Capsid protein mediated evasion of IRAK1-dependent signalling is essential to Sindbis virus neuroinvasion and virulence in mice

ABSTRACT

Alphaviruses are arthropod-borne, single-stranded positive-sense RNA viruses that are recognized as rapidly emerging pathogens. Despite being exquisitely sensitive to the effects of the innate immune response alphaviruses can readily replicate, disseminate, and induce pathogenesis in immunologically competent hosts. Nonetheless, how alphaviruses evade the induction of an innate immune response prior to viral gene expression, or in non-permissive infections, is unknown. Previously we reported the identification of a novel host/pathogen interaction between the viral Capsid (CP) protein and the host IRAK1 protein. The CP/IRAK1 interaction was determined to negatively impact IRAK1-dependent PAMP detection in vitro, however, the precise importance of the CP/IRAK1 interaction to alphaviral infection remained unknown. Here we detail the identification of the CP/IRAK1 interaction determinants of the Sindbis virus (SINV) CP protein and examine the importance of the interaction to alphaviral infection and pathogenesis in vivo using an interaction deficient mutant of the model neurotropic strain of SINV. Importantly, these interaction determinants are highly conserved across multiple Old-World alphaviruses, including Ross River virus (RRV), Mayaro virus (MAYV), Chikungunya virus (CHIKV), and Semliki Forest virus (SFV). In the absence of a functional CP/IRAK1 interaction, SINV replication is significantly restricted and fails to disseminate from the primary site of inoculation due to the induction of a robust type-I Interferon response. Altogether these data indicate that the evasion of IRAK1-dependent signalling is critical to overcoming the host innate immune response and the in vivo data presented here demonstrate the importance of the CP/IRAK1 interaction to neurovirulence and pathogenesis.

KEYWORDS:

• Alphavirus
• Sindbis virus
• capsid
• IRAK1
• neurovirulence
• pathogenesis
• innate immunity

Introduction

Alphaviruses are rapidly emerging arthropod-borne positive-sense RNA viral pathogens with a significant capacity to cause severe illness in otherwise healthy individuals. The clinical diseases caused by alphaviruses are divided into two categories, those which cause febrile arthritis, and those which cause encephalitis. The arthritogenic alphaviruses, including Sindbis virus (SINV); Chikungunya virus (CHIKV); and Ross River virus (RRV), cause moderate to severe multi-joint febrile arthritis in otherwise healthy individuals [1–3]. Unfortunately, arthritis caused by these viruses has the potential to develop into chronic arthritis which may persist or worsen over a period of several months, or years, after the resolution of acute infection [4,5]. This prolonged chronic disease results in a long-term reduction of quality of life due to ongoing disability and limited use of affected joints. The encephalitic alphaviruses, which include members such as Western Equine Encephalitis virus (WEEV); Venezuelan Equine Encephalitis virus (VEEV); and Eastern Equine Encephalitis virus (EEEV), have the capacity to cause viral encephalitis in young and elderly patients [6–8]. Despite being comparatively rare clinically, these viruses exhibit high morbidity as they can cause severe encephalitis that may result in the death of the infected individual. Furthermore, individuals who survive clinical alphaviral encephalitis often experience life-long cognitive impairments. Due to their capacity to cause severe disease in otherwise healthy individuals, the lack of safe and effective therapeutics or vaccines for the treatment or prevention of alphavirus infection, and the rapid expansion of competent vector mosquito geographic range, the alphaviruses are recognized as emerging health concerns [9-12]. Accordingly, research that defines alphaviral pathogenesis is critical to the development of innovative mitigation strategies by which the burdens of alphaviral disease may be alleviated.

The alphaviruses are exceptionally sensitive to the host innate immune system and are readily controlled by the antiviral effects of the type-I interferon (IFN) response [13–15]. To overcome host restriction of viral replication, the alphaviruses have evolved means by which the host innate immune response may be limited to enable viral replication and spread. The evasion of the innate immune response by the alphaviruses has been thought to be predominantly driven by the shutoff of host macromolecular synthesis, which effectively precludes the production of IFN, and interferon-stimulated genes (ISGs) [16]. However, the shutoff of host macromolecular synthesis brought about by transcriptional and translational repression requires viral gene expression and does not occur in earnest until mid- to late infection in highly permissive models of alphaviral infection [17]. The delayed circumvention of the host response and the necessity for viral gene expression constitutes a window of opportunity for the host to detect and elicit a controlling innate immune response to infection in permissive and nonpermissive cell hosts. A major mechanism by which the cellular host may detect and respond to viral infection are the Toll-like Receptors (TLRs) which sense Pathogen Associated Molecular Patterns (PAMPs) including viral nucleic acids. Nonetheless, prior studies have revealed that the TLRs have a minimal role during alphaviral infection, suggesting that the capacity of the host to sense viral infection via TLR-dependent mechanisms is greatly reduced during infection; however, the precise mechanisms as to how this occurs, especially in nonpermissive cells exposed to alphaviral PAMPs (as they lack the host shutoff afforded by viral gene expression), is unknown [18].

Recently our lab defined the host/pathogen Protein:Protein interactions of the Sindbis virus (SINV) Capsid (CP) protein utilizing an innovative BioID approach [19]. These efforts demonstrated that the alphavirus CP protein interacts with the host Interleukin-1 Receptor Associated Kinase 1 (IRAK1) protein. The identification and validation of the CP/IRAK1 interaction was notable, as the IRAK1 protein plays a key role in the signalling pathways of all TLRs (with the notable exception of TLR3) [20]. Further assessments using in vitro models of alphaviral infection revealed that the CP/IRAK1 interaction inhibits IRAK1-dependent signalling in a highly specific manner. Importantly, it was found that the CP proteins delivered by the disassembly of the incoming nucleocapsid cores were sufficient to perturb IRAK1-dependent signalling, and even the CP proteins derived from non-infectious viral particles were capable of repressing IRAK1-dependent sensing of PAMPs. Altogether these data provided a potential explanation as to why TLRs failed to significantly contribute to the restriction of alphaviral infections in knockout mice, in that IRAK1-dependent TLR sensing may already be inhibited during alphaviral infection and that loss of TLR sensing provides no specific additional contribution to replication/infection.

While our previous efforts have demonstrated the capacity of the CP/IRAK1 interaction to negatively affect host PAMP sensing via IRAK1-dependent signalling pathways in cellular models of infection, the precise importance and biological impact of the interaction to alphaviral infection and pathogenesis in vivo has yet to be described. Here we present data obtained by way of leveraging a nanoluciferase-based Bi-molecular complementation (BiMC) approach to determine the necessary and sufficient CP/IRAK1 interaction determinants of the SINV CP protein. The knowledge gained from these efforts led to the development of a mutant virus unable to inhibit IRAK1-dependent signalling, allowing us to test the importance of the CP/IRAK1 interaction in both in vitro and in vivo settings. From these efforts, we conclude that the CP/IRAK1 interaction is crucially important for viral replication and dissemination as mutants lacking in the CP/IRAK1 interaction are severely limited by the induction of type-I IFNs. The elicitation of the host innate immune response, as per the effects of IFN, is achieved by way of an IRAK1- and Myd88-dependent mechanism, such as TLR sensing of viral PAMPs.

Methods

Tissue culture cells

BHK-21 (ATCC CCL-10, VA, USA) and HEK293 (ATCC CRL-1573, VA, USA) cells were cultured in Minimal Essential Media (MEM; Cellgro, NY, USA), supplemented with 10% Fetal Bovine Serum (FBS; Corning, 35-010-CV, NY, USA), 1x Penicillin/Streptomycin (Pen/Strep; Corning, 30-002-CI, NY, USA), 1x Non-essential Amino Acids (NEAA; Corning, 25-025-CI, NY, USA), and L-glutamine (Corning, 25-005-CI, NY, USA). HEK293-derived reporter cells, HEK-Blue hTLR7 (Invivogen, hkb-htlr7, CA, USA), were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Corning, 10-017-CV, NY, USA) supplemented with 4.5 g/L glucose, 10% FBS, 1x Pen/Strep, and 1x Normocin (Invivogen, ant-nr-1, CA, USA). HEK-Blue cells were kept at low passage numbers to maintain genetic homogeneity and given selection antibiotics every other passage to maintain genomic integrity. Immortalized wild-type C57Bl/6-derived (NR-9456, BEI resources, VA, USA), and C57Bl/6 congenic MyD88^-/- macrophages (NR-15633, BEI resources, VA, USA) were maintained in DMEM supplemented with 4.5 g/L glucose, 10% FBS (Corning), 2 mM L-glutamine (Corning), 1 mM Sodium Pyruvate (Corning. 25-000-CI, NY USA), and 10ug/ml ciprofloxacin (Corning, 61-277-RF).

All cells were cultured in a humidified incubator at 37°C in the presence of 5% CO₂, and passaged according to standard practices for each cell line.

Expression plasmid construction

The nanoluciferase BiMolecular Complementation expression plasmids used in this study were generated through site-directed mutagenesis or Gibson Assembly, using the previously described parental pSplit.Nanoluc.C67.SINV CP plasmid or the pSplit.Nanoluc.C67 plasmid, as described in detail below [19].

The development of SINV CP truncation mutants was developed utilized site-directed mutagenesis via the Q5 Site-Directed Mutagenesis Kit (NEB, E0554S, MA, USA) to excise the individual domains of the CP individually in a stepwise fashion to generate a full battery of mutant constructs. The deletions of the individual domains were accomplished using primers- C67.SINV.DRI.F (5’-AAGCCTAAGAAGCCCAAGAC-3’) and C67.SINV.DRI.R (5’- CATGCCGCTTCCACCTCC-3’) for the deletion of the RI domain to generate pSplit.Nanoluc.C67.SINV.ΔRI; C67.SINV.DRIandII.F (5’-AGACTGTTCGACGTGAAGAACG-3’) and C67.SINV.DRIandII.R (5’-CATGCCGCTTCCACCTCC-3’) for the deletion of the RI and RII domains to generate pSplit.Nanoluc.C67.SINV.ΔRI/RII; C67.SINV.DPRO.F (5’- TAACTTAAGCTTGGTACCGAG-3’) and C67.SINV.DPRO.R (5’-TCTGTCGGCTTCCAGCTT-3’) for the deletion of the Protease domain to generate pSplit.Nanoluc.C67.SINV.ΔPRO. The nanoluciferase BiMC expression constructs consisting of the RI and RII domains in isolation were generated by site-directed mutagenesis and via- C67.SINV.DRI.F and C67.SINV.DRI.R site-directed mutagenesis of the pSplit.Nanoluc.C67.SINV.ΔPRO plasmid to generate the RII-only construct pSplit.Nanoluc.C67.SINV.RII; and to generate the pSplit.Nanoluc.C67.SINV.RI construct primers C67.SINV.D2andPRO.F (5’-TAACTTAAGCTTGGTACCGAGCTC-3’) and C67.SINV.D2andPRO.R (5’-AGGAGGCTGCTTAGGGGC-3’) were used to modify the parental pSplit.Nanoluc.C67.SINV CP plasmid. All reactions were performed according to the manufacturer's instructions.

The nanoluciferase BiMC expression plasmid encoding the RII fragments of the CP proteins of Ross River (RRV), Mayaro (MAYV), Chikungunya (CHIKV), and Semliki Forest (SFV), were cloned into the original pSplit.Nanoluc.C67 plasmid using Gibson Assembly approaches. All synthetic DNA fragments for these efforts were obtained from Genewiz (NJ, USA) and the restriction enzyme SfoI (NEB, R0606S, MA, USA) was used for the Gibson Assembly reaction. All DNA fragments used in this study were generated by GeneScript (NJ, USA) and assembled using the Gibson Assembly mastermix available form Synthetic Genomics, Inc. (GA1100-10, CA, USA) according to the manufacturer’s instructions.

All plasmids were cultured overnight in E. coli DH5α (or comparable) bacteria under antibiotic selection. Plasmids were purified by miniprep or midiprep purification kits (Omega Bio-Tek, D6943-02, D6904-04, GA, USA).

Nanoluc-based bimolecular complementation analysis (Nanoluc BiMC)

NanolucBiMC was performed as previously described. Briefly, HEK293 cells were seeded into a flat white bottom 96 well plate at a density of 1.25 × 10⁴ cells per well. After overnight incubation wells were transfected with pSplit.Nanoluc.N67 and one of the pSplit.Nanoluc.C67 constructs using lipofectamine 3000 (Invitrogen, L3000001, MA, USA). A total of 0.1 µg of DNA was transfected consisting of 50 ng of each construct. After allowing transfected cells to incubate for 48 h under normal growth conditions the media was carefully removed and replaced with 100μl of Optimem (Gibco, 31985070, MA, USA) supplemented with Furimazine (AOBIOUS, AOB36539, MA, USA) at a concentration of 10μm. The reactions were briefly incubated for two minutes at room temperature prior to the measurement of luminescence activity via a Synergy H1 microplate reader (BioTek, VT, USA).

Western blotting/immunodetection of SINV CP truncation constructs

To quantitatively assess SINV CP truncation construct expression transiently transfected cells, prepared as described above, were washed with 1xPBS prior to harvesting and lysis in RIPA buffer (50 mM Tris HCl, pH 7.4 / 150 mM NaCl / 1% NP-40 / 0.5% Sodium Deoxycholate / 0.1% Sodium Dodecyl Sulfate). Protein concentrations were determined using a Qubit 4 Fluorometer (ThermoFisher Scientific, MA, USA) using broad-range protein assay reagents and equal amounts of protein were assessed via Slot-blot using a PR 648 Slot Blot Manifold (Cytiva, MA, USA) and activated 0.2 micron PVDF membrane (Cytiva, MA, USA). The membrane was briefly rinsed twice with 1xPBS, and after a brief incubation in methanol to purge excess water from the membrane, the PVDF membrane was thoroughly dried prior to incubating in a primary antibody solution containing anit-NanoLuc Luciferase antibody (Clone 965853, MAB100261; R&D Systems, MN, USA) at a dilution of 1:1000. Antibody specificity to the C-terminal fragment of the Nanoluc protein was confirmed by standard western blotting approaches prior to use in this assay. After washing the slot blot membrane was detected using HRP-conjugated Goat anti-Mouse IgG (#31460, ThermoFisher Scientific, MA, USA) diluted 1:40,000. Detection was achieved using SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher Scientific, MA, USA) via a UVP GelStudio (Analytik Jena US, MA, USA). Protein abundance was determined by densitometry.

Generation and preparation of SINV

This study utilized a series of SINV infectious clones, including p389, a Toto1101-derived SINV strain with a GFP reporter in the nsP3 protein; p389_P726G, a derivative of p389 that includes a point mutation in the nsP2 protein known to prevent the cessation of host gene expression; the AR86 neurotropic strain, and a derivative of AR86 encoding the mCherry protein and an FMV2A protease cleavage site following the CP protein. For all strains referred to as SINV.ΔRII, the derivative mutant was generated by Gibson assembly using the corresponding fragment lacking amino acids 81–95 of the SINV CP protein. The restriction enzyme sites used to generate the ΔRII clones were HpaI and ZraI for the p389-derived mutants or BstBI and StuI for the AR86 clones. Synthetic DNA fragments containing the regions of interest for Toto1101 and AR86 were obtained from Genscript (NJ, USA), and Genewiz (NJ, USA), respectively.

The use of two strains of SINV was intentional, with the AR86 strain being utilized for in vivo studies as well as the macrophage studies, whereas the Toto1101-derived strain was used for all other studies. Infectious viral stocks were generated as previously described [21]. Briefly, RNA transcripts were generated in vitro then 10 µg were electroporated into ∼3 × 10⁶ BHK-21 cells using a single pulse at 1.5 kV, 25 mA, and 200Ω. Viral titres were determined by way of standard plaque assays in BHK-21 cells.

Analysis of SINV mutant growth kinetics

The growth kinetics SINV.WT and the SINV.ΔRII mutant were assessed using one-step growth kinetic assays involving BHK-21 cells cultured in 12-well tissue culture dishes. Briefly, confluent monolayers of BHK-21 cells were infected with either SINV.WT or SINV.ΔRII at an MOI of 5 infectious units per cell. After a one-hour adsorption period, the inoculum was removed, and the cell monolayers were extensively washed with 1x PBS (Corning, 21-040-CMR, NY, USA) prior to the addition of whole medium supplemented with 25 mM HEPES (gibco, 15630080, MA, USA) to enable the use of an automated liquid handling system lacking a CO₂ atmosphere. The infected cells were incubated at 37°C, and gently mixed prior to the harvesting of the tissue culture supernatants. The cell supernatant was collected at the indicated times post-infection and stored at 4°C, and fresh media was added to the cell monolayers.

Viral titres were determined using standard plaque assays using BHK-21 cells overlaid with 2% Avicel (in whole media). After a 30-hour incubation period, the cell monolayers were fixed with 3.7% formaldehyde (diluted in 1x PBS), and the plaques were enumerated following crystal violet staining.

To evaluate the capacity of the SINV.ΔRII mutant to replicate and disseminate in highly permissive and TLR-expressing tissue culture models of infection, BHK-21 or HEK-hTLR7 cells were infected with either SINV.WT or SINV.ΔRII at an MOI of 0.01 infectious units per cell. To aid in the detection of viral infection/dissemination the SINV strain used for these studies included an mCherry reporter expressed as part of the structural ORF using an FMV2A-based protease strategy. At 24 and 48hpi the cells were harvested, and a single-cell suspension was created via vigorous micropipetting of the harvested cells. After brief centrifugation, the cells were resuspended in 1% Formaldehyde (in 1xPBS) and fixed at 4°C for at least 16 h prior to centrifugation, washing, and resuspension in 1xPBS. The percentage of mCherry positive cells was determined via flow cytometry via a BD Accuri C6 flow cytometer (Excitation 488 nm laser, with a 670LP filter). Single cells were gated for using light scattering and the percent of mCherry positive cells was determined using mock-infected cells to establish the limits of mCherry detection/expression.

Analysis of SINV particle production, infectivity, vRNA content, and thermal stability

To quantitatively asses viral particle production, in terms of Genome Equivalents (GE) per ml volume, aliquots of SINV.WT and SINV.ΔRII P(0) stocks were used as the input materials for qRT-PCR. Briefly, 5μl of viral culture was used to generate cDNA via reverse transcription reactions using OneScript Plus Reverse Transcriptase (AbmGood, G237, BC, Canada) as per the manufacturer’s instructions. The resulting cDNA was used as the input materials for standard curve qRT-PCR using PerfeCTa SYBR Green FastMix (Quantabio, 95074-250, MA, USA) to detect coding sequences in the SINV nsP1 and E1 genes SINV.nsP1.F (5’-AAGGATCTCCGGAACCGTA-3’) / SINV.nsP1.R (5’-AACATGAACTGGGTGGTGTCGAAG) and SINV.E1.F (5’-TCAGATGCACCACTGGTCTCAACA-3’) / SINV.E1.R (5’- ATTGACCTTCGCGGTCGGATACAT-3’), respectively. Absolute RNA abundance per ml was determined via comparison to internal standards made from the corresponding infectious clones.

Viral-specific infectivity was determined by obtaining the Particle-to-PFU ratios for each individual virus preparation. This was achieved by comparing the total number of particles per ml, as determined by particle assay, to the corresponding infectious units per ml, as determined by standard plaque assays.

To measure the selectivity and specificity by which the vRNA cargo was packaged into nascent particles, the absolute quantities of nsp1- and E1-containing vRNAs were compared for preparations of SINV.WT and SINV.ΔRII. This approach allows for the rapid and quantitative assessment of genomic RNA packaging by comparing the relative abundances of sequences specific to the genomic (nsP1 and E1) and subgenomic (E1 only) vRNAs. Increased ratios of E1 signal respective to nsP1 signals are indicative of subgenomic RNA packaging.

Thermal stability of AR86-derived SINV.WT and SINV.ΔRII particles was determined via heating aliquots of viral stocks to 50°C for a set period of time prior to quenching the reaction on ice. The incubated samples were then subjected to standard plaque assays to determine the extent to which the viral particles were inactivated by incubation at 50°C.

Quantitative analysis of TLR7 dose responsiveness

To define the impact of the individual alphaviral CP protein domains on IRAK1-dependent TLR7 signalling, HEK-Blue hTLR7 cells cultured to 75% confluence were transfected with the aforementioned CP protein expression plasmids. Transfection conditions were identical to those described above for the nanoluciferase BiMC assays, with the exception that the IRAK1 encoding plasmid was omitted in lieu of increased SINV CP expression plasmids. After a 24-hour incubation period, the supernatant was carefully removed and replaced with whole media supplemented with the TLR7-specific agonist CL307 (Invivogen, tlrl-c307, CA, USA) at the indicated concentrations prior to returning the cell to incubate under normal conditions. At 16 h post-treatment, 20μl samples of the supernatant were carefully removed and transferred to a new 96-well containing 180μl of HEK-Blue Detection media (Invivogen, hb-det2, CA, USA) and after careful mixing the plate was quantitatively assessed for SEAP expression via colorimetric assay.

The quantitative detection of IRAK1-dependent TLR7 signalling was accomplished by incubating the above detection plate at 37°C in a plate reader while regularly taking absorbance readings at 620 nm for a period of three hours, or until the absorbance curves of the highest concentration of agonist reached saturation. Readings from pre-saturation time points were comparatively assessed to determine agonist detection with respect to concentration.

Briefly, the quantitative analysis of TLR7 signal transduction was determined by comparing the measurable SEAP activity of the control and experimental conditions over the CL307 agonist dose range after the subtraction of non-agonist-treated wells. The control agonist treatment with the highest level of SEAP activity was then standardized to 100%, and all other wells within an experimental replicate were normalized accordingly to determine their relative activity. The quantitative data from multiple biological replicates for a given dose were then averaged and plotted with respect to agonist concentration. Non-linear regression analyses of the data, via GraphPad Prism 7.0.2, using the log(agonist) vs. response variable slope (four parameters) non-linear curve fit function, was used to determine the activation profiles and 95% confidence intervals of the data. The concentrations of agonist required to reach 50% of maximal activity of control-treated reactions (EC50_MAX) were determined using these non-linear regression equations.

To determine the impact of SINV.WT and SINV.ΔRII infections on IRAK1-dependent TLR7 signalling a similar approach to that described above was utilized, with the notable exception being that the experimental conditions involved SINV p389_P726G-derived infections at an MOI of 10 infectious units per cell. Quantitative analysis of TLR7 signalling was identical to that described above.

Mouse experiments

Four-week-old C57BL/6 mice were obtained from Jackson Laboratory (ME, USA) and acclimated in UofL vivarium facilities for a period of no less than 48 h. For the footpad infection model, the mice were infected via rear foot pad injection with either 1000 PFU of SINV.WT or SINV.ΔRII, or mock-infected, in a final volume of 10μL using a 30G syringe after sedation via isoflurane inhalation (502017, Vet One, ID, USA). The mice were returned to their cages and monitored to ensure recovery after isoflurane anaesthesia, and the mice were monitored daily for weight gain/loss and the development of neurological disease. Individual animal weights were monitored with respect to time, and the development of neurological disease was scored on a 5-point scale as follows- 0, no signs of overt disease and normal behavioural activity; 1, abnormal trunk curl, grip, or tail weakness (1 of 3); 2, abnormal trunk curl, grip, or tail weakness (2 of 3); 3, absent trunk curl, lack of gripping, tail paralysis; 4, pronounced dragging of one or more limbs; 5, hind or fore limb paralysis.

The Intracranial infection (IC) model was largely identical to that described above, with the major exception being that the virus inoculum was directly delivered to the left-brain hemisphere via a 30G syringe to a depth of 3 mm. Recovery after IC injection was prolonged relative to that of the footpad model; however, all mice recovered after a period of no more than 60 s.

Once mice reached humane endpoint criteria (clinical score of 4 or 5 or 20% weight loss), or at the designated time for tissue collection, the mice were euthanized via isoflurane inhalation overdose followed by cervical dislocation or by thoracotomy and the collection of vital tissues. Tissues collected for virological and biological analysis were rapidly frozen and stored at −80°C, and/or preserved in 10% formalin for later use. Serum was collected using blood collection tubes containing heparin sulfate, and brief centrifugation prior to aliquoting.

For the detection of infectious virus loads in vivo, harvested tissues were first homogenized by bead beating using a Bead Ruptor 4 (Omni International, 25-010, GA, USA) in a standardized volume of 1xPBS. After brief centrifugation to clarify the lysates, the supernatant was transferred to a fresh tube and viral loads were assessed using TCID50 assays in a 384-well format. The results of which were then converted to PFU per unit volume/mass by way of calibrated TCID50 assays using standardized samples.

Animal ethics and research

This study was performed under strict accordance with the recommendations described in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. This protocol was approved by the Institutional Animal Care and Use Committee of the University of Louisville. Isoflurane anaesthesia was used for any manipulations that could result in pain or distress.

Transcriptomic analysis of infected brain tissue

Transcriptomic analyses were conducted in parallel to the virological analyses described above, with the notable exception that the tissues were solubilized in TRIzol (Invitrogen, 15596026, MA, USA) rather than 1xPBS. Total RNA was extracted from the tissue lysates using the Direct-zol-96 MagBead RNA extraction kit (Zymo Research, R2102, CA, USA) per manufacturer’s instructions using a KingFisher Duo Prime Purification System (ThermoFisher Scientific, MA, USA). The resulting purified total RNAs were then used as the input materials for the synthesis of cDNA libraries for next-generation sequencing.

Libraries for next-generation sequencing were prepared using the NEBNext Ultra RNA Library Prep kit (NEB #E7530L, MA, USA) according to the manufacturer’s instructions using NEBNext Mutliplex Oligos for Illumina Index primer sets #1 and #4. Information regarding specific samples and their indices may be found in the Supplemental Data accompanying this manuscript. Libraries were quantified by Qubit Flex Fluorometry using the Qubit 1X dsDNA HS Assay Kit (ThermoFisher, Cat. No. Q33231, MA, USA). The average Fragment length of each sample was assessed using an Agilent Fragment Analyzer 5200, utilizing either the Standard Sensitivity Next Generation Sequencing kit, or the High Sensitivity Next Generation Sequencing Kit, depending on the mass as determined by the Qubit assay. Equimolar amounts of each sample were pooled, and the mass of the pool was quantified by Qubit using a 1X dsDNA HS assay kit. The Qubit concentration reading, and the average fragment length were used to determine molar concentration for loading. The library and PhiX standards were diluted using the standard normalization method following the manufacturer's directions. The total volume of the library was 20 μL at 750 pM, with 2% PhiX spike in. Further denaturing and dilution of the library to 75pM is done on the instrument (NextSeq 2000). Sequencing was performed on a NextSeq 2000 using a P3 200-cycle reagent kit with a P3 flow cell.

The raw FastQ sequence data was downloaded from Basespace and trimmed using Fastqc and Trimmomatic prior to alignment to the Mus musculus genome via STAR. Differential expression analysis was performed using Cuffdiff2 and DESeq2. For the Cuffdiff2 analysis, Cuffnorm was used to produce FPKM (Fragments Per Kilobase Million) normalized counts. The counts were then filtered to include only genes with minimum expression of one FPKM in three or more samples and an average expression of at least one FPKM. For the DESeq2 analysis, raw read counts were obtained from the STAR aligned bam format files using HTSeq version 0.10.0. The raw counts were normalized using the Relative Log Expression (RLE) method and then filtered to exclude genes with fewer than 10 counts across the samples.

The raw unprocessed next-generation sequencing data associated with this study has been deposited as BioProject PRJNA1001310. Processed data tables listing the differential transcriptomic data obtained during these studies may be found in the Supplemental Data Files associated with this manuscript.

Ontological analysis of next-generation sequencing data

The data obtained from the next-generation sequencing of IC-infected whole mouse brain homogenates were processed to identify enriched Biological Process Ontological categories using standard approaches. Briefly, the differential transcriptomics data for the SINV.WT vs Mock Infected and SINV.ΔRII vs Mock Infected data sets were parsed into lists of up-regulated and down-regulated transcripts which fit a criterion of >2-fold enrichment relative to Mock and a corrected p-Value of < 0.05. This resulted in the identification of 1388 up-regulated and 352 down-regulated genes for the SINV.WT vs Mock Infected data set; and 1004 up-regulated and 1234 down-regulated for the SINV.ΔRII vs Mock Infected data set. These lists were assessed using DAVID to identify enriched Biological Process ontological groups. The biological process ontological groups enriched for the parsed data sets were compared across the SINV.WT and SINV.ΔRII conditions to identify common and unique enrichment categories. In all cases, the ontological enrichment data was reassessed using Revigo to eliminate redundancy amongst the enrichments and group the enriched biological process ontological categories into simplified parent ontological groups to aid comparative analysis.

Full ontological tables and accompanying quantitative information may be found in the Supplemental Data Files associated with this manuscript.

Quantification of IFNβ expression in macrophages

C57BL/6-derived macrophages were seeded into a 24-well plate. After growing to 80% confluency, they were mock infected (via PBS) or infected with equal particle numbers of either SINV.WT or SINV.ΔRII or 1 h with occasional shaking. After the 1-hour absorption period, the media was carefully removed, and the cells were washed 1x with PBS to remove any unattached virus prior to the replacement of the media and further incubation. After a 3-hour incubation period, the media was again carefully removed, and the cells were washed with 1x PBS and then harvested via the addition of TRIzol and stored at −80°C. Analyses of MyD88^-/- macrophages were identical to that described above for wild-type macrophages. To determine the specificity to which IRAK1-dependent signalling contributed to IFNβ expression after SINV infection wild type macrophages were co-incubated in the presence of the IRAK1 kinase inhibitor JHX-119-01 (MedChemExpress, HY-103017, NJ, USA) at a concentration of 20 mM during the above experimental process.

Total RNA was extracted from the TRIzol-containing samples using Direct-zol-96 MagBead RNA extraction kit (Zymo Research) per manufacturer’s instructions using a KingFisher Duo Prime Purification System (ThermoFisher Scientific). The extracted total RNA was then used as the input materials for cDNA synthesis via Reverse Transcriptions using OneScript Plus Reverse Transcriptase (AbmGood, G237, BC, Canada) as per the manufacturer’s instructions. The resulting cDNA was used as the input materials for qRT-PCR using PerfeCTa SYBR Green FastMix (Quantabio, 95074-250, MA, USA) to detect the levels of IFNβ mRNA expression relative to the control gene GUSB. The relative quantities of mouse IFNβ RNA, as per the ΔΔCt method, were detected using primers Mus.IFNβ.F (5’-AAGAGTTACACTGCCTTTGCCATC-3’) and Mus.IFNβ.R (5’-CACTGTCTGCTGGTGGAGTTCATC-3’) with the mouse GusB housekeeping gene being detected using Mus.GusB.Fi (5’-GGAGGTACTTCAGCTCTGTGAC-3’) and Mus.GusB.Ri (5’- TGCCGAAGTGACTCGTTGCCAA-3’).

Statistical analysis

All quantitative data shown are from a minimum of three independent biological replicates, unless more replicates are specifically stated. All in vivo data involved the use of two independent preparations of viral inoculum to control for prep-to-prep effects. In all cases, the data shown represents the quantitative mean with the error bars representing the standard deviation of the means. Quantitative data obtained from in vivo experiments are represented by the geometric means and their respective errors. Where appropriate, the statistical analysis of ratios was performed using variable bootstrapping, as previously described [21]. Pairwise statistical analyses were conducted first using ANOVA analyses followed post hoc by unpaired Student’s t-tests, with a minimum threshold p-value of < 0.05 being accepted as statistically significant. The statistical interpretation of next-generation sequence and ontological enrichment data relied on the use of Bonferroni corrections to counteract the multiple comparisons problem.

Results

The CP/IRAK1 interaction is mediated by domain RII of the alphaviral capsid protein

As we have previously demonstrated that the alphaviral CP protein interacts with the host IRAK1 protein, we sought to determine the necessary and sufficient interaction determinants of the CP protein to better understand the interaction with the long-term goal of developing interaction-deficient viruses. The alphavirus CP protein contains three functional domains. The first N-terminal domain, the RI domain, is a polybasic and proline-rich domain that is responsible for RNA packaging into the nucleocapsid core, presumably by electrostatic interactions between the polybasic residues of the CP protein and the negative phosphodiester backbone of the RNA cargo [22]. The second domain, the RII domain, is responsible for vRNA specificity when packaging the RNA cargo into the nucleocapsid core, and the release of the RNA cargo during disassembly [23–25]. Finally, the third domain, the C-terminal Protease domain is responsible for cleaving the CP protein off the structural polypeptide which consists of the remaining structural genes synthesized during viral structural protein expression [26]. In addition to the functional protease activity, the Protease domain is also responsible for forming the bulk of the structure of the nucleocapsid cores via interprotein interactions, and responsible, at least in part, for the assembly and release of infectious particles via an interaction with the cytosolic tail of the E2 protein [26, 27].

To define the interaction determinants, the previously described SINV CP nanoluciferase BiMolecular Complementation (BiMC) system was used to assess a series of mutant SINV CP truncation constructs, resulting in the creation of a panel of SINV CP constructs with deletions of one or two domains in a stepwise fashion (as shown in Figure 1(A)). This approach allows for the qualitative and quantitative assessment of the CP/IRAK1 interaction via luminescence detection, as nanoluciferase activity is restored via BiMC when, and only if, a cognate SINV CP protein truncation mutant is capable of interacting with the corresponding IRAK1 BiMC construct.

Figure 1. Functional mapping and identification of the CP/IRAK1 interaction determinants. (A) Schematics of the SINV capsid protein truncation mutants generated and evaluated via nanoluciferase-based BiMolecular Complementation (BiMC) assays. (B) Expression plasmids encoding the human IRAK1 protein with an N-terminal fusion of a nanoluciferase protein fragment, and either wild type SINV CP protein, or a truncation mutant, N-terminally fused to a complementing nanoluciferase fragment were co-transfected into HEK293 tissue culture cells. After incubation, the capacity of the human IRAK1 protein and the corresponding SINV CP protein to interact was detected via luminescence following the reconstitution of the nanoluciferase enzyme via BiMC. (C) Essentially identical to panel A, with the notable difference that the RII domains of SINV, RRV, MAYV, CHIKV, and SFV capsid proteins were assessed for their capacity to interact with the human IRAK1 protein. The quantitative data shown are the means of at least three independent biological replicates, with the error bar representing the standard deviation of the means. Statistical significance was determined by one-way ANOVA analysis and post hoc Student’s T-tests, with p-values of < 0.01 and < 0.0001 denoted by ** and ****, respectively.

As shown in Figure 1(B), the SINV CP/IRAK1 interaction is mediated by specific determinants within the N-terminus of the SINV CP protein. The RI domain was dispensable to the CP/IRAK1 interaction as the deletion of the RI domain (as per the ΔRI mutant) increased the BiMC of nanoluciferase by approximately two-fold relative to the full-length SINV CP protein (as per WT SINV). Nonetheless, extending the ΔRI truncation mutant to include the RII domain, as per the ΔRI/II mutant, resulted in the near complete loss of BiMC. Together these data strongly implicate the RII domain as the primary interaction determinant. Nonetheless, from these data, it remained possible that the Protease domain contributed to the interaction, as the ΔRI construct exhibited increased BiMC relative to full-length SINV CP protein. To determine whether the Protease domain contributed to the CP/IRAK1 interaction the capacity of a SINV CP protein construct lacking the Protease domain (as per ΔPro) to engage in BiMC with IRAK1 was evaluated. As shown by Figure 1(B), the deletion of the protease domain resulted in no significant change in BiMC activity compared to full-length CP protein, indicating that the protease domain does not contribute substantially to the interaction with IRAK1. Lastly, to confirm the specifics of the individual contributions of the N-terminal domains to the CP/IRAK1 interaction the above efforts were followed up with constructs expressing either the RI or RII domains in isolation. As shown in Figure 1(B), the CP/IRAK1 interaction was entirely dependent on residues in the RII domain, and the RI domain, in its entirety, was dispensable to the CP/IRAK1 interaction (as per RI and RII constructs, respectively). To ensure that the capacities of the individual SINV CP truncation mutants to engage or not engage in BiMC with IRAK1 were genuine the protein abundances of the SINV CP protein constructs were quantified during ectopic expression following transfection. As shown in Supplemental Figure 1, the expression levels of the SINV CP protein constructs were more or less equivalent indicating that the BiMC activities observed above were not simply due to low protein expression levels.

As the above data indicates that the RII domain of the SINV CP protein was the sole essential IRAK1 interaction determinant, we sought to determine whether the capacity of the RII domain to instigate the CP/IRAK1 interaction was functionally conserved across multiple arthritogenic alphavirus species. To this end, nanoluciferase BiMC constructs encoding the isolated RII domains of Ross River (RRV), Mayaro (MAYV), Chikungunya (CHIKV), and Semliki Forest (SFV) were assessed for their capacity to interact with IRAK1 and restore nanoluciferase activity via BiMC. As shown in Figure 1(C), all aforementioned arthritogenic alphavirus RII domains were capable of restoring nanoluciferase activity via IRAK1 BiMC relative to the control condition.

Collectively, these data indicate that the RII domain of the SINV CP protein contains the necessary and sufficient CP/IRAK1 interaction determinants; and that the conservation of the RII domains’ capacity to elicit the CP/IRAK1 interaction is indicative of not only the importance of the CP/IRAK1 relationship but also the ubiquitous nature of this interaction amongst the arthritogenic alphaviruses. More importantly, identifying the elements of the SINV CP protein responsible for the CP/IRAK1 interaction enables the further evaluation of the repression of IRAK1-dependent signalling via the development of SINV CP mutants.

The SINV RII domain is insufficient to impart total loss of IRAK1-dependent signalling

After determining that residues in the RII domain were the primary CP/IRAK1 interaction determinant, we set out to determine if the RII domain alone was sufficient for the disruption of IRAK1-dependent signalling activity; or alternatively, if the consequences of the CP/IRAK1 interaction were dependent on the whole CP protein or a subset of CP protein domains. To accomplish this, the aforementioned SINV CP truncation mutants were evaluated for their capacity to interfere with IRAK1-dependent signalling in a tissue culture reporter model system. Specifically, a HEK293-derived TLR7-responsive reporter cell line that expresses Secreted Embryonic Alkaline Phosphatase (SEAP) when stimulated with the TLR7-specific agonist CL307 was used to determine the contributions of the SINV CP protein domains by way of assessing the maximal activation and dose-responsiveness of the TLR7 receptor after agonist treatment in the presence and absence of SINV CP protein [19].

To define the specific contributions of the individual SINV CP domains, the TLR7 reporter cells were transfected with the SINV CP protein truncation mutants and treated with the TLR7-specific agonist CL307 over a broad dose range to trigger TLR7 activation in a quantifiable manner. As previously demonstrated, the transfection of an expression plasmid encoding full-length wild-type SINV CP protein caused a substantial reduction in IRAK1-dependent TLR signalling, as evidenced by significantly reduced maximum SEAP production/activity and a significant shift in dose responsiveness as evidenced by the EC50_MAX (as shown in Figure 2(A)). As BiMC identified the RII domain as containing the primary interaction determinant, the assessment of the capacity of the RII domain (in isolation) to inhibit IRAK1-dependent signalling was prioritized. As shown in Figure 2(B), the expression of the RII domain alone was sufficient to repress IRAK1-dependent signalling; however, the overall effects were muted relative to those observed with the full-length SINV CP protein. Specifically, expression of the RII domain elicited a minor decrease to maximal activation levels, and a ∼5-fold shift of the amount of agonist required to reach EC50_MAX. While these differences are biologically significant, they fall far short of the near-total inhibition of TLR7 signalling observed with the expression of the full-length CP protein. Thus, to define whether other domains of the SINV CP protein contribute to the repression of IRAK1-dependent signalling, these efforts were expanded to include the other truncation mutants that include the RII domain. Interestingly, the inclusion of either the RI domain or the Protease domain alongside the RII domain did not result in increased suppression of IRAK1-dependent signalling (as per Figure 2(C and D), respectively). Consistent with this observation is the fact that neither the RI domain nor the Protease domain individually contributed to the repression of IRAK1-dependent signalling (Figure 2(E and F), respectively).

Figure 2. Suppression of IRAK1-dependent signalling requires the RII domain. Panels A through F represent quantitative analysis of signalling the TLR7 responsiveness of HEK293 TLR7 reporter cells that were transfected with expression plasmids encoding either full length wild type SINV CP protein, or a truncation mutant as specified above the corresponding graph. TLR7 activation was accomplished using the TLR7-specific agonist CL307. The quantitative data shown are the means of at least four independent biological replicates, with the error bar representing the standard deviation of the means. The connecting line represents a non-linear regression of the underlying data, and the shaded region indicates the 95% confidence interval of the non-linear regression. Thus, data points where the shaded regions do not intersect are statistically significant by a p-value of at least <0.05, as determined by ANOVA analysis and post hoc Student’s T-tests.

Taken together these data agree with the above BiMC analyses and demonstrate that elements of the RII domain are necessary for both the interaction with IRAK1 and the inhibition of IRAK1-dependent signalling. Interestingly, while the RI and Protease domains do not contribute towards facilitating the CP/IRAK1 interaction, the presence of both domains was necessary for the exceptionally robust repression of IRAK1-dependent TLR signalling by the SINV CP protein.

Mutation of the RII domain of the SINV CP protein ablates the inhibition of IRAK1-dependent signalling

The identification of the necessary and sufficient CP/IRAK1 interaction determinants enables the use of a reverse genetics approach to develop CP/IRAK1 interaction deficient viruses by which the significance of the interaction to infection and pathogenesis could be assessed. Nonetheless, the RII domain has several important roles during alphaviral infection, and thus the capacity to make mutations is constrained by alphaviral biology. Prior examinations of the SINV RII domain have demonstrated that the leading 15 amino acids of the domain are dispensable for viral replication and assembly in highly permissive tissue culture cell models of infection [28]. Armed with this knowledge a SINV CP deletion mutant lacking amino acids 81 through 95, SINV.ΔRII, was generated using existing Toto1101 and AR86 infectious clones (Figure 3(A)).

Figure 3. The N-terminal residues of the SINV RII domain are dispensable to replication but essential to repressing IRAK1-dependent signalling. (A) A schematic diagram of the CP/IRAK1 interaction deficient SINV.ΔRII mutant relative to wild type SINV. (B) One-step growth kinetics of the SINV.ΔRII mutant relative to wild type SINV. Briefly, BHK-21 cells were infected with the aforementioned SINVs at an MOI of 5 infectious units per cell, and the release of mature infectious viral particles into the tissue culture supernatant was followed with respect to time. The quantitative data shown are the means of three independent biological replicates, with the error bar representing the standard deviation of the means. (C) The capacity of the SINV.ΔRII mutant to repress IRAK1-dependent signalling during infection was assessed similarly to that described in Figure 2, with the exception that the HEK293 TLR7 reporter cells were infected with wild type SINV_P726G, SINV.ΔRII_P726G, or mock infected prior to the addition of TLR7-specific agonist. The quantitative data shown are the means of at least four independent biological replicates, with the error bar representing the standard deviation of the means. The connecting line represents a non-linear regression of the underlying data, and the shaded region indicates the 95% confidence interval of the non-linear regression. Thus, data points where the shaded regions do not intersect are statistically significant by a p-value of at least <0.05, as determined by ANOVA analysis and post hoc Student’s T-tests.

As the RII domain has been previously identified to mediate several critical structural aspects of alphaviral infection, the one-step growth kinetics of the SINV.ΔRII mutant were evaluated in BHK-21 cells which are highly permissive to alphaviral infection. As demonstrated by the data in Figure 3(B), the SINV.ΔRII mutant exhibited no over-defects relative to wild-type SINV AR86, reaffirming the prior observations that these residues were unessential in highly permissive models of infection. Similar observations were made using Toto1101-derived SINV.ΔRII mutants.

To determine whether the mutation of the SINV CP RII domain negatively impacted viral particle production or function the SINV.ΔRII mutant was further characterized in tissue culture models of infection. Briefly, and as shown in Supplemental Figure 2, mutation of the SINV CP RII domain did not negatively impact viral particle production or the specific infectivity of the particles themselves (as measured by the ratio of particles-to-infectious units). As the RII domain has been previously implicated in the selection of cargo for packaging, the vRNA content of the particles was evaluated using qRT-PCR to determine the selectivity and specificity to which the genomic vRNA was packaged. As with particle production and infectivity there were no over differences between cargo packaging between SINV.WT and SINV.ΔRII. Finally, the thermal stability of the wild type and mutant viral particles was compared to determine whether the RII mutant particles were inherently more fragile. As expected, as per the growth kinetics, particle production, infectivity, and cargo specificity data, the SINV.ΔRII mutant particles exhibited nearly identical stability relative to wild type viral particles. After confirming that SINV.ΔRII mutant virus had no defects regarding replication and growth kinetics, we next sought to demonstrate that this deletion was sufficient to restore IRAK1-dependent signalling activity during infection. To this end, IRAK1-dependent signalling was assessed in TLR7 reporter cells which had been either mock infected or infected with either SINV.WT or SINV.ΔRII. As shown in Figure 3(C), IRAK1-dependent signalling remains largely intact during SINV.ΔRII infections, as the concentration of TLR7 agonist required to reach EC50_MAX during SINV.ΔRII infection was indistinguishable from that of mock-infected cells. Curiously, despite a complete restoration of dose-responsiveness, the maximal activation observed during SINV.ΔRII infection was consistently reduced by approximately 20% relative to the control. While not currently understood, the reduction in maximal activation appears to be specific to SINV infections, as the exogenous expression of a SINV CP protein mutant where the region of interest has been replaced with alanine residues fails to negatively impact IRAK1-dependent signalling (Supplemental Figure 3).

As the SINV.ΔRII mutant failed to repress IRAK1-dependent signalling, we nest sought to determine whether the SINV.ΔRII mutant differed in its capacity to replicate in the presence of IRAK1-dependent TLRs. To this end, multi-step growth kinetics in highly permissive and TLR7-expressing cell lines were assessed using flow cytometry. As shown in Figure 4(A), both SINV.WT and SINV.ΔRII replicated identically in highly permissive cells. Nonetheless, in the presence of TLR7, viral replication kinetics were significantly reduced for both SINV.WT and SINV.ΔRII, with SINV.ΔRII failing to appreciably disseminated (Figure 4(B)). Therefore, the SINV.ΔRII mutant, which exhibits no overt replication defects despite failing to repress IRAK1-dependent signalling, is highly sensitive to IRAK1-dependent TLR responses in tissue culture models of infection.

Figure 4. The CP/IRAK1 interaction is required for replication in TLR-expressing tissue culture cells. (A) Multi-step growth kinetics and dissemination of the SINV.ΔRII mutant relative to wild type SINV as measured by flow cytometry using an AR86-derived strain of SINV encoding an mCherry reporter. Briefly, BHK-21 cells were infected with the aforementioned SINVs at an MOI of 0.01 infectious units per cell, and the increase in the proportion of tissue culture cells expressing a viral encoded mCherry reporter with respect to time was detected by flow cytometry. (B) Identical to panel A, with the exception that the multi-step growth kinetics were assessed using HEK-hTLR7 cells. The quantitative data shown are the means of three independent biological replicates, with the error bar representing the standard deviation of the means.

From these data, we conclude that the repression of IRAK1-dependent signalling is dependent on residues within the N-terminal region of the RII domain, and that the repression of IRAK1-dependent signalling can be relieved by breaking the interaction between CP and IRAK1 through mutation of the SINV CP protein. Furthermore, these data indicate that the CP/IRAK1 interaction deficient mutant virus SINV.ΔRII is viable despite being highly sensitive to TLR sensing and represents a means by which the importance of the CP/IRAK1 interaction to infection and pathogenesis may be tested during genuine viral infections in vivo.

The SINV ΔRII mutant is significantly attenuated in vivo

Knowing that the CP protein of SINV.ΔRII is incapable of repressing IRAK1-dependent signalling leading we next sought to determine the role of the CP/IRAK1 interaction, and the importance of the inhibition of IRAK1-dependent signalling, to alphaviral pathogenesis in vivo. To this end, 4-week-old C57BL/6J mice were inoculated via rear footpad injection with either sterile PBS (as a mock infection), SINV.WT, or SINV.ΔRII at a dose of 10³ PFU in a volume of 10ul. After inoculation, the mice were monitored for signs of neurological pathogenesis, including weakness; limb dragging; paralysis; and weight loss or gain. As shown in Figure 5(A), mice that were infected with SINV.WT had a mean survival time (MST) of approximately 6 days post-infection (dpi) and showed significant weight loss compared to mock-infected animals (Figure 5(B)). In contrast to infections of SINV.WT, all mice infected with SINV.ΔRII via footpad injection survived the infection, suggesting that the SINV.ΔRII virus is heavily attenuated in vivo. In parallel with the survival analyses, the development of clinical disease was quantified via an established scoring system. Like the aforedescribed survival and weight-loss observations, mice infected with SINV.WT exhibited pronounced neurological disease as all mice exhibited limb paralysis in at least one limb, whereas mice infected with SINV.ΔRII showed no signs of neurological clinical disease (Figure 5(C)).

Figure 5. Disruption of IRAK1-dependent signalling by the SINV CP protein is critical to dissemination, neuroinvasion, and pathogenesis. Four-week-old male and female C57BL/6J mice were mock infected or infected with 10³ PFU of either SINV.WT or SINV.ΔRII via subcutaneous injection into the left rear footpad and the infected animals were monitored regularly for signs of morbidity and mortality over a 14-day period. (A) Survival curves of animals in the experimentally infected groups (SINV.WT n = 10; SINV.ΔRII n = 6; Mock n = 6). Statistical significance was determined by Kaplan-Meier analysis of the median survival time (MST), with a p-value of < 0.0001 denoted by ****. (B) Percent weight data of all mice in the experimental groups involved in these studies relative to their initial starting weight, with the line representing the mean of the underlying data. Each point represents an individual mouse. (C) Clinical scores of mice relative to a neurological scale outlined in the materials and methods section. The data presented in panels (D through I) present quantitative analysis of viral titres in target tissues with respect to time for cohorts of 2 male and 2 female mice infected with either SINV.WT or SINV.ΔRII. Panels D and E represent titres of ankle tissues at 1 and 3 dpi, respectively. Panel F represents viremia as detected at 1 dpi. Panels G and H represent titres of quadricep muscle tissues at 1 and 3 dpi, respectively. The data in panel I represents the viral titre observed in brain tissue homogenate at 5 dpi. Statistical significance, as determined by Student’s T-test is denoted above the pairwise comparisons, with p-values of < 0.05 and <0.01 denoted by * and **, respectively.

To better understand the phenotypic consequences of allowing IRAK1-dependent signalling during SINV infection, the capacity of SINV.ΔRII to disseminate from the site of inoculation to the brain was assessed through the quantitative analysis of viral loads in target tissues at several times post-infection. Specifically, ankle, quadricep, blood, and brain samples were taken at 1-, 3-, and 5 dpi, and viral titres were quantitatively assessed. As shown in Figure 5(D and E), the viral titres of SINV.ΔRII were approximately 100-fold lower than those of SINV.WT in infected ankle tissues at 1- and 3-dpi, respectively. Despite reduced replication in tissues proximal to the site of inoculation, SINV.ΔRII was able to disseminate into the blood, albeit to levels approximately 100-fold lower than those observed for SINV.WT (Figure 5(F)). Nonetheless, despite being able to cause viremia, SINV.ΔRII failed to disseminate into other tissues as evidenced by the absence of detectable infectious virus in either the quadriceps muscle or the brain, at 1- and 3-; and 5 dpi, respectively (Figure 5(G, H and I)).

These data illustrate the importance of the SINV RII domain to SINV neurovirulence, as the SINV.ΔRII mutant exhibited reduced titres in vivo and a failure to disseminate into other target tissues following the development of viremia. Accordingly, a major conclusion from these data is that the CP/IRAK1 interaction, and by extension the loss of IRAK1-dependent signalling, is critical to SINV neuroinvasion and pathogenesis.

Bypassing neuroinvasion via intracranial infection results in neuropathogenesis

As the above data indicated that SINV.ΔRII was failing to disseminate from the site of inoculation to the brain, we sought to determine whether neurovirulence would be restored following direct inoculation of the virus into the brain. To test this, 4-week-old C57BL/6J mice were given intracranial (IC) injections with either PBS (as Mock infected), SINV.WT, or SINV.ΔRII at a dose of 10³ PFU in 10μl. As above, the experimentally infected animals were monitored for signs of morbidity and mortality, including signs of neurological disease and weight loss. As shown in Figure 6(A), SINV.WT-infected animals exhibited an MST of 4 dpi, whereas SINV.ΔRII infected animals had an MST of 5.5 dpi. Despite exhibiting mortality in these studies, the SINV.ΔRII infected mice displayed a different course of clinical illness from those infected with SINV.WT. As depicted in Figure 6(B), the SINV.ΔRII infected animals had a more gradual weight loss profile than SINV.WT infected animals, and whereas all SINV.WT animals exhibited pronounced neurological disease only moderate clinical signs (yet moribund as per our euthanasia criteria) were observed in two experimentally infected SINV.ΔRII animals (Figure 6(C)). The remaining SINV.ΔRII infected animals met weight loss criteria despite demonstrating no outward signs of neurological disease other than limited limb weakness. The weight loss observed in these experimentally infected mice may be alternatively explained by dehydration rather than infection per se (as the animals displayed signs of clinical dehydration, including tenting of the skin when gently pinched).

Figure 6. Direct inoculation of CP/IRAK1 interaction deficient mutant SINV into the brain partially restores virulence. Three-week-old male and female C57BL/6J mice were mock infected or infected with 10³ PFU of either SINV.WT or SINV.ΔRII via intracranial injection into the left hemisphere of the brain and the infected animals were monitored regularly for signs of morbidity and mortality over a 14-day period. (A) Survival curves of animals in the experimentally infected groups (SINV.WT n = 4; SINV.ΔRII n = 4; Mock n = 4). Statistical significance was determined by Kaplan-Meier analysis of the median survival time (MST), with a p-value of < 0.01 denoted by **. (B) Percent weight data of all mice in the experimental groups involved in these studies relative to their initial starting weight, with the line representing the mean of the underlying data. Each point represents an individual mouse. (C) Clinical scores of mice relative to a neurological scale outlined in the materials and methods section. The data in panels D and E represent the viral titres observed in right hemisphere brain tissue homogenate at 1 and 5 dpi, respectively. Statistical significance, as determined by Student’s T-test is denoted above the pairwise comparisons, with p-values of < 0.05 denoted by *.

While these data indicate that SINV.ΔRII has the capacity to be neurovirulent if directly introduced into the brain, to be thorough in our analyses we quantitatively evaluated the brains of the infected animals for viral replication. Specifically, at 1- and 3 dpi the brains of the experimentally infected animals were harvested and measured for viral titre. Interestingly, following IC infection the titres observed for SINV.ΔRII were still approximately 100-fold less than that of SINV.WT (Figure 6(D and E)); however, it is notable that the levels of SINV.ΔRII in the brain of the infected animals were roughly equivalent to that observed in the brains of animals infected with SINV.WT via footpad injection (Figure 5(I)).

Transcriptomic analyses reveal similar inflammatory profiles despite differing viral burdens

To better understand the consequences of, and inflammatory response to, SINV infection, next-generation sequencing of IC-infected brains 3 dpi was performed. As shown in Figure 7(A), SINV.WT infection resulted in the differential expression of many host transcripts; and, as expected, SINV.ΔRII infection also resulted in a similar expression profile relative to wild-type SINV (Figure 7(B)). Indeed, comparative differential transcriptomic analysis between the SINV.WT and SINV.ΔRII IC infections reveals a high degree of overall similarity between the experimentally infected groups, as comparatively few host transcripts were differentially expressed beyond the customary 2-fold change window relative to the comparisons involving mock-infected animals (Figure 7(C)). This observation is particularly striking as viral loads of the SINV.ΔRII mutant were approximately 100-fold less than wild type.

Figure 7. Differential transcriptomic analysis reveals enhanced inflammatory gene expression in SINV.ΔRII infected mice. Total RNA was extracted from brain tissue homogenates from mock, SINV.WT, and SINV.ΔRII infected mice (per group n = 4 comprised of two male and two female mice) and assessed using next-generation sequencing approaches. Volcano plots of differentially expressed transcripts of SINV.WT infected relative to Mock (Panel A), SINV.ΔRII relative to Mock (Panel B), and SINV.ΔRII relative to SINV.WT (Panel C) with thresholds of greater than 2-fold change and Bonferroni adjusted p-value of < 0.01. (D) A heatmap plot of transcript abundances (in FKPM as determined by Cuffnorm) of type-I IFN transcripts and transcripts belonging to the Response to Type-I Interferon gene ontology group (GO:0034340). Transcripts were triaged if they did not meet the standard criteria of greater than 2-fold enrichment relative to mock infected animals, and an FDR-corrected statistical significance p-value of < 0.05. Within each group the four squares represent the transcript specific FKPM values of the individual animals used in the study.

To better understand the antiviral response to SINV.WT and SINV.ΔRII infection at the level of host gene expression, the RNA abundances of the type-I IFNs and genes belonging to the Response to Type-I Interferon gene ontology group (GO:0034340) were evaluated. As shown in the heatmap presented in Figure 7(D), the experimentally infected animals exhibited unique profiles relative to mock-infected controls. The overall abundances of type-I IFN transcripts were reduced in SINV.ΔRII infections relative to SINV.WT infections, on average by approximately 5-fold. Nevertheless, abundances of antiviral interferon-stimulated genes such as IRF7, OAS1, Mx2, and many of the IFITM-family of ISGs are equivalent between SINV.ΔRII and wild-type infections. Thus, and interestingly, despite decreased levels of IFNβ and IFNα transcripts in the SINV.ΔRII infected animals, IFN-stimulated genes are broadly expressed at levels similar to or in excess of SINV.WT despite decreased viral loads and type-I IFN transcript expression.

Ontological analyses of the up-regulated transcripts detected in whole brain homogenates from SINV.WT and SINV. ΔRII experimentally infected mice revealed similar biological process enrichment profiles (Supplemental Figure 4). Unsurprisingly, the majority of the statistically enriched biological process ontological categories found to be upregulated in both SINV.WT and SINV.ΔRII infections centred around the host antiviral and innate immune response. While the majority of upregulated biological process categories were common, both SINV.WT and SINV.ΔRII infections resulted in unique upregulation and downregulation profiles.

The high degree of similarity between the transcriptomes of SINV.WT and SINV.ΔRII infected brains is suggestive of differential inflammatory stimulation potentials with respect to viral burden. This assertion is based upon the fact that while a ∼100-fold difference in viral titre is observed between the two experimental conditions the transcriptomic differences are exceptionally modest. Nonetheless, from these data, it cannot be directly concluded as to whether the SINV.ΔRII mutant elicits a more pronounced antiviral response in vivo, or whether the response is equivalent between SINV.WT and SINV.ΔRII and yet the SINV.ΔRII is acutely more sensitive to the innate immune response.

The SINV ΔRII mutant induces increased type-I IFN expression in infected/exposed macrophages

As SINV.ΔRII was largely incapable of repressing IRAK1-dependent signalling during infection and exhibited greatly restricted in TLR-expressing models of infection, we hypothesized that the inability of SINV.ΔRII to disseminate in vivo may be due to increased induction of Type-I IFN in an IRAK1-dependent manner. As IRAK1-dependent innate immune sensing relies on the detection of extracellular or vesicular PAMPs, the experimental design focused on evaluating the response to viral particle-associated PAMPs during wild-type and CP/IRAK1 interaction deficient infections.

To test the above hypotheses, C57BL/6-derived macrophages were either mock infected, infected with SINV.WT, or infected with SINV.ΔRII using equal numbers of viral particles to ensure equivalent delivery of any virion-associated PAMPs. At 3 h post-infection the cells were harvested, and the total RNA was extracted via TRIzol for analysis of IFNβ expression levels by way of qRT-PCR. As shown in Figure 8(A), macrophages infected with SINV.WT showed expression of IFNβ mRNA at levels comparable to mock-infected cells. In contrast, SINV.ΔRII infections induced significantly higher levels of IFNβ mRNA induction compared to SINV.WT (Figure 8(A)). These data suggest that without the CP/IRAK1 interaction afforded by the RII domain of the SINV CP protein, the viral particle-associated PAMPs are more readily recognized and responded to by IRAK1-dependent innate immune sensors of viral infection, including the TLRs.

Figure 8. Disruption of IRAK1-dependent signalling by the SINV CP protein is essential to evading the induction of an IRAK1- and MyD88-dependent IFN response. (A) Wild type C57BL/6-derived macrophages were mock infected or infected with either SINV.WT or SINV.ΔRII with equal numbers of viral particles equivalent to a wild type infection at an MOI of 10 infectious units per cell, and at 3 h post infection total RNA was extracted and IFNβ expression was quantified via qRT-PCR. (B) Essentially identical to panel A, with the exception that the macrophages were treated with the IRAK1-specific inhibitor JH-X-119-01 to selectively repress IRAK1-dependent signalling. (C) Also identical to panel A with the notable exception that the macrophages were derived from MyD88^-/- congenic C57BL/6J mice. The quantitative data shown are the means of at least three independent biological replicates, with the error bar representing the standard deviation of the means. Statistical significance was determined by one-way ANOVA analysis and post hoc Student’s T-tests, with p-values of < 0.01 denoted by **.

To confirm that the stimulation of IFNβ expression observed during SINV.ΔRII infection was specifically due to IRAK1-dependent signalling, we treated wild-type macrophages with JH-X-119-01, a small chemical inhibitor of IRAK1 that covalently binds to the active site and permanently blocks kinase activity and reassessed the capacity of the host to sense viral PAMPs and induce type-I IFN expression. As shown in Figure 8(B), mock, SINV.WT, and SINV.ΔRII infections of wild-type macrophages treated with JH-X-119-01 elicited similar levels of IFNβ mRNA expression, indicating that the induction of the innate immune response by viral particle-associated PAMPs was indeed IRAK1-dependent.

To further confirm that the trigger of IFN-β induction observed in SINV.ΔRII infected macrophages was indeed an IRAK1-dependent TLR, MyD88^-/- macrophages were assessed using the same experimental conditions as the WT macrophages. As depicted by the data presented in Figure 8(C), MyD88^-/- macrophages infected with SINV.ΔRII showed no increase in IFNβ expression when compared to mock or SINV.WT-infected macrophages. These data suggest that all IFNβ expression observed in the WT macrophages infected with SINV.ΔRII comes from a MyD88-dependent signalling pathway.

In conjunction with the in vivo data above, these in vitro data support the hypothesis that the CP/IRAK1 interaction is essential to evading the induction of an innate immune response through the detection of viral-associated PAMPs. Given that the differential induction of an innate immune response was largely lost in the absence of MyD88^-/-, it is probable that viral PAMP detection is occurring via one, or more, of the host TLRs responsive to positive-sense RNA virus PAMPs, such as TLR7 and TLR8. Importantly, these in vitro observations are in firm alignment with the restricted dissemination of SINV.ΔRII as alphaviruses are exceptionally sensitive to the impacts of the host innate immune response.

Discussion

Here we have demonstrated that the CP/IRAK1 interaction, via the inhibition of IRAK1-dependent signalling, is a novel virulence determinant in alphaviral infection that is critical to the evasion of the innate immune response. This conclusion is supported by the body of evidence indicating that disruption of the functional consequence of the CP/IRAK1 interaction resulted in viral PAMP detection leading to the loss of neuroinvasion due to a severe restriction of SINV infection in vivo. Classically, it is believed that the inhibition of host macromolecular synthesis is the primary means by which the alphaviruses evade the effects of the type-I IFN system [17, 29]. Indeed, host shutoff is believed to be the mechanism by which other innate immune sensing systems that detect the products of viral replication, such as RIG-I and PKR, are limited in regard to the induction of an innate immune response. Nonetheless as this evasion mechanism requires viral gene expression, a window of opportunity for the host to detect viral PAMPs and mount an antiviral response existed early in infection and is ever present in non-permissive host cells [16, 29–32]. The data presented here represents one means by which the alphaviruses have closed this window, as the inhibition of IRAK1-dependent signalling effectively precludes the sensing of PAMPs by host TLRs without the need for viral gene expression. Importantly, it has been previously demonstrated that the CP proteins derived from incoming SINV particles are sufficient to repress IRAK1-dependent signalling; therefore, even the meagre quantities of CP protein released into the host following the delivery and disassembly of the nucleocapsid core to the host cytoplasm are effectively capable of assisting in the evasion of the innate immune response [19]. As viral dissemination in tissues consists of relatively large boluses of viral particles delivered locally to neighbouring cells (in terms of multiplicities of infection), the disassembled CP proteins are liable to effectively mask IRAK1-dependent PAMP detection. Thus, on the basis of all available data, we propose that the CP/IRAK1 interaction represents a means by which the detection of viral-associated PAMPs may be actively evaded prior to cessation of host macromolecular synthesis (Figure 9).

Figure 9. Proposed model regarding the biological impact of the CP/IRAK1 interaction.

The evasion of IRAK1-dependent PAMP sensing is critically important to SINV neuroinvasion and pathogenesis

The data presented here indicates that the alphaviruses have evolved a means by which the activation of the innate immune response can be evaded prior to viral gene expression. As the CP/IRAK1 interaction mediated masking of IRAK1-dependent PAMP detection does not require viral gene expression, the CP/IRAK1 interaction represents a means by which the detection of viral-associated PAMPs may be masked in permissive and non-permissive cells. The net result of this activity is the prevention of an infection-limiting IFN response induced by exposed but uninfected host cells [19]. Crucially, in the absence of the CP/IRAK1 interaction, the recognition of viral particle-associated PAMPs resulted in the induction of a robust type-I IFN response which correlated with the loss of neuroinvasiveness in vivo. While it is currently unclear as to which specific host PAMP receptors are detecting which specific viral PAMPs, the receptors responsible for IFN induction during SINV.ΔRII infection are IRAK1- and Myd88-dependent, leading us to conclude that they are members of the TLR family. For the reasons of alphaviruses being positive-sense RNA viruses, we hypothesize that TLRs 7 and 8 are the most likely TLRs that are sensing PAMPs associated with alphaviral particles [33–35].

When the need for dissemination to the brain was bypassed via intracranial inoculation, the CP/IRAK1 interaction deficient virus exhibited a unique pathological profile, suggesting that the CP/IRAK1 interaction may play additional roles in alphaviral pathogenesis beyond TLR evasion. While the control of SINV infection via PAMP detection in the brain likely contributes to this phenotype, it should be noted that the host IRAK1 protein is also a critical component of the signal transduction pathways of the host IL-1 superfamily cytokine receptors [36, 37]. Thus, an additional function of the CP/IRAK1 interaction may be to dysregulate the IL-1 response during infection, contributing towards alphaviral pathogenesis via the establishment of an aberrant pro-inflammatory state. Assessing the impact of the CP/IRAK1 interaction on IL-1 signalling during alphaviral infection is important, as IL-1 is known to be a key contributor to alphaviral pathogenesis [38–41]. Efforts aimed at teasing apart the importance of the CP/IRAK1 interaction to IL-1 dysregulation and alphaviral pathogenesis are a key ongoing focus of the Sokoloski Lab.

The RII domain of the alphavirus CP protein is functionally complex and involves the interactions with host factors

These efforts identified a novel function of the alphaviral CP protein RII domain. Prior work in the field has determined that the RII domain is responsible for the selection of the nucleocapsid cargo during assembly, and the disassembly of the nucleocapsid core during viral entry [25, 28, 42, 43]. These earlier forays on the RII domain delineated the residues required for these nucleocapsid-associated functions of the CP protein, and largely concluded that the N-terminal residues were unimportant for viral infection in highly permissive cell models of infection [28]. Thus, despite the importance of the RII domain to critical assembly and disassembly events in the viral lifecycle, we were still able to develop a mutant virus that abrogated the interaction with IRAK1 while having no appreciable impact on viral growth kinetics in highly permissive tissue culture models of infection.

Using BiMC, the RII domains of several arthritogenic Old World alphaviruses were assessed for their capacity to interact with the host IRAK1 protein. Together these efforts revealed that the capacity of the RII domain to instigate the CP/IRAK1 interaction was conserved across multiple members of the genus. Whether individual alphavirus species differ in their capacity to restrict IRAK1-dependent PAMP detection, perhaps in a manner that correlates with clinical severity or viremia, is unknown at this time. One could envision a scenario where the affinity of the CP/IRAK1 interaction, or the tenacity with which the CP protein negatively impacts IRAK1-dependent signalling, contributes meaningfully to the development of high viral loads and disease in vivo. This possibility is particularly worthy of evaluation as it would represent a novel virulence determinant that may be fine-tuned towards the development or improvement of vaccine candidates.

Implications for the evasion of the invertebrate innate immune system

Though the alphaviral CP protein RII domains exhibit limited sequence conservation, the homologs of the TLRs and IRAK1 kinase demonstrate a high degree of structural and functional similarity across the vertebrate and invertebrate alphavirus hosts [44–46]. Indeed, the mosquito IRAK1 homolog Pelle is synonymous with the IRAK1 protein regarding its role as a critical component of the Toll signal transduction pathway [47, 48]. The conservation of the CP/IRAK1 interaction across multiple alphaviruses, the robustness of its consequence to IRAK1-dependent signalling, and the conservation of the IRAK1 protein across multiple host species are all highly suggestive of a potentially important role for the alphaviral CP protein regarding the evasion of mosquito innate immunity. As invertebrates lack an interferon response, they are heavily reliant on the Toll and IMD pathways to respond to microbial pathogens [49–52]. It is known that alphaviruses are sensitive to Toll and IMD effector proteins, however, the Toll pathway is not appreciably activated during alphaviral infection [53–55]. Thus, given the parallels between the narratives observed for the vertebrate and invertebrate hosts, we postulate that the alphaviral CP protein similarly interacts with the mosquito Pelle protein to evade the induction of the Toll innate immune pathway.

Potential mechanistic insight into the impairment of IRAK1-dependent signalling by the CP/IRAK1 interaction

The assessment of the contributions of the individual domains of the SINV CP protein to the inhibition of IRAK1-dependent signalling provides valuable insight into the potential mechanism of inhibition. While the RII domain is a necessary interaction determinant, expression of the RII domain alone is unable to cause the full robust inhibition of IRAK1-dependent signalling that is observed with the full-length SINV CP protein. Inclusion of either the RI or Protease domains alongside the RII domain failed to recapitulate the phenotype associated with the expression of the full-length SINV CP protein. All together, these data imply that the mechanism of inhibition is facilitated by the presence of both the RI and Protease domains; and although further studies are needed to elucidate the exact mechanism(s) by which the SINV CP protein inhibits IRAK1-dependent signalling, the need for a full-length CP protein suggests that steric hindrance of IRAK1 activity / Protein:Protein interactions are at least partially responsible. Alternatively, it remains possible that the flanking domains of the CP protein serve to stabilize the primary interaction of the RII domain to block the catalytically active site of IRAK1.

Acknowledgements

We thank the members of the laboratories of K.J. Sokoloski, D. Chung, I.S. Lukashevich, T.C. Mitchell, and J. Bagaitkar for their invaluable input during the development and execution of this project, and the preparation/editing of this manuscript. We also thank Jenna Olson for assistance with the analysis of flow cytometry data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was funded by grants from the National Institute of Allergy and Infectious Diseases (NIH-NIAID), specifically R01 AI153275 to K.J.S.; and by a COBRE programme grant from the National Institute of General Medical Sciences (NIGMS), P20 GM125504 to K.J.S. and R. Lamont. Sequencing and bioinformatics support for this work provided by National Institutes of Health (NIH) grants P20GM103436 (Martha Bickford, PI) and P20GM106396 (Donald Miller, PI). V.D.L was supported by an NIH-NIAID funded predoctoral fellowship, T32 AI132146. Additional support was received from the Integrated Programs in Biomedical Sciences (IPIBS) to C.M.I, and a generous startup package from the University of Louisville to K.J.S. This work was supported in part by a grant from the Jewish Heritage Fund for Excellence in Research Enhancement Grant Program at the University of Louisville School of Medicine.