An emergent Ebola virus nucleoprotein variant influences virion budding, oligomerization, transcription and replication

To investigate how Ebola virus phenotypes changed during the 2013‒2016 Western African Ebola virus disease epidemic, we examined a key viral mutation that rose to high frequency: an R111C substitution in the viral nucleoprotein (NP). Though NP plays many essential roles during infection, there are a limited number of assays for studying these functions. We developed new reporter assays to measure virion-like particle (VLP) production and NP oligomerization in live cells under biosafety level 2 conditions. We found that NP-R111C significantly enhanced VLP production and slightly increased NP oligomerization without impairing viral transcription and replication. By contrast, a synthetic charge-reversal mutant, NP-R111E, greatly increased oligomerization but dramatically reduced transcription and replication. We detected an interaction of NP with the cellular clathrin adaptor protein-1 (AP-1) complex, which may explain how NP facilitates VLP production. Our study provides enhanced methods to study NP and indicates a complex interplay between NP’s roles in virion budding, protein structure, and transcription and replication.

Given the past and current public health threats caused by Ebola virus disease (EVD) 2 outbreaks, rapidly evaluating whether Ebola virus (EBOV) genomic mutations change viral 3 phenotypes is critically important. As an RNA virus, EBOV generates many mutations over the 4 course of an outbreak. The vast majority of these mutations likely will not be adaptive and will 5 instead have negligible or negative effects on EBOV viability and replication (Holmes 2009). Makona clades, which emerged as the outbreak was accelerating in May 2014. Each of the 20 clades, termed "SL1" through "SL4," descended from one another sequentially (e.g., SL2 21 derived from SL1), with a subsequent increase in EVD cases. Because non-synonymous 22 mutations directly change protein sequence, we focused on the only two non-synonymous 23 clade-defining mutations-one each defining the SL1 and SL2 clades. 24 25 membrane around them into EBOV virions, which exit the cell by budding (Luke D. Jasenosky 1 and Kawaoka 2004). Expression of VP40 in the absence of other viral proteins generates 2 similarly shaped particles, dubbed virion-like particles (VLPs) (Harty et al. 2000; L. D. Jasenosky 3 et al. 2001;Noda et al. 2002). Co-expression of VP40 with NP or other viral proteins 4 significantly increases the number of VLPs in cellular supernatant (Licata et al. 2004), 5 suggesting that NP plays a structural role in assembling and stabilizing VLPs. Mutations in NP 6 that affect its ability to enhance VLP production are not known. 7 EBOV NP also plays an essential role in viral transcription and replication. By directly 8 interacting with EBOV RNA, VP35, and VP30, NP recruits L to enact both of these essential 9 functions (Groseth et al. 2009). Based on EBOV NP structural data, homology modeling versus 10 other viral nucleoproteins, and site-directed mutagenesis experiments, key EBOV NP residues 11 that interact with EBOV RNA (Dong et  However, the oligomerization domain (OD) at the very N-terminus of NP has eluded 20 crystallization, presumably due to its disordered structure (Su et al. 2018). Alterations to NP 21 oligomerization could affect virion assembly and transcription and replication, but the interplay of 22 these functions is not obvious. Aside from deletion of the oligomerization domain (∆OD), no 23 other mutations are known to affect EBOV NP oligomerization, in part because oligomerization 24 is a challenging phenotype to assay in cell culture. 25 To illuminate why NP-R111C increases VLP production, we developed an assay to measure 2 intracellular NP oligomerization using bioluminescence resonance energy transfer (BRET) . 3 Traditional oligomerization assays in cell culture involve tagging a protein separately with two 4 different tags, co-expressing both, and then co-immunoprecipitation (co-IP) targeting one tag 5 and WB targeting the other tag (Watanabe, Noda, and Kawaoka 2006;Ng et al. 2012;Ortiz-6 Riano et al. 2012). However, WB often has linear dynamic range issues; furthermore, co-IPs 7

Ebola virus nucleoprotein position 111 significantly affects oligomerization of NP
can introduce non-specific or spurious protein-protein interactions under different cell lysis and 8 binding buffers. To overcome these deficits of co-IPs and WBs, we used BRET to study NP 9 oligomerization in live cells. We tagged NP with either NLuc or HaloTag (which covalently binds 10 to an acceptor fluorophore), co-expressed the tagged NPs in cells, and activated the NP- NLuc 11 with substrate, resulting in emission of light at 465 nm. Spatial proximity of NP-NLuc to NP-12 HaloTag due to NP oligomerization results in energy transfer and a second light emission at a 13 longer wavelength, 625 nm ( Figure 3A) (Machleidt et al. 2015). 14 To verify that our assay was truly measuring NP oligomerization, we generated NP LOF 15 mutants and disrupted oligomerization with biologically relevant EBOV VP35. We first generated 16 NP-∆OD, an LOF mutant that size exclusion chromatography and multiangle light scattering 17 indicate to be defective in oligomerization (Kirchdoerfer et al. 2015). Then, we confirmed that 18 the mutant lost oligomerization capability using the traditional dual-tag co-IP-WB strategy 19 ( Figure S3). We then performed our BRET assay in live cells and, as expected, the lack of NP-20 NLuc or NP-HaloTag, expression of NP-∆OD, or free NLuc reduced BRET signal appreciably 21 ( Figure 3B). To confirm our assay in a biologically relevant context, we additionally expressed 22 the NP-binding peptide (NPBP) of EBOV VP35, which disrupts NP oligomerization (Kirchdoerfer 23 et al. 2015). To quantitatively detect VP35 NPBP expression, we fused eGFP to NPBP via a 24 bridging porcine teschovirus 1 2A 'self-cleaving' peptide (Kim et al. 2011) (eGFP-P2A-25 VP35(NPBP)). Titrating increasing amounts of VP35 NPBP led to a quantitative decrease in 1 BRET oligomerization signal (residual ÷ total sum of squares = 0.95) ( Figure 3C). 2 Next, we measured the propensity of NP-R111 variants to form oligomers and found that 3 NP-R111C, and to an even greater extent NP-R111E and NP-K109E/K110E/R111E, increased 4 NP oligomerization. To quantify changes in oligomerization, we titrated an increasing 5 concentration of acceptor NP-HaloTag to saturate the donor NP-NLuc signal. The resulting 6 binding curves fit well to Michaelis-Menten kinetics, described by the parameters v max (maximum 7 signal) and K m (concentration of NP-HaloTag needed to reach half v max ) ( Figure 3D). As 8 expected, control eGFP substituted for NP resulted in no detectable BRET signal, and NP-∆OD-9 HaloTag resulted in background signal detected only at high concentrations ( Figure 3D). 10 Relative to NP-R111, NP-R111C increased oligomerization slightly (12% lower K m ; adj. p < 11 0.003, Dunnett's test), whereas the charge-reversed NP-R111E (36% lower K m ; adj. p < 0.0001, 12 Dunnett's test) and NP-K109E/K110E/R111E (28% lower K m ; adj. p < 0.0001, Dunnett's test) 13 mutants oligomerized at even lower concentrations ( Figure 3D). These results support our 14 hypothesis that the NP 111 residue affects NP-NP interactions ( Figure 1D, S1A). Indeed, 15 deuterium exchange mass spectrometry indicates that this residue is partially buried in wild-type 16 NP compared to an oligomerization-incompetent NP (Su et al. 2018 To build upon these previous results, we performed immunoprecipitation tandem mass 3 spectrometry (IP-MS/MS) using myc-tagged NP from EBOV/Makona bearing either R111 or NP-4 K109E/K110E/R111E. Our approached yielded multiple members of the adaptor related protein 5 1 (AP-1) complex as strong candidate interactors ( Figure 4A, right), which were identified 6 previously (García-Dorival et al. 2016) but were not further confirmed in that study. Here, we 7 confirm that both NP-R111 and NP-K109E/K110E/R111E strongly interact with AP-1 subunit M1 8 (AP1M1) and AP1G1 by reciprocal IP-WB ( Figure 4B). Yet, NP-R111C, NP-R11E, and NP-9 K109E/K110E/R111E all bind to the AP-1 complex with similar affinity as NP-R111 ( Figure 4C), 10 suggesting that the AP-1 interaction does not explain why the epidemic mutation NP-R111C 11 produced more VLPs than the ancestral NP-R111. The mechanism by which changes in NP's structural phenotypes (budding, 16 oligomerization) affect viral transcription and replication is not obvious because NP is highly 17 multi-functional. We quantified viral transcription and replication using a minigenome reporter 18 assay (Luke D. Jasenosky, Neumann, and Kawaoka 2010). In this assay, we express the 19 components of the EBOV RNP complex (NP, VP35, VP30, and L) in the presence of a 20 'minigenome' consisting of a reporter FLuc-encoding gene flanked by the EBOV promoter-like 21 genomic leader and trailer sequences. Transcription is essential for minimal FLuc activity; 22 replication is further required to achieve maximum signal (T. Hoenen et al. 2010). 23 Intriguingly, we found that NP-R111C caused similar transcription and replication activity 24 as NP-R111, whereas the charge-reversal NP-R111E abrogated these activities. As expected, 25 absence of VP30, L, or the minigenome resulted in <5% normalized luminescence compared to 1 cells expressing the minigenome and the entire RNP complex with NP-R111 ( Figure 5). 2 Substitution of NP-R111C in place of NP-R111 yielded similar activity (99%). On the other hand, 3 the charge-reversal mutants NP-R111E (23% reporter activity; adj. p < 0.003; Dunnett's test) 4 and NP-K109E/K110E/R111E (44% reporter activity; adj. p < 0.017; Dunnett's test) greatly 5 attenuate transcription and/or replication. These results indicate that the NP 111 residue is 6 connected to both the structural and transactivation roles of EBOV NP. 7

1
Here, we developed and modified BSL-2 assays to study in-depth a key EBOV NP mutant, NP-2 R111C, which arose during the 2013-2016 Western African EVD epidemic. Though the NP-3 R111 residue has not been previously annotated as functional, NP-R111C increases VLP 4 production and NP oligomerization, and the charge-reversal mutation NP-R111E dramatically 5 increases NP oligomerization while hindering viral transcription and replication. 6 Many viral proteins are highly multi-functional, making study of individual mutations 7 challenging without high-throughput, robust assays that are sensitive to subtle changes in viral 8 phenotype. Since luciferase-based reporter systems fit the aforementioned requirements, we 9 took advantage of these systems to develop VLP detection assays and BRET assays for NP 10 oligomerization. As new assays always require thorough testing, we have verified that luciferase 11 activity in these assays indeed reflects phenotype using LOF mutations (NLuc-VP40-L117R for 12 VLP production, NP-ΔOD to assess oligomerization) and biologically relevant disruptions 13 (heating to denature VLPs, VP35 expression to reduce NP oligomerization). These BSL- 2 14 assays are simple and flexible for testing new viral mutations as they emerge during epidemics. 15 With more rigorous screening and quantification of key metrics of variability, like Z-factor, these 16 assays could potentially be used for high-throughput screens of hundreds of EBOV NP mutants, 17 interactions with host factors, or antagonism by drug candidates. 18 Although these reporter assays show that NP-R111C increases VLP production, the 19 mechanism behind this increase remains unclear. In this study, we used co-IP-MS/MS to 20 identify many new putative NP binding partners, since there has only been a single previous NP-R111 and NP-R111C bound to AP-1 with similar affinity, suggesting that this interaction was 4 not the reason behind NP-R111C's increased ability to promote VLP production. To identify 5 protein-protein interactions in this study, we used a standard label-free co-IP-MS/MS approach. 6 To gain deeper mechanistic insight into this complex NP surface, researchers of future studies 7 could apply more powerful label-based quantification approaches to identify differential 8 interacting partners between NP-R111, NP-R111C, and NP-R111E. 9 Charge reversal at the NP residue R111, NP-R111E, further demonstrates the 10 importance of this position to multiple essential viral life cycle functions. We show that NP-11 R111E is not aberrantly misfolded and degraded since it expresses similarly as NP-R111 and 12 NP-R111C. In fact, we find that NP-R111E oligomerizes at significantly lower protein 13 concentration compared to NP-R111, yet NP-R111E is unable to support normal levels of viral 14 replication and transcription. Basicity at residues K109, K110, and R111 is highly conserved 15 among nearly all immediate relatives of EBOV in the genus Ebolavirus, including the newly 16 discovered Bombali virus (Goldstein et al. 2018 Because NP-R111E phenocopies the triple charge-reversal mutant, it is possible that NP-25 R111E disrupts the K110-E349 interaction as well. Intriguingly, the epidemic substitution NP-26 R111C slightly increases oligomerization compared to NP-R111, but to a lesser extent than NP-1 R111E. Biochemically, 14% of cysteine residues (pK a = 8.18) (Nelson, Lehninger, and Cox 2 2008) will be negatively charged at a typical intracellular pH of 7.4, in between lysine (100% 3 positively charged) and arginine (100% negatively charged). Yet, NP-R111 and NP-R111C 4 produce similar levels of minigenome transcription and replication, while NP-R111E and NP-5 K109E/K110E/R111E are significantly ablated. This correlation between side chain charge, 6 oligomerization, and transcription and replication hints at the possibility of subtle shifts in the NP 7 structure and/or additional electrostatic interactions that coordinate NP's ability to influence all 8 these different functions simultaneously. 9 Findings from reporter-based assays are chiefly limited because such assays must still 10 be supported by live virus experiments. Viral proteins like NP have multiple essential and 11 accessory roles during infection, and despite our best efforts, we did not assay every function. 12 There are innumerable molecular phenotypes that could not all be assayed (e.g., alterations to 13 NP protein structure, localization, interaction binding sites, immune epitopes). Even for the potential global threat of EVD epidemics, using BSL-2 model systems allows initial and more 7 rapid and comprehensive exploration of any potentially consequential EBOV mutations. 8 Our finding of the importance of the R111 residue and experimental systems established 9 represent steps towards characterizing key EBOV mutations that arose during the 2013-2016 10 Western African EVD epidemic. These findings provide additional insight into the interplay 11 between the many functions of NP in viral assembly and budding, oligomerization, and 12 transcription and replication. 13   6 We performed all assays with the same mammalian expression vector for EBOV NP and its 7 mutants. We synthesized EBOV NP-R111 in 2 dsDNA gBlocks (Integrated DNA Technologies 8

Constructs and cloning
[IDT], Coralville, IA) and cloned these into pGL4. 23 Opti-MEM for 20 min at room temperature, and then added the mixture dropwise onto cells in [6][7][8][9][10][11][12][13][14][15][16][17][18][19] or 12-well plates. 20 21 Virion-like particle (VLP) budding assay 22 We grew cells to near confluency, harvested following trypsinization, reverse-transfected, and We prepared samples for EM based on our previously described protocol (Gao and 11 Hendricks 2012). Briefly, we performed all spreads on freshly prepared Carbon stabilized 12 Formvar Support films on 200 mesh copper grids. We adsorbed VLPs onto a carbon-coated 13 Formvar support films for 30 s. We removed excess liquid with filter paper and negatively 14 stained the samples immediately by running 6 drops of 1% uranyl acetate over the grid to 15 contrast the VLPs. We removed excess stain and air-dried the samples in a controlled humidity 16 chamber. We then examined the samples using a FEI Tecnai 12 Spirit BioTwin transmission 17 electron microscope (Thermo Fisher Scientific) using an accelerating voltage of 120 Kv. We Co-immunoprecipitation (co-IP) 23 We washed cells in 6-well plates with PBS, harvested by scraping, pelleted, and resuspended For reciprocal co-IP experiment in Figure 4B, we instead captured protein complexes 12 with Protein A/G PLUS-Agarose beads (Santa Cruz Biotechnology, Dallas, TX). In this setup, 13 we first incubated cleared cell lysate with 1-2 µg of primary antibody, rotated at 4 °C for 2-4 h, 14 and then added 40 µL of protein A/G agarose beads, and rocked at 4 °C overnight. We washed 15 bead-antibody complexes four times with wash buffer, twice with PBS, and eluted proteins as 16 described above. 17 For the dual-tag co-IP-WB for NP oligomerization, we used RIPA buffer (50 mM Tris pH 18 6.8, 150 mM NaCl, 0.5% (w/v) sodium deoxycholate, 1% (w/v) Triton X-100 (Sigma-Aldrich)) as 19 both the lysis and wash buffer because the NP-NP interaction is very strong (Watanabe, Noda, 20 and Kawaoka 2006). 21 22 Western blot (WB) 23 We loaded the specified amount of input into 10% acrylamide SDS gels, and ran at 180 V until 24 complete. We transferred protein to Immun-Blot PVDF membranes (Bio-Rad) in a wet tank 25 either at 200 mA for 1.5 h at 4 °C, or at 40 V overnight at 4 °C. We blocked membranes by 26 rocking in blocking buffer consisting of 5% non-fat dry milk (Santa Cruz Biotechnology) 1 dissolved in tris-buffered saline with 0.1% Tween 20 (TBS-T) buffer for 1 hr at room 2 temperature. We incubated membranes with primary antibody in blocking buffer for 45 mins, 3 washed the membrane three times in TBS-T, incubated with HRP-conjugated secondary 4 antibody in blocking buffer for 1 hr, and washed the membrane three times. We detected 5 chemiluminesence with SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher 6 Scientific), and imaged with an AlphaInnotech ChemiImager (ProteinSimple, San Jose, CA) or 7 FluorChem E (ProteinSimple) CCD camera. 8 9 10 BRET NP oligomerization assay 11 We grew cells to near confluency, harvested by trypsinization, reverse-transfected, and plated To ensure that each well received the same total amount of DNA, we also co-transfected 11 increasing amounts of control pcDNA3.3/eGFP plasmid. We serially diluted pGL4.23-CMV/NP-12 HaloTag or pcDNA3.3/eGFP in control pcDNA3.3/eGFP plasmid, as described in the 13 manufacturer's protocol. We collected 6 biological replicates for each NP mutant and controls. 14 15 Minigenome assay 16 Screening of NP mutants was done as described in (Luke D. Jasenosky, Neumann, and 17 Kawaoka 2010). We seeded HEK 293T cells into 12-well plates, grew to 70% confluence, and To assess protein-protein interactions of NP, we scaled up our co-IP protocol. We grew 14 two 15-cm 2 plates of cells to 40-60% confluence and transfected with 32 µg of pGL4.  CMV/NP-myc encoding either the NP-R111 or the NP-K109E/K110E/R111E mutants. After 48 16 h, we harvested cells by scraping and lysed in 2.5 mL of mild lysis buffer. 17 We performed co-IP of myc-tagged NP complexes using 25 µg of mouse α-myc IgG or 18 irrelevant normal mouse IgG at 4° C overnight, and bound complexes to 250 µL of protein A/G 19 agarose beads at 4° C for 2 h. We washed beads as described above and eluted proteins in 20 120 µL of Laemmli sample buffer at 95 °C for 10 min. We separated proteins by SDS-PAGE, 21 visualized with PageBlue Protein Staining Solution (Thermo Fisher), and excised lanes 22 excluding IgG chains. 23 We cut gel bands into approximately 1-mm 3 pieces and performed a modified in-gel 24 trypsin digestion procedure (Shevchenko et al. 1996). We dehydrated pieces with acetonitrile for 25 10 min, dried them completely in a speed-vac pump, and rehydrated with 50 mM ammonium 26 bicarbonate solution containing 12.5 ng/µl of modified sequencing-grade trypsin (Promega) at 4 1 °C for 45 min. To extract peptides, we replaced the solution with 50 mM trypsin-free ammonium 2 bicarbonate solution and incubated at 37 °C overnight. We washed once with 50% acetonitrile 3 and 1% formic acid, dried in a speed-vac pump for ~1 h and then stored at 4 °C. On the day of 4 analysis, we reconstituted peptides in 5-10 µl of high-performance liquid chromatography 5 (HPLC) solvent A (2.5% acetonitrile, 0.1% formic acid). We packed nano-scale reverse-phase 6 HPLC capillary columns with 2.6 µm C18 spherical silica beads into fused silica capillary tubes 7 (100 µm inner diameter x ~25 cm length) using flame-drawn tips ( 10 analysis (Snel 2000;Szklarczyk et al. 2015) and visualized interactions with Cytoscape 25 (Shannon et al. 2003). See Supplementary file S1 for raw and filtered peptide/protein PSM 1 counts. 2 3 Statistical analysis 4 We performed all hypothesis testing using Prism 7 (GraphPad Software, La Jolla, CA) and all 5 non-linear curve fitting using R (R Core Team 2016) and the 'nlstools' package (Baty et al. 6 2015). We generated most plots using the 'ggplot2' package in R (Wickham 2016). 7 We quantified raw NLuc intensities from VLPs produced from expression of NLuc-VP40, 8 NLuc-VP40-L117R, or NLuc alone with n = 6 biological replicates each. To assess statistical 9 significance, we performed a repeated measures ANOVA (rANOVA) with Dunnett's post-test in 10 which each condition was compared to NLuc-VP40 to generate an adjusted p-value. 11 To measure the impact of NP genotype on VLP production, we co-expressed NLuc-12 For the VLP35 inhibition using BRET, we expressed varying amounts of VP35 NP 4 binding peptide in the presence of NP-NLuc and NP-HaloTag with n = 3 biological replicates for 5 each VP35 expression level. We fit the inverse function to data using the 'nls' function in R: 6 BRET ~ scale/(VP35 + max) + min (2)  7 , where VP35 expression is the independent variable, BRET is the dependent variable, and 8 scale, max, and min are constants to be fitted. In the non-linear regression, scale = 1.9 x 10 5 , , where the concentration of NP-HaloTag is the independent variable, BRET is the dependent 24 variable, and v max and K m are constants to be fitted. For NP-R111 or each NP mutant or eGFP 25 control, we inferred v max and K m and generated 95% confidence intervals using 'nlstools' as 1 described above. Data points from eGFP and NP-ΔOD controls failed to generate appropriate 2 curve fits. To determine whether the remaining curve fits were significantly different from each 3 other, we performed ANOVA with Dunnett's post-test in which NP-R111C, NP-R111E, and NP-4 K109E/K110E/R111E were compared to NP-R111 to generate an adjusted p-value using Prism 5 7.   2 3 Supplementary files 4 • Supplementary file S1. List of all proteins identified by co-immunoprecipitation tandem mass 5 spectrometry (co-IP-MS/MS), related to Figure 4A. We expressed V5-tagged Ebola virus 6 NP-R111 or NP-K109E/K110E/R111E in HEK 293FT cells, and performed co-IP-MS/MS 7 with either normal mouse IgG control or ɑ-V5 mouse IgG. We matched MS/MS 8 fragmentation spectra to human forward protein databases and against reverse databases 9 to a 1-2% false discovery rate using the SEQUEST database search program (Thermo 10 Fisher) (Eng, McCormack, and Yates 1994). We computed unique and total peptide spectra Data availability 24 25 All data generated or analyzed during this study are included in the manuscript and supporting 1 files. 2 domain has yet to be determined by crystallography (orange dashed line), the R111 residue 15 (yellow) is located on the same face as residues proximal to the oligomerization domain 16 (orange), but opposite to the VP35 (magenta) and RNA (red) interaction interfaces. 17

Figure 2. Ebola virus nucleoprotein mutation R111C increases budding of virion-like 19
particles. 20 (A) Schematic of the virion-like particle (VLP) budding assay. We transfect plasmid encoding 21 NLuc-VP40 to form luminescent VLPs, and co-transfect NP-expressing plasmids to measure the 22 impact of NP genotype on VLP budding. 23 (B) VLP budding assay control. VP40 loss-of-function mutant L117R fails to form VLPs. n = 6 24 near x-axis) did not produce data suitable for curve fitting. Shading indicates 95% confidence 1 intervals based on 999 bootstrap pseudoreplicates. N = 6 biological replicates. We expressed NLuc-VP40 in cells, collected total supernatant, and heated at a gradient of 15 temperatures. We then either measured luminescence directly to assess NLuc thermal stability 16 (blue), or pelleted VLPs and then measured luminescence to assess VLP stability (tan). Data (n 17 = 3 biological replicates) were normalized to 4 °C, log-transformed, and fit to sigmoidal curves.  Co-immunoprecipitation (co-IP) western blot (WB) NP oligomerization assay. Deletion of NP 24