key: cord-103430-x6zzuu7v
authors: Contu, Lara; Balistreri, Giuseppe; Domanski, Michal; Uldry, Anne-Christine; Mühlemann, Oliver
title: Characterisation of the Semliki Forest Virus-host cell interactome reveals the viral capsid protein as an inhibitor of nonsense-mediated mRNA decay
date: 2020-10-12
journal: bioRxiv
DOI: 10.1101/2020.10.12.335497
sha: 
doc_id: 103430
cord_uid: x6zzuu7v

The positive-sense, single-stranded RNA alphaviruses pose a potential epidemic threat. Understanding the complex interactions between the viral and the host cell proteins is crucial for elucidating the mechanisms underlying successful virus replication strategies and for developing specific antiviral interventions. Here we present the first comprehensive protein-protein interaction map between the proteins of Semliki Forest Virus (SFV), a mosquito-borne member of the alphaviruses, and host cell proteins. Among the many identified cellular interactors of SFV proteins, the enrichment of factors involved in translation and nonsense-mediated mRNA decay (NMD) was striking, reflecting the virus’ hijacking of the translation machinery and indicating viral countermeasures for escaping NMD by inhibiting NMD at later time points during the infectious cycle. In addition to observing a general inhibition of NMD about 4 hours post infection, we also demonstrate that transient expression of the SFV capsid protein is sufficient to inhibit NMD in cells, suggesting that the massive production of capsid protein during the SFV reproduction cycle is responsible for NMD inhibition.

suggesting that the massive production of capsid protein during the SFV reproduction cycle is 13 responsible for NMD inhibition. 14 15

As we live through the current SARS-COV2 pandemic, the world is reminded of the unpredictable 17 nature of viral epidemics and the importance of studying potential emerging viral threats. Recent 18 studies present valid arguments for the worldwide epidemic threat of alphaviruses (among other 19 arboviruses) that currently circulate endemically in particular regions 1,2 . The outbreak potential of 20 alphaviruses has already been showcased by the two worldwide epidemics caused by Chikungunya 21 virus (CHIKV) that affected more than 8 million people in over 50 countries and could be attributed to 22 a single point mutation leading to a 100-fold increase in infectious virus in the salivary glands of urban 23 mosquitoes 1,2 . This demonstrates that small genetic alterations can cause dramatic changes in human 24 transmissibility and infection. Semliki Forest Virus (SFV) is closely related to CHIKV, both evolutionarily 25 grouped within the Semliki Forest (SF) clade of the Old World alphaviruses (Family: Togaviridae) 3 . SFV 26 causes lethal encephalitis in mice 4 . Though mostly associated with mild febrile illness or 27 asymptomaticity in humans, SFV is endemic to African regions 1 and a handful of studies indicate 28 serious disease relevant symptoms associated with SFV in humans, including encephalitis, myalgia and 29 arthralgia 5-8 . 30 31 SFV is a small (~70 nm in diameter), enveloped virus comprising a nucleocapsid core made up of 240 32 copies of capsid protein that surrounds its positive-sense single-stranded RNA genome (~11.8 kb). The 33 genome contains a 5´ cap (N7mGppp) and poly(A) tail and is organised into two distinct open reading frames (ORFs). The first ORF encodes the non-structural proteins (nsP1, nsP2, nsP3 and nsP4) ( Figure  1 1a), which are translated as one polyprotein (P1234) immediately upon exposure of the viral mRNA-2 genome to the cytoplasm [9] [10] [11] [12] . The polyprotein is then proteolytically cleaved by the protease activity 3 of nsP2 to yield functional viral replicase complexes 13 . The first protein to be cleaved from the 4 polyprotein is nsP4, comprising RNA-dependent RNA polymerase activity. The resulting P123 5 polyprotein in complex with nsP4 forms the viral replication complex (RC), responsible for synthesizing 6 minus strand template RNA from the genomic viral (v)-RNA early during infection 9 . The ensuing 7 double-stranded vRNA intermediates can trigger the activation of Protein Kinase double-stranded 8 RNA-dependent (PKR), resulting in phosphorylation of the α-subunit of the eukaryotic translation 9

Initiation Factor 2 (eIF2) and thus causing a decrease in global translation of host cell messenger RNAs 10 (mRNAs) 10,14,15 . As proteolytic cleavage of P123 by nsP2 progresses, individual nsPs form new viral RCs 11 of altered composition, resulting in a shift from synthesis of the minus strand template, to synthesis 12 of new viral genomes and viral subgenomic RNA (sgRNA) from the 26S promoter ( Figure 1a ) 9, 16 . 13

Alphavirus replication occurs in membrane invaginations called 'spherules', where high 14 concentrations of RCs are present 9, 11 . Binding of host cell proteins to RCs has been reported, though 15 the abundances and the functions thereof are still not fully understood 11,17, 18 . In addition, individual 16 SFV proteins localise independently of the RC to perform functions separate from viral replication 17 9, 11, 19, 20 . One example is the nsP2 protein, which translocates to the nucleus 21 Here, we investigated the virus-host protein interactome of SFV. A greater understanding of the 1 repertoire of host proteins that may be exploited by viruses is a vital first step toward developing 2 antiviral strategies aimed at targeting or interfering with interactions that may be critical for the 3 infection. While previous studies have reported host interactors of SFV from isolations of RCs from 4 lysosome fractions, as well as affinity purifications and localisation studies of nsP3-tagged 5 recombinant virus 18, [29] [30] [31] , there are so far no SFV studies that assess the complete set of viral-host 6 protein-protein interactors (PPI). Using affinity purifications followed by high-throughput quantitative 7 mass spectrometry, we identified host protein interactors of individual SFV proteins in human cells. In 8 addition, using a genome-wide siRNA screen we assigned pro or antiviral functions to some of the 9 identified SFV interactors. Gene ontology (GO) enrichment analyses of protein complexes that could 10 form between the identified host interactors revealed highly significant GO terms related to 11 translation and Nonsense-mediated mRNA Decay (NMD). NMD is known to restrict infection of 12

alphaviruses, but whether and how the virus counteracts this cellular intrinsic defence is still not clear 13 32 . Here we show that during the course of infection SFV suppresses NMD. We present evidence that 14 the capsid protein of SFV is sufficient to suppress NMD independently of translation inhibition. 15 16 17

As obligatory parasites, all viruses exploit the host cell to favour their own replication. In turn, cells 19

have evolved mechanisms to protect against viral infections. To gain insight into the repertoire of host 20 proteins that could be exploited by SFV, we systematically mapped the interactions between the 21 individual SFV proteins, nsP1-4, C, and the envelope polyprotein, Env (which includes E3, E2, 6K and 22 E1), and the host cell proteome using high-throughput quantitative mass spectrometry (Figure 1a) . 23

The SFV proteins were N-terminally tagged with 3xFLAG and transiently expressed in HeLa cells, a cell 24 type susceptible to infection by SFV 33 . The proteins were then affinity purified from the respective 25 lysates using anti-FLAG antibodies, with and without treatment with RNase A to distinguish RNA-26 mediated from protein-mediated interactions (Figure 1b) . Western blot analysis of the eluates from 27 each anti-FLAG affinity purification revealed the successful pulldown of all six transiently expressed 28 SFV proteins (Figure 1c) . 29

The protein compositions of the eluates, from three biological replicates of each affinity purified SFV 31 protein, were analysed by quantitative mass spectrometry (Figure 2a and Suppl. Figure 1 ). Significant 32 interactors (see Methods) (Figure 2b, purple circles and Table 1 ) were further filtered by abundance, 33 such that proteins whose abundance made up at least 0.5 % of the relevant SFV bait protein were 5 1 2 Figure 1 . Strategy for creation of Semliki Forest Virus (SFV) -host protein-protein interaction map. a, Schematic illustration 3 of the genomic organisation of SFV. The first ORF encoding the non-structural proteins (nsp1, nsp2, nsp3, nsp4) is highlighted 4 in green. The ZSG tag inserted within the nsp3 protein is also depicted. The second ORF encoding the structural proteins 5 (capsid, E3,E2,6K,E1) is highlighted in orange. Other viral features depicted include the 5´ cap, poly(A) tail, as well as the 6 position of the 26S subgenomic viral promoter. b, Flowchart outlining the experimental approach to transiently express N-7 terminally FLAG-tagged (yellow rectangle) SFV proteins in mammalian cells in order to construct a SFV-host protein-protein 8 interactome. Nsp3-Z refers to the nsp3 protein with the ZSG tag, C refers to the capsid, and E3 E2 6K E1 refer to the envelope 9 proteins (Env) that were expressed as one polyprotein. c, Anti-FLAG western blot of SFV proteins after transient transfection 10 and affinity purification from HeLa cells (without RNase A treatment). Red asterisks indicate 3xFLAG tagged SFV proteins at 11 their expected sizes. Untransfected cells (Untr) and cells transfected with a plasmid encoding only the 3xFLAG tag with no 12 additional coding region (empty) were included as controls. The expected sizes of the 3xFLAG-tagged proteins were: empty 13 ~8kDa; nsp1 ~63kDa; nsp2 ~92kDa; nsp3-Z ~82kDa; nsp4 ~72kDa; capsid ~33kDa and Env (polyprotein ~111kDa, cleavage 14 intermediates ~57kDa / ~64kDa). The affinity purifications were conducted in triplicate (± RNase A treatment), and eluates 15 analysed by mass spectrometry. Table 1 ). In the case of the nsP3 bait (here fused with the 18 fluorescent protein ZsGreen, nsP3-Z), which was very lowly abundant in the sample as it proved 19 difficult to elute from the beads (Suppl. Figure 2) , we retained proteins whose abundance made up at 20 least 5 % of the bait (Table 1 ). The heat map in figure 2c summarises the most abundant significant 21 interactors of each SFV protein in the -RNase A samples, with their corresponding abundance in the 22 +RNase A samples alongside them. Many of the host interactors identified in the -RNase A sample 23 were lost upon treatment with RNase A, indicating that these interactions were likely mediated by 1 RNA. This was clear for many nsP2, nsP3-Z and capsid interactors, where the heat map ( Figure 2c ) 2 corroborated the observations in the analytical silver stain gel (Figure 2a) . In both the heat map and 3 the gel, we noted proteins in the nsP2 eluate that were enriched in the +RNase A sample compared 4 to the -RNase A sample. Also reflected in the heat map were proteins observed in the gel that were 5 more than or as abundant as nsP2 (~92 kDa) (Figure 2a and c, +RNase A samples). These agreements 6 between the quantified lists obtained to create the heat map and the observations in the analytical 7 silver stain gel gave us confidence in the strategy employed to collect host protein interactors from 8 the mass spectrometry datasets. Since many SFV-host protein interactions were dependent on RNA, 9

we chose to focus on the lists of interactors from the -RNase A datasets going forward. A summary of 10 these revealed that a large fraction of the interactions for nsP2, nsP3-Z and capsid consisted of 11 ribosomal proteins (Figure 2d ). 12

These stringent lists of interactors are displayed as SFV-host interactome networks (Figure 3a and 14 Suppl. Figure 3 ). Affinity purified SFV proteins are displayed as black circles, while host cell proteins 15 are displayed as smaller, colour-coded circles. Host proteins that were identified as unique interactors 16 to one of the SFV proteins are connected to the relevant host protein with a grey line. Many of the 17 host proteins were identified as interactors to more than one of the SFV proteins. For simplicity, these 18 non-unique interactors are grouped into grey boxes with grey lines connecting the whole group of 19 proteins to the SFV proteins for which they were identified as interactors (Figure 3a ). Considering this 20 overlap, the total number of host proteins that were identified as interactors was 251 (Figure 3a and  21 Suppl. Figure 3) , 77 of which were ribosomal proteins and are shown separately (Suppl. Figure 3 ). Host 22 proteins displayed in the networks were manually curated and categorised into colour-coded groups 23 based on descriptions gathered from both Gene Ontology (GO) and STRING analyses ( Figure 3a) . 24

Interactions that stood out included subunits of the chaperonin-containing t-complex polypeptide 1 25 (CCT complex) (pink) that interacted with both nsP2 and nsP4, a number of cytoskeletal proteins or 26 proteins involved in cytoskeletal signalling (grey) interacting with nsP2 and nsP1 (tubulins), ER 27 chaperones (pink) bound uniquely to Env, and a large number of RNA binding proteins (violet) 28 interacting with nsP2, nsP3-Z and capsid. In addition, a striking presence of rRNA processing / 29 ribosome biogenesis factors emerged as interactors, many of which were found bound uniquely to 30 the capsid (dark pink) (Figure 3a ). Previously reported human protein-protein interactions were 31 analysed by STRING and additionally displayed on the networks (pink dashed lines). The dense 32 network of edges (pink dashed lines) that emerged among the rRNA processing / ribosome biogenesis 33 factors (dark pink) reflects the known protein-protein interactions that have been reported between depicted. c, Using the list of -RNase A interactors, a threshold of abundance of at least 0.5 % of the bait protein (in the case 5 of nsp3-Z, at least 5 % of the bait protein) was applied. A heat map summarising either the top10 most abundant interactors 6 (or all if there were fewer than 10 interactors identified) for each SFV bait protein without RNase A treatment (-) is shown.

The corresponding abundance as % of bait of the interactors that were statistically significantly enriched when treated with 8 RNase A (+) is also shown in the heat map. Grey blocks indicate proteins that did not appear or were not statistically 9 significantly enriched in the +RNase A samples. Note that due to the low abundance of nsp3-Z, the abundance as % of bait 10 of nsp3-Z and all its interactors are presented as a factor of 10 less than what was calculated (for better visual representation 11 of the heat map as a whole). For all values and the complete list of interactors, see Table 1 . d, Bar graph summary of the 14 15 them ( Figure 3a ). This indicates that the capsid protein (and to a lesser extent nsP2/nsP3-Z) may 16 interact with a complex of proteins involved in rRNA processing and/or ribosome biogenesis. 17 18 We next set out to determine whether the identified host proteins have pro-or antiviral effects in the 19 context of infection. To systematically address this question, we used a genome-wide fluorescence 20 microscopy-based siRNA screen 33 to identify host proteins that affected SFV replication ( Table 2 ). In product, compared to the control, non-specific siRNAs (set as 1). A low multiplicity of infection (MOI) 27 of 0.3 was used to allow for detection of both reduced or increased infection levels. The maximum II 28 obtained in the screen was 1.85 (Table 2) . We therefore chose to set an II threshold of 1.3 to identify 29 proteins having a potential antiviral role against SFV, and an II threshold of 0.5 to indicate proteins 30 having a potential proviral role for SFV (Table 2 ). When we compared the proteins identified in the 31 siRNA screen with the SFV-host protein interaction networks, a sizable fraction of the interactors 32 overlapped. Those with potential pro-or antiviral roles during infection are depicted with turquoise 33 or green outlines, respectively (Figure 3a and Suppl. Figure 3 ) and collected in two lists: 'Proviral' and 34 'Antiviral', with their II values shown alongside them ( Figure 3b ). As expected, all the identified 35 ribosomal proteins affecting SFV replication had a proviral effect (Suppl. Figure 3 ) and are listed 36 separately ( Figure 3b ). In addition to ribosomal proteins, many other RNA binding proteins were also 37 identified as having a potential role in SFV replication (Figure 3a (nsP1, nsP2, nsP3-Z, nsP4, capsid and Env) as well as for the merged list (All baits) was also applied. 28

This allowed us to compare the significance of the GO terms collected in Figure 4a for the different 29 SFV proteins (Figure 4b ). We observed that the most significantly enriched GO terms were those 30 related to translation, NMD and ribosome biogenesis. In addition, it became clear that it was mainly 31 the interactors of nsP2, nsP3-Z and capsid that contributed to the enrichment of these GO terms 32 Because of the enriched GO terms, we chose to investigate the effect of SFV on translation, NMD and 9 ribosome biogenesis. We reported previously that the NMD machinery could target the SFV genome 10 independently of the 3´ UTR 33 . Half-life measurements of the genome of a replication incompetent 11 SFV mutant suggested that this occurred early during infection, upon entry of the viral genome into 12 cells 33 . Since viruses are known to evade cellular defence responses against viral infection, we 13 wondered whether the virus could inhibit NMD at later stages of infection. Since NMD depends on 14 translation 34 and viruses are known to inhibit translation, it was important to carefully analyse the 15 time course of infection in our system in an attempt to disentangle these two tightly linked cellular 16 processes. We used anti-ZsG or anti-nsP3 antibodies to detect nsP3-Z, as a representative for the 17 presence of early produced non-structural proteins expressed from the gRNA, and anti-capsid 18 antibodies to detect the capsid, as a representative for the presence of structural proteins that are 19 expressed from sgRNAs later during the virus replication cycle (Figure 5a and b). The nsP3-Z protein 20 was reproducibly detected at 3-4 hours post infection (p.i.) and as early as 2 hours p.i., while the capsid 21 was reproducibly detected at 4 hours p.i. (Figure 5a and b). We measured the presence of 22 phosphorylated (p-)eIF2α compared to total eIF2α as an indication of virus-induced translation 23 inhibition and showed that a virus-dependent accumulation of p-eIF2α was reproducibly detected at 24 3-4 hours p.i. (Figure 5a ). In addition, we performed time course puromycin incorporation assays to 25 assess global translation activity using a more direct method 35 . This assay involves a puromycin pulse 26 for 10 minutes, which causes the release of nascent polypeptides and results in many puromycin-27 labelled polypeptides of different lengths that can then be visualised by western blotting using anti-28 puromycin antibodies. The decrease in puromycin 3-4 hours p.i. is therefore indicative of a decrease 29 in global translation, in agreement with the observed increase in p-eIF2α ( Figure 5b ). 30

We set aside samples from the time course infections in Figure 5a to assess NMD activity by measuring 32 relevant RNA levels by RT-qPCR. To assess NMD activity, we adapted an assay described in 36 , which 33 measures the relative amounts of a NMD-sensitive splice isoform (NMD target) versus a NMD 34 insensitive protein coding isoform (non-NMD target) of the same gene. We showed that at 4 hours started to accumulate at 3 hours p.i., whereas the increase of BAG1_NMD only became apparent at 4 1 hours p.i. (Suppl. Figure 4 ). In addition, we measured the RNA levels of the well-known endogenous 2 NMD targets, RP9P, IRE1α and GADD45, which also accumulated 3-4 hours p.i. (Figure 5c and Suppl. 3 Figure 4 ). Together, these data are indicative of reduced NMD activity 3-4 hours p.i., suggesting that 4 SFV can indeed inhibit NMD at later stages of infection. Since the timing of the NMD inhibition 5 correlated with that of eIF2α-dependent inhibition of cellular mRNA translation, we were unable to 6 pull apart the effect the viral infection had on the two cellular processes independently. Translation 7 inhibition by SFV, and RNA viruses in general, is well described to occur through induction of p-eIF2α, 8

which occurs upon host cell detection of the double-stranded viral RNA intermediate that arises during 9 its replication cycle 10,14,15 . We therefore reasoned that, if one of the SFV proteins was responsible for 10 the NMD inhibitory phenotype, we would be able to disentangle the effect of the virus on the two 11 cellular processes. Taking the mass spectrometry data analysis ( Figure 4b ) into account, we reasoned 12 that nsP2, nsP3-Z or capsid could be responsible for the NMD inhibitory phenotype. First, in order to 13 confirm the interactions identified between nsP2, nsP3-Z and capsid with cellular UPF1 (Figure 3 were identified in the UPF1 eluates of the infected sample ( Figure 5d ). It should be noted that their 23 abundance was higher than that of the NMD factors, UPF3A and SMG6. Taken together, we were 24 therefore able to confirm the RNA-mediated interactions of nsP3-Z and capsid with UPF1. 25

The results above suggested that nsP3-Z and capsid were the most likely candidates to influence NMD 27 activity in cells. Nevertheless, we decided to analyse the effect of all individual SFV proteins on NMD 28 activity. To do this, the SFV proteins (from plasmids described in Figure 1b suppress NMD in cells. Since translation-related GO terms were also enriched for capsid interactors 1 (among other SFV proteins) (Figure 4b) , it was important to investigate whether expression of capsid 2 influenced translation in cells, as this would in turn influence NMD activity. We showed that 3 expression of capsid did not induce p-eIF2α (Figure 6c ) nor, as judged from the puromycin 4 incorporation assays, effect changes in global translation (Figure 6d ). In addition, polysome profile 5 gradients revealed that cells expressing the SFV capsid retained intact polysomes (Suppl. Figure 5) , 6

indicative of unperturbed translation. These three independent sets of data convincingly show that 7 expression of the capsid protein did not influence global translation in cells. We therefore concluded 8 that the SFV capsid suppresses NMD through a mechanism independent of translation inhibition. Since 9 ribosome biogenesis related GO terms were most highly enriched among capsid interactors compared 10 to the other SFV proteins, we used capsid expressing cells to look for any indication of altered 11 ribosomes/rRNA that could give us any hints on the mechanistic action of capsid. Though capsid could 12 be trapped in the nucleus upon blocking of export, we were unable to find any phenotypic changes in 13 polysome gradients (Suppl. Figure 5) (Figure 2a and c) . As such, a large number of RNA-binding proteins were 28 identified as host interactors of these three SFV proteins (Figure 3a) , raising the question of whether 29 they could play a role in altering and exploiting the compositions of mRNPs during infection. Some of 30 the identified RNA-binding proteins that were previously found in SFV RCs include HNRNPC, HNRNPA1, 31 SFPQ, DHX9, DDX3X, PABPC1, G3BP1 and G3BP2 18 . In addition to being found in RCs, the nsP3: G3BP 32 interaction has been well characterised 29-31 . SFV nsP3 has been reported to bind G3BP and suppress 33 the formation of stress granules, thought to have antiviral activity 30 . Consistent with these previous findings, we identified G3BP1, G3BP2 and USP10, a deubiquitinase protein known to bind G3BP 30 , as 1 interactors of nsP3-Z. USP10 was identified as a unique interactor of nsP3, while G3BP1 and G3BP2 2 were identified as also interacting with nsP2 and capsid. Thanks to a nuclear translocation signal, nsP2 3 shuttles between cytoplasm and nucleus during infection and interferes with transcription 22 . In line 4 with this, and as observed by others, we observed localisation of the 3xFLAG-nsP2 in the nucleus of 5 HeLa cells at steady state (data not shown). Our study identified a number of nuclear proteins 6 interacting with nsP2, many of which are splicing regulators, including HNRNPC, HNRNPA1, HNRNPA3, 7 HNRNPF, HNRNPH3 and SFPQ. HNRNPC was the most abundant significantly enriched protein in the 8 nsP2 pulldown. It was also significantly enriched and abundant in the RNaseA-treated nsP2 sample, 9

indicating that this interaction may be either direct or mediated by another protein. Interestingly, 10 SFPQ was found to be one of the most abundant interactors not only of nsP2, but also of capsid ( Figure  11 2c). SFPQ was additionally identified through the siRNA screen, along with the mRNA export factor, 12 NXF-1, as being among the strongest proviral interactors (i.e. depletion of these factors inhibited viral 13 infection). The binding of nsP2 and capsid to these nuclear proteins could influence the regulation of 14 RNA processing or mRNA modification steps in the nucleus or re-localise nuclear proteins to the 15 cytoplasm in order to achieve a productive infection. Some cytoplasmic viruses, for example, hijack 16 nuclear proteins including splicing factors (hnRNPs and SFPQ) from the nucleus to the cytoplasm, 17 increasing infectivity 40-42 . We also identified RNA-binding interactors exhibiting strong antiviral effects 18 activity, including RPS27a, which was interestingly the only ribosomal protein to be found bound to 28 (in addition to nsP2) nsP1, nsP4 and Env (Figure 2c ). We were surprised by the presence of newly 29 identified nuclear interactors involved in rRNA processing and ribosome biogenesis, many of which 30 exhibited antiviral activity. Interestingly, many of these were uniquely bound to capsid (Figure 3a) . 31

Evidence of capsid in the nucleus has previously been reported 47 and we were able to trap 3xFLAG-32 capsid in the nucleus of HeLa cells upon blocking of export (data not shown). Even so, little is known 33 about the role of the capsid in the nucleus and how this could affect the cellular ribosome. We therefore assessed polysome gradients (Suppl. Figure 5) and measured 18S, 28S and 45S precursor 1 rRNAs in nuclear fractions of capsid expressing cells (data not shown), but found no obvious 2 phenotypic changes compared to cells expressing the 'empty' vector. Perhaps the effects of these 3 interactions on the ribosome are more subtle or only affect a small pool of 'specialised' ribosomes, 4 making changes difficult to detect. Since viruses rely on the host cell ribosome for translation of their 5 own genomes, a better understanding of the involvement of the viral proteins in recruiting or 6 potentially altering the host cell ribosomes through interaction with specific ribosomal proteins and 7 this novel set of ribosome biogenesis factors definitely warrants further investigation. 8 9

To obtain additional hints about possible functional consequences of the detected interactions 10 between SFV and host cell proteins, we used MCODE to analyse the protein complexes that could form 11 between all host interactors, including the ribosomal proteins (Figure 4 and Suppl. Figure 3 ). GO 12 enrichment analyses of the protein complexes reinforced many of the cellular processes discussed 13 above and revealed that the most highly enriched GO terms were related to translation and NMD 14 (Figure 4a and b) . As a counter defence strategy, viruses are known to inhibit cellular mRNA decay 15 factors that can degrade viral RNAs and restrict infection 32, 48, 49 . We therefore hypothesized and 16 decided to investigate, whether SFV was able to inhibit the NMD pathway, which has antiviral activity 17 against alphaviruses 33 . Indeed, we found that starting from 3-4 hours after infection, SFV antagonises 18 the NMD pathway, with consequent stabilisation of bona fide NMD mRNA transcripts (Figure 5c and  19 Suppl. Fig. 4) . Viruses known to inhibit mRNA decay pathways do so by different mechanisms. Often a 20 viral protein counteracts a key cellular regulator. Here, we show novel data that in the case of SFV, it 21 is the capsid protein that inhibits NMD (Figure 6b) . Therefore, inhibiting this function of the viral capsid 22 could lead to novel avenues for therapeutic intervention. 23 24 Using both SFV protein affinity purifications in transient expression experiments and UPF1 IPs in SFV 25 infected cells, we show that the core NMD factor UPF1 binds to SFV capsid among other SFV proteins 26 in an RNA-dependent manner. Together, this indicates that the capsid, and potentially other SFV 27 proteins, associate with mRNP molecules that also contain UPF1. The large number of ribosomal 28 proteins pulled down by capsid (Figures 2d, S3, and 4b) reported 50,51 . It was also postulated that the capsid or 'core' protein of Hepatitis C Virus (HCV) may 33 be responsible for the NMD inhibitory phenotype that was reported upon HCV infection 52 . Our findings therefore add SFV as the first alphavirus to a growing list of viruses of which the capsid protein 1 is responsible for an NMD inhibitory effect. Though the stability of the SFV genome has thus far been 2 attributed to evasion of deadenylation through binding to HuR 53 , the virus may require additional 3 strategies to protect itself in order to ensure efficient translation of viral genes and packaging of 4 genomes into new progeny viruses. Perhaps the SFV capsid plays a protective role against degradation 5 of its RNAs by NMD. 6 7 In summary, we present here two valuable resources that will aid in the study of SFV: a SFV-host 8 protein interactome as well as a genome-wide siRNA screen for host factors influencing SFV infection. The coding sequences for the SFV proteins were cloned into pcDNA5_FRT_TO_3xFlag(N), to yield 4 pcDNA5.3xFLAG-nsp1, pcDNA5.3xFLAG-nsp2, pcDNA5.3xFLAG-nsp3-ZSG, pcDNA5.3xFLAG-nsp4, 5 pcDNA5.3xFLAG-capsid and pcDNA5.3xFLAG-Env, which were used for all subsequent transfections. 6 pcDNA5_FRT_TO_3xFlag(N) was linearised using BamHI. PCR products for each of the SFV proteins, 7

were generated from either SFV-ZSG(-3'UTR) 33 , SFV-capsid or SFV-Envelope 58 plasmids using the 8 following primers: buffer and 500 μL of the cleared cell lysate was added, and the mixture was then incubated at 4 °C for 22 one hour with rotation. Beads were collected on a magnet and washed three times with lysis buffer. 23

At the third wash, each sample was split into two (for treatment or no treatment with RNase A). For 24 +RNase A samples, RNase A treatment was performed on the beads. The supernatant was removed 25 using the magnet, and 50 μL of RNase A (0.8 mg/mL, Sigma Aldrich) containing lysis buffer was added 26 to the beads and incubated at 25 °C for 15 minutes, shaking. Thereafter, the RNase A treated beads 27 were washed with lysis buffer, and then samples eluted from the beads using 3xFLAG peptide 28 (sciencepeptide.com): 20 μg of FLAG peptide was incubated with the beads at 25 °C for 15 minutes, 29

shaking. The FLAG peptide elution was also performed for the -RNase A samples. Thereafter, the 30 eluates were collected and 10 μL of loading buffer (4xLDS + DTT [75 mM]) was added to each eluate, 31 while 30 μL of loading buffer was added to the remaining beads samples. The samples were then 32 incubated at 75 °C for 10 minutes, ready for analysis by western blot, silver stain, and coomassie gel 33 for mass spectrometry sample preparation.

Samples were electrophoresed on 26-well NuPAGE™ 4-12 % Bis-Tris gradient gels (Thermo Fisher 3 Scientific) in 1xMOPS running buffer at 200V for approx. 1 hour. The gels were fixed in 50 % methanol 4 / 12 % acetic acid for one hour at room temperature, followed by three 5 minutes washes in 35 % 5 ethanol. The gels were sensitised for 2 minutes in 0.02 % sodium thiosulfate, followed by three 5 6 minutes washes in Milli-Q H2O. The gels were stained in 0.2 % silver nitrate / 0.03 % formaldehyde for 7 20 minutes, followed by two 1 minute washes in Milli-Q H2O. The gels were developed in 0.57 M 8 sodium carbonate + 0.02 % formaldehyde / 0.0004 % sodium thiosulfate and then incubated in 50 % 9 methanol / 12 % acetic acid for 5 minutes. Thereafter, the gels were placed in 1 % acetic acid for short 10 term storage at 4 °C. Images were taken using the gel documentation system (www.vilber.com) 11 12

The protein compositions of the eluates from the SFV affinity purifications were analysed by label-free 14 quantitative mass spectrometry. The eluates (20 μL) were electrophoresed in 1xMOPS running buffer 15 about 1 cm into the 26-well NuPAGE™ gels and then stained with coomassie-blue (10 % phosphoric 16 acid, 10 % ammonium sulfate, 0.12 % coomassie G-250, 20 % methanol) as described previously 59 . 17

Images were taken using the gel documentation system. Rectangular segments (10 mm x 3 mm) for 18 each lane were cut from the gels using sterile blades. The gel pieces were reduced, alkylated and 19 digested by trypsin as described elsewhere 60 to custom nsP1, nsP2, nsP3-Z, nsP4, capsid and Env sequences. The following MaxQuant settings were 32 used: separate normalisation groups for the +Rnase A and -Rnase A samples, mass deviation for 33 precursor ions of 10 ppm for the first search, maximum peptide mass of 6000Da, match between runs activated with a matching time window of 0.7 min only allowed across replicates; cleavage rule was 1 set to strict trypsin, allowing for 3 missed cleavages; allowed modifications were fixed 2 carbamidomethylation of cysteines, variable oxidation of methionines, deamination of asparagines 3 and glutamines and acetylation of protein N-termini. Protein intensities are reported as MaxQuant's 4 iTop3 63 values (sum of the intensities of the three most intense peptides). Peptide intensities were 5 first normalised by variance stabilisation normalisation and imputed. Imputation values were drawn 6 from a Gaussian distribution of width 0.3 centred at the sample distribution mean minus 1.8x the 7 sample standard deviation, provided there were at least 2 evidences in the replicate group. In order 8 to perform statistical tests, iTop3 values were further imputed at the protein level, following the rule: 9

'if at least two detections in at least one group' and using the following protein impute parameters: 10 imputation values were drawn from a Gaussian distribution of width 0.3 centred at the sample 11 distribution mean minus 2.5x the sample standard deviation. Potential contaminants and proteins 12 marked 'only identified by site' were removed prior to performing Differential Expression (DE) tests, 13 which were done by applying the empirical Bayes test 64 followed by the FDR-controlled Benjamini and 14

Hochberg 65 correction for multiple testing. A significance curve was calculated based on a minimal 15 log2 fold change of 1 and a maximum adjusted p-value of 0.05. Proteins that were consistently 16 reported as DE throughout 20 imputation cycles were flagged as "persistent". In addition, SAINT 17 analysis was performed according to 66 and a significance threshold of FDR<0.05 was applied. 18

Significant interactors for each SFV bait protein were defined as those that were significantly 21 differentially expressed (enriched) compared to the untransfected control and the significance 22 persisted throughout the imputation cycles. In addition, the proteins taken as "significant interactors" 23 had to be considered "true interactors" as determined using SAINT analysis (with a threshold FDR of 24 ≤ 0.05). To simplify the lists and attempting to retain potentially biologically relevant interactors, we 25 further filtered the lists, retaining proteins whose abundance made up at least 0.5 % of the relevant 26 bait protein. In the case of the nsp3-Z bait, which was very lowly abundant in the sample as it proved 27 difficult to elute from the beads, we retained proteins whose abundance made up at least 5 % of the 28 SFV bait. 29 30

The final lists of proteins were used to create SFV-host protein interaction networks using 32 Cytoscape_v3.7.2. STRING analysis (string-db.org, version 11.0) was performed using a minimum 33 required interaction score of highest confidence (0.900) for both database and experimental evidence and the known protein-protein interactions were overlaid onto the SFV-protein interactome 1 networks, using Cytoscape_v3.7.2. were infected for 6 hours with SFV-ZsG at a concentration giving an infection rate of approx. 30 % in 10 control siRNA-treated cells. Following fixation in 4 % formaldehyde and Hoechst staining, nine images 11 per well were acquired using high-content automated fluorescence microscopes (ImageX-press, 12

Molecular Devices). Infected cells were detected using Cell Profiler (www.cellprofiler.org) and 13

Advanced Cell Classifier (www.acc.ethz.ch/acc.html) and the % of infected cells per well was 14

determined. An "Infection Index" value was calculated for each gene, indicating the fold change of 15 infection upon depletion of the gene product, compared to the control siRNAs, which was set as 1. An 16 Infection Index threshold of 1.3 was chosen to indicate proteins having a potential antiviral role against 17 SFV, and an Infection Index threshold of 0.5 to indicate proteins having a potential proviral role for 18 SFV ( Table 2 , raw data). The proteins identified in the siRNA screen were overlaid with the SFV-host 19 protein interaction networks. 20 21

Protein-protein interaction enrichment analysis was performed using the Metascape online tool 23 (www.metascape.org) according to 67 . Metascape allows for the input of multigene lists. For each 24

given list (ie: interactors of nsp1, nsp2, nsp3-Z, nsp4, capsid and Env) as well as for the merged list, 25 protein-protein interaction enrichment analysis was carried out using the following databases: identified for the merged list of interactors. GO / pathway and process enrichment analysis was carried 1 out with the following ontology sources: KEGG Pathway, GO Biological Processes, Reactome Gene 2 Sets, Canonical Pathways and CORUM, using the Metascape tool. All genes in the genome were used 3 as the enrichment background. From the top10 most significant GO term descriptions gathered for 4 each MCODE network (Table 3 , Supplementary Information) , between 1-4 terms were chosen to 5 assign biological meaning to each MCODE term (Figure 5a ). The log(q-value) of the terms indicates the 6 significance calculated using the merged list of interactors (Table 3, 

Total RNA was extracted from TRI-reagent by isopropanol precipitation and resuspended in disodium 5 citrate buffer (pH 6.5). Contaminating DNA was degraded by treatment with Turbo DNase (Ambion). 6

Thereafter, reverse transcription was carried out using AffinityScript Multiple Temperature Reverse 7

Transcriptase (Agilent), followed by qPCR using Brilliant III Ultra-Fast SYBR® Green qPCR Master Mix 8 (Agilent), according to manufacturer's instructions. The following primers were used: 9

Hnrnpl_PROT 5'-CAATCTCAGTGGACAAGGTG -3' for one hour at 4 °C. The coupled beads were washed three times with lysis buffer and then 500 μL of 16 the cleared cell lysate was added to the beads pellet. The cleared lysates and coupled beads were 17 incubated at 4 °C for one hour with rotation. The beads were collected on a magnet and washed three 18 times with lysis buffer. At the third wash, each sample was split into two (for treatment or no 19 treatment with RNaseA). The supernatant was removed and the beads were resuspended with 20 µL 20 lysis buffer + 10 µL loading buffer (4x LDS+DTT). The samples were incubated at 75 °C for 10 minutes. 21

Using the magnet, the supernatants (eluates) were transferred to new tubes, ready to load onto gels 22 for coomassie staining and mass spectrometry sample preparation. The compositions of the eluates 23 were quantified by mass spectrometry, following the same protocol as above. Triplicate samples were 24 processed against the same sequence database as above, by Transproteomics pipeline (TPP) 68 tools. 25

Four database search engines were used (Comet 69 , Xtandem 70 , MSGF 71 and Myrimatch 72 , with search 26 parameters as above. Each search was followed by the application of the PeptideProphet 73 tool; the 27 iprophet 74 tool was then used to combine the search results, which were filtered at the false discovery 28 rate of 0.01; furthermore, the identification was only accepted if at least two of the search engines 29 agreed on the identification. Protein inference was performed with ProteinProphet 75 . For those 30 protein groups accepted by a false discovery rate filter of 0.01, a Normalized Spectral Abundance 31 Factor (NSAF) 76 was calculated based on the peptide to spectrum match count; shared peptides were 32 accounted for by the method of 77 , giving normalised spectral abundance factor (dNSAF) values. The dNSAF values were used to calculate abundance as a % of the UPF1 bait protein, which were reported 1 for the NMD factors and SFV proteins that were detected in the samples. 2 3 Small-scale SFV protein expression and puromycin incorporation assays 4 1.5x10 5 HeLa cells per well were seeded into 6-well plates. The following day, 1.25 μg of the relevant 5 pcDNA5.3xFLAG plasmids were transfected using Dogtor transfection reagent (OZ Biosciences), as per 6 manufacturer's recommendations. Cells were harvested 48 hours later, after which 1x10 6 cells per 7 condition were collected to make lysates for western blot analysis, while the remaining cells were 8 collected for RNA analysis. The cells were harvested at 4 °C for 5 minutes at 250 x g, washed once with 9

1x PBS, and resuspended in either 2x SDS-PAGE Loading Buffer (for protein analysis) or 900 μL of TRI 10 reagent (for RNA analysis). For the puromycin incorporation assays, cells were transfected as above 11 and prior to harvesting, the medium was aspirated and medium containing either DMSO or 100 µg/mL 12 cycloheximide (CHX) was added to the cells and incubated for 2 hours at 37 °C. Thereafter, the cells 13 were pulse labelled for 10 minutes with puromycin (10 μg/mL). After the 10 minutes pulse, as a 14 recovery step, medium containing either DMSO or CHX was re-added to the cells for 30 minutes at 37 15 °C. Cells were harvested as above and 1x10 6 cells per condition were resuspended in 2x SDS-PAGE 16 loading buffer and incubated at 95 °C for 5 minutes for western blot analyses. 17

18

Lysates were loaded into either 10 % SDS-PAGE gels or pre-casted 4-12 % Bis-Tris 26-well NuPAGE™ 20 or 10-well Bolt™ gels (Invitrogen, Thermo Fisher Scientific) and electrophoresed in either 1x SDS-PAGE 21 running buffer or 1x MOPS running buffer. Proteins were transferred onto nitrocellulose membranes 22 using either the BioRAD Semi-Dry Blot system (30 minutes in 1xBjerrum buffer + 20 % methanol) for 23 SDS-PAGE gels or the iBlot2® (P0, 7mins, using iBlot® 2NC regular stacks) (Invitrogen, Thermo Fisher 24 Scientific) for pre-casted gels. Membranes were blocked for 1 hour at room temperature (RT) in 5 % 25 milk-TBS-T (0.1 % Tween20), or in the case of Rb anti-eIF2α, Ms anti-eIF2α and Rb anti-p-eIF2α blots, 26 5 % milk-TBS, and then incubated with the indicated primary antibodies at 4°C overnight. Primary 27 antibodies were diluted in 5 % milk-TBS-T, apart from Rb anti-eIF2α, Ms anti-eIF2α and Rb anti-p-28 eIF2α, which were diluted in 5 % BSA-TBS-T (0.1 % Tween20). After three 10 minutes washes in TBS-29 T, membranes were incubated with the indicated secondary antibodies in 5 % milk-TBS-T for 1.5 hours 30 at RT, followed by three 10 minutes TBS-T washes. The blots were then visualised using the Odyssey 31 System (LICOR). 32

The protocol for polysome fractionations was adapted from 78 . HeLa cells were transfected with the 2 'empty' (pcDNA5_FRT_TO_3xFlag) or 'capsid' (pcDNA5.3xFLAG-capsid) plasmids as described for the 3 large scale transfections above. The cells were treated with 100 µg/mL CHX for 4 mins at 37 °C prior 4 to harvesting. Cells were washed once with ice-cold PBS containing 100 µg/mL CHX, scraped off the 5 surface of the dish with 750 µL per 15cm dish of PBS-CHX and then transferred into tubes. Cells were 6 collected by centrifugation at 4 °C for 5 minutes at 500 x g and lysed in 600 µL of lysis buffer (10 mM 7

Tris-HCl [pH 7.5], 10 mM NaCl, 10 mM MgCl2, 1 % Triton X-100, 1 % sodium deoxycholate, Red asterisks indicate 3xFLAG tagged SFV proteins (or the 3xFLAG alone) at respective sizes. 'Untr' refers to an 3 untransfected control condition that underwent the affinity purification procedure, while 'empty' denotes transfection with 4 a plasmid construct containing only the 3xFLAG tag (with no additional coding region) followed by affinity purification. The 5 expected sizes of the proteins (3xFLAG included) were: empty ~8kDa; nsp1 ~63kDa; nsp2 ~92kDa; nsp3-Z ~82kDa; nsp4 6 ~72kDa; capsid ~33kDa and Env ~111kDa. The left side of the WBs indicate the Eluates (after flag peptide elution from the 7 beads) of each SFV affinity purification, which were sent for mass spectrometry analysis. The right hand side of the WBs 8 indicate the SFV proteins that were still present on the beads after the flag peptide elution step. identified as interactors to more than one of the SFV proteins. In these cases, the solid grey lines connect the grouped set of 5 host proteins to the SFV proteins for which they were identified as interactors. Host-host PPI ascertained through STRING 

Mosquito-3 borne arboviruses of African origin: review of key viruses and vectors

Arthritogenic Alphaviruses: A Worldwide Emerging Threat?

The molecular pathogenesis of Semliki Forest 9 virus: a model virus made useful?

Semliki forest virus: cause of a fatal case of human encephalitis

An Outbreak of Human Semliki Forest Virus Infections in Central African 13

Isolation of a 15 newly recognized alphavirus from mosquitoes in Vietnam and evidence for human infection 16 and disease

Me Tri virus: a Semliki Forest virus strain from Vietnam?

Microtubule-Dependent Transport of Semliki Forest Virus Replication Complexes from the

Plasma Membrane to Modified Lysosomes

Alphavirus Infection: Host Cell Shut-Off and Inhibition of Antiviral 23 Responses

Polyprotein Processing as a 25 Determinant for in Vitro Activity of Semliki Forest Virus Replicase

The alphaviruses: gene expression, replication, and evolution

Semliki Forest virus-specific non-structural polyprotein by nsP2 protease

The eIF2α kinases: their structures and 32 functions

Eukaryotic aspects of translation initiation brought into focus

Independent Mechanisms Are Involved in Translational Shutoff during Sindbis Virus Infection

Alphavirus RNA synthesis and non-5 structural protein functions

Magnetic Fractionation and Proteomic Dissection of Cellular Organelles

Occupied by the Late Replication Complexes of Semliki Forest Virus

Novel Functions of the Alphavirus Nonstructural Protein 10 nsP3 C-Terminal Region

Induces Filopodia and Rearrangement of Actin Filaments

Nuclear localization of Semliki Forest 14 virus-specific nonstructural protein nsP2

Evasion of the Innate Immune Response: the Old 16

World Alphavirus nsP2 Protein Induces Rapid Degradation of Rpb1, a Catalytic Subunit of RNA 17

A structural and functional perspective of alphavirus 19 replication and assembly

Translation of Sindbis Virus mRNA: Effects of Sequences 21 Downstream of the Initiating Codon

Translation of Sindbis virus mRNA: analysis of sequences 23 downstream of the initiating AUG codon that enhance translation

Alphavirus vectors and vaccination

The Regulation of Translation in Alphavirus-28 Infected Cells

The Alphavirus Exit Pathway: What We Know and What 30

We Wish We Knew

The C-Terminal

Old World Alphaviruses Bind Directly to G3BP

Viral and Cellular Proteins Containing FGDF Motifs Bind G3BP to Block 38

Structure and Function of a protein folding machine: the eukaryotic cytosolic chaperonin CCT

Function and Regulation of Cytosolic Molecular Chaperone CCT

hnRNPs Relocalize to the Cytoplasm following 23 Infection with Vesicular Stomatitis Virus

Exploitation of nuclear functions by 25 human rhinovirus, a cytoplasmic RNA virus

Mapping of Chikungunya Virus Interactions with Host Proteins Identified 2 nsP2 as a Highly Connected Viral Component

Karyophilic properties of Semliki Forest virus nucleocapsid protein

Cytoplasmic Viruses: Rage against the (Cellular RNA Decay) Machine

The Role of Stress Granules and the

Nonsense-mediated mRNA Decay Pathway in Antiviral Defence

Interplay between 11 coronavirus, a cytoplasmic RNA virus, and nonsense-mediated mRNA decay pathway

The Cellular NMD Pathway Restricts Zika Virus Infection and Is Targeted 14 by the Viral Capsid Protein

A Combined Proteomics/Genomics Approach Links Hepatitis C Virus 16 Infection with Nonsense-Mediated mRNA Decay

Sindbis Virus Usurps the Cellular HuR Protein to Stabilize Its Transcripts 18 and Promote Productive Infections in Mammalian and Mosquito Cells

Tracking and Elucidating Alphavirus -Host Protein Interactions

New World and Old World Alphaviruses Have Evolved to Exploit Different 23 Components of Stress Granules, FXR and G3BP Proteins, for Assembly of Viral Replication 24

Eastern Equine Encephalitis Virus nsP3 Redundantly Utilizes Multiple Cellular Proteins for 27

New World alphavirus 29 protein interactomes from a therapeutic perspective

Two-Helper RNA System for Production of Recombinant Semliki 31

Blue silver: A very sensitive colloidal Coomassie G-250 staining for 33 proteome analysis

Proteome 1 remodelling during development from blood to insect-form Trypanosoma brucei quantified 2 by SILAC and mass spectrometry

MaxQuant enables high peptide identification rates, individualized p.p.b.-4 range mass accuracies and proteome-wide protein quantification

UniProt: a worldwide hub of protein knowledge

Detecting significant changes in protein 11 abundance

Controlling the False Discovery Rate: A Practical and Powerful 13 Approach to Multiple Testing

Analyzing Protein-Protein Interactions from Affinity Purification-Mass 15

Metascape provides a biologist-oriented resource for the analysis of systems-17 level datasets

A guided tour of the Trans-Proteomic Pipeline

A Deeper Look into Comet-Implementation and Features

A method for reducing the time required to match protein sequences 23 with tandem mass spectra

MS-GF+ makes progress towards a universal database search tool for 25 proteomics

Highly Accurate Tandem Mass

Spectral Peptide Identification by Multivariate Hypergeometric Analysis

Statistical Validation of Peptide Identifications in Large-30

Scale Proteomics Using the Target-Decoy Database Search Strategy and Flexible Mixture 31

Multi-level Integrative Analysis of Shotgun Proteomic Data

Improves Peptide and Protein Identification Rates and Error Estimates

Interpretation of Shotgun Proteomic Data

Quantitative shotgun proteomics using a 4 protease with broad specificity and normalized spectral abundance factors

Quantitation: How to Deal with Peptides Shared by Multiple Proteins

Studying the Translatome with Polysome Profiling

Methods in Molecular Biology