Virus–Host Protein Interaction Network of the Hepatitis E Virus ORF2-4 by Mammalian Two-Hybrid Assays

Throughout their life cycle, viruses interact with cellular host factors, thereby influencing propagation, host range, cell tropism and pathogenesis. The hepatitis E virus (HEV) is an underestimated RNA virus in which knowledge of the virus–host interaction network to date is limited. Here, two related high-throughput mammalian two-hybrid approaches (MAPPIT and KISS) were used to screen for HEV-interacting host proteins. Promising hits were examined on protein function, involved pathway(s), and their relation to other viruses. We identified 37 ORF2 hits, 187 for ORF3 and 91 for ORF4. Several hits had functions in the life cycle of distinct viruses. We focused on SHARPIN and RNF5 as candidate hits for ORF3, as they are involved in the RLR-MAVS pathway and interferon (IFN) induction during viral infections. Knocking out (KO) SHARPIN and RNF5 resulted in a different IFN response upon ORF3 transfection, compared to wild-type cells. Moreover, infection was increased in SHARPIN KO cells and decreased in RNF5 KO cells. In conclusion, MAPPIT and KISS are valuable tools to study virus–host interactions, providing insights into the poorly understood HEV life cycle. We further provide evidence for two identified hits as new host factors in the HEV life cycle.


Introduction
During its life cycle, viruses interact with a variety of host proteins for their survival, pathogenesis and virus spread.Unraveling virus-host interactions at the protein level aids in the understanding of a virus' life cycle, its molecular virology and immunology.Moreover, insight into these complex interactions will indicate potential antiviral targets and therefore, protein-protein interactions (PPIs) offer a wide range of possibilities to develop new and more effective therapies.
Many technologies to discover PPIs have been developed, ranging from two-hybrid systems to mass spectrometry, phage display and protein chip technology; all function in high-throughput formats [1,2].Two described analogous high-throughput mammalian two-hybrid assays are the MAmmalian Protein-protein Interaction Trap (MAPPIT) and the KInase Substrate Sensor (KISS) [2,3].MAPPIT and KISS are complementation assays based on (parts) of the JAK-STAT signaling pathway.In MAPPIT, a bait protein of interest is fused to a type I transmembrane cytokine receptor that is mutated at its STAT recruitment sites.A prey protein is fused to the cytoplasmic portion of another cytokine receptor containing intact STAT docking sites.Upon the functional association of bait and prey, the JAK proteins can phosphorylate the prey-linked moiety in trans, leading to subsequent STAT binding, dimerization and translocation to the nucleus to induce reporter gene activation.In KISS, a bait protein of interest is fused to the tyrosine kinase 2 (TYK2) domain, which after interaction with the prey is able to phosphorylate the prey-linked moiety.This results in the same downstream signaling as with MAPPIT, but it overcomes the tethering of the bait moiety to the plasma membrane.The advantage of those mammalian systems is that they allow for temporal, spatial and functional modulation and that necessary co-factors or regulatory proteins to study those dynamics of PPIs are present.Moreover, proteins can undergo the proper modifications in their native cellular environment, which is often necessary to allow interactions.The sensitivity and specificity of the described methods are comparable to that of other well-known assays such as Yeast-Two-Hybrid (Y2H) [4].Both MAPPIT and KISS have shown in the past to be robust methods for detecting PPIs for a variety of applications, including in the field of virus research [5,6].
In this study, both systems were used to identify novel protein binding partners for the proteins of the hepatitis E virus (HEV).HEV is a positive-sense, single-stranded RNA virus and currently the major cause of acute hepatitis worldwide, affecting about 20 million people each year [7].No specific therapy is available today and a vaccine is only licensed in China [8], highlighting the need for new treatment options.HEV is a virus infecting a range of species, but variants infecting humans can mainly be classified into four genotypes (gts), which differ in geographical distribution, route of transmission and pathogenicity [9].Gt-1 and gt-2 solely infect humans, are predominant in developing countries and mostly result in self-limiting infections.Nevertheless, mortality in pregnant women reaches up to 25%.Gt-3 and gt-4 infections are mostly zoonotic and are predominant in the industrialized world.Chronic gt-3 and gt-4 infections have been described in immunocompromised patients, which represent an important risk group.
The viral genome translates into three or four proteins, depending on viral genotype (gt).ORF1 encodes a nonstructural polyprotein consisting of eight putative domains and is important for viral replication [10].Whether this polyprotein is cleaved into several non-structural proteins by either a viral and/or cellular proteases is still a matter of debate.ORF2 and ORF3 are translated from a bicistronic subgenomic replicon.Different forms of ORF2 are produced during the life cycle: an infectious ORF2 that is the component of viral particles, as well as secreted/glycosylated ORF2 and cleaved ORF2 [11,12].The ORF3 protein is a small phosphoprotein playing a role in multiple cellular processes such as modulating immune responses and is important for viral egress [13][14][15].A fourth ORF is only observed in gt-1 HEV and is expressed under cellular stress conditions and influences viral replication [16].
The continued effort to define the interactions between the virus and host in general and in particular for HEV, will certainly lead to information about the life cycle of a virus that is far from understood.Previous work on this topic has already led to interesting findings, but overall our knowledge remains limited.To better understand the HEV-host interface as a whole and to compare different genotypes, we generated a protein-protein interaction network by subjecting ORF2 and ORF3 of both gt-1 and gt-3 and ORF4 of gt-1, to two mammalian two-hybrid assays utilizing a prey library of 15,000 proteins.We used bioinformatic databases to analyze these interactions and found significant enriched processes in relation to the respective protein(s).Two proteins from the hit lists, SHARPIN and RNF5, were selected for their interaction with ORF3 and their effect on the ubiquitination status of target proteins involved in the antiviral interferon signaling pathway.Using knockout (KO) cells for the respective proteins, we show they influence HEV-induced interferon (IFN) signaling.Moreover, HEV infection is affected in the KO cells.

High Throughput Screens
The MAPPIT and KISS bait were constructed as follows: the coding sequences of both gt-1 ORF2-4 were amplified from a fecal sample of a HEV Sar-55-infected liver-chimeric mouse, and gt-3 ORF2-3 were amplified from the gt-3 Kernow-C1 passage 6 plasmid.These were fused to the EpoR-LR-F3 chimeric receptor in the pSEL (+2L) vector via SalI-NotI restriction cloning (MAPPIT) or N-terminally to a fragment of human TYK2 in the pSVSport vector via EcoRI/MfeI-NotI restriction cloning (KISS).Primers are listed in Table 1.
Viruses 2023, 15, 2412 3 of 50 The obtained products were sequence-verified.The construction of prey constructs and the reverse transfection method was performed as described previously [2,3,17].Highthroughput MAPPIT and KISS screens were performed in microarray and retested in luciferase assay as described previously, using a prey library generated using the hOR-Feome v8.1 and ORFeome Collaboration clones (Center for Cancer Systems Biology, Boston, MA, USA) [17][18][19][20].

PPI Analysis
Jvenn was used to create Venn diagrams of overlapping proteins [21].The Database for Annotation, Visualization and Integrated Discovery (DAVID) was used to perform a functional annotation analysis regarding the different GO terms (biological process, molecular function, cellular compartment).DAVID's modified Fisher Exact p-value was set to 0.05, as a p-value equal or smaller to this number is considered as strongly enriched in the annotation category.Additionally, functional annotation clustering for the terms Functional Annotation and Gene ontology was performed with default settings and a medium classification stringency.

CRISPR-Cas9 Knockout of SHARPIN and RNF5
Single-guide RNAs (sgRNA) targeting the DNA sequence of SHARPIN and RNF5 were designed using the CRISPR gRNA design tool of Atum https://www.atum.bio/eCommerce/cas9/input (accessed on March-April 2020).Oligo pairs encoding the guide sequences were annealed and ligated into the sgRNA expressing plasmid, pX458 (kindly provided by Feng Zhang, Addgene plasmid #48138) [22].The sequences of the oligo pairs are represented in Table 2. PLC3 cells were transfected with the CRISPR plasmids using lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA).In brief, 3 × 10 5 cells were seeded per well in 6-well plates and transfected with 2500 ng DNA and 12.5 µL lipofectamine.Two days after transfection, single-cell sorting of GFP+ cells was performed using a BD FACS Aria III cell sorter (Becton Dickinson, NJ, USA).Clonal cell populations were expanded and checked for absence of protein expression by Western blot.

HEV ORF3 Transfection and IFN qRT-PCR Assays
An HEV ORF3 plasmid was constructed as follows.The ORF3 sequence was amplified from the pSK-E2 cDNA clone with following primers: 5 -aagcttaccatgaataacatgtcttttgctgcgcccatgggttc-3 and 5 -gcggccgcttagcggcgcggccccag-3 .Purified PCR products were cloned into pcDNA3.1 using HindIII and NotI restriction sites and resulting construct was verified by sanger sequencing.

HEV Infection and HEV ORF2 Immunofluorescence Staining
PLC3 WT , PLC3 SHARPIN-KO and PLC3 RNF5-KO were seeded in black 96-well plates and exposed to a dilution of HEV preparation with an MOI of 22,000, defined as 1 IU/cell, and IU (International unit) was quantified by qPCR as described previously [23].To determine the infectivity titer, cells were fixed by methanol, followed by incubation with 0.5% Triton X-100 and blocking with PBS containing 10% goat serum.HEV ORF2 staining was performed using a mouse ORF2-specific monoclonal antibody (1E6, Merck, Darmstadt, Germany) in combination with an Alexa Fluor 488-conjugated goat anti-mouse antibody.Nuclei were stained with DAPI (4 ,6-diamidino-2-phenylindole).Microscopy was performed using a Leica TSC-SPE confocal microscope (10× or 20× objective) (Wetzlar, Germany).Infectivity of the respective well was calculated as a percentage with automated cell counting as well as ORF2 fluorescence with ImageJ software (v1.54d).

Reverse Transcription Quantitative PCR (RT-qPCR)
Total RNA was extracted from cultured cells using the RNeasy plus mini kit (Qiagen, Hilden, Germany).cDNA synthesis was performed with SuperScript III reverse transcriptase (Thermo Fisher Scientific, Waltham, MA, USA).Target and reference gene transcripts were detected using pre-designed Taqman primer-probe assays (Thermo Fisher Scientific, Waltham, MA, USA).All used assays are listed in Table 4. cDNA was diluted 2.5× (target genes) or 10× (reference genes) and 2 µL was used as input for qPCR reaction using TaqMan Fast Advanced Master mix (Thermo Fisher Scientific, Waltham, MA, USA).Cycles of quantification were generated on a LightCycler 480 (Roche, Bazel, Switzerland) using the second-derivative maximum method.Calibrated normalized relative quantities (CNRQs) were calculated for each target gene with qBasePlus Software v3.1 (CellCarta, Montreal, QC, CA), using the most stable reference genes as selected by GeNorm analysis in the qBasePlus Software.

Statistical Analyses
Statistical analysis was performed using GraphPad Prism 8.0.1 software.Statistical significance of differences between groups was evaluated with t-tests.

MAPPIT and KISS Are Functional Tools to Study HEV-Host Protein-protein Interactions
ORF2 and ORF3 from both gt-1 and gt-3 and ORF4 (gt-1) were used as bait in the two related assays MAPPIT and KISS.
Using this approach, we had four constructs for ORF2: gt-1 ORF2 MAPPIT, gt-1 ORF2 KISS, gt-3 ORF2 MAPPIT and gt-3 ORF2 KISS.Microarrays were performed and revealed 37 hits in total for this protein (Figure 1A).These were uncovered solely with the MAPPIT constructs, as the KISS assay failed to reveal any interactions.Comparing both genotypes, we found that two proteins were shared, specifically CA12 and ADCY3 (Figure 1A) (Appendix A Table A1).Likewise, four constructs were created for ORF3: gt-1 ORF3 MAPPIT, gt-1 ORF3 KISS, gt-3 ORF3 MAPPIT and gt-3 ORF3 KISS, revealing 187 hits in total (Figure 1B).Most hits were shared between the KISS assay of the two different genotypes, and this initial screening showed an overlap of 25 hits between gt-1 and gt-3 (Figure 1B) (Appendix Table A1).
For ORF4, combining MAPPIT and KISS revealed 91 hits, of which 11 are shared between the two assays (Figure 1C) (Appendix Table A1).
To obtain a general overview of all the HEV hits discovered with MAPPIT and KISS, we compiled all the different hits per HEV protein, irrespective of the used method or genotype, and looked whether there are particular hits shared between HEV ORF2, ORF3 and ORF4, as HEV proteins often interact with each other during the viral life cycle.Following this approach, the Venn diagram shows that a substantial amount of hits were shared between both ORF3 and ORF4, STUB1 was shared between ORF2 and ORF4; and ORF2 and ORF3 shared FTL, TMEM154 and FCGR2A (Figure 1D) (Appendix Table A1).Likewise, four constructs were created for ORF3: gt-1 ORF3 MAPPIT, gt-1 ORF3 KISS, gt-3 ORF3 MAPPIT and gt-3 ORF3 KISS, revealing 187 hits in total (Figure 1B).Most hits were shared between the KISS assay of the two different genotypes, and this initial screening showed an overlap of 25 hits between gt-1 and gt-3 (Figure 1B) (Appendix A Table A1).
For ORF4, combining MAPPIT and KISS revealed 91 hits, of which 11 are shared between the two assays (Figure 1C) (Appendix A Table A1).
To obtain a general overview of all the HEV hits discovered with MAPPIT and KISS, we compiled all the different hits per HEV protein, irrespective of the used method or genotype, and looked whether there are particular hits shared between HEV ORF2, ORF3 and ORF4, as HEV proteins often interact with each other during the viral life cycle.Following this approach, the Venn diagram shows that a substantial amount of hits were shared between both ORF3 and ORF4, STUB1 was shared between ORF2 and ORF4; and ORF2 and ORF3 shared FTL, TMEM154 and FCGR2A (Figure 1D) (Appendix A Table A1).

Functional Annotation Clustering for the HEV Proteins
To gain a comprehensive overview of the hits and relate biological relevance to those identified, we first performed gene ontology (GO) annotation for each protein with the DAVID bioinformatics tool.
For ORF2, five significant clusters could be observed for biological processes (BP) (Figure 2A, Appendix A Table A2), all containing two proteins or 5.4% of the ORF2 dataset.From the cellular compartment (CC) clusters, it is obvious that almost half of the amount of hits are associated with the (plasma) membrane, and about 14% is associated with the cell surface (Figure 2B, Appendix A Table A2).Three clusters could be observed regarding the molecular function (MF) annotation, being TPR-domain binding, carbonate dehydratase activity and hydro-lyase activity, which all contained two proteins or 5.4% of the dataset (Figure 2B, Appendix A Table A2).
Viruses 2023, 15, x FOR PEER REVIEW 8 of 49 For ORF2, five significant clusters could be observed for biological processes (BP) (Figure 2A, Appendix Table A2), all containing two proteins or 5.4% of the ORF2 dataset.From the cellular compartment (CC) clusters, it is obvious that almost half of the amount of hits are associated with the (plasma) membrane, and about 14% is associated with the cell surface (Figure 2B, Appendix Table A2).Three clusters could be observed regarding the molecular function (MF) annotation, being TPR-domain binding, carbonate dehydratase activity and hydro-lyase activity, which all contained two proteins or 5.4% of the dataset (Figure 2B, Appendix Table A2).For ORF3, 28 BP clusters were observed, with among them processes such as viral entry into host cells, transmembrane transport, GPCR signaling and immune response (Figure 3A, Appendix Table A3).As with the ORF2 hits, a high proportion of the total ORF3 hit list reside in membranes or plasma membranes (Figure 3B, Appendix Table A3).
For ORF4, 13 BP clusters could be observed with regulation of alternative mRNA splicing, with the positive regulation of proteolysis and negative regulation of transcription among the most significant ones (Figure 4A, Appendix Table A4).More than half of the proteins are clustered in the cellular compartment cytoplasm, but a great deal of proteins also cluster to the nucleus.Regarding the MF ontology, 88% of the hits have proteinbinding capacities, and about 15% are (m)RNA-binding proteins (Figure 4C, Appendix Table A4).For ORF3, 28 BP clusters were observed, with among them processes such as viral entry into host cells, transmembrane transport, GPCR signaling and immune response (Figure 3A, Appendix A Table A3).As with the ORF2 hits, a high proportion of the total ORF3 hit list reside in membranes or plasma membranes (Figure 3B, Appendix A Table A3).
For ORF4, 13 BP clusters could be observed with regulation of alternative mRNA splicing, with the positive regulation of proteolysis and negative regulation of transcription among the most significant ones (Figure 4A, Appendix A Table A4).More than half of the proteins are clustered in the cellular compartment cytoplasm, but a great deal of proteins also cluster to the nucleus.Regarding the MF ontology, 88% of the hits have protein-binding capacities, and about 15% are (m)RNA-binding proteins (Figure 4C, Appendix A Table A4).
As mentioned in the previous section, HEV proteins often interact with each other and therefore, we were interested to see if the identified candidate interactors would cluster together in certain (upregulated) GO terms.Functional annotation clustering of the total HEV hit list was performed in DAVID and revealed 22 clusters (Appendix A Table A5).We could again observe that most of the proteins are observed in a membrane cluster.The second most enriched cluster involved proteins important for viral entry or receptor activity.Proteins related to GPCR signaling were again observed.Cluster 5 and 6 relate to ion channel terms and may be related to the function of ORF3 as an ion channel.Inflammatory response and immunity also emerged and molecular functions relating to post-translational modifications such as protein phosphorylation and ubiquitination are also present.As mentioned in the previous section, HEV proteins often interact with each other and therefore, we were interested to see if the identified candidate interactors would cluster together in certain (upregulated) GO terms.Functional annotation clustering of the total HEV hit list was performed in DAVID and revealed 22 clusters (Appendix Table A5).We could again observe that most of the proteins are observed in a membrane cluster.The second most enriched cluster involved proteins important for viral entry or receptor activity.Proteins related to GPCR signaling were again observed.Cluster 5 and 6 relate to ion channel terms and may be related to the function of ORF3 as an ion channel.Inflammatory response and immunity also emerged and molecular functions relating to posttranslational modifications such as protein phosphorylation and ubiquitination are also present.
Additionally, we also performed a search for hits that have previously been described in the context of viral infections (Appendix Table A6).Specifically, we found that 15 of the 37 ORF2-interacting proteins identified with MAPPIT and KISS are in some way involved in viral infections.Likewise, 108 of the 187 identified ORF3-interacting host proteins and 33 ORF4-interacting protein hits have previously been described to play a role in viral infections.Additionally, we also performed a search for hits that have previously been described in the context of viral infections (Appendix A Table A6).Specifically, we found that 15 of the 37 ORF2-interacting proteins identified with MAPPIT and KISS are in some way involved in viral infections.Likewise, 108 of the 187 identified ORF3-interacting host proteins and 33 ORF4-interacting protein hits have previously been described to play a role in viral infections.
To further scrutinize the proteins, we focused on hits that were identified either for both genotypes, or in both assays, or shared among HEV proteins.The selection was further narrowed down by investigating if potential hits are known to be important in viral infections (Appendix A Table A6).One of the hits we observed in the clusters of protein ubiquitination is SHARPIN or SHANK-Associated RH Domain Interactor.In our analysis, we observed this as a hit of gt-1 ORF3, identified by both MAPPIT and KISS assay, and as a hit of ORF4, also identified by both MAPPIT and KISS assay.We speculated that gt-3 ORF3 could also bind SHARPIN.Therefore, we performed a retest of the gt-3 ORF3 in the KISS configuration, as this proved the most effective for this genotype in our initial analysis.Using this approach, we could confirm that gt-3 ORF3 also bound SHARPIN (Table 5).Results are compiled from the luciferase reporter read-out.The positive score of an interaction is related to the lowest value of BP/PIB and BP/BIP.In MAPPIT, a fold induction of a specific bait-prey interaction (BP) is calculated by dividing the average of EPO stimulated and non-stimulated controls.The obtained value of the specific BP is then compared to a fold induction of a bait with irrelevant prey interaction (BIP) and to a fold induction of an irrelevant bait and prey interaction (PIB).The BIP in this case was 1.34.To score an interaction as positive, the obtained lowest value of PB/BIP and PB/PIB needs to be above a threshold of 9.In KISS, stimulation is not possible and the average value of the bait-prey interaction is divided by the average of the BIP or PIB and needs to be above a threshold of 5. BIP in this case was 7414.The chosen thresholds depend on positive and negative reference sets and in this case relate to a false positivity rate of 2%.
In the same cluster of protein ubiquitination, we identified the Ring Finger Protein 5 (RNF5).During our analysis, this appeared a candidate hit of gt-1 ORF3.Along with the experiments we performed for SHARPIN, we speculated that gt-3 ORF3 could also bind RNF5.Upon retest, we could indeed confirm the interaction of gt-3 ORF3 with RNF5 (Table 5).

SHARPIN Affects the Induction of Interferon upon ORF3 Transfection
SHARPIN forms the linear ubiquitin chain assembly complex (LUBAC) which mainly targets proteins in the Retinoic acid-Inducible Gene I (RIG-I)-like receptor (RLR)-Mitochondrial Antiviral Signaling Protein (MAVS) Pathway.
To understand the roles of this protein in the context of HEV infection, a CRISPR-Cas9 knockout (KO) cell line was generated (Figure 5A).We used PLC3 cells, a subclone of PLC/PRF/5, that are permissive to HEV replication [11].
To understand a potential interaction of this protein with ORF3, we expressed this protein in PLC3 WT and PLC3 SHARPIN-KO (Figure 5A) and investigated its effect on the induction of IFN, since LUBAC has been shown to regulate signaling downstream of TLR3, RIG-I and MDA5.According to what was previously published [14], we could show that ORF3 also boosted poly (I:C) mediated-type I IFN induction in PLC3 cells.Viral protein-transfected cells showed a response twice as high as the mock control (p = 0.017), while unstimulated cells had an absent or only very minimal response (Figure 5B, left panel).Strikingly, in the SHARPIN KO, the ORF3-induced enhancement of type I IFN was abolished (p = 0.24) and rather tended to show an inverse effect (Figure 5B, left panel).Likewise, we investigated the type III IFN response.Although the trend seemed to be the same as for type I IFN, we could not show a significant induction of type III IFN upon ORF3 addition in PLC3 WT cells (p = 0.13) (Figure 5B, right panel).We next wondered if these changes in interferon response could have an impact on HEV infection.To this end, WT and SHARPIN-KO cells were infected and the level of infection was measured by immunostaining 6 days later.Strikingly, infection increased from 17% in WT cells to 25% in the SHARPIN-KO cells (p = 0.0357) (Figure 5C).Together, these results show SHARPIN has an effect on the ORF3-mediated IFN induction, while infection is stimulated in the KO cells.changes in interferon response could have an impact on HEV infection.To this end, WT and SHARPIN-KO cells were infected and the level of infection was measured by immunostaining 6 days later.Strikingly, infection increased from 17% in WT cells to 25% in the SHARPIN-KO cells (p = 0.0357) (Figure 5C).Together, these results show SHARPIN has an effect on the ORF3-mediated IFN induction, while infection is stimulated in the KO cells.

RNF5 Interferes with IFN Induction upon ORF3 Transfection
The Ring Finger Protein 5 (RNF5) is a membrane-bound ubiquitin ligase that is known to inhibit IFN type I in the context of virus infections.To investigate a potential effect of RNF5 in the HEV life cycle by interacting with ORF3, we performed a similar experiments as for SHARPIN above.PLC3 RNF5-KO cells were generated and successful KO was verified by Western blot; ORF3 was expressed in these cells and compared to PLC3 WT (Figure 6A).ORF3 enhanced the type I IFN response upon poly I:C stimulation (p = 0.022), but this was not observed in the RNF5-KO cells.Moreover, the trend seems to be that IFN-β diminishes upon ORF3 transfection in these cells (p = 0.28) (Figure 6B, left panel).For type III IFN responses, we observed the same trend: there was an increase upon ORF3 transfection in WT cells (p = 0.023), which was not observed in the KO cells (p = 0.26) (Figure 6B, right panel).

RNF5 Interferes with IFN Induction upon ORF3 Transfection
The Ring Finger Protein 5 (RNF5) is a membrane-bound ubiquitin ligase that is known to inhibit IFN type I in the context of virus infections.To investigate a potential effect of RNF5 in the HEV life cycle by interacting with ORF3, we performed a similar experiments as for SHARPIN above.PLC3 RNF5-KO cells were generated and successful KO was verified by Western blot; ORF3 was expressed in these cells and compared to PLC3 WT (Figure 6A).ORF3 enhanced the type I IFN response upon poly I:C stimulation (p = 0.022), but this was not observed in the RNF5-KO cells.Moreover, the trend seems to be that IFNβ diminishes upon ORF3 transfection in these cells (p = 0.28) (Figure 6B  Furthermore, we checked if there is an influence on HEV infection in PLC3 RNF5-KO cells compared to PLC3 WT .Cells were infected and the amount of infection was measured by immunostaining 6 days later.Infection reduced from 15% in the WT population to 6% in the KO population (Figure 6C).These results show that RNF5 interferes with the ORF3increased IFN response and influences HEV infection.

Discussion
In this study, we used two PPI methods to identify interactions of the HEV ORF2-4 to aid in the urgent need to better understand the HEV pathogenesis, life cycle and virus-host interactions.Some HEV PPI studies have been performed in the past, most of them using yeast-two-hybrid and affinity purification in combination with mass spectrometry [24][25][26][27][28][29].One method can only detect a certain fraction of all PPIs, illustrating that in order to comprehensively map the interactome of one particular protein of interest, it is important to use different methods and consider them as complementary rather than as independent assays [30].The advantage of the MAPPIT and KISS screenings used in this study is their mammalian background and high-throughput capability [2,3,19].We validated this approach for HEV and identified 37 candidate interactors for ORF2, 187 for ORF3 and 91 for ORF4.We did not investigate ORF1 due to the controversy and complexity about the processing of this protein, and the studies conducted so far with this protein have always used the individual domains as bait [31,32].Additionally, ORF1 interacts with JAK2 [32], which means it could bind the JAK2 proteins that are constitutively associated with the chimeric MAPPIT receptor, posing an additional difficulty.
For ORF2, our candidate interactors were identified only in the MAPPIT configuration.In initial tests of the KISS screening, positive controls that bind the Tyk2 domain did not result in acceptable signals and a pre-screen with known ORF2 interactions did not reveal any hits.It is unclear why this happened.It could point to a technical issue of the assay but it is more likely that it is attributed to the nature of the HEV ORF2 protein, as this is the only one of the HEV proteins that failed.In MAPPIT, the bait moiety is tethered to the plasma membrane, whereas in KISS, the bait can shuttle in the cytosol.It is known that ORF2 translocates to the nucleus and reticular compartments; therefore, it is possible that by residing in these subcellular compartments, the construct is hidden for potential interactions with prey proteins in the cytosol [33].Before screening, we anticipated that the subcellular localization of HEV proteins could impede some analyses, especially since it is known, for example, that replication of viruses can induce membrane rearrangements to shield for immune responses [34].For this reason, we decided to perform screenings with both methods.
In our 37 identified ORF2 hits, we could confirm four previously described interactions of this HEV protein: HSP90, TMBIM4, FTL and CYB5A [24,26].Only two hits were shared between gt-1 and gt-3 in our assay, being CA12 and ADCY3 (Figure 1A); the former protein is part of the zinc metalloenzymes catalyzing the reversible hydration of carbon dioxide, whereas ADCY3 catalyzes the formation of the secondary messenger cyclic adenosine monophosphate (cAMP).Gene ontology annotation analysis revealed that most of the identified ORF2 hits are membrane-associated proteins, which is in agreement with what was published in another ORF2 screen using Y2H [26].Looking at the biological processes, there were two proteins that were often observed: FLOT1 and COLEC12 (Figure 2A).FLOT1 is an integral membrane component of caveolae and plays a role in vesicle trafficking and cell morphology.Additionally, flotillin-dependent endocytosis is an alternative pathway, independent of clathrin or dynamin that could in theory be used by viruses.However, mechanisms remain unclear and many studies report no involvement of the flotillin in virus entry, including for hepatitis C virus or hepatitis A virus (HAV) [35][36][37].However, FLOT1 is present in HAV exosomes [38], which might also be the case for HEV, and a potential interaction with FLOT1 might be of importance during membrane fusion in the process of viral release [39].COLEC12 is a member of the C-lectin family: it is a scavenger receptor that may function as pattern recognition molecule (PRM) by binding carbohydrate antigens on microorganisms (particularly bacteria) to facilitate their removal.However, no clear studies have yet confirmed this role in the context of viral infection, except one study showing no binding of the SARS-CoV-2 spike to COLEC12, while it did bind other related humoral pattern recognition molecules [40].The expression of COLEC-12 in human hepatocytes was confirmed in a study involving the hepatitis C virus, but the role of the protein was not further investigated [41].
For ORF3, we confirmed the previously described interactions with TSG101, NAT1, MAT1A, HSPA8, HPX, FTL, ALDOB, C151, FGB, APOH and AMBP and an additional 176 hits that were not previously described.Strikingly, the most significant biological process we observed was the one containing proteins involved in 'viral entry into host cells' (Figure 3A).EFNB2 is a receptor for Hendra and Nipah virus [42], CD4 for human immunodeficiency virus [43], CXADR is the coxsackievirus and adenovirus receptor [44], and BSG has been shown to be involved in infection of measles and human cytomegalovirus [45,46].AGTR1 also plays a role during SARS-CoV-2 endocytosis and NECTIN-4 is a receptor for measles virus [47,48].These proteins mostly bind a viral protein to mediate viral entry, so it is intriguing why ORF3 is observed as an interacting partner of these proteins.Only HAVCR1 is a phosphatidylserine receptor that mediates the entry of a variety of viruses that acquire PS in their viral envelope through a process called 'apoptotic mimicry' [49].We very recently studied the role of this protein in the entry of HEV and could demonstrate the involvement of PS acquired in the enveloped form of the HEV virions [50].The observation of all these viral receptors might indicate that the role of ORF3 during entry of enveloped hepatitis E virus is of more importance than we think and should be further studied.We also observed some BPs and MFs related to ion transmembrane transport (Figure 3), which could be related to the function of ORF3 as a functional ion channel in the release of virions [13].We also observed some BPs regarding the immune response.For ORF3, it was previously described that it can interfere with cellular host defenses and its interferon-induced effectors, sometimes by inhibiting IFN induction, or by induction of IFN [14,51,52].Most of the proteins here were also membrane proteins (Figure 3B).Intriguingly, we observed an annotation term of extracellular exosome, containing 31 proteins.It is well known that ORF3 uses the exosome pathway, a process orchestrated by the interaction of ORF3 and TSG101 to dock virions in the multivesicular bodies leading to viral release [53].
For ORF4, only two previous studies have been performed to screen for interacting proteins [16,54].The hits we observed during our MAPPIT and KISS analysis were not reported before.We observed that ORF3 and ORF4 shared 10 protein hits.Previously, it was reported that ORF4 forms a protein complex with the RdRp, helicase and X, of which the assembly is inhibited by interaction with ORF3 [16].It is possible ORF3 and ORF4 form another complex together with host proteins to exert this and/or other functions.Interestingly, TSG101 seems to be one of the proteins that is shared between ORF3 and ORF4.Another multivesicular body protein, CHMP4C, interacted with ORF4, illustrating it might be worthwhile to study this interaction in the context of gt-1 virus assembly and release.For ORF4 BPs, we could also observe proteins related to virus responses.Both ADAR and PRKRA (PACT), part of the RNA sensing pathway, were observed to interact with ORF4.ADAR has both pro-and antiviral activities, one of them by limiting the amount of dsRNA and thereby inhibiting PKR activation.Furthermore, PKR is activated by PACT and its activation inhibits cellular mRNA translation through the phosphorylation of EIF2A [55,56].It is intriguing that ORF4 binds both these proteins.It was also reported that ORF4 is a target of the proteasome due to ubiquitination of Lysine at the 51st amino acid position [16].Several proteins in our hit list were linked to a (poly)ubiquitination status (Figure 4), most of them are (part of) E3 ubiquitin ligases (complex).FBXW11 and BTRC are even part of the same complex.It would be interesting to see if one of the observed interactors is able to ligate ubiquitination to ORF4.On the other hand, these proteins often ubiquitinate target proteins involved in innate defense pathways such as NF-kB pathways.Additionally, SHARPIN and TRAF1 were also identified as ORF4 interactors, and previous work illustrated that TRAF1 is a target of the LUBAC complex, consisting of HOIP, HOIL-1L and SHARPIN, to enhance NF-kB activation.On the contrary, it was also reported that TRAF1 sequesters LUBAC to limit ubiquitination of another target NEMO, thereby limiting NF-kB activation [57].It remains to be elucidated if a similar process could be observed upon ORF4 interaction.
Many of the proteins we discovered as HEV-interacting proteins were already described to be implicated in other viral infections (Appendix A Table A6).This clearly indicates the relevance of the interactome we generated in this work.For an abundance of these host proteins, this role was linked to an increased or decreased expression upon viral infection or relates to data generated by other PPI-screens.The specific role that the identified host protein played in the viral life cycle was not necessarily further investigated.For other host proteins, including the already-mentioned HAVCR1, ADAR, EFNB2, CXADR and NECTIN4, their role has already been more thoroughly studied, making a potential relationship to the HEV life cycle more easy.A disadvantage of the here-applied PPI methods, but by extension of similar approaches such as yeast-two-hybrid, is that they rely on interactions between proteins that are artificially brought together in the same cell by ectopic expression.Confirming biological relevance of the uncovered interactions is therefore of the utmost importance.We selected SHARPIN and RNF5 for further characterization.
Initially, we only observed an interaction of both SHARPIN and RNF5 with gt-1 ORF3.By retesting the constructs in a second MAPPIT and KISS screening experiment, we could confirm the interaction for both ORF3 (Table 5).Again this illustrates that additional hits may be found upon repetition of a certain PPI method.That we confirmed the interaction with gt-3 for both SHARPIN and RNF5 is not unusual: amino acid sequence homology between the used gt-1 Sar55 strain and gt-3 Kernow-C1 p6 strain is more than 80%, but gt-specific regions exist.Similarly, Geng et al., identified 32 shared human proteins interacting with ORF3 from human isolates of gt-1 and gt-4, and with ORF3 from a gt-3 rabbit isolate [28].
SHARPIN, together with HOIL-1 and HOIP, forms LUBAC, the only complex known so far able to polyubiquitinate target proteins in a linear fashion.Studies regarding this complex in viral infections have led to opposing data.For Sendai virus (SeV) and Vesicular stomatitis virus (VSV), it was shown that LUBAC inhibits virus-induced interferon, whereas a study using norovirus showed that parts of the LUBAC complex are essential for the induction of both type I and type III IFN [58][59][60].Here, we also investigated both the type I and type III IFN response.In addition, it is known that HEV can induce both types as well [61][62][63].In agreement with the previous literature, we could confirm that ORF3 enhances poly (I:C)-mediated type I IFN induction 1.9-2.3fold compared to the mock control (Figures 5B and 6B) [14].Remarkably, when PLC3 SHARPIN-KO cells were transfected with ORF3, the poly(I:C)-mediated induction of IFN diminished (Figure 5B).It is thought the ORF3-mediated IFN enhancement is caused by RIG-I activation through K63-linked ubiquitination.Two ubiquitin ligases can mediate this activation: TRIM25 and Riplet [14].Interestingly, TRIM25 has been described as a target for LUBAC [60].During SeV, LUBAC suppresses RIG-I ubiquitination and activation by inducing TRIM25 degradation, thereby suppressing subsequent type I IFN production.However, our data would suggest an inverse relationship as absence of SHARPIN/LUBAC might suppress IFN induction upon ORF3 (Figure 5B).Still, since opposing effects of LUBAC (components) have been described for a variety of viruses, the targets may be differentially regulated depending on the virus of interest.These effects were mainly seen upon type I IFN induction, but the trend seems to be the same for type III IFN induction.Interestingly, HEV infection is slightly higher in SHARPIN-KO cells compared to control (Figure 5C), and might be related to the lower level of ORF3-mediated IFN induction in these cells.
Future research should explore which exact proteins belonging to the RLR-MAVS pathway leading to IFN induction are targeted by SHARPIN/LUBAC in the context of HEV.Taking things together, a protein complex between RIG-I, TRIM25, SHARPIN and ORF3 might be a possibility, as mentioned above.Additionally, LUBAC can also target NEMO to attenuate IFN response upon VSV infection, via disruption of the MAVS-TRAF3 complex.The hepatitis B virus induces the protein Parkin to recruit LUBAC to MAVS and disrupt downstream signaling [59,64].
A second candidate interactor of ORF3 that we selected was RNF5, especially because this protein has also been described to influence IFN induction in context of viral infections.For example, the V protein of Newcastle disease virus and PB1 protein of Influenza A virus recruit RNF5 to polyubiquitinate and degrade MAVS, which inhibit type I IFN production [65,66].Next to MAVS, the protein STING has also been described as a target for RNF5.SeV induces the ubiquitination and degradation of STING by RNF5 affecting IFN signaling [67].We showed that IFN induction is affected in RNF5-KO cells upon ORF3 transfection (Figure 6B).We also show that HEV infection is affected in these KO cells, with a significantly lower infection level compared to wild-type cells (Figure 6C).It would be interesting to speculate if HEV (ORF3) could interfere with the IFN induction by recruiting RNF5 to its target proteins.One previous study showed that HEV does not induce MAVS degradation; further research should explore whether STING would be a more reasonable target in case of HEV infection [61].

Conclusions
In conclusion, our data demonstrate that MAPPIT and KISS are valuable tools to probe HEV-host protein interactions.Moreover, our obtained hit list for HEV ORF2-4 provides a wealth of information to find new host factors implicated in the HEV life cycle, which remains far from understood.Confirmation of biological relevance of obtained hits in high-throughput PPI-methods is important.We specifically identified SHARPIN and RNF5 as interactors for both gt-1 and gt-3 ORF3 and showed they interfere with IFN signaling and HEV infection in general.

Screen
Shared Protein Hits

Viruses 2023 , 49 Figure 1 .
Figure 1.Venn diagrams depicting HEV ORF interactors identified by MAPPIT and KISS analyses.(A) Interactors of the ORF2 protein identified by MAPPIT analyses.(B) Interactors of the ORF3 protein identified by both MAPPIT and KISS analyses.(C) Interactors of ORF4 protein identified by both MAPPIT and KISS analyses.(D) Venn diagram shows distinct and overlapping interacting proteins of all HEV proteins analyzed.

Figure 1 .
Figure 1.Venn diagrams depicting HEV ORF interactors identified by MAPPIT and KISS analyses.(A) Interactors of the ORF2 protein identified by MAPPIT analyses.(B) Interactors of the ORF3 protein identified by both MAPPIT and KISS analyses.(C) Interactors of ORF4 protein identified by both MAPPIT and KISS analyses.(D) Venn diagram shows distinct and overlapping interacting proteins of all HEV proteins analyzed.

Figure 2 .
Figure 2. Gene ontology (GO) annotation of the ORF2 hit list.GO annotation for the ORF2 interacting proteins of both gt-1 and gt-3 as identified by MAPPIT analysis; different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF2 hit list (=37).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Figure 2 .
Figure 2. Gene ontology (GO) annotation of the ORF2 hit list.GO annotation for the ORF2 interacting proteins of both gt-1 and gt-3 as identified by MAPPIT analysis; different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF2 hit list (=37).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Viruses 2023 , 49 Figure 3 .
Figure 3. Gene ontology annotation of the ORF3 hit list.GO annotation for the ORF3 interacting proteins of both gt-1 and gt-3 as identified by MAPPIT and KISS analyses.Different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF3 hit list (=187).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Figure 3 .
Figure 3. Gene ontology annotation of the ORF3 hit list.GO annotation for the ORF3 interacting proteins of both gt-1 and gt-3 as identified by MAPPIT and KISS analyses.Different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF3 hit list (=187).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Figure 4 .
Figure 4. Gene ontology annotation of the ORF4 hit list.GO annotation for the gt-1 ORF4 interacting proteins as identified by MAPPIT and KISS analyses, different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF4 hit list (=91).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Figure 4 .
Figure 4. Gene ontology annotation of the ORF4 hit list.GO annotation for the gt-1 ORF4 interacting proteins as identified by MAPPIT and KISS analyses, different categories are depicted: biological process (A), cellular compartment (B) and molecular function (C).Red circles indicate the number of proteins belonging to each term and shown as percentage of the total ORF4 hit list (=91).p-value was set to equal to or smaller than 0.05 (indicated by dashed line).

Figure 5 .
Figure 5. SHARPIN affects the induction of IFN response.(A) PLC3 WT and PLC3 SHARPIN-KO cells (indicated, respectively, with + and −), are mock-or HEV ORF3-transfected (indicated, respectively, with − and +).Successful knockout and ORF3 transfection verified by Western blot.B-actin is displayed as a control.Two replicates of each condition are displayed.(B) PLC3 WT and PLC3 SHARPIN-KO cells were mock-or HEV ORF3-transfected and two days later transfected with poly (I:C).IFN-β (left) and IFN-λ mRNA levels measured by qRT-PCR.Transcripts normalized to reference genes (ATP5B, CYC1, YWHAZ, HPRT1, RPL30) and scaled to PLC3 WT -MOCK (ORF3-).(C) PLC3 WT and PLC3 SHARPIN-KO cells infected with HEV (MOI 22000) and infection levels analyzed 6 days later by HEV ORF2 immunostaining.Representative confocal images on the right.Nuclei in blue and ORF2 in green.Results are represented as mean ± SEM compiled from three independent experiments.Statistical significance calculated by (multiple) t-test.Differences were considered as statistically significant when p < 0.05 (*).

Figure 5 .
Figure 5. SHARPIN affects the induction of IFN response.(A) PLC3 WT and PLC3 SHARPIN-KO cells (indicated, respectively, with + and −), are mock-or HEV ORF3-transfected (indicated, respectively, with − and +).Successful knockout and ORF3 transfection verified by Western blot.B-actin is displayed as a control.Two replicates of each condition are displayed.(B) PLC3 WT and PLC3 SHARPIN-KO cells were mock-or HEV ORF3-transfected and two days later transfected with poly (I:C).IFN-β (left) and IFN-λ mRNA levels measured by qRT-PCR.Transcripts normalized to reference genes (ATP5B, CYC1, YWHAZ, HPRT1, RPL30) and scaled to PLC3 WT -MOCK (ORF3-).(C) PLC3 WT and PLC3 SHARPIN-KO cells infected with HEV (MOI 22000) and infection levels analyzed 6 days later by HEV ORF2 immunostaining.Representative confocal images on the right.Nuclei in blue and ORF2 in green.Results are represented as mean ± SEM compiled from three independent experiments.Statistical significance calculated by (multiple) t-test.Differences were considered as statistically significant when p < 0.05 (*).
, left panel).For type III IFN responses, we observed the same trend: there was an increase upon ORF3 transfection in WT cells (p = 0.023), which was not observed in the KO cells (p = 0.26) (Figure 6B, right panel).

Figure 6 .Figure 6 .
Figure 6.RNF5 affects the induction of IFN response.(A) PLC3 WT and PLC3 RNF5-KO cells (indicated, respectively, with + and −), are mock-or HEV ORF3-transfected (indicated, respectively, with − and +).Successful knockout and ORF3 transfection verified by Western blot (upper band in the RNF5 blot is aspecific).B-actin is displayed as a control.Two replicates of each condition are displayed.(B) PLC3 WT and PLC3 RNF5-KO cells were mock-or HEVORF3-transfected and two days later transfected with poly (I:C).IFN-β (left) and IFN-λ mRNA levels measured by qRT-PCR.Transcripts Figure 6.RNF5 affects the induction of IFN response.(A) PLC3 WT and PLC3 RNF5-KO cells (indicated, respectively, with + and −), are mock-or HEV ORF3-transfected (indicated, respectively, with − and +).Successful knockout and ORF3 transfection verified by Western blot (upper band in the RNF5 blot is aspecific).B-actin is displayed as a control.Two replicates of each condition are displayed.(B) PLC3 WT and PLC3 RNF5-KO cells were mock-or HEVORF3-transfected and two days later transfected with poly (I:C).IFN-β (left) and IFN-λ mRNA levels measured by qRT-PCR.Transcripts normalized to reference genes (ATP5B, CYC1 and YWHAZ) and scaled to PLC3 WT -MOCK (ORF3-).(C) PLC3 WT and PLC3 RNF5-KO cells infected with HEV (MOI 22,000) and infection levels analyzed 6 days later by HEV ORF2 immunostaining.Representative confocal images on the right.Nuclei in blue and ORF2 in green.Results represented as mean ± SEM from three independent experiments.Statistical significance calculated by (multiple) t-test.The level of significance is indicated with p < 0.01 (**) or p < 0.05 (*).

Table 1 .
PCR primers for amplification of ORF3 for MAPPIT and KISS assays.

Table 2 .
Sequences of oligos used for annealing and cloning into pX548.

Table 4 .
List of predesigned primer/probe pairs used for RT-qPCR.

Table 5 .
Overview of the interaction of SHARPIN and RNF5 with gt-1 and gt-3 ORF3 in MAPPIT and KISS.

Table A2 .
GO analysis of hits identified during MAPPIT and KISS screens against the HEV ORF2 protein.

Table A3 .
GO analysis of hits identified during MAPPIT and KISS screens against the HEV ORF3 protein.

Table A5 .
Functional annotation clustering of the HEV ORF candidate interactors.

Table A6 .
Described functions in viral infection of HEV ORF candidate interactors.