Abstract
Coronaviruses (CoV), such as severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), are of medical importance with high mortality rates and significant zoonotic and pandemic potential. Here, we apply ribosome profiling and parallel RNASeq to globally analyse changes in the host cell translatome and transcriptome upon infection with mouse hepatitis virus, strain A59 (MHV-A59), a model murine coronavirus in the same genus as SARS-CoV and MERS-CoV. We observed translational upregulation of ATF4, ATF5 and Ddit3 and activation of the unfolded protein response (UPR). Phosphorylation of eIF2α led to the global inhibition of translation and a substantial increase in empty 80S ribosomes. A drug that inhibits the UPR attenuates virus growth suggesting that MHV may have evolved to subvert the UPR to its own advantage. We also investigated an artefact of cycloheximide pretreatment in ribosome profiling whereby ribosomes accumulate at the 5’ end of coding sequences in stressed cells but not in unstressed or untreated cells, thus extending earlier studies in yeast to mammalian cells. The study sheds light on the mechanisms of CoV translational shutoff and reveals a potential new therapeutic strategy.
Importance
Ribosome profiling is an emerging technique that reveals a global snap shot of protein synthesis by precisely mapping the positions of translating ribosomes. It has particular applicability to virology where it allows not only monitoring of viral gene expression, but also host response at the level of individual genes and globally. In this work, we use ribosome profiling and parallel transcriptome sequencing (RNASeq) to monitor changes in transcription and translation in cells infected with the model coronavirus, murine coronavirus strain MHV-A59, a virus in the same genus as the medically important SARS-CoV and MERS-CoV. The sensitivity and precision of the approach permit us to obtain a high-resolution analysis of different biological processes – such as the activation of the unfolded protein response. At late time points of infection, translation is inhibited globally and, surprisingly, viral transcripts are not resistant to this effect. Nonetheless, relief of translational inhibition with a drug that interferes with the unfolded protein response is detrimental to virus replication. This study sheds new light on the complex interactions between virus and host during infection and provides new potential targets for antiviral intervention.
Introduction
The Coronaviridae are a family of enveloped viruses with positive-sense, monopartite, single-stranded RNA genomes. At 27–32 kb, coronaviruses (CoVs) have the largest known RNA genomes. CoVs cause a broad range of diseases in animals and humans, ranging from the common cold to severe acute respiratory syndrome (SARS) [1]. Amongst CoVs of medical importance with high mortality rates and pandemic potential are SARS-CoV and MERS-CoV, both members of the genus Betacoronavirus. Murine coronavirus, a betacoronavirus more commonly refered to as mouse hepatitis virus (MHV), has been used as a model to study the replication and biology of other members of the genus.
Virus infection alters cellular gene expression to facilitate replication of the viral genome and the assembly of virus particles. As with all viruses, CoVs rely on the host cell translational machinery for viral protein synthesis. Many viruses have evolved mechanisms to shut-off host mRNA translation, which can increase the availability of the translational machinery for non-canonical modes of viral protein synthesis, and at the same time inhibit host antiviral responses [2]. Exactly how CoVs induce host translational shut-off and its significance in relation to the synthesis of virus proteins, particularly at later times of infection, is still poorly understood. During CoV replication, the massive production and modification of viral proteins, as well as virion budding-related endoplasmic reticulum (ER) membrane depletion, can lead to overloading of the folding capacity of the ER and, consequently, ER stress [3]. This activates the unfolded protein response (UPR) which returns the cell to homeostasis and mitigates the major risk that protein misfolding poses for correct cellular function [4]. In mammalian cells, the UPR is controlled by three ER-resident transmembrane sensors: the inositol-requiring enzyme-1 (IRE1), the PKR-like ER kinase (PERK), and the activating transcription factor-6 (ATF6). These sensors recognize unfolded/misfolded proteins inside the ER and transmit a signal to the nucleus to transcribe specific genes whose products act to lower protein synthesis and increase ER folding capacity [4]. Previous studies (reviewed in [5]) have aimed to establish how the different UPR pathways are involved during CoV infection. Ribosome profiling (RiboSeq) allows global monitoring of cellular translation by mapping the positions of translating ribosomes on the transcriptome [6–8]. RiboSeq reveals the location and abundance of ribosomes on specific mRNA species with single-nucleotide precision. In conjunction with RNASeq, to determine the corresponding transcriptome, RiboSeq has been used to elucidate changes in translation, transcription and translation efficiency in viral and host gene expression during the course of infection [9–19].
Here, we use RiboSeq and parallel RNASeq to globally analyse changes in the host translatome and transcriptome throughout a time course of CoV infection. We observe activation of different pathways of the UPR leading to eIF2α phosphorylation and translational shut-off at the level of initiation which we confirm by polysome profiling. Surprisingly, however, a pharmacological inhibitor of the UPR was found to attenuate virus replication. We also investigate a remarkable perturbation in ribosome footprint distributions on coding sequences previously observed at late time points of infection and determine that it is a methodological artefact arising from the use of the elongation inhibitor cycloheximide. This detailed analysis of cellular translation during MHV infection provides new insights into the mechanism of CoV translational shutoff and the complex interactions between virus and host during infection, and may aid the identification of new targets for antiviral intervention.
Results
Effects of MHV-A59 infection on cellular gene expression
To survey genome-wide changes in host translation and transcription during CoV-infection, we interrogated sequencing data generated previously in this laboratory [16]. In these experiments, we generated RNASeq and RiboSeq data from two independent biological repeats of murine 17 clone 1 cells (17 Cl-1) infected with recombinant MHV-A59 at a multiplicity of infection (MOI) of 10, and harvested at 1, 2.5, 5 and 8 h post-infection (p.i.), with mock-infected cells harvested at 1 and 8 h.
To assess the effects of MHV infection on cellular transcript abundance, differential expression analyses were performed with DESeq [20], EdgeR [21], NOISeq [22] and BaySeq ([23]). The different analyses had a large number of overlapping predictions (Fig 1A). The volcano plot (Fig 1B) derived from BaySeq analysis of four uninfected (1 and 8 h) and two infected (5 h p.i.) RNASeq libraries revealed that some of the most differentially expressed cellular transcripts (red points) are related to the host translational apparatus (Eef1a1 – eukaryotic elongation factor 1A-1; Eif3f – eukaryotic initiation factor 3 subunit F; Eif1 – eukaryotic initiation factor 1; Eif2s3x – eukaryotic initiation factor 2 subunit 3 gamma, and Eif2b3 – eukaryotic initiation factor 2B GDP-GTP exchange factor subunit gamma), the UPR (Herpud1 – homocysteine inducible ER protein with ubiquitin like domain 1 and Chac1 – glutathione-specific gamma-glutamylcyclotransferase 1), and the gene coding for the largest subunit of RNA polymerase II, Polr2a.
To further validate changes in the transcript abundance of these genes, total RNA was extracted from three biological replicates of MHV-infected and mock-infected cells at 5 h p.i. and the levels of selected up-regulated (Fig 1C, left panel) and down-regulated (Fig 1C, middle panel) transcripts assessed by quantitative real-time PCR (qRT-PCR), normalized by a ‘housekeeping gene’, ribosomal protein L19 (RPL19), which has been reported to be unaffected by ER stress [24,25]. qRT-PCR was also used to assess the abundance of viral nucleocapsid (N) transcripts (Fig 1C, right panel). Up-regulated transcripts had qRT-PCR values broadly consistent with the RNASeq measurements (Fig 1B and 1C) whereas down-regulated transcripts did not, with the exception of Eef1a. However, the Rpl19 transcript was, in fact, moderately down-regulated at 5 h p.i. (relative to the global transcriptome mean) (Fig 1B, yellow), which may account for these discrepancies. A comparison of differential expression between the 1 and 8 h mocks showed much smaller differences, with only four transcripts having changes with a false discovery rate (FDR) ≦ 0.05 (S1A Fig).
CoVs induce host translational shut-off [26–31] though the mechanisms have not been well-defined. We reasoned that some host genes may be resistant to virus-induced shut-off and that identifying such genes might give new insights into the shut-off mechanism(s). To evaluate differences at the level of translation as a result of MHV infection, we calculated relative translation efficiencies (TE) – defined herein as the ratio of ribosome-protected-fragment (RPF) and total RNA density in the coding region (CDS) of a given gene – at 5 and 8 h p.i. TE was estimated using Bayseq in the paired library mode, Babel [32] and Xtail [33]. These analyses had few overlapping predictions of genes significantly differentially expressed, with Xtail producing a much higher number of predictions (Fig 2). These differences are likely due to variations in model assumptions between the different approaches and reflect the greater difficulty inherent in modeling a quotient distribution (i.e. RiboSeq/RNASeq). To proceed, we focused on genes that showed more than a 2-fold difference in mean TE between infected and mock samples, and a FDR ≤ 0.05 (upper right and upper left quadrants; Fig 2B – Babel 5 h p.i.; Fig 2C – Xtail 5 h p.i. and Xtail 8 h p.i.). Translational differences between 1 and 8 h mock-infected cells are shown in S1B Fig.
Among the mRNAs found to be preferentially translated during MHV-infection were those whose translation is known to be induced by eIF2α phosphorylation, including those for transcription factors ATF4 (activating transcription factor 4), ATF5 (activating transcription factor 5) and CHOP (DDIT3/GADD153) (Fig 2B and 2C; S2A Fig). The mRNAs encoding these proteins contain (multiple) inhibitory upstream open reading frames (uORFs) in their 5’ leaders and undergo selective re-initiation of the main ORF under conditions of eIF2α phosphorylation [34–39]. Another gene previously shown to be resistant to eIF2α phosphorylation-induced translational attenuation is Slc35a4 (Solute Carrier Family 35, Member A4; Fig 2B; [39]). In our study, the vast majority of RPFs derived from this mRNA mapped to the 102-codon uORF (S2B Fig). These results are consistent with eIF2α phosphorylation (leading to inhibited initiation) being a major cause of host translational shut-off during MHV infection.
An artefact of cycloheximide treatment explains a remarkable perturbation in the distribution of elongating ribosomes
Previous meta-analyses of RPFs mapping to host mRNAs in MHV-infected cells indicated a remarkable perturbation in host cell translation at late time points [16] with a greatly increased proportion of reads mapping to the first ~30–40 codons of CDSs in MHV-infected cells at 5 and 8 h p.i. (e.g. Fig 3A compares 8 h p.i. with 8 h mock; data from [16]). Other ribosome profiling studies have revealed a similar accumulation of ribosomes in the 5’ portion of CDSs after heat shock, proteotoxic and oxidative stress, and attributed this to an early post-initiation inhibition of elongation under cell stress [40–42]. More recently, these results have been called into question by the discovery that cycloheximide pre-treatment in yeast leads to an accumulation of ribosomes in the 5’ portion of CDSs specifically in stressed cells [43,44]. Moreover, this accumulation was not seen if the yeast cells were flash frozen without cycloheximide pre-treatment, or if very high concentrations of cycloheximide were used.
In order to determine whether this experimental caveat might extend to stressed mammalian cells and perhaps explain the observed accumulation of ribosomes downstream of initiation sites during virus infection [16], 17 Cl-1 cells were infected with MHV A-59 at MOI of 10 and at 8 h p.i. harvested either by flash freezing in the absence of cycloheximide, or following supplementation of the growth medium with 100 μg/ml cycloheximide (CHX 1X) or 10 mg/ml cycloheximide (CHX 100X) for minutes prior to harvesting. A similar experiment was performed with 17 Cl-1 cells incubated with a pharmacological inducer of ER stress, tunicamycin, for 6 h which activates all UPR signalling pathways. Ribosomal occupancy profiles and the effects of stress and drug treatment are shown in Fig 3B (for MHV-infected cells) and Fig 3C (for tunicamycin-treated cells).
A modest increase in ribosome occupancy downstream of initiation sites occured in tunicamycin-stressed cells under normal cycloheximide treatment (compare “tunicamycin CHX 1X” with “mock CHX 1X” plots; Fig 3C). Similar to findings in stressed yeast cells [43], this could be alleviated by high concentrations of cycloheximide or flash freezing without cycloheximide pre-treatment (Fig 3C, CHX 100X and “flash frozen” plots). In virus-infected cells, however, a much more pronounced accumulation of ribosomes was observed downstream of initiation sites under cycloheximide treatment, and this was not alleviated even when cycloheximide was used at 100-fold the normal concentration (Fig 3B, CHX 1X and 100X plots). However, this ribosome peak completely disappeared when cells were flash frozen without cycloheximide pre-treatment (Fig 3B, “flash frozen” plots). Thus we conclude that the cycloheximide-induced accumulation of ribosomes in the 5’ region of CDSs also occurs in stressed mammalian cells and, further, that this effect is not necessarily negated simply by increasing the cycloheximide concentration.
As in previous studies [41,42], we quantified excess 5’ ribosome density for individual transcripts by means of a 5’ loading ratio statistic where we compared the density of RPFs in the 5’ approximately 30 codons of each CDS to the density of RPFs in the 3’ remainder of the CDS. As expected, when harvested at 8 h in the presence of cycloheximide the 5’ loading ratios of cytoplasmic mRNAs showed a substantial increase in MHV-infected cells compared to mock-infected cells (Fig 4, left panel), whereas there was no such effect when cells were harvested with flash freezing (Fig 4, right panel). Interestingly, but not unexpectedly, this effect was not seen for mitochondrial mRNAs, which had similar 5’ loading ratio statistics over all conditions (Fig 4, red crosses).
In view of the cycloheximide effect, data from the flash frozen 8 hpi mock and MHV-infected samples were used for the individual gene plots shown in the following sections.
MHV infection and activation of the unfolded protein response
Although several studies [29, 45–47] have aimed to establish how each of the three UPR sensor-pathways may be involved during CoV infection, we wanted to take advantage of the high-resolution that ribosome profiling can provide to analyse in detail each specific arm of the UPR response during MHV infection.
Monitoring ATF6 and IRE1α activity
Upon induction of ER stress, ATF6 translocates from the ER to the Golgi apparatus where it is cleaved by the proteases Site-1 (S1P) and Site-2 (S2P) [48]. After cleavage, the amino-terminus of ATF6, containing a basic leucine zipper (bZIP) transactivating domain, translocates to the nucleus to upregulate the ER chaperone immunoglobulin heavy chain binding protein (BiP/Grp78), whose mRNA or protein levels serve as a proxy for activation of the ATF6 pathway [29,49] 17 Cl-1 cells were infected with MHV-A59 or incubated with tunicamycin. Cells were harvested at 2.5, 5, 8 and 10 h p.i. and analyzed by qRT-PCR (Fig 5A). An induction of BiP/Grp78 transcription was observed in tunicamycin-treated (purple) and MHV-infected cells (orange) from 2.5 to 8 h p.i. followed by a modest decline, whereas mock-infected cells (blue) showed no induction. Surprisingly, whereas Western blot analysis (Fig 5B) confirmed induction of Bip/Grp78 protein in tunicamycin-treated cells by 8 h p.i., no such induction was seen in MHV-infected cells.
At 8 h p.i., MHV-infected cells were harvested by flash freezing in the absence of cycloheximide. As a positive control for UPR activation, cells treated with tunicamycin for 6 h were also harvested by flash freezing. RNASeq and RiboSeq read counts of BiP/Grp78 (Mock 8 h and MHV 8 h p.i.; Fig 5C) revealed an increase in RNASeq reads in MHV-infected cells (Mock RNA and Inf RNA panels) consistent with the qRT-PCR results. Although an expected increase in RPFs was seen (Inf Ribo panel), ribosome density was not as high in MHV-infected cells as in tunicamycin treated cells (Fig 5C, Tunica Ribo panel). A lower ribosome density in the BiP/Grp78 ORF for MHV-infected cells likely explains why this protein was not immunodetected at later times of infection.
ER stress induction activates endonuclease IRE1α which cleaves X-box binding protein-1 (Xbp-1) mRNA [50,51]. Activated IRE1α removes a 26-nt intron from unspliced Xbp-1 (Xbp-1u) mRNA leading to a translational reading frame shift and a longer protein. The product of spliced Xbp-1 mRNA (XBP-1s) is an active transcription factor that upregulates the expression of ER-associated degradation (ERAD) components and ER chaperones. Determination of Xbp-1 splicing was done by reverse transcriptase PCR (RT-PCR) of total RNA extracted from 17 Cl-1 cells infected with MHV-A59 or incubated with tunicamycin, using specific primers flanking the Xbp-1 splice site (Fig 5D). At all timepoints, Xbp-1u was the predominant form in mock-infected cells whereas Xbp-1s was the major species in tunicamycin-treated cells. In virus-infected cells, Xbp-1u was predominant at 2.5 h p.i. but Xbp-1s became predominant at 5 h p.i. In order to analyse translation of Xbp-1u and Xbp-1s in virus-infected cells, we inspected the ribosome profiling data (Fig 5E). For 8 h p.i. MHV-infected cells (Inf Ribo panel) and tunicamycin-treated cells (Tunica Ribo), an increased number of reads mapped in the +2 reading frame (yellow peaks) of the Xbp-1u sequence, and downstream of the annotated stop codon. These reads result from translation of the Xbp-1s frameshifted isoform and indicate a dramatic increase in production of the active transcription factor.
Monitoring PERK-eIF2α -ATF4 activity
In response to ER stress, PERK oligomerizes and auto-phosphorylates [52]. Activated PERK phosphorylates the α-subunit of eukaryotic initiation factor 2 (eIF2α) which in turn impairs recycling of inactive eIF2-GDP to active eIF2-GTP resulting in a general shutdown of protein synthesis [53]. However, as previously described, translation of ATF4 is increased in this situation [34,35,54] leading to the induction of its target gene Chop. To monitor activation of this pathway, we analysed CHOP, ATF4 and p-eIF2α expression by qRT-PCR and Western blotting. 17 Cl-1 cells were infected with MHV-A59 or incubated with tunicamycin for 2.5, 5, 8 and 10 h. As shown in Fig 6A, Chop mRNA levels (measured as the Chop/RpL19 ratio) increased five-fold in tunicamycin-treated cells (purple) compared to mock-infected cells (blue), and were stable over the time course. In MHV-infected cells (orange), the ratio also increased from 2.5 to 8 h p.i. although not to the level seen in tunicamycin-treated cells. Protein expression was determined by immunoblotting using antibodies specific for ATF4, p-eIF2α and N proteins with GAPDH and eIF2α as loading controls (Fig 6B). ATF4 and p-eIF2α were detected at all time points in tunicamycin-treated cells and all of the proteins were detected in MHV-infected cells from 5 h p.i. onwards. Subsequently, we analysed profiles of the RiboSeq and RNASeq reads mapping to ATF4 in the virus-infected and tunicamycin-treated cells (Fig 6C). Consistent with previous studies [34,35], translation of the short (three codon) uORF1 (frame +2, nucleotides 399 to 407) was observed under all conditions. In mock-infected cells, uORF2 was efficiently translated (Mock Ribo panel; reads in yellow mapping to uORF2 indicated by a yellow rectangle, frame +2) thus diverting scanning preinitiation ribosomes from accessing the main ORF (blue rectangle, frame 0, reads in purple) to which very few RPFs mapped. In contrast, in MHV-infected cells at 8 h p.i. (Inf Ribo panel), a substantial fraction of preinitation ribosomes were able to scan past uORF2 to translate the main ORF, leading to a reduced density of ribosomes on uORF2 and a greatly increased number of RPFs mapping to the main ORF. Tunicamycin-treated cells showed an intermediate ribosome distribution, but again with efficient translation of the main ORF.
Fig 6D displays the RiboSeq and RNASeq profiles for Chop. Consistent with the qRT-PCR analysis (Fig 6A), RNASeq reads (Inf RNA panel) were increased at 5 h p.i. compared to mock and there was an increase in the number of RPFs mapping to the main ORF (blue rectangle, purple reads). This increase in translation of the main ORF was also accompanied by an increase in the number of RPFs derived from ribosomes initiating on the Chop uORF (yellow rectangle, blue reads). This uORF occupancy was especially strong in tunicamycin-treated cells (Tunica Ribo panel).
Polysome profiling of 17 Cl-1 cells infected with MHV-A59
Since total read counts are normalized by library size, ribosome profiling does not provide information on total global translation levels. To further investigate virus-induced inhibition of translation as a consequence of UPR activation and eIF2α phosphorylation, analytical polysome profiling (Fig 7A) was done for mock- and MHV-infected 17 Cl-1 cells. Cytoplasmic extracts were prepared in the presence of cycloheximide to retain intact monosomes and polysomes and analyzed by sucrose density gradient centrifugation. This revealed an accumulation of monosomes (80S) in MHV-infected cells from 5 h p.i. onwards, consistent with inhibition of initiation. To investigate whether the 80S ribosomes accumulating during MHV infection contain mRNA (as an indicator of a translating ribosome), polysome profiling was repeated using a higher salt buffer (400 mM KCl; Fig 7B), conditions in which 80S ribosomes lacking mRNA dissociate into constituent subunits. In mock-infected cells, a modest diminution of 80S levels was observed at 400 mM KCl (mock 5 h, compare Fig 7A panel 2, and Fig 7B left panel), but a much greater reduction in 80S was observed in MHV-infected cells (MHV 5 h p.i., compare Fig 7A panel 5 and Fig 7B right panel), indicating that the vast majority of 80S ribosomes accumulating at this time point are not mRNA-associated. These data indicate that MHV-infection leads to translational shut-off via inhibited initiation, consistent with the effects of eIF2α phosphorylation.
Effect of the PERK inhibitor GSK-260614 on MHV replication
GSK-260614 is a potent and selective high affinity ligand of the PERK kinase that interferes with kinase activity by competing for ATP [55,56]. In MHV-infected 17 Cl-1 cells at 5 h.p.i., the drug prevented phosphorylation of the PERK substrate, eIF2α, in a dose-dependent manner (Fig 8A), effectively blocking this branch of the UPR, with only a minor effect on cell viability (S4A Fig). Pulse labelling of infected cells for one hour at 5 h.p.i. revealed, as expected, that prevention of eIF2α phosphorylation increased modestly both viral (Fig 8A) and host protein synthesis (Fig 8B), without effect on mock-infected cells (Fig 8B).
Despite the increased virus protein synthesis, 17 Cl-1 cell monolayers infected with MHV-A59 in the presence of the PERK inhibitor remarkably showed delayed formation of syncytia in comparison to untreated cells at 8 h p.i. (Fig 8C), and quantification of released virions through TCID50 assays revealed an ~four-fold reduction in virus titre in cells incubated with GSK-260614 compared to control cells (P = 0.064; S4B Fig). These observations suggest that relieving inhibition of protein synthesis – affecting both cellular and viral proteins – is detrimental to virus production and the development of syncytia in virus infected cells.
Discussion
We have used ribosome profiling and parallel RNASeq to investigate changes in the cellular translatome and transcriptome in response to infection with MHV, a representative of the Betacoronavirus genus of the Coronaviridae family. These studies provide the highest resolution data to date on the translatome of cells during coronavirus-induced stress.
RNASeq libraries revealed that some of the most significantly upregulated cellular transcripts in virus-infected cells were part of the UPR (Herpud1 and Chac1) and changes in the translation efficiency of cellular proteins were consistent with uORF-regulated responses to eIF2α phosphorylation, including those previously implicated as effectors of the UPR such as in Atf4, Atf5 and Chop [34–39]. These high-resolution data confirm again that there is a close interplay between virus infection and the UPR, with the host activating the UPR to combat the effects of virus infection, and viruses sometimes manipulating the UPR to promote replication and pathogenesis [57–60]. The intimate association of CoVs with the ER during replication results in ER stress responses as the cell attempts to return to homeostasis [46, 61–65; reviewed in [66]).
The relative modulation of UPR branches differs between different CoVs [3,5,66]. For example, SARS-CoV infection does not lead to Xbp-1 splicing [46] whereas the IRE1 pathway is activated by infectious bronchitis virus or MHV infection or by MHV S protein overexpression [29,67]. In spite of the observed Xbp-1 mRNA splicing during MHV infection [29], Xbp-1s protein had not previously been detected in coronavirus-infected cells. In our study (8 h p.i. data set), an increased number of RPF reads mapped in the +2 reading frame of the Xbp-1u transcript corresponding to translation of the Xbp-1s frameshifted isoform.
Activation of the ATF6 pathway by CoV infection has not yet been fully addressed. ATF6 cleavage into its active form is observed during MHV infection but is significantly reduced at late time points [29]. On the other hand, the trimmed ATF6 form is not detected in SARS-CoV infected cells [68]. Furthermore, ER stress-responsive promoters exhibit little activity under these conditions. In the present study, an induction of BiP/Grp78 transcription due to ATF6 activation was observed to a similar extent in both tunicamycin-treated and MHV-infected cells, whereas BiP/Grp78 protein expression was only detected by western blotting in tunicamycin-treated cells. Ribosome profiling data revealed that, in virus-infected cells, the amount of RPFs corresponding to the BiP CDS was not as high as in tunicamycin-treated cells and this was probably the reason why this protein was not detected by western blot analysis.
With respect to the UPR-related inhibition mediated by eIF2α phosphorylation, it has been shown that infectious bronchitis virus activates or suppresses protein kinase RNA-activated (PKR) and PERK during the course of an infection [69] whereas transmissible gastroenteritis virus protein 7 emulates the function of DNA damage-inducible protein 34 (GADD34) to dephosphorylate eIF2α [70]. Our study has now shown that MHV-A59 infection increases the level of p-eIF2α and ATF4 from 5 h p.i. onwards. The RiboSeq data also revealed decreased translation of the Atf4 uORF2 at 8 h p.i. and a concomitant increase in translation of the main ORF. Although [29] failed to detect the products of ATF4 target genes, Gadd34 and Chop, during MHV infection by western blotting, we found evidence supporting an increase in transcription and translation of Chop at later time points p.i.
We tested the effect of the selective PERK inhibitor GSK-260614 on MHV replication [55,56]. GSK-260614 (IC50 = 0.4 nM) exhibits >1000-fold selectivity for PERK over heme-regulated eIF2α (HR1) and PKR. Up to 5 μM of this inhibitor was well tolerated by 17 Cl-1 cells and, in MHV-infected cells, the prevention of eIF2α phosphorylation alleviated the inhibition in translation of cellular and viral proteins as expected. Surprisingly, the higher content of viral proteins did not lead to a more prominent cytopathic effect but instead delayed syncytia formation and reduced viral titre. Therefore, we conclude that UPR-mediated eIF2α phosphorylation may be favourable to MHV replication – perhaps by preventing translation of various anti-viral factors – and the pharmacological manipulation of this UPR branch can be explored as a potential target for antiviral intervention.
Our previous analysis of RPFs mapping to host mRNAs during MHV infection [16] indicated a remarkable perturbation in ribosome distributions at late time points, with a greatly increased proportion of RPFs mapping to the first ~30–40 codons of coding regions. In yeast, it has been reported that an accumulation of reads in the 5’ region of coding sequences specifically under cell stress is an effect of cycloheximide pretreatment, and it has been hypothesized that stressed cells might have a slower rate of cycloheximide uptake leading to increased mean post-initiation ribosome run-on times prior to cycloheximide-induced arrest of elongation [43,44]. Thus, it was unclear whether this artefact might also occur in mammalian cells given that mammalian cells lack a cell wall. Our results show that this does indeed extend to mammalian cells and thus call into question the interpretations put forward by [41] and [42] for the similar perturbations in ribosome density observed when cells were subjected to heat shock or proteotoxic stress. Further, we find that the magnitude of this artefact depends on the type of stress and, in the case of MHV infection, is not alleviated by increasing the concentration of cycloheximide to 100 times the standard amount, but can be alleviated completely by flash freezing without cycloheximide pretreatment. Since the artefact also occurs in mammalian cells, it is possible that it is not related simply to cycloheximide uptake rates. Potentially it may instead be related to dysregulation of the translational machinery under stress in a way that reduces cycloheximide efficiency (e.g. due to alterations in ribosomal exit (E)-site tRNA occupancy).
Ribosome profiling provides information on initiating and elongating 80S ribosomes but it does not account for free monosomes or early stages in initiation prior to formation of the 80S complex. Analytical polysome profiling showed an accumulation of 80S monosomes in MHV-infected cells from 5 h p.i. with the vast majority not being associated with mRNA which is a typical outcome of impaired translation [71]. This suggests that protein translation was inhibited at the stage of initiation probably due to the activation of the PERK branch in response to ER stress and the concomitant phosphorylation of eIF2α. Phosphorylated eIF2α (p-eIF2α) forms a stable complex with eIF2B – the guanine exchange factor responsible for recycling inactive eIF2-GDP to eIF2-GTP – which rapidly reduces the pool of available eIF2B. This prevents recycling of the ternary complex of eIF2, GTP and Met-tRNAi and formation of the 43S pre-initiation complex, and thus leads to a general shutdown of protein synthesis by inhibition of initiation [72].
Viruses commonly employ translational shutoff mechanisms to facilitate viral replication. On the one hand, shut-off of host cell translation can redirect the translation machinery towards viral gene expression if the virus has evolved non-canonical modes of translation, such as internal ribosome entry site (IRES) mediated initiation. On the other hand, the shut-off of host cell protein synthesis will inhibit a range of cellular anti-viral responses. Previous studies have shown that MHV can induce host translational shutoff and mRNA decay in LR7 cells with the concomitant formation of stress granules and processing bodies [28]. Furthermore, a number of reports have demonstrated that CoV nsp1, the most N-terminal product of the replicase polyprotein, modulates host protein synthesis. In different CoVs, nsp1 has been shown to associate with the 40S ribosomal subunit thus preventing viral and cellular mRNA translation; induce cellular mRNA degradation via an endonucleolytic mRNA cleavage in the 5’ region of capped mRNA; and selectively target nuclear host mRNAs and transport them to the cytoplasm for degradation [30–31,73–74]. The involvement of nsp1 in host protein translation could not be ruled out in this study without a comparison with a mutant virus lacking nsp1. However, the UPR-related translational modulation and the CoV nsp1-related modification of translation (and mRNA degradation) testify to the complexity of cellular translational shutoff mechanisms utilized by CoVs. How MHV proteins can be synthesized in a state of global translation inhibition has been the subject of previous speculation. Viral mRNAs contain a common 5’-leader sequence (65–90 nucleotides long) that could bind to the nucleocapsid (N) protein to form a complex that might act as a strong translation initiation signal [75], or the leader RNA sequence may bind to nsp1, protecting the viral mRNAs from nsp1-induced RNA cleavage [74,76]. However, we found previously that virus mRNAs 2–7 were translated with generally similar efficiencies during infection and, importantly, were not preferentially translated relative to host mRNAs. Thus we concluded that the synthesis of large quantities of virus proteins, especially N, was achieved mainly through high levels of transcription [16].
In conclusion, this study provides a survey of coronavirus effects on the cellular transcriptome and translatome, complementing previous investigations on the UPR and host cell shutoff during MHV infection. The results of our analyses will help inform further investigations on host-CoV interactions and may help identify new targets for antiviral intervention.
Materials and Methods
Ribosomal profiling and RNASeq data: Virus sequencing data (RNASeq and RiboSeq with cycloheximide pretreatment) were generated previously [16] and have been deposited in the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress) under the accession number E-MTAB-4111.
Data for the tunicamycin and flash freezing sequencing experiments have been deposited in ArrayExpress under the accession numbers E-MTAB-5391 and E-MTAB-6278 respectively.
For the tunicamycin experiments, 17 Cl-1 cells were incubated in the presence of tunicamycin (2 μg/ml) and, after 6 h, cells were treated with CHX (Sigma-Aldrich; to 100 μg/ml or 10 mg/ml; 2 min). Cells were rinsed with 5 ml of ice-cold PBS and then submerged in a reservoir of liquid nitrogen for 10 s. For flash freezing experiments, cells were rinsed with 5 ml of ice-cold PBS and then submerged in liquid nitrogen for 10 s. Cell lysates were subjected to RiboSeq and RNASeq as previously described [16,77].
Computational analysis of RiboSeq and RNASeq data: Reads were trimmed for adaptor sequences, filtered for length ≥ 25 nt, and reads mapping to Mus musculus rRNA (with up to 2 mismatches) removed, as previously described [16]. The remaining reads were aligned directly to the mouse genome (UCSC, assembly mm10) (with up to 2 mismatches) using TopHat (parameters: --no-novel-juncs --bowtie1 --prefilter-multihits --max-multihits 500, with --transcriptome-index defined using the genes.gtf file from the UCSC mm10 annotation available from the TopHat website) [78]. Reads were tabulated using htseq-count (parameters: -t CDS -m intersection-strict -i gene_id -s yes), using the NCBI RefSeq mRNAs as the gene feature annotation. For differential expression analyses, the htseq-count parameter “-type CDS” was used to count only those reads (both RiboSeq and RNASeq) that mapped to the annotated coding regions of transcripts [79]. Thus the differential expression analyses exclude reads mapping to uORFs or non-annotated coding sequences (unless such sequences overlap the main annotated ORF).
Differential expression analyses on RNASeq or RiboSeq count data were performed with DESeq [20], EdgeR [21], NOISeq [22] and BaySeq [23]. Read counts were normalized by library size prior to comparison and low count genes (genes with fewer than 1 count per million reads mapping to the CDS in either the infected or mock libraries) were discarded. For BaySeq analyses, the sample size used to calculate priors was set to 200000. For each comparison, two replicates from infected cells at a specific timepoint were compared to four libraries from uninfected cells (two each at 1 and 8 h). Changes in translation efficiency (TE) were estimated using Bayseq in the paired library mode, Babel [32] and Xtail [33]. A given gene was considered to be differentially expressed if the FDR was less than 0.05 and the fold change between the averages of infected and mock replicates was greater than two. Volcano plots and inter-replicate consistency plots were generated using standard R plotting features and FDR and log2(fold change) values from the BaySeq, Babel and Xtail analyses.
As reads were mapped to the genome, genomic read mapping locations had to be converted to transcript-specific coordinates to generate RNASeq and RPF profiles for specific transcripts. Tophat bam files were sorted by genomic coordinate and indexed using SAMTools [80]. A custom R script using Rsamtools [81] was developed to extract reads at a given series of genomic loci from the bam file (corresponding to the exon sequences of a given transcript isoform), convert the genomic positions of reads to positions relative to the transcript sequence, calculate the phasing of the reads, and generate plots showing the distribution of reads on the transcript. In some cases it is not possible to definitively distinguish usage of alternative exons or changes in transcription start sites or initiation codon utilization. The advantage of the aforementioned visualization process is that it retains all reads mapping to a given exon allowing for easy manual inspection of data which may originate from different isoforms. In individual gene plots, read positions for RPFs were offset +12 nt to map the approximate P-site.
Histograms of 5’ end positions of host mRNA reads relative to initiation and termination codons were derived from reads mapping to RefSeq mRNAs with annotated CDSs ≥450 nt in length and with annotated 5’ and 3’ UTRs ≥60 nt in length. All figures are based on total numbers of mapped reads, rather than weighted sums for highly expressed mRNAs.
Ribosome accumulation at the 5’ end of coding sequences was assessed via a 5’ loading ratio statistic. Specifically, the density of RPFs with estimated P-sites mapping within nucleotide positions 16–90 of a given ORF was compared with the density of reads in the downstream portion of the ORF (ending 15 nt before the ORF stop codon). These windows were designed to exclude the accumulation of ribosomes involved in initiation and termination. We then compared the loading ratio in infected samples with the loading ratio in mocks for each gene.
Cells and virus: Murine 17 clone 1 (17 Cl-1) ([82], a kind gift of Dr Stanley Sawicki, University of Toledo) cells were maintained in Dulbecco’s modification of Eagle’s medium supplemented with 10% (vol/vol) fetal calf serum (FCS). Recombinant MHV strain A59 (MHV-A59) was derived as previously described ([83], a kind gift of Dr Stanley Sawicki, University of Toledo, ATCC VR764). Upon reaching 70–80% confluence, 17 Cl-1 cells were infected with MHV-A59 at MOI 10 in infection medium [Hank’s balanced salt solution (HBSS) containing 50 μg/ml DEAE-dextran and 0. 2% bovine serum albumin (BSA)]. After 45 min at 37 °C, the inoculum was removed and the cells were incubated in DMEM containing 10% FCS, 100 U/ml penicillin and 100 μg/ml streptomycin at 37 °C until harvest.
17 Cl-1 mock and MHV-infected cells were treated with different concentrations (1–5 μM) of the PERK-inhibitor GSK-2606414, a kind gift of Dr Edward Emmott and Prof Ian Goodfellow. GSK-2606414 was added to the cells just after the adsorption time and maintained until cells were harvested.
Quantitative real-time PCR assays: Total RNA was isolated as described previously [76] for RNA-Seq analysis, and cDNA was synthesized from 1 μg total RNA. Transcript levels were determined by quantitative real-time PCR using a Rotor-Gene 3000 (Corbett Research). Reactions were performed in a final volume of 20 μl containing Hot Start Taq (1 U; QIAGEN), 3.5 mM MgCl2, 2.5 mM deoxynucleotides, SYBR Green dye, 500 nM forward and reverse specific primers and 1 μl of cDNA. After enzyme activation (95 °C, 15 min), amplification was carried out in a three-step PCR procedure (50 cycles: 15 s at 95 °C for denaturation, 20 s at 55 °C for annealing and 20 s at 72 °C for extension). Non-template controls were included for each primer pair, and each PCR reaction was carried out in triplicate.
Immunoblotting: Proteins were separated by 10% or 12% SDS-PAGE and transferred to nitrocellulose membranes. These were blocked (5% non-fat milk powder in PBST [137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.5 mM KH2PO4, pH 6.7, and 0.1% Tween 20]) and probed with mouse monoclonal antibodies raised against N (1:1,000), S (1:500) - kind gifts of Dr Helmut Wege, University of Würzburg -, GAPDH (G8795, Sigma-Aldrich, 1:20,000), S6 (1:500, Cell Signaling); rabbit monoclonal antibodies against Grp78/BiP (1:1,000, Abcam) and RPL10a (1:500, Abcam); or polyclonal rabbit anti-ATF4 (1:500, Proteintech), anti-eIF2α and anti-p(Ser-51)-eIF2α (1:1,000, Cell Signaling). Membranes were incubated in the dark with an IRDye-conjugated secondary antibody in phosphate-buffered saline (PBS) and 0.1% Tween 20 [IRDye 800CW Donkey Anti-Mouse IgG (H+L), IRDye 800CW Donkey Anti-Rabbit IgG (H+L), IRDye 680RD Goat Anti-Mouse IgG (H+L) and IRDye 680RD Goat Anti-Mouse IgM (μ chain specific)]. Blots were scanned using an Odyssey Infrared Imaging System (Licor).
Polysome profiling: 17 Cl-1 cells were infected as previously described. Ten minutes prior to harvesting, cells were treated with cycloheximide (100 μg/ml), washed with PBS and lysed in a buffer containing 20 mM Tris HCl pH 7.5, 100 mM KCl, 5 mM MgOAc, 0.375 mM CHX, 1 mM DTT, 0.1 mM PMSF, 2U/μl DNase I, 0.5% NP-40, supplemented with protease and phosphatase inhibitors (ThermoFisher Scientific). Following trituration with a 26-G needle (ten passes), lysates were cleared (13,000 g at 4 °C for 20 min) and the supernatants layered onto 12 mL sucrose density gradients (10–50% sucrose in TMK buffer – 20 mM Tris-HCl pH 7.5, 100 mM KCl, 5 mM MgCl2) prepared in Beckman SW41 polypropylene tubes using a Gradient Master (Biocomp). Following centrifugation (200,000 g for 90 min at 4 °C), fractions were prepared using an ISCO fractionator monitoring absorbance at 254 nm. Proteins were concentrated from fractions using methanol-chloroform extraction [84] and subjected to immunoblotting analysis. Polysome profiling in higher salt conditions was carried out as described above except that the lysis buffer and sucrose density gradient contained 400 mM KCl.
Metabolic labelling: 17 Cl-1 cell monolayers were infected with MHV A-59 at a MOI of 10 PFU/cell. At 5 h p.i., cells were washed twice with PBS and labelled for 1 h in methionine-free DMEM supplemented with 125 μCi/ml [35S] methionine. After this period, cells were harvested, washed twice with PBS and resuspended in lysis buffer (50 mM Tris pH 7.5, 100 mM NaCl, 5 mM EDTA, 0.5% NP40). Cell lysate aliquots were mixed with Laemmli’s sample buffer to a final concentration of 1× and subjected to 10% SDS-PAGE followed by autoradiography.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.
- 8.↵
- 9.↵
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.↵
- 17.
- 18.
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.
- 37.
- 38.
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.
- 59.
- 60.↵
- 61.↵
- 62.
- 63.
- 64.
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵