The functional domains and genome organization of HEV (Figure 1A) were first suggested based on bioinformatic alignments of the genotype 1 Burma reference strain with other well-characterized viruses in 1992 (Koonin et al., 1992), and the putative protease domain was proposed based on limited sequence identity with the distantly related rubella virus. These analyses suggested the existence of a putative papain-like protease within HEV ORF1 and that the proposed catalytic dyad were residues C483 (which is highly conserved across all known HEV genotypes) and H590 - which is variable across all eight known HEV genotypes (Figure 1—figure supplement 1). Since, tools have been developed to systematically interrogate the HEV genome, such as the development of infectious clones of cell-culture adapted strains (Johne et al., 2014; Shukla et al., 2012; Shukla et al., 2011), and reporter replicons utilizing green fluorescent protein (GFP; Emerson et al., 2004) or Gaussia luciferase (Gluc; Graff et al., 2005; Figure 1A).

To determine the importance of the residues that have been proposed as the putative catalytic dyad, we mutated C483 and/or Y590 to chemically similar amino acids, alanine, or in the case of Y590, residues found in other genotypes of HEV. Huh7.5 human hepatoma cells were transfected with in vitro transcribed RNA from a recombinant version of an HEV genome derived from the KernowC1/p6 strain (Shukla et al., 2011) in which ORF2 and ORF3 are replaced by a secreted version of Gluc (Shukla et al., 2011), termed Kc1/p6 Gluc (Figure 1A). Gluc activity as a measure for the efficiency of RdRp-mediated viral replication was quantified in the culture supernatants over 4 days post RNA transfection (4 d.p.t.). Transfection of the wild type (wt) Kc1/p6 Gluc into naïve Huh7.5 human hepatoma cells led to a ca. 34,000-fold increase in luminescence over mock cells. Transfection of a polymerase deficient genome harboring a mutation in the highly conserved catalytic triad (GDD motif of the RdRp (deemed pol [-])) expectedly did not augment Gluc activity (Figure 1B–C). Notably, genomes harboring mutations in the C483 and/or Y590 positions were incapable of establishing stable replication (Figure 1B–C).

We next sought to understand if this lack of viral replication brought on by mutating the highly conserved C483 was unique to the Kc1/p6 cell culture adapted strain of HEV, or if it translated to other known human-tropic HEV strains. Thus, we mutated the C483 residues in the Gluc reporter genome configurations of HEV strains SAR55 (genotype 1), SHEV3 (genotype 3), and TW6196E (genotype 4; Ding et al., 2018b), and transfected in vitro transcribed RNA into HepG2C3A human hepatoma cells similar to Figure 1B–C. In line with our observations using Kc1/p6 replicon, HEV RNA replication was severely impaired in SAR55, SHEV3, and TW6196E genomes harboring the C483A mutation (Figure 1D–G). Notably, Gluc levels were equivalently as low as the pol (-) versions of the reporter replicons.

To determine if this deficiency was due to a disruption in RNA folding, we mutagenized C483 into each available codon for cysteine and alanine (Figure 1H). These analyses demonstrated that encoding the alternate cysteine had a mild negative affect on viral replication efficiency, while any alanine codon usage brought replication levels down to those of the Pol (-) mutant (Figure 1H), suggesting that the deficiency primarily lies in protein folding or function. Collectively, these data demonstrate the necessity of these residues for viral fitness, despite Y590 being heterogeneous across HEV viral genotypes.

To probe further the mechanism underlying the functional impairments of the C483 mutants, we devised an experimental system in which HEV RNA replication is uncoupled from protein translation. Following our previously established successful transcomplementation approach for studying HEV ORF3’s viroporin function (Ding et al., 2017) and cis-regulatory elements responsible for regulating transcription of the subgenomic RNA (Ding et al., 2018b), we lentivirally expressed a wt, pol (-), or C483A version of ORF1, or the pPCP alone in HepG2C3A human hepatoma cells (Figure 2A). These cells were subsequently transfected with in vitro transcribed RNA from Kernow C1/p6 Gluc wt, pol (-), or C483A genome (Figure 2A). Gluc activity as a measure of the efficiency for HEV replication was quantified in the culture supernatants over 4 d.p.t. Notably, when the mock signal fold change over WT luciferase signal was examined, cells expressing a mutant form of ORF1 demonstrate a deleterious effect on WT replicon replication, likely due to competitive inhibitory binding of the mutant protein with the replicon RNA (Figure 2B). Impairments in viral genome replication due to the pol (-) or C483A mutations could be rescued in trans by expression of WT ORF1 near to levels of those following transfection of WT replicon RNA (Figure 2D–E). Of note, expression of the pPCP failed to restore replication of Kc1/p6 Gluc C483A, suggesting that the functions of this region of ORF1 are not adequately maintained outside of the context of the ORF1 polyprotein. This transcomplementation platform provides further means to uncouple the putative functions of the pPCP, or larger intramolecular regions of ORF1, e.g., polyprotein processing or modulation of the host cellular environment, from viral genome replication.

We examined the pPCP sequences across all eight known HEV genotypes (Figure 1—figure supplement 1) and noticed an octa-cysteine motif highly conserved across all HEV genotypes (Figure 3A). Previous work has shown the core hexa-cysteine motif encompassing cysteines 457–483 in SAR55 genotype 1 HEV were necessary for viral replication (Parvez, 2013). To determine the necessity of each cysteine to the viral replication cycle in the Kc1/p6 strain of HEV, we mutated each in turn to alanine within the Kc1/p6 Gluc reporter replicon and quantified the luciferase signal 4 d.p.t. (Figure 3B). We noticed that of the eight conserved cysteines, only the core six that form a CxC[x₁₁]CC[x₈]CxC hexa-cysteine motif are vital for viral replication, with the first cysteine at position C434 being completely dispensable for replication, and the final cysteine at position C563 having a slight detriment to replication when mutated to alanine (Figure 3B).

To determine more broadly which other regions of the HEV pPCP are indispensable for viral replication, we sought to conduct an unbiased genetic mutagenesis screen of the entire pPCP region of 160 amino acids. Site-directed triplicate alanine scanning mutagenesis was conducted across the entirety of the HEV pPCP, identifying several triplicates that offer pro-viral activity, as well as identifying the majority of triplicates vital for viral replicative fitness (Figure 3C). Notably, triplicates containing any of the conserved cysteines, as well as the variable amino acid at position 590 were indispensable for viral replication, despite point mutations at positions C434 and C563 being tolerated (Figure 3B). Additionally, the amino acid stretches 55–63 and 73–81 directly downstream of the hexa-cysteine motif tolerate mutagenesis quite well, suggesting a possible structural/linker function of these amino acids; these triplicates are not highly conserved across the eight HEV genotypes (Figure 1—figure supplement 1).

Interrogating the results of point mutations to the eight highly conserved cysteines within the pPCP led us to further bioinformatic analysis to determine the function of the vital hexa-cysteine motif. Utilizing motif searches with ScanProSite (de Castro et al., 2006) and TrEMBLE (Bairoch and Apweiler, 2000; Boeckmann et al., 2003; O’Donovan et al., 2002), we began by searching for a relaxed expression of the HEV hexa-cysteine motif (CxC[x]_3–20CC[x]_3–20CxC, where x can be any amino acid). From this analysis over 33,000 proteins emerged. To further refine our approach, we identified proteins that matched the HEV CxC motif exactly (CxC[x]₁₁CC[x]₈CxC, hereafter referred to as the HEV motif); however, all of the protein hits that emerged are as of now uncharacterized, offering little insight as to the function of this motif (Figure 3—figure supplement 1A). Relaxing the criteria to (±) 1 for each of the stretches of [x] (CxC[x]_(10–12)CC[x]_(7–9)CxC) brought forth 26 proteins with known functions (Figure 3—figure supplement 1A), enriched for proteins with divalent metal ion binding activity (Figure 3—figure supplement 1B). We hypothesized that this region within the pPCP is necessary for metal ion coordination.

The dearth of structural information of the HEV ORF1 protein has hampered the complete understanding of the viral replication cycle. To glean more information about the domain organization and protein folding of HEV ORF1, we turned to AlphaFold (Jumper et al., 2021) to predict the structure of ORF1. The complete sequence of HEV ORF1 was fed into the AlphaFold algorithm, and the best ranked model (Figure 4A) was chosen for further analysis. To gain confidence in the best ranked model, we analyzed the AlphaFold prediction in several ways. First, we analyzed the confidence levels produced by AlphaFold (predicted local distance difference test (pLDDT) score) for each residue across the ORF1 structure prediction (Figure 4B–C). This analysis revealed varying levels of confidence throughout the entirety of ORF1, and importantly, low confidence throughout much of the pPCP. We further looked at how AlphaFold predicts the pPCP outside of the context of ORF1 and found that the pLDDT averages are very similar (Figure 4C), with the pPCP alone averaging a pLDDT score of 65.92, and the pPCP within the context of ORF1 scoring a slightly better average of 66.05. Second, we tested how closely the AlphaFold prediction aligned with two separate solved structures of fragments of HEV ORF1, as well as the known solved structures of the macro domains in other distantly related viruses. To this end, we used a combinatorial approach of sequence-based alignments with structure-based alignments to gain the most accurate distance matrices of relevant atomic coordinates within each alignment of regions of ORF1 and the corresponding known structures (approach reviewed in Carpentier et al., 2019). To accomplish this, we took the ORF1 structure prediction, and using the tool TM-Align (Zhang and Skolnick, 2005) we aligned: a region of HEV ORF1 that spans parts of the pPCP and HVR (AAs 510–691) of the SAR55 strain of genotype 1 HEV (PDB: 6NU9; Proudfoot et al., 2019), the amphipathic ‘thumb’ of HEV genotype 3 strain 83-2-27 RdRp (amino acids 1628–1647; ORF1 kc1/p6 residues 1684–1709; Oechslin et al., 2022), and the macro domains of Sindbis virus (SINV amino acids 1342–1509 of PDB: 4GUA; Shin et al., 2012), and Chikungunya virus (CHIKV; PDB: 3GPG; Malet et al., 2009; Figure 4—figure supplement 1A–C, respectively).

Alignment of 6NU9 with the corresponding region of Kc1/p6 ORF1 shows high local alignment identity, with the average distance of this superimposition aligning at 0.67 Å (Figure 4—figure supplement 1A). Alignment of the amphipathic RdRp ‘thumb’ domain similarly shows incredibly high local alignment identity with a 0.48 Å average differential (Figure 4—figure supplement 1B). Understandably, alignment of the macrodomains of SINV and CHIKV also show high local alignment, though less robust than that of other HEV strains to Kc1/p6 ORF1 of HEV. Notably, the HEV macro domain and the macro domains of SINV and CHIKV share little amino acid similarity and identity, with the Kc1/p6 HEV macro domain sharing only 36.45% sequence identity and 54.21% sequence similarity with the SINV macro domain, and sharing only 35.58% sequence identity and 56.73% sequence similarity with the CHIKV macrodomain (Figure 4—figure supplement 2; Figure 4—figure supplement 2—source data 1). Despite this sequence disparity, these viral macrodomains share high structural identity, with the ORF1 and SINV superimposition aligning at 1.79 Å on average, and similarly, aligning with the CHIKV superimposition at 1.89 Å on average (Figure 4—figure supplement 1C).

Importantly, these aforementioned highly ordered secondary structures within HEV ORF1 all exhibit much higher AlphaFold pLDDT scores on average than the whole of the pPCP (Figure 4C). This prompts several considerations: (1) AlphaFold is not able to accurately predict the amino acid backbone or side chains of the pPCP with a high amount of confidence. (2) If this region were a protease domain, it would likely have a highly ordered and readily predicted secondary structure. To test this latter point, we analyzed AlphaFold’s ability to accurately predict the structure of a known protease, which it does very well. When comparing the known structure of the hepatitis A virus (HAV) 3 C protease (Bergmann et al., 1999) with the AlphaFold prediction, they align with an average superimposition distance of 1.03 Å (Figure 4—figure supplement 1D). Interestingly, AlphaFold was able to fold the HAV protease with very high confidence, with an average pLDDT score of 97.14 across the entirety of the prediction, giving high confidence of interpretation to both the amino acid backbones and side chain chains (Figure 4—figure supplement 1D). Due to the low pLDDT score of the pPCP, we visualized the predicted folding of this region by color-coding the outcomes of the alanine-scanning mutagenesis to better understand whether the secondary structures predicted by AlphaFold stand. We found that the majority of the predicted alpha-helices and beta-sheets do not tolerate mutations well, while the majority of the tolerated mutants lie within predicted regions of disorder (Figure 4D; Video 1). Taken together, these results suggest the structure predictions of HEV ORF1 by AlphaFold demonstrate high confidence within known regions that possess a high level of secondary structure. AlphaFold is able to predict the structure of a known viral protease with high confidence, suggesting further that the HEV pPCP domain does not possess the necessary secondary structure of a protease. This prediction of the HEV ORF1 pPCP is bolstered by the enrichment of replication tolerant triplets from the alanine scanning mutagenesis to preferentially localize to predicted areas of disorder (Figure 4D; Video 1), suggesting as a whole that the pPCP does not fold as a protease with a catalytic pocket.

Upon demonstrating that mutations within the pPCP domain prevent HEV from replicating, we sought to further elucidate potential mechanisms by which this deficiency occurs. Utilizing the structural predictions of HEV ORF1 generated with AlphaFold, we began by analyzing the predicted folding structure of WT ORF1, and two of the point mutants within the pPCP: C483A and C563A (chosen based on the heterogeneity of their phenotypes). C483A is fully replication deficient, while C563A is blunted in replication at two orders of magnitude lower than WT 4 d.p.t. (Figure 3B).

We next fed the sequences of the C483A and C563A mutant sequences of ORF1 into the AlphaFold algorithm. Upon inspection of the best ranked models for each of these versions of ORF1 (WT, C483A, and C563A), we noticed a novel pseudo-zinc-finger formed by the amino acids underlying the HEV hexa-cysteine motif, despite the low pLDDT scores within this region (Figure 5A, left; Videos 2–4, respectively). Mutation of C483A, but not C563A, is predicted to disrupt the molecular architecture of this pseudo-zinc-finger; however, mutation of either causes a relaxation of several predicted bond lengths to beyond biological relevance (Zheng et al., 2008; Figure 5—figure supplement 1, Videos 2–3). Cysteine side chain residues at positions 457 and 459 at the end of the predicted beta sheet leading into the pseudo-zinc-finger were shown to be projecting into inter-domain space. These residues are predicted to form a potential tetrahedral divalent ion binding pocket (Laitaoja et al., 2013) with D248 and H249 in the upstream Y-domain (Figure 5A, middle), which had pLDDT scores of 82.99 and 87.67, respectively, giving high confidence in the predicted location of the amino acid backbones and potentially their side-chains (Figure 4C). Cysteine, histidine, and aspartic acid residues are well known to bind divalent metal ions and form tetrahedral geometry (Laitaoja et al., 2013; Zheng et al., 2008). To further test whether this predicted interaction between C457, C459, D248, and H249 was vital to viral replication, we mutated either D248 or H249 to alanine in the Kc1/p6 Gluc replicon and quantified Gluc expression 4 d.p.t. D248 was dispensable for viral replication while H249 is not Figure 5—figure supplement 1; D248 is variable across all eight known HEV genotypes while H249 is highly conserved (Figure 5B). This observation led us to inquire as to what the disruptions these point mutations could have on ORF1 globally. We noticed that by comparing the structure of ORF1 WT (Video 2) next to the mutants while highlighting domains of ORF1 with well-defined functions such as the methyltransferase, helicase, and RdRp, structural differences emerge. Mutating C483A, C563A, or H249A are predicted to cause regions of the methyltransferase, helicase, and RdRp to reconfigure (Figure 5A, right; Videos 3 and 4, 6 respectively). Furthermore, mutating C563A, D248A, or H249A causes a predicted membrane association domain that is exposed in the WT ORF1 protein to become buried, possibly preventing the association with intracellular membranes and preventing/hindering the formation of a replication compartment (Metzger et al., 2022; Szkolnicka et al., 2019; Videos 4–6, respectively). Taken together, these results suggest that by interfering with divalent metal ion binding domains within the pPCP, structural domains vital to viral replication form aberrantly, and prevent HEV ORF1 from efficiently replicating.

Outside of our own analysis that suggests the pPCP of HEV ORF1 has metal ion binding activity, other groups have identified regions within ORF1 predicted to harbor Ca²⁺ and Zn²⁺ ion binding sites (Parvez and Khan, 2014; Proudfoot et al., 2019). However, the low abundance of ORF1 protein within infected cells and the lack of well-characterized ORF1 specific antibodies has hampered attempts at purifying a replication competent ORF1 protein. This barrier has only recently been overcome when tolerant epitope tag insertion sites were discovered within the HVR region of ORF1 and characterized (Metzger et al., 2022; Szkolnicka et al., 2019).

Utilizing an HA-tag insertion site flanked on either side by the linker sequence (AAAPG-HA-AAPG, hereafter referred to as HA-tagged) within the HVR of ORF1 (Szkolnicka et al., 2019), we generated an overexpression system of WT or mutant HA-tagged ORF1 via lentiviral transduction of Huh7 human hepatoma cells (Figure 6A). To determine if the tolerable or intolerable mutations within the pPCP at residues D248A or E489A, or at C483A, C563A, and H249A, respectively (Figure 3C), or the Pol (-) mutation in the RdRp affected ORF1’s ability to bind divalent ions, we immune purified each protein via the HA-tag mediated immunopurification (Figure 6B) and subjected each purified protein eluant to ICP-MS (Figure 6C; Figure 6—figure supplement 1). Element signatures indicate that while each mutant is heterogenous with each other in binding activity for most divalent ions, all mutations within the pPCP bind less Zn²⁺ than WT. Unsurprisingly, the element signature for the Pol (-) mutant bound all elements similar to WT (Figure 6—figure supplement 1). These results suggest that divalent ion binding capacity likely affects proper structural confirmation of ORF1, leading to differential replication capacity based on mutation position.

To determine whether the differential in divalent ion binding potential due to mutations within the pPCP or upstream Y-domain is responsible for a loss of subcellular localization, we turned to confocal microscopy. Cells bicistronically expressing zsGreen as a marker of transduction as well as WT ORF1 without an epitope tag, WT ORF1-HA-tagged, C483A-HA-tagged, C563A-HA-tagged, D248A-HA-tagged, or H249A-HA-tagged were imaged for zsGreen, and using antibodies against the HA-tag and nuclei were imaged to visualize ORF1 subcellular localization patterning. Cells expressing ORF1 WT-HA showed significant expression of ORF1 in both the cytoplasm and nucleus as previously reported (Metzger et al., 2022) and showed many puncta aggregates throughout the cytoplasm (Figure 7). In contrast, the replication deficient ORF1 C483A-HA expressing cells lost the ability for ORF1 to localize to the nucleus and was found dispersed throughout the cytoplasm (Figure 7). Cells expressing the C563A-HA mutant of ORF1 showed similar localization patterns to ORF1 WT-HA, with puncta forming in the cytoplasm and maintaining the ability to localize to the nucleus (Figure 7).

Cells expressing mutations in the novel upstream interacting Y-domain were also varied in their localization when compared to ORF1 WT-HA. Replication competent ORF1 D248A-HA, like C563A-HA, shared localization patterns with ORF1 WT-HA, forming cytoplasmic puncta and localizing within the nucleus (Figure 7). In contrast, replication deficient ORF1 H249A-HA showed very disperse cytoplasmic localization, no aggregate formation, and a lesser ability to localize to the nucleus (Figure 7). Taken together with the replication data of these epitope-tagged ORF1 mutants (Figure 5—figure supplement 1), localization of ORF1 seems to correlate with replicative capacity, with nuclear localization being lost or diminished in mutants that cannot replicate (C483A-HA and H249A-HA), and mutants able to establish replication (C563A-HA and D248A-HA) share localization patterns with ORF1 WT-HA. These results shed light on a potential novel mechanism by which mutants in the pPCP and upstream Y-domain interfere with HEV’s replication cycle.

HepG2C3A cells (ATCC and CRL-10741), Huh7, and Huh7.5 cells (kindly provided by Charles Rice, The Rockefeller University) were maintained in Dulbecco’s modified Eagle medium (Gibco, NY, USA) supplemented with 10% (vol/vol) fetal bovine serum, 50 IU/mL penicillin and streptomycin, in a humidified 5% (vol/vol) CO₂ incubator at 37°C.

The following HEV strains were used for sequence alignment: GenBank identifiers (IDs) or accession numbers M73218 (genotype 1 a, Burma strain), M74506 (2 a, Mex), JQ679013.1 (3,Kernow-C1/p6), AB197673 (4 a), AB573435 (5 a), AB856243 (6, wbJNN_13), KJ496144 (7, 180 C), and KX387866 (8, 48XJ). M73218 (genotype 1 a, Burma strain) was used as the reference strain for numbering. Sequence alignments were conducted with the SnapGene software (from Insightful Science; available at snapgene.com) using the multiple sequence alignment tool. HEV genotype alignments were conducted using CLUSTALW alignment algorithms embedded within the software.

Alignment of the HEV macro domain with the macrodomain of SINV (PDB: 4GUA) and CHIKV (PDB: 3GPG) were conducted with the SnapGene software (from Insightful Science; available at snapgene.com) using multiple sequence alignment tool with the local alignment Smith-Waterman algorithm.

The HEV hexa-cysteine motif CxC[x]₁₁CC[X]₈CxC motif sequence identified was searched using ScanProsite (de Castro et al., 2006) on all UniProtKB/Swiss-Prot (release 2020_02 of 22-Apr-20: 562253 entries), UniProtKB/TrEMBL (release 2020_02 of 22-Apr-20: 0 entries) databases sequences using the regular expression [C-X-C-X(3,20)-C-C-X(3,20)-C-X-C]. The regular expression allows from 3 up to 20 residues in the two long stretches of amino acid residues where X can be any amino acid shown by previous bioinformatic analysis and sequence alignments across HEV genotypes (Figure 1—figure supplement 1). The search produced 429 hits in SwissProt and 32,764 hits in TrEMBL. We further refined the regular expression [C-X-C-X(11)-C-C-X(8)-C-X-C] to match the HEV hexa-cysteine motif exactly and produced 25 protein hits with no known function. We refined this search yet again to include the regular expression [C-X-C-X(10,12)-C-C-X(7,9)-C-X-C], which allows (±) one residue in each long stretch of amino acids where X can be any amino acid, and found 26 proteins hits with known functions (Figure 3—figure supplement 1) and were used to predict the function of the motif sequence.

FASTA files of each species of ORF1 (WT or mutants) were submitted to the AlphaFold algorithm (Jumper et al., 2021; DeepMind, United Kingdom, v. 2.0.0--model preset = monomer; or in the case of the HAV 3 C protease, v. 2.1.1--model preset = monomer) run on the Princeton Research Computing DELLA Cluster at Princeton University. Five models of each protein prediction were produced, and the best ranked model for each was used for subsequent analysis.

Atomic distance calculations between the HEV ORF1 structure predications produced via AlphaFold (WT or point mutants) and known structures within ORF1 or distantly related viral macrodomains were done via the structure based alignment tool TM-Align (Zhang and Skolnick, 2005). Briefly, the known crystal structure of a region of the ORF1 pPCP/HVR (PDB: 6NU9 [Proudfoot et al., 2019]) was directly fed into the TM-Align software. The amphipathic ‘thumb’ of the RdRp solved via nuclear magnetic resonance (Oechslin et al., 2022), encountered a multi-mapping problem due to its alpha-helical nature and short amino acid sequence when fed directly into TM-Align; to generate the correct distance plot, the AlphaFold ORF1 prediction was trimmed of amino acids not corresponding to the region of the amphipathic RdRp thumb domain (Kc1/p6 AAs 1690–1708) utilizing UCSF Chimera. The trimmed ORF1 structure was aligned with the amphipathic thumb domain PDB, kindly provided by Jérôme Gouttenoire, using TM-Align. The distance plot generated between the AlphaFold ORF1 prediction and the macro domain of SINV was done by trimming a single chain of the trimer to remove amino acids outside the macrodomain of the SINV prediction P23pro-zbd (Shin et al., 2012; utilizing AAs 1342–1,509 of PDB: 4GUA) and aligning it to the ORF1 prediction using TM-Align. The CHIKV macrodomain PDB file 3GPG (Malet et al., 2009) was edited to remove the three additional chains comprising the hetero-tetramer and aligned with the ORF1 prediction.

Molecular graphics and analyses performed with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311 (Pettersen et al., 2004).

Note: Kernow c1/p6 strain contains an s17 insertion within the HVR, so amino acid positions shift downstream of this insertion at amino acids 751–806.

To construct lentiviral constructs encoding ORF1 of Kernow C1/p6 (GenBank accession number JQ679013), the Kernow C1/p6 ORF1 cDNA was amplified by PCR from a plasmid encoding the FL infectious HEV clone Kernow C1/p6 (kindly provided by Suzanne Emerson, NIH) and then cloned into pLVX-IRES-zsGreen1 vector using an In-Fusion HD cloning kit (Clontech, Mountain View, CA, USA). The GAD mutant of ORF1 inactivating the polymerase was generated by QuikChange (Stratagene) site-directed mutagenesis. The HEV Kernow-C1 p6-Gluc (Shukla et al., 2011) and pSAR55-GLuc were kindly provided by Suzanne Emerson and Patricia Farci. pGEM-9zf-pSHEV3 and pGEM-7Zf(-)-TW6196E encoding the infectious pSHEV3 (gt 3) and TW6196 (gt 4) clone, respectively were gifts from X.J. Meng. Site-directed mutagenesis of these plasmids for C483A or Pol(-) mutants was obtained with the QuikChange kit (Stratagene) using primers listed in Supplementary file 1. HEV Kernow-C1 p6-Gluc was used to generate the triplicate mutants for the alanine scanning mutagenesis of the entire pPCP domain, as well as the C434A, C457A, C459A, C471A, C472A, C481A, C483A, C563A, C483S, C483C (TGT), Y590A, Y590F, C483A-Y590A, C483G-Y590L, C483A (GCT), C483A (GCA), C483A (GCG), D248A, and H249A point mutants by QuikChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA). All primers used for site-directed mutagenesis can be found in Supplementary file 1. AAAPG-HA tag-AAAPG insert was generated by amplifying out the HA tag sequence from pLVX ORF2-HA using primers listed in Supplementary file 1. pLVX Kernow C1/p6 ORF1 AAAPG-HA tag-AAAPG IRES zsGreen was generated via XhoI and XbaI digestion of pLVX IRES zsGreen plasmid and In-Fusion HD cloning of ORF1 AAAPG-HA tag-AAAPG from p6/BSR-2A-ZsGreen AAAPG-HA tag-AAAPG HVR using primers listed in Supplementary file 1. Overexpression constructs for ORF1 AAAPG-HA tag-AAAPG point mutants C483A, C563A, D248A, and H249A were generated from the parent pLVX Kernow C1/p6 ORF1AAAPG-Ha tag-AAAPG IRES zsGreen construct via QuikChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) using primers listed in Supplementary file 1.

The generation of p6/BSR-2A-ZsGreen was described previously (Nimgaonkar et al., 2021). The generation of pSK-SAR55-Gluc, pGEM-9Zf-pSHEV3-Gluc, and pGEM-7Zf(-)-TW6196E-Gluc reporter constructs were described previously (Ding et al., 2018b). Generation of pSK-Kernow AAAPG-HA tag-AAAPG HVR GLuc was conducted via PCR linearization of pSK Kernow WT GLuc and In-Fusion HD cloning of AAAPG-HA tag-AAAPG cloned out of p6/BSR-2A-ZsGreen AAAPG-HA tag-AAAPG HVR using primers listed in Supplementary file 1.

All DNA fragments were cloned into the respective vectors using an In-Fusion HD cloning kit (Clontech, Mountain View, CA, USA). All constructs or primers used to construct the HEV reporter genomes in Supplementary file 1 have been validated through Sanger sequencing and are available upon request.

HEV Kernow-C1 p6-Gluc, HEV Kernow-C1 p6 C483A-Gluc, HEV Kernow-C1 p6 GAD-Gluc, and all Kernow point mutant and alanine scanning triplite construct plasmids were linearized by MluI. pSAR55-Gluc, pSAR55 C483A-Gluc, and pSAR55 GAD-Gluc were linearized by BglII, pGEM-9Zf-pSHEV3-Gluc, pGEM-9Zf-pSHEV3 C483A-Gluc, and pGEM-9Zf-pSHEV3 GAD-Gluc were linearized by XbaI, and pGEM-7Zf(-)-TW6196E-Gluc, pGEM-7Zf(-)-TW6196E C483A-Gluc, and pGEM-7Zf(-)-TW6196E GAD-Gluc were linearized by SpeI. All capped viral RNA was in vitro transcribed from the corresponding linearized plasmid using the HiScribe T7 antireverse cap analog mRNA kit (New England Biolabs, Ipswich, MA, USA) according to the manufacturer’s protocol. In vitro transcribed viral RNA was purified by LiCl precipitation following DNAse1 digestion. In vitro transcribed viral RNA was transfected into HUH7, HUH7.5, or HepG2C3A cells via the TransIT-mRNA transfection reagent (Mirus Bio LLC, Madison, WI, USA) according to the manufacturer’s instructions.

Gluc activity was determined using Luc-Pair Renilla luciferase HS assay kit (GeneCopoeia, Rockville, MD, USA). Specifically, 10 µl of harvested cell culture medium was added per well of a 96-well solid white, flat-bottom polystyrene microplate (Corning, NY, USA), followed by the addition of Renilla luciferase assay substrate according to manufacturer protocol, and the detection of luminescence was performed using a Berthold luminometer (Bad Wildbach, Germany).

Immunopurification of HA tagged proteins for ICP-MS analysis was conducted using the Pierce MS-Compatible Magnetic IP Kit, protein A/G (Thermo Scientific, Waltham, MA, USA, Catalog number 90409). 750 µg of crude protein lysates of stably expressing ORF1 cells (WT or mutants) were subjected to each round of IP for subsequent analyses (ICP-MS or western blot). IP was done with rabbit anti-HA antibody (Cell Signaling Technology, Danvers, MA, USA, catalog number C29F4) at a ratio of 1:50 in a laminar flow tissue culture hood that was sterilized with 70% ethanol and washed with MilliQ water prior to IP. The IP was conducted according to the manufacturer’s instructions, and each collected fraction (crude cell lysate, unbound fraction, and IP eluate) was split into large (80% total volume) and small (20% total volume) fractions. The large fractions were used for downstream nitric acid/hydrogen peroxide digestion and ICP-MS analysis, whereas the small fraction was used for protein quantification and western blot analysis. This was necessary to reduce chances of contaminating divalent ions being introduced into the sample during non-ICP-MS characterization.

Eluted protein samples were freeze dried and placed at –80°C until nitric acid digestion. Briefly, protein samples were resuspended in 1.5 mL Nitric Acid 67–69% Optima, for Ultra Trace Elemental Analysis (Fisher Chemical, Fair Lawn, NJ, USA) in ICP-MS grade Teflon at 80°C for 24 hr in a laminar flow hood. 500 µL Hydrogen Peroxide (Optima, Fisher Chemical, Fair Lawn, NJ, USA) was added to each sample and incubated at 80°C for 24 hr in a laminar flow hood, when the samples were subsequently dried for 24 hr at 40°C. Fully digested samples were then resuspended in 1 mL 2% (vol/vol) nitric acid dilutes with MilliQ water, diluted in 1×, 2×, and 10× dilution series, and interrogated via ICP-MS (iCap, Thermo Scientific, Waltham, MA, USA).

Samples were quantified via a single quadrupole iCap ICP-MS (Thermo Scientific, Waltham, MA, USA). Briefly, 1 mL of 2% (vol/vol) nitric acid running buffer blank was measured, followed by 1 mL per dilution of a dilution series (1×, 2×, 5×, 10×, and 50×) of the certified elemental standard 1643 F (National Institute of Standards and Technology, Gaithersburg, MD, USA) to generate an elemental calibration curve. All elements were measured in standard (STD) mode with the exception of lithium and iron, which were measured in kinetic mode to remove unwanted polyatomic interferences with the argon plasma. Samples were bracketed by an additional blank and STD curve to monitor instrument drift and ensure consistency throughout each experimental run. Samples were run via a series of three dilutions per sample, per run. Any measurement for a sample within the dilution series that fell outside of the dilution series range was excluded from further analysis. Contamination of trace elements from reagents and MilliQ water was monitored by processing MilliQ ‘samples’ through the entire immunopurification and protein digestion protocol on a run-by-run basis. Any trace metal analyses that were not statistically different from the MilliQ control ‘sample’ were excluded from further consideration. Iron and zinc ions were binned prior to statistical analysis for ease due to having with multiple isotopes.

Naïve HUH7 cells or HUH7 cells expressing ORF1 WT, WT-HA-tag, C483A-HA-tag, C563-HA-tag, D248A-HA-tag, or H249A-HA-tag were seeded onto separate glass coverslips (#1.5; 10 mm; Thomas Scientific, Swedesboro, NJ, USA) in a 24-well plate at 100,000 cells per well. 2 days post seeding, the cells were fixed with 4% PFA for 15 min and subsequently permeabilized in 0.25% Triton x-100 for 15 min. The rabbit anti-HA tag, C29F4 (Cell Signaling Technology, Danvers, MA, USA) primary antibody was used at a ratio of 1:1000 (V/V), and the AlexaFluor647 (goat anti-rabbit IgG [H+L], ThermoFisher Scientific, Waltham, MA, USA) secondary antibody was used at a final concentration of 1 µg/mL. All antibodies were diluted with PBS and incubated for 40 min at room temperature (RT). Hoechst 33342 (ThermoFisher Scientific, Waltham, MA, USA) was incubated at a final concentration of 1 µg/mL for 10 min at RT. The coverslips were then mounted onto glass microscopic slides (VWR International, Radnor, PA, USA) with 5 µL of ProLong gold antifade reagent (ThermoFischer Scientific, Waltham, MA, USA). The stained samples were imaged using the Nikon A1R-Si microscope (Nikon, Melville, NY, USA) in the Princeton University Confocal Microscopy Facility. The images were taken at 40× magnification. Images were then analyzed using Fiji (ImageJ2) image analysis software.

All statistical analyses were performed using GraphPad Prism software version 9.3.1. One-way ANOVA with Dunnett’s multiple comparison analysis or Brown-Forsythe one-way ANOVA with Dunnett’s T3 multiple comparison analysis tests were used to test for statistical significance of the differences between the different group parameters in experiments utilizing the Gluc reporter replicon. p Values of less than 0.05 were considered statistically significant. All data sets were analyzed for and cleaned of outliers using the robust regression and outlier removal (ROUT) method.

All materials generated by the Ploss lab will be available upon request from the corresponding author.