key: cord-0810337-wk1np33o
authors: Kruse, Thomas; Benz, Caroline; Garvanska, Dimitriya H.; Lindqvist, Richard; Mihalic, Filip; Coscia, Fabian; Inturi, Ravi Teja; Sayadi, Ahmed; Simonetti, Leandro; Nilsson, Emma; Ali, Muhammad; Kliche, Johanna; Morro, Ainhoa Moliner; Mund, Andreas; Andersson, Eva; McInerney, Gerald; Mann, Matthias; Jemth, Per; Davey, Norman E; Överby, Anna K; Nilsson, Jakob; Ivarsson, Ylva
title: Large scale discovery of coronavirus-host factor protein interaction motifs reveals SARS-CoV-2 specific mechanisms and vulnerabilities
date: 2021-04-19
journal: bioRxiv
DOI: 10.1101/2021.04.19.440086
sha: 2fdbbfc814b50f4a2d9f66fa8af75b7f868bd990
doc_id: 810337
cord_uid: wk1np33o

Viral proteins make extensive use of short peptide interaction motifs to hijack cellular host factors. However, current methods do not identify this important class of protein-protein interactions. Uncovering peptide mediated interactions provides both a molecular understanding of viral interactions with their host and the foundation for developing novel antiviral reagents. Here we describe a scalable viral peptide discovery approach covering 229 RNA viruses that provides high resolution information on direct virus-host interactions. We identify 269 peptide-based interactions for 18 coronaviruses including a specific interaction between the human G3BP1/2 proteins and an ΦxFG peptide motif in the SARS-CoV-2 nucleocapsid (N) protein. This interaction supports viral replication and through its ΦxFG motif N rewires the G3BP1/2 interactome to disrupt stress granules. A peptide-based inhibitor disrupting the G3BP1/2-N interaction blocks SARS-CoV-2 infection showing that our results can be directly translated into novel specific antiviral reagents.

RNA viruses such as the Ebola, dengue and coronaviruses cause a variety of diseases and constitute a continuous threat to public health. The coronaviruses are the largest single stranded RNA viruses known and their genomic RNA encodes around 30 viral proteins 1 . During infection each viral protein performs unique functions and interacts with a range of cellular protein host factors to allow viral proliferation and immune escape [2] [3] [4] [5] . Precise disruption of viral-host factor interactions is an attractive strategy for developing novel antiviral reagents.

The advantage of targeting these interactions is that resistance is less likely to develop and furthermore as the same host factor can be used by multiple viruses such reagents may provide broader spectrum activity. For this reason, numerous large scale mass spectrometry-based interaction screens 2,3,5 as well as CRISPR based screening approaches [6] [7] [8] [9] [10] have been used to uncover host factor interactions and dependencies for SARS-CoV-2 allowing repurposing of drugs 11, 12 . Although these methods have been transformative in our understanding of SARS-CoV-2 biology the molecular detail provided by these methods is not always sufficient to readily transform the results into novel antiviral reagents. Experimental approaches that would complement the existing powerful methods and provide a more detailed view of viral interactions with host factors could accelerate development of new antivirals.

An attractive class of protein interactions that can be inhibited for therapeutic purposes are viral short linear interaction motifs (SLiMs) that bind to defined pockets on globular domains of the host factor 13, 14 . SLiMs are short peptide motifs in unstructured regions of proteins and contain 2-3 amino acid binding determinants within a 10 amino acid stretch 15, 16 . Viruses extensively use SLiMs to hijack cellular host factors and SLiMs can readily evolve through mutations in unstructured regions allowing viruses to interact with novel host factors [17] [18] [19] .

Despite the importance of SLiMs for understanding viral biology they are not uncovered by current methods 15, 16 . Proteomic peptide-phage display (ProP-PD) provides the opportunity to identify novel SLiM-based interactions and binding sites at amino acid resolution 20 . As shown in a small scale pilot study on C-terminal peptides of viral proteomes, it can be used to faithfully capture SLiM-based host-pathogen interactions 21 . Here we describe a novel phage-based viral peptide library to map SLiMs from 229 RNA viruses (Riboviria) mediating host factor interactions (Fig. 1A) . This approach allows the simultaneous pan-viral identification of SLiMbased interactions with amino acid resolution of the binding sites. We document the power of this approach by identifying novel SARS-CoV-2 specific SLiM mediated host factor interactions and directly translate our screening results into novel mechanistic insights and pinpoint a potential target for antiviral intervention.

Exploiting recent developments in proteomic phage display (ProP-PD) technology 20, 22 to identify RNA virus SLiMs binding to host factors. We designed a unique phage display library (RiboVD library) that tile the unstructured regions of 1074 viral proteins from 229 RNA viruses including SARS-CoV-2, SARS-CoV, MERS-CoV and 19 additional coronavirus strains. This library represents 19,549 unique 16 amino acid long peptides that are multivalently displayed on the major coat protein of the M13 phage. Given the recurrent pandemic potential of coronaviruses we used this library to understand interactions for this class of virus but note that it can be used for any RNA virus. We scanned published host factor interactomes for the SARS-CoV-2 viral proteins and recombinantly produced 57 domains from 53 cellular proteins reported to interact with SARS-CoV-2 2,3,5 . As transient SLiM based protein interactions might be lost during purifications of viral proteins for subsequent mass spectrometry analysis, we screened an additional set of 82 peptide binding domains, of which at least 27 have previously been reported to act as viral host factors and to be hijacked by SLiMs from viral proteins 23 . In total, 139 recombinantly expressed and purified human bait proteins (Table S1 ) were used in selections against the RiboVD library. Enriched phage pools were analyzed by next generation sequencing (NGS) to identify viral peptides that bound to the bait. This uncovered 269 putative SLiM-based interactions between 44 human protein domains (42 proteins) and 64 viral proteins from 18 coronavirus strains (Table S2) . Of these, 117 (43%) interaction pairs involved human coronavirus proteins. We validated 27 out of 27 tested interactions using fluorescence polarization (FP) affinity measurements (Fig 1B, Fig S1, Table S3 ). We visualized the information generated for human and bat coronavirus proteins in an extensive network ( Fig   S2) . We also generated a map of the viral proteins mediating SARS-CoV-2, SARS-CoV and MERS-CoV interactions with human host factors (Fig 1C) . The map reveals common as well as unique interactions with host factors for these three coronaviruses. For instance, NSP14 of all three strains has a YxxL motif that binds to the clathrin coat adaptor protein AP2M1 with high affinity (Fig 1B; Fig S1, Table S3 ), which may be linked to trafficking of the viral protein or blocking of endocytosis of host proteins 24, 25 . The N-terminal region of the E protein from all three strains binds to the FERM domains of Ezrin and Radixin via a recently established [FY] x[FILV] SLiM 26 . Interestingly, our data show that the FERM domains also bind to NSP3 of SARS-CoV and SARS-CoV-2, thus, they can be targeted by distinct viral proteins. The SARS-CoV NSP3 FERM binding site overlaps with a [FWY]xx [ILV] binding site for the ATG8 domains of the autophagy related MAP1LC3A-C proteins. As an example of strain specific interactions, we found that an N-terminal peptide from the Nucleocapsid (N) proteins from SARS-CoV-2 and SARS-CoV bound to the NTF2 domain of the homologous G3BP1 and G3BP2 proteins (G3BPs) with high affinity (Fig 1B-C and S1, Table S3 ). This N peptide contains an FxFG SLiM (where F is a hydrophobic residue) that resembles motifs in USP10 and UBAP2L and in the alphavirus nsP3 protein known to bind a hydrophobic pocket in the NTF2 domain of G3BP 27-30 . The FxFG SLiM is also present in the N proteins from bat betacoronaviruses and consistently the corresponding bat HKU5 peptide was identified in our screen ( Fig S2; Table S2 ).

To pinpoint therapeutically relevant host protein-viral SLiM interactions we screened three of the identified peptide motifs for antiviral activity. To this end, we generated lentiviral vectors expressing GFP fused to four copies of one viral SLiM reasoning that this would inhibit binding of the corresponding full-length SARS-CoV-2 protein to the specific host factors through competition. As a control we used GFP fused to SLiMs containing mutations in the binding motif. The host proteins targeted by viral peptides were G3BPs (SARS-CoV-2 N), Ezrin and Radixin (SARS-CoV-2 E and NSP3), and the MAP1LC3s (NSP3). VeroE6 cells were first transduced with the lentiviruses and 3 days later infected with SARS-CoV-2 and viral titer determined after 16 hours. This revealed that the G3BP-binding peptide from the N protein decreased viral titer 3.4-fold (Fig 2A) . To obtain a more potent inhibition of the SARS-CoV-2 N-G3BP interaction, we used a 25 amino acid residue peptide from Semliki Forest vius (SFV) nsP3 containing two continuous FGDF like SLiMs that has previously been shown to bind G3PBs with high affinity 31 . Remarkably, this peptide binds approximately 10-fold stronger than the SARS-CoV-2 N peptide to both G3BP1 and G3BP2 (Kd= 4 µM vs Kd= 0.3 µM, Fig   2B and S3A ). We constructed "G3BP inhibitors (G3BPi)" by fusing sequences encoding one or three copies of wild type (wt) or mutated (ctrl) SFV nsP3 SLiMs to GFP. As expected, mass spectrometry analysis confirmed that the major cellular targets of the G3BPi are the G3BPs (Fig 2C, S3B and Table S4 ). Furthermore, expression of the G3BPi wt but not G3BPi ctrl prevented binding of SARS-CoV-2 N to G3BP1 in cells (Fig 2D) . Consistent with these binding and competition data, lentiviral mediated expression of the G3BPi in VeroE6 cells potently inhibited SARS-CoV-2 proliferation after 16 hours of infection (Fig 2A) . An effect of the G3BPi was also evident in assays monitoring viral infection rates or replication (Fig 2E-F) .

In a cell based transfection assay monitoring assembly and release of virus-like particles mutating the G3BP binding motif in N had no effect ( Fig S3D) .

Thus, the approach presented here is useful for identifying important virus-host factor interactions which inhibit viral proliferation when disrupted.

The above results prompted us to further investigate the N-G3BP interaction and its function during infection. The coronavirus N protein is important for viral replication as well as packaging of the viral RNA 32-34 . The G3BPs are multi-functional RNA-binding proteins best known for their essential roles in innate immune signaling and the assembly and dynamics of cytosolic stress granules 35-38 . Stress granules are large protein-RNA assemblies formed in response to various stresses and viral infections 39-41 . Perhaps unsurprisingly, the G3BPs have turned out to be major targets for viral interference and several viral proteins have been shown to recruit G3BP1 to support viral replication and/or to inhibit stress granules formation 42 . Of note, the herpesviruses and alphaviruses have been shown to recruit G3BPs by SLiMs having resemblance to the sequence in N 28,43-45 . However, a deeper mechanistic understanding for how viral proteins affect G3BP biology is missing. Given that the N-G3BP interaction was important for SARS-CoV-2 infection and presents a novel antiviral strategy we investigated this interaction in more detail. We first confirmed that the interaction between N and G3BP1 takes place in SARS-CoV-2 infected cells ( Fig 2G) . We also confirmed binding of recombinant full length N protein to G3BP1 using FP, which revealed an affinity similar to the N peptide ( Fig 2H) . To confirm the SLiM mediated interaction in cells, we compared the interactome of N wild type (N wt) to an N protein where we mutated two amino acids in the FxFG motif (N 2A) using label free quantitative mass spectrometry. This confirmed a highly specific N-G3BP1/2 interaction fully dependent on an intact FxFG motif ( Fig. 2I and S3C, Table S4 ).

Using a similar approach, we quantitatively compared the interactomes of the N protein from MERS-CoV with that of N from SARS-CoV-2, revealing specific binding of G3BPs to N from SARS-CoV-2 ( Fig 2J, Table S4 ). This is in line with our observation that the N peptides containing the FxFG motif from SARS-CoV-2 and SARS-CoV bind to G3BPs with high affinity, while the corresponding MERS-CoV peptide bound weakly (Kd= 2.8 µM vs Kd= 26 µM, Fig 1B) . Similar results were obtained by western blot, which also showed that the N protein from HKU1-CoV did not bind G3BPs, consistent with the ProP-PD results ( Fig. S2 and   S3E ). Consistently, the FxFG motif resulted in the specific co-localization of mCherry tagged N protein from SARS-CoV and SARS-CoV-2 to arsenite induced stress granules in cells expressing YFP tagged G3BP1 (Fig 3A) .

Recent publications have reported that SARS-CoV-2 N induces stress granule disassembly 46,47 . but the mechanistic basis of this is unclear. To investigate the effect of the SARS-CoV-2 N FxFG motif on endogenous stress granule formation we overexpressed YFP tagged N wt or N 2A in HeLa cells and stained for endogenous G3BP1 following arsenite treatment. Quantifying the intensity of cytoplasmic G3BP1 foci in cells positive for YFP revealed that N WT expression disrupted stress granule formation more efficiently when the G3BP binding motif was intact (Fig 3B-C) . Thus, the N-terminal FxFG motif of the N protein constitutes the main determinant of G3BP binding and stress granule disassembly. We next analyzed G3BP1 foci formation and cellular localization of viral dsRNA in relation to N protein expression levels in VeroE6 cells after six hours of SARS-CoV-2 infection (Fig 3D-E) . At this timepoint, a mixture of early and later stage infected cells is observed. In mock treated cells we detected no cells with G3BP1 foci while infected cells with low levels (early-stage infection) of N protein had G3BP1 foci in about 60% of the cells (Fig 3E) . In cells with low levels of N, this protein and viral dsRNA co-localized with G3BP1 to stress granules ( Fig 3D) . However, in cells with high levels of N protein (late-stage infection) we detected no G3BP1 foci.

Collectively our results suggest that during early stages of infections the levels of N protein are insufficient to disrupt stress granule formation and instead N and viral dsRNA co-localize with G3BP1 in these structures. Once N concentrations reaches a certain threshold, this disrupts stress granules, and this depends on the FxFG motif. A possible interpretation of these observations is that during the earlier stages of infection SARS-CoV-2 takes advantage of the stress granule RNA machinery. Consistently, dsRNA and N co-localize with G3BP1 foci and when the N-G3BP interaction is inhibited a reduction of viral replication is observed (Fig 2F   and 3D ).

To understand how N could affect stress granule formation and G3BP function through the FxFG motif we set out to identify cellular G3BP interactors with similar binding motifs. To this end, we screened a novel ProP-PD library that displays the intrinsically disordered regions of the human proteome 22 against the NTF2 domains from G3BP1 and G3BP2. The combined data set includes 72 peptides from 57 proteins with the majority of sequences containing a FxFG motif (F: [FILV] ), thus resembling the sequence in the N protein ( Fig 4A-B , Table S5 ).

Nineteen of the proteins uncovered by the screen are in core stress granule proteins -including known peptide motifs in USP10 and UBAP2L, but also peptides from stress granule proteins that have not previously been reported to contain FxFG motifs. The screen also uncovered a peptide from Caprin-1 which has been shown to bind G3BPs but does not match the consensus sequence 48 (Fig. 4A-B ). This suggests that G3BPs serve as major hubs for stress granule biology in part by interacting with FxFG like motifs residing in several stress granule components. However, the screen also returned many peptides in proteins with roles outside of stress granule biology such as TRIM25 and IRF7 (anti-viral interferon signaling) 42 , and DDIT3 (endoplasmic reticulum stress) 49 . FP affinity measurements were used to confirm binding between the purified NTF2 domain of G3BP2 and several identified peptides originating from TRIM25, DDIT3, UBAP2L, Caprin-1, USP10 and PRRC2B ( Fig S4A) . Furthermore, we biochemically validated a number of the G3BP binding motifs in the context of the full-length proteins ( Fig S4B) . Caprin-1 and UBAP2L co-localized with G3BP1 in stress granules after arsenite treatment in a manner dependent on intact SLiMs ( Fig S4C) . Conversely, no stress granule localization was observed for TRIM25 and DDIT3 (Fig S4C) further supporting the notion that the G3BPs also have cellular roles beyond stress granule biology 50 .

The N protein is a highly expressed viral protein 51 during infection so we hypothesized that it would compete with host cell proteins containing FxFG SLiMs for binding to G3BPs.

Consistently, FP affinity measurements revealed competition between the N FxFG peptide and all of the 7 peptides we tested for interaction with G3BP2 ( Fig 4C) . Next, immunopurifications of full length YFP-tagged TRIM25, DDIT3, Caprin-1 and UBAP2L in the presence of either a N wt peptide or a N 3A peptide where the FxFG motif is mutated to AxAA were performed.

As expected, the N wt peptide disrupted interactions to G3BP1 thus validating a direct competition between the N FxFG peptide motif and four G3BP binding proteins ( Fig 4D) . The observed competition between the viral N FxFG peptide and UBAP2L for binding to G3BP1 is particularly interesting since UBAP2L is required for stress granule assembly through a direct interaction to the G3BPs via its FxFG like motif 27,29 . This suggests a mechanistic basis for the ability of the N protein to inhibit stress granule formation.

Given the high levels of N we speculated that it could mediate a general rewiring of the G3BP interactome through its FxFG motif. To test this on a global scale, we purified G3BP1-YFP from HeLa cells and added either N wt or the N 3A mutated peptide as competitors for cellular proteins. Quantitative label free mass spectrometry allowed us to determine the proteins being specifically displaced by the N wt peptide (Fig 4E and Table S4 ). This revealed specific displacement of 59 proteins including several core stress granule components. In addition, the N peptide also displaced a large number of nuclear pore complex components, heat shock chaperones of the Hsp70 family and proteins of the ASC-1 and CTLH complexes. Except the CTLH components all of these proteins have been reported to localize to stress granules 27,52,53 .

The displacement of nucleoporins from G3BP1 by the N peptide suggests that FG motifs, which are abundant in nucleoporins 54 might recruit them to stress granules through direct interaction to the NTF2 domain of the G3BPs. Consistently, two FxFG like motifs from nucleoporins were selected in the G3BP ProP-PD screen ( Figure 4A -B; Table S5 ). Importantly, Nup62, which we identify in our MS competition screen ( Fig 4E) has been shown to be required for efficient SARS-CoV-2 infection 3 . It is possible that N displaces Nup62 from G3BPs to make it accessible for other viral processes. Together, we show that the N protein modulates the G3BP1/2 host interactome through its FxFG motif by competing with numerous cellular FxFG containing proteins. Our G3BP motif and mass spectrometric screens provide a rich resource for the future dissection of basic stress granule biology and G3BP signaling in general.

CoV-2 N and the G3BP proteins. Our results reveal that the N protein during infection hijacks G3BPs to viral replication centers likely to facilitate replication and possibly other aspects of viral RNA metabolism. The disruption of stress granules at later stages of infection could also dampen the cellular anti-viral response. Consistent with this idea we identify FxFG motifs in TRIM25, MEX3C and IRF7 that are key components of the G3BP-RIG-1 antiviral interferon pathway 42,55-57 .

By screening the intrinsically disordered regions of 229 RNA viruses against a host factor in one go our ProP-PD screening approach uncovers both common principles shared by several viruses as well as interactions specific for a given strain. We show that the SLiMs can be screened for antiviral activity to pinpoint therapeutically relevant interactions. Given the amino acid resolution provided by the ProP-PD this can guide the development of agents targeting these interactions specifically. Our approach is scalable and easily applied to other relevant host factors and the library readily updated to incorporate novel RNA viruses emerging in the future. We thus foresee that this approach can be a powerful tool for future investigations of virus interactions with cellular host factors and for developing novel antivirals. 

JN is on the scientific advisory board for Orion Pharma.

The other authors declare no conflict of interest. 

The RiboVD library was designed using a previously described design 

Phage display selections against the immobilized bait proteins were performed in triplicate selections for four rounds of selection following a protocol described in detail elsewhere 26 . The

RiboVD library constructed in this study was used in selections against 139 human bait proteins purified (Table S1 ). The NTF2 domains of G3BP1 and G3BP2 were further used in To assess the quality of RiboVD library we analyzed the coverage percentage of the phage library over the library design using the NGS results of non-challenged naive library phage aliquotes. 95.50% of the peptide sequences designed to be in the library were confirmed by the NGS analysis of the naïve phage library.

Following a recently outlined ProP-PD data analysis approach 22 four metrics were used to rank the peptides i) NGS read counts, ii) peptide occurrence in replicated selections, iii) number of overlapped peptides, iv) motif match. The four metrics were then combined into a single score called 'Confidence level' forming 3 categories: high (4 metric criteria matched); medium (2 to 3 criteria matched); and low (only 1 metric is matched). Due to the relatively small size of the RiboVD library, the coronavirus dataset was further filtered for target specific ligands (occurring in less than 10 unrelated selections) ( Table S2 ). The G3BPs HD2 P8 data was combined and a joint confidence score was calculated (Table S5 ). The peptide data was combined with available information on SG localized proteins from mass spectroscopy of purified mammalian stress granules 27,52,59 and from other studies based on the information listed in the HIPPIE database. Each ProP-PD selected peptide was annotated with the above data and a count of the number of sources of evidence of SG localisation.

Network's visualization was done using Cytoscape 60 using the data provided in Table S2 and   Table S5 .

FP experiments were carried out using a SpectraMax iD5 multi-mode microplate reader (Molecular Devices) using a black half area 96-well plate (Corning, USA #3993) with a total volume of 50 µL. The settings were set to 485 nm for excitation and 535 nm for emission.

Peptides were from GeneCust (France) with a purity of over 95%. All measurements were performed at least in triplicates. For saturation experiment, proteins were serially diluted in the plate with phosphate buffer in a volume 25 µL, and then supplemented with a master mix of 10 nM FITC-labeled peptide in phosphate buffer. In the displacement assay, unlabeled peptides were serially diluted in the plate in 25 µL phosphate buffer, and 25 µL of displacement master mix was added (10 nM FITC-labeled peptide mixed with the protein of interest at a concentration 4 x the KD determined through direct binding using the FITC-labeled probe peptide).

HeLa and HEK293 cells were maintained in DMEM GlutaMAX containing 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% FCS (all from Thermo Fisher Scientific).

Stable HEK cell lines were generated using the T-Rex doxycycline inducible Flp-In system 

Lenti-X 293T cells (Takara bio) grown in a 100 mm plate format were co-transfected with 4. 

Total RNA (400 ng) was used to synthesize cDNA using High Capacity cDNA Reverse

Transcription Kit (Applied Biosystems) according to the manufacturer's instructions. 

For immunofluorescence microscopy, HeLa cell lines were seeded in eight-well Ibidi dishes 

Partial on-bead digestion was used for peptide elution from GFP-Trap Agarose (Chromotek).

Briefly 

Liquid chromatography mass spectrometry (LC-MS) analysis was performed with an EASY-nLC-1200 system (Thermo Fisher Scientific) connected to a trapped ion mobility spectrometry quadrupole time-of-flight mass spectrometer (timsTOF Pro, Bruker Daltonik GmbH, Germany) with a nano-electrospray ion source (Captive spray, Bruker Daltonik GmbH).

Peptides were loaded on a 50 cm in-house packed HPLC-column (75 µm inner diameter packed with 1.9 µm ReproSilPur C18-AQ silica beads, Dr. Maisch GmbH, Germany). Peptides were separated using a linear gradient from 5-30% buffer B (0.1% formic acid, 80% ACN in LC-MS grade H2O) in 43 min followed by an increase to 60% buffer B for 7 min, then to 95% buffer B for 5min and back to 5% buffer B in the final 5min at 300nl/min. Buffer A consisted of 0.1% formic acid in LC-MS grade H2O. The total gradient length was 60 min. We used an in-house made column oven to keep the column temperature constant at 60 °C. 

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository 67 with the dataset identifier PXD025410. (Table S2) . 

Interactions between indicated coronavirus (human and bat) proteins and host factors based on the results from the ProP-PD screen. 

Table S1

Table of expression constructs and information on proteins screened 

The Architecture of SARS-CoV-2 Transcriptome

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Comparative host-coronavirus protein interaction networks reveal panviral disease mechanisms

COVID-19 and the human innate immune system

Multilevel proteomics reveals host perturbations by SARS-CoV-2 and SARS-CoV

Functional interrogation of a SARS-CoV-2 host protein interactome identifies unique and shared coronavirus host factors

Genome-wide CRISPR screening identifies TMEM106B as a proviral host factor for SARS-CoV-2

Genome-wide CRISPR Screens Reveal Host Factors Critical for SARS-CoV-2 Infection

Genome-Scale Identification of SARS-CoV-2 and Pan-coronavirus Host Factor Networks

Identification of Required Host Factors for SARS-CoV-2 Infection in Human Cells

Comparative host-coronavirus protein interaction networks reveal panviral disease mechanisms

Plitidepsin has potent preclinical efficacy against SARS-CoV-2 by targeting the host protein eEF1A

The Ebola Virus Nucleoprotein Recruits the Host PP2A-B56 Phosphatase to Activate Transcriptional Support Activity of VP30

Small molecules, big targets: drug discovery faces the protein-protein interaction challenge

Short linear motifs -ex nihilo evolution of protein regulation

The functional importance of structure in unstructured protein regions

How pathogens use linear motifs to perturb host cell networks

Protein-Protein Interactions in Virus-Host Systems

How viruses hijack cell regulation

Discovery of short linear motif-mediated interactions through phage display of intrinsically disordered regions of the human proteome

Large-scale interaction profiling of PDZ domains through proteomic peptide-phage display using human and viral phage peptidomes