERK5 signaling, we used the specific ATP-competitive ERK5 inhibitor, XMD8-92 (XMD) (Yang et al., 2010). XMD treatment potently reduced phosphorylation of an ERK5 substrate (MEF2A) in cardiomyocytes and significantly increased caspase-9 and caspase-3 cleavage upon CVB3 infection (p < 0.05) (Figures 3A and 3B). CVB3-induced apoptosis also increased when endogenous ERK5 was downregulated with shRNA (Figures 3C and 3D). We performed a reciprocal gain-of-function experiment by establishing stable lines expressing a doxycycline (DOX)-inducible mutant of MEK5 that was constitutively active (MEK5-DD) (Figure 3E). Upon low-level infection with CVB3 (M.O.I. = 1.5), we found that DOX treatment of MEK5-DD-expressing cells caused a significant decrease in caspase-3 cleavage (p < 0.05) (Figure 3F). Interestingly, the drop in caspase-3 cleavage was associated with changes in caspase-8 activation rather than caspase-9 activation as with XMD. We attribute this difference to the kinetics of ERK5 activation with MEK5-DD (~8 hours) versus ERK5 inhibition with XMD (< 1 hour). The MEK5-DD, XMD, and shRNA results together indicate that CVB3-induced ERK5 signaling inhibits cardiomyocyte apoptosis. This link between ERK5 and host-cell survival is unique, because virtually all other CVB3-stimulated pathways described thus far promote apoptosis rather than inhibit it (Table S1).
Next, we examined the p38 pathway by using a similar set of approaches. Consistent with an earlier report (Si et al., 2005), we found that p38 inhibition via SB profoundly reduced caspase-3 cleavage during CVB3 infection (Figure 4A). We reinforced the SB result by showing that p38 inhibition with BIRB phenocopied SB in its blockade of initiator and effector caspases (Figures S4A and S4B). The intuitive conclusion from these experiments is that p38 promotes CVB3-induced apoptosis. However, when we attempted the reciprocal gain-of-function experiment with a DOX-inducible, constitutively active mutant of MKK6 (MKK6-EE), there was no detectable change in caspase activation (Figures 4B and 4C). The apparent contradiction prompted us to re-evaluate our experiments considering the cross-communication between p38 and ERKs (Figures 2E–J).
We reasoned that secondary inhibition of ERK5 would partially offset the observed SB–BIRB phenotype rather than cause it (Figures 2H–J and and3).3). Therefore, our attention turned to ERK1/2, which becomes hyperactivated upon prolonged p38 inhibition (Figures 2F and 2G). This negative regulation of ERK1/2 by p38 was further strengthened by the reduced p-ERK1/2 observed in DOX-treated MKK6-EE cells (Figure 4D). To determine whether the consequences of p38 inhibition were mediated through ERK1/2, we combined SB with PD to block ERK1/2 hyperactivation and found that CVB3-induced apoptosis occurred normally (Figures 4E and S4C). Remarkably, ERK1/2 inhibition by itself did not substantially affect apoptosis of CVB3-infected cells (Figures 4F and S4C), suggesting that ERK1/2 acted as a pro-survival signal only when p38 function was blocked. The p38-specific role of ERK1/2 was re-emphasized in ERK5-inhibited cells, where PD+XMD increased apoptosis as with XMD alone (Figures 3B, S3A, and S3B). Upon this re-evaluation of earlier studies using SB (Si et al., 2005), we conclude that p38 signaling does not directly control CVB3-induced apoptosis.
Apoptosis is but one facet of the host-cell response to CVB3 infection, raising the question of whether other aspects of pathogenesis could require p38 signaling (Carthy et al., 1998; Yuan et al., 2003). In the original dataset, overall CVB3 cytotoxicity was measured via tetrazolium reduction (see Experimental Procedures). However, this method was inadequate to read out cytotoxicity in the presence of signaling perturbations, which could also affect proliferation and metabolism. We therefore switched to a fluorescent amine-reactive dye that intensely labels cells with compromised plasma-membrane integrity irrespective of the mechanism of cell death (Perfetto et al., 2006).
We found that CVB3 infection caused a dramatic increase in the percentage of dye-labeled, non-viable cells as compared to sham infection (Figures 5A–D). The actual extent of cytotoxicity was much greater than the flow-cytometry estimate (compare Figures 5B and 5D), because many infected cells were so damaged that they were unavoidably lost during the suspension preparation. As with the earlier apoptosis experiments (Figure 4), we found that SB pretreatment strongly decreased the extent of CVB3-induced cytotoxicity, whereas PD did not have a significant impact (p > 0.05) (Figures 5B–F). Surprisingly, when SB and PD were combined, we observed a clear improvement in overall cell viability even though caspase activation was unaffected under these conditions (Figures 4E, 5B, and 5G). The pronounced result of dual p38–ERK1/2 inhibition was also reflected in significantly reduced titers of released viral progeny (p < 0.001) (Figures 6A and 6B). This raised the possibility that p38 could control alternative death pathways that were distinct from apoptosis but critically important for CVB3 pathogenesis.
We closely examined the morphology of CVB3-infected cells by microscopy and noted a mixture of phenotypes indicative of discrete single-cell outcomes (Figure 5D). Some cells had a rounded appearance with condensed nuclei, suggesting an apoptotic fate. Others, however, remained fully spread and had aberrant lamellipodia-like projections (Figure 5D, right). These cells also had an intact nucleus along with intracellular vesicles that remained dye impermeant. Our observations suggested that a fraction of CVB3-infected cells undergo a vesiculated form of cell death with certain hallmarks of necrosis (Yuan et al., 2003).
To determine whether CVB3 infection was associated with biochemical readouts of necrosis, we examined the chromatin protein HMGB1, which is released extracellularly by necrotic cells (Scaffidi et al., 2002). We validated the marker by stimulating cardiomyocyte necrosis with hydrogen peroxide and observing pronounced HMGB1 release (Figure S5A). Importantly, we found that HMGB1 was clearly detected in supernatants from CVB3-infected cells (Figure 6C). HMGB1 release was unaffected by the apoptosis inhibitor DEVD-CHO but was slightly reduced by the necrosis inhibitor Necrostatin-1, likely as a result of CVB3-induced autocrine TNF signaling (Degterev et al., 2005; Garmaroudi et al., 2010). Thus, HMGB1 is a reliable marker of necrosis stimulated by CVB3.
Upon p38 inhibition with SB or BIRB, we observed near-complete blockade of HMGB1 release, suggesting potent inhibition of necrosis (Figures 6D and S5B). Conversely, necrosis was negligibly affected in CVB3-infected cells treated with PD to inhibit ERK1/2, consistent with the earlier labeling results (Figures 5B, 5F, and and6E).6E). In stark contrast to the apoptotic readouts (Figure 4E), we did not observe any reversion of necrosis when CVB3-infected cells were pretreated with PD + SB (Figure 6F). Last, to test whether p38 signaling was sufficient to drive virus-induced necrosis, we returned to the inducible MKK6-EE cells and found that DOX treatment substantially augmented HMGB1 release during CVB3 infection (Figure 6G). We conclude that p38 signaling is a critical component of a necrosis pathway, which promotes CVB3 propagation independently of ERK-dependent apoptosis.
Viruses such as CVB3 activate many host-cell signaling pathways and evoke many host-cell responses. Our study here began with a holistic approach to monitor these events dynamically and as a function of CVB3 titer. By analyzing the data to make quantitative predictions of host-cell outcome, we quickly converged on ERKs and p38 as key pathways for CVB3 pathogenesis. Early-phase ERK1/2 activation stems directly from CVB3 docking to host membranes, whereas late-phase activation occurs due to cleavage of upstream signaling molecules by viral proteases (Huber et al., 1999; Luo et al., 2002). Late-phase p38 and ERK5 signaling probably lies downstream of autocrine proinflammatory cytokines, which are induced during the final stages of the viral life cycle (Figure S6) (Garmaroudi et al., 2010). Despite differences in activation, our work here shows that ERKs and p38 are strongly interconnected (Figure 7). These dependencies are important for interpreting the results of “single-pathway” perturbations that propagate through the network (Luo et al., 2002; Si et al., 2005).
Notably, we were able to uncover a role for ERK5 in CVB3 pathogenesis by modeling a dataset that did not measure ERK5 explicitly. We have shown elsewhere that quantitatively accurate signaling measurements are critical for data-driven models to reflect underlying biological mechanisms (Janes et al., 2008; Janes and Yaffe, 2006). Our results here using a pan-ERK p-ELISA indicate that measurements of specific proteins may not be as important. This is encouraging, because many modern signaling assays increase overall throughput by relaxing the specificity constraints of traditional approaches (Albeck et al., 2006).
Similarly, our work shows that agglomerated cell-outcome data may be sufficient for viral-host modeling and discovering overlooked phenotypes. The importance of CVB3-induced necrosis as a host-cell fate was revealed here without direct necrotic readouts in the model (Figure 7). This information was presumably embedded in the overall cytotoxicity measure, which depends strongly on the level of necrosis (Figures 5A–D). Interestingly, the associated RVP titers appear to be influenced by apoptosis and necrosis reciprocally. When both apoptosis and necrosis are blocked upon p38 inhibition with SB, there is a slight reduction in RVP. However, when apoptosis is restored in p38-inhibited cells by blocking ERK1/2 hyperactivation, RVP is dramatically reduced (Figure 6B). Thus, necrosis may be the preferred outcome for CVB3, which is counteracted by the host-cell drive to die by apoptosis. To isolate necrosis specifically requires targeting an upstream mediator (p38) and resetting the other secondary consequences of pathway inhibition (e.g., ERK1/2) (Figure 7). Such combinatorial anti-viral strategies would be difficult to predict without the aid of a systems model for the host-cell response to CVB3 infection.
DOX-inducible MEK5 (Boehm et al., 2007) and MKK6-EE (Raingeaud et al., 1996) were cloned by PCR into the Tet-tight entry vector pEN_TTmiRc2 (Shin et al., 2006). To generate MEK5-DD, S311 and T315 of MEK5 were both mutated to glutamate by site-directed mutagenesis (Quikchange II XL, Stratagene). All entry vectors were verified by sequencing, and lentiviral vectors were cloned by LR recombination into pSLIK neo (Shin et al., 2006). pLKO.1 puro shERK5 lentiviral vectors (TRCN0000023234 and TRCN0000023236) were obtained from Open Biosystems.
HL1 cells were provided by Dr. William Claycomb (Louisiana State University Health Sciences Center, New Orleans, USA) (Claycomb et al., 1998). 293T cells were obtained from ATCC. CVB3 (Kandolf strain) was propagated in HeLa cells, and virus titers were determined by plaque assay. Retroviruses and lentiviruses were packaged as previously described (Wang et al., 2011). Stably transduced HL1 cells were selected with 4 μg/ml puromycin or 150 μg/ml G418 until control plates had cleared.
HL1 cells were sham-infected with PBS or infected with CVB3 at M.O.I. = 0.5, 1.5, 4.5, 9, or 18 and cell extracts were prepared at 0, 0.17, 1, 8, 16, and 24 hr. For perturbation experiments, the following chemical inhibitors were added one hour before infection: SB203580 (20 μM, Tocris Biosciences), BIRB796 (5 μM, Selleck Chemicals), XMD8-92 (5 μM, Axon Medchem), U0126 (20 μM, Tocris Biosciences), PD184352 (2 μM, Santa Cruz Biotechnology), DEVD-CHO (0.1 μM, EMD), and Necrostatin-1 (50 μM, Calbiochem).
CVB3 titers from triplicate cell supernatants were determined on monolayers of HeLa cells by an agar overlay plaque assay as described elsewhere (Garmaroudi et al., 2010).
Cell lysates were normalized to protein concentration and analyzed by p-ELISA (Biosource) for the phosphorylation levels of Akt (S473), ATF2 (T69/T71), CREB (S133), ERK1/2 (T185/Y187), GSK3β (S9), Hsp27 (S82), IκBα (S32), and p38 MAPK (T180/Y182) according to the manufacturer's instruction.
Caspase activities were measured according to the manufacturer's instruction (R&D Systems) as described elsewhere (Si et al., 2005). Fluorescence was measured at excitation and emission wavelengths of 485 nm and 535 nm, respectively, using a Tecan GENios fluorescent reader.
Immunoblotting was performed as described previously (Garmaroudi et al., 2010) with one of the following primary antibodies: anti-p-ERK1/2 (T202/Y204, Cell Signaling, 1:1000), anti-p-ERK5 (T218/Y220, Cell Signaling, 1:1000), anti-p-p38 (T180/T182, Cell Signaling, 1:1000), anti-p-MAPKAPK2 (T334, Cell Signaling, 1:1000), anti-VP1 (Dako, 1:1000), anti-cleaved caspase-8 (Cell Signaling, 1:1000), anti-caspase-9 (Cell Signaling, 1:1000), anti-caspase-3 (Cell Signaling, 1:1000), anti-β-actin (Sigma, 1:5000), anti-p-Akt (S473, Cell Signaling, 1:1000), anti-p-GSK3β (S9, Cell Signaling, 1:1000), anti-p-ATF2 (T69/T71, Cell Signaling, 1:1000), anti-p-CREB (S133, Cell Signaling, 1:1000), anti-p-IκBα (S32, Cell Signaling, 1:1000), anti-p-Hsp27 (S82, Cell Signaling, 1:1000), anti-p-RSK (T359/S363, 1:1000), anti-MEK5 (StressGen, 1:1000), anti-HA (Roche, 1:1000), anti-HMGB-1 (Epitomics, 1:1000), anti-p-MEF2A (T312, Abcam, 1:1000), anti-cleaved caspase-8 (Cell Signaling, 1:1000), or anti-tubulin (Cell Signaling, 1:5000 or Abcam, 1:20000) for 1 hr or overnight, followed by incubation for 1 hr with horseradish peroxidase-conjugated secondary antibodies (Santa Cruz) or infrared dye-conjugated secondary antibodies (Licor). Immunoreactive bands were visualized by enhanced chemiluminescence (Pierce, Rockford, IL) on ChemiGenius2 or ChemiDoc MP camera-based detection systems or by infrared fluorescence on an Odyssey infrared imaging system. Where indicated, band intensities were quantified by densitometry with ImageJ, and all blotting results were replicated with at least one additional set of independent biological samples.
HL1 cells were grown in 12-well plates and infected with CVB3 (M.O.I. = 9) for 16 and 24 hr after pretreatment with inhibitors. The MTS solutions (1:5) were added to wells for 2.5 hr and then transferred to 96-well plates. Cell viabilities of infected cells and non-infected were assessed by MTS assay (CellTiter 96; Promega, Inc., Madison, WI). Amine-reactive labeling was performed with the LIVE/DEAD fixable violet dead stain (Invitrogen) according to the manufacturer's recommendations. For flow cytometry, cells were labeled in suspension, washed with PBS + 0.1% Tween-20, and analyzed on a BD FACSCalibur equipped with 407 nm violet laser excitation. For microscopy, adherent cells were labeled, washed with PBS, permeabilized with 0.3% Triton X-100 in PBS, and counterstained with DRAQ-5 before imaging by widefield microscopy as described previously (Wang et al., 2011).
Phospho-proteins (predictor variables) and readouts (response variables) were standardized as z-scores, and the phospho-protein time course was time-integrated over early (0–8 hr) and late (8–24 hr) phases. Partial least squares regression was performed with the “plsregress” function in MATLAB by standard approaches (Janes et al., 2005; Janes and Yaffe, 2006). The stability of the model was assessed by fivefold leave-one-out cross-validation.
All hypothesis testing was performed with Welch's one- or two-sided t test at a significance level of α = 0.05.
We thank Benjamin Kuhn for technical assistance with immunoblotting and image analysis. This work was supported by the National Institutes of Health Director's New Innovator Award Program (1-DP2-OD006464 to K.A.J.), the Pew Scholars Program in the Biomedical Sciences (to K.A.J.), and the David and Lucile Packard Foundation (to K.A.J.), the Heart and Stroke Foundation of British Columbia & Yukon (to B.M.M.), and the Canadian Institutes of Health Research (CIHR) (to B.M.M.). F.S.G. is supported by a Doctoral Award from Tehran University of Medical Science-Iran. K.J.J. is partly supported by a predoctoral award from the ARCS Foundation.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.