SUMMARY
Type III or lambda interferons (IFNλs) form a critical barrier to infection by diverse pathogens. However in humans, production of IFNλ4 is associated with decreased clearance of hepatitis C virus (HCV). How an antiviral cytokine came to promote infection and whether this phenomenon occurs in other species is unknown. Here we show that, compared to chimpanzee IFNλ4, the human orthologue has reduced activity due to a single amino acid substitution (E154K). IFNλ4s with E154 restrict virus infection more potently and induce more robust antiviral gene expression. Remarkably, E154 is the ancestral residue in mammalian IFNλ4s but altered in representatives of the Homo genus. Nonetheless, the more active E154 form of IFNλ4 can be found in African Congo rainforest ‘Pygmy’ hunter-gatherers. We postulate that evolution of an IFNλ4 with attenuated activity in humans has been exploited by pathogens such as HCV, which could explain distinct host-specific outcomes of infection.
MAIN TEXT
Vertebrates co-ordinate antiviral defences through the action of signalling proteins called interferons (IFNs). IFNs induce expression of hundreds of ‘interferon-stimulated genes’ (ISGs), which establish a cell-intrinsic ‘antiviral state’ and regulate inflammation (Randall and Goodbourn 2008). Thus, IFNs are pleiotropic in activity and modulate aspects of protective immunity and pathogenesis (Schoggins 2014). Three groups of IFNs have been identified (types I – III), with the type III family – or IFNλs - being the most recently discovered (reviewed in Lazear et al. 2015b). Emerging evidence highlights the critical and non-redundant role IFNλs play in protecting against diverse pathogens, including viruses, bacteria, and fungi (Dixit et al. 2010, Nice et al. 2015, Lazear et al. 2015a, Galani et al. 2017, Odendall et al. 2017, Espinosa et al. 2017).
Although important in host defence, some IFNs are highly polymorphic (Manry et al. 2011). A single nucleotide insertion converting the ‘ΔG’ allele to a ‘TT’ allele (rs368234815) yields a frameshift leading to loss of active human IFNλ4 (HsIFNλ4) (Prokunina-Olsson et al. 2013). Although IFNλ4 is highly conserved among mammals, the TT allele has evolved under positive selection in some human populations (Key et al. 2014). Despite their broad antimicrobial functions, genome-wide association studies have convincingly demonstrated a correlation between ΔG IFNλ4 and reduced spontaneous clearance of hepatitis C virus (HCV) infection, i.e individuals homozygous for TT clear HCV infection with greater frequency (Ge et al. 2009, Prokunina-Olsson et al. 2013). HsIFNL4 has also been linked to protection from liver inflammation (Eslam et al. 2015) and reduced HCV treatment response (Prokunina-Olsson et al. 2013). The mechanism underlying this contribution of HsIFNλ4 to viral persistence and pathogenesis is not well understood but is associated with differences in ISG induction (Terczyńska-Dyla et al. 2014).
A major remaining question is how IFNL4 evolution has led to it occupying a ‘pro-viral’ role during HCV infection. To address this question, we explored whether differences in IFNλ4-mediated antiviral signalling exist between closely-related host species (humans versus chimpanzees, Pan troglodytes) in response to a common pathogen (HCV), taking advantage of the historical use of experimental infection of chimpanzees.
To this end, we compared intrahepatic gene expression during early HCV infection in humans and chimpanzees using published transcriptomic data (see Experimental Procedures). This revealed distinct host responses in humans and chimpanzees as well as overlapping differentially-regulated genes (Figure 1A and Supplementary Data File 1). In chimpanzees, the transcriptional profile was dominated by ISGs known to restrict HCV infection (RSAD2, IFI27 and IFIT1) (Schoggins et al. 2011), as well as genes involved in antigen presentation and adaptive immunity (HLA-DMA and PSMB8). These genes were not significantly differentially expressed in humans, whose response was mainly directed towards up-regulation of pro-inflammatory genes (for example, CXCL10, CCL18 and CCL5) (Figures 1A and 1B). ‘Chimpanzee-specific’ differentially-expressed genes were induced early in infection and remained significantly up-regulated during the acute phase (Figure 1B). We hypothesised that the more robust antiviral response to HCV infection in chimpanzees compared to humans could arise from host genetic differences.
Given the marked up-regulation of antiviral ISGs during acute infection in chimpanzees and relevance of IFNλ4 during HCV infection in humans, we undertook genetic and functional comparisons of natural human IFNλ4 coding variants and between human and chimpanzee IFNL4 orthologues. In humans, we identified 15 non-synonymous HsIFNλ4 variants in the 1000 Genomes Project database (Figure S1A and Supplementary Data File 2), including the only three previously described variants (C17Y, P60R and P70S; >1% frequency, ‘common’) (Prokunina-Olsson et al. 2013). The remaining 12 variants were classified as rare (<1% frequency). Variants were located in functional regions such as the predicted signal peptide (amino acids 1-24), surrounding the single glycosylation site (N61), and helix F that interacts with the IFNλR1 receptor (variants 151–158; Figure 1C). The chimpanzee IFNL4 gene (encoding PtIFNλ4) differs from the human orthologue at six amino acid (aa) positions (data not shown). However, only one, aa154, differed both within humans and between species.
Screening the entire panel of HsIFNλ4 variants in antiviral (EMCV, which is a highly IFN-sensitive virus [Figure 1D]) and ISG mRNA induction assays (MX1, Figure S1B and ISG15, data not shown) revealed three out of 15 variants substantially affected activity. Consistent with previous data (Terczyńska-Dyla et al. 2014), P70S had reduced activity; a similar decrease in activity was also observed for L79F. By contrast, K154E enhanced antiviral activity by ~10-fold. These effects on activity did not arise from differences in the levels of HsIFNλ4 production or glycosylation (Figures S1C and S1D) and for K154E, enhanced secretion could not explain its higher activity (Figure S1E).
Remarkably, glutamic acid (E) is encoded at position 154 in most mammals with an IFNL4 orthologue, including chimpanzees (Figure 1E). Therefore, we compared wt HsIFNλ4 and its K154E variant to wt PtIFNλ4 and an equivalent ‘humanised’ E154K mutant in functional assays. wt PtIFNλ4 was significantly more active than HsIFNλ4 in each assay and had approximately equivalent potency to the HsIFNλ4 K154E variant (Figures 1F-1H). Moreover, PtIFNλ4 E154K had decreased activity similar to wt HsIFNλ4 (encoding lysine at aa154). Extending the analysis to include rhesus macaque IFNλ4 (Macaca mulatta, MmIFNλ4) gave the same pattern whereby wt MmIFNλ4 with E154 had greater activity than its K154 variant. By contrast, introducing a lysine into HsIFNλ3 had less of an effect on its activity (Figure 1F to H). Thus, we conclude that HsIFNλ4 has weaker activity compared to primate orthologues principally because of a single amino acid change at position 154.
We next examined the impact of HsIFNλ4 K154E on HCV infection in vitro as well as infectious assays with influenza A virus (IAV) and Zika virus (ZIKV). We also included the less active P70S and L79F HsIFNλ4 variants in these assays. Using the HCVcc infectious system, HsIFNλ4 K154E decreased both viral RNA abundance (Figure 2A) and the number of infected cells (Figure S2A), whereas P70S and L79F were less active than wt HsIFNλ4. To determine the stage in the HCV life cycle targeted by K154E, we performed assays examining virus entry (HCV pseudoparticle system [HCVpp]), initial viral RNA translation and RNA replication (both assessed by the HCV sub-genomic replicon). HsIFNλ4 K154E did not significantly alter HCVpp entry or translation of replicon RNA compared to wt protein whereas HCV RNA replication was significantly reduced by the K154E variant (Figure 2B and Figure S2B and C). HsIFNλ4 K154E also reduced titers of IAV and ZIKV to a greater extent than wt protein (Figure S2D and S2E). Correspondingly, the P70S and L79F HsIFNλ4 variants were less active than wt protein.
To examine effects of HsIFNλ4 on global transcription, cells were stimulated with HsIFNλ4s and their transcriptomes analysed, which revealed that K154E induced the broadest profile of up-regulated genes (n = 149) compared with either the wt protein (n = 88) or the P70S variant (n = 71; Figures 2D/E and Figures S2F/G). Many of the shared differentially-expressed genes included known restriction factors (IFI27, MX1, ISG15), and several unique K154E ISGs with antiviral activity, such as IDO1 and ISG20, alongside signalling activators such as STING and IRF1. From pathway analysis, all HsIFNλs induced similar transcriptional programmes with differences in the overall significance of these pathways, most notably enhancement of the antigen presentation and protein ubiquitination pathways with K154E (Figure S2H). The majority (20/32) of the chimpanzee-specific differentially-regulated genes (Figures 1A and B) were induced by HsIFNλ4 stimulation, with approximately half of those being significantly up-regulated with K154E compared to wt, including MX1, IFITM1, IFIT1 and IFIT3, TRIM22 and IFI44L (Figure 2F). Together, these data show that similar to PtIFNλ4, the HsIFNλ4 K154E, which is rarely found in humans, has greater activity and antiviral potential compared to the wt protein that is common in the human population.
From further interrogation of human genome datasets (Lachance et al. 2012), the rare HsIFNλ4 K154E variant was present in two individuals from different African rainforest ‘Pygmy’ hunter-gatherer populations (Baka and Bakola) in Cameroon (Figure 2G). The Bakola individual was homozygous for the ΔG allele, indicating that the K154E variant can give rise to functional HsIFNλ4. The Baka subject was heterozygous at rs368234815 (ΔG/TT) and thus could produce either wt or the more active K154E form of HsIFNλ4. Each of these individuals also had additional non-synonymous HsIFNλ4 variants (V158I and R151P, Baka and Bakola individuals respectively); these variants were included in our functional screen of HsIFNλ4 variants but did not significantly alter activity (Figures 1D and S1B). K154E was not found in other African hunter-gatherer populations (such as Hadza and Sandawe, Figure 2H) nor in the African San, who have the oldest genetic lineages among humans, nor was it identified in Neanderthal and Denisovan. Notably, the human TT allele encodes a K154 codon (data not shown) suggesting that the less active E154K substitution emerged very early during human evolution but not in chimpanzees, our closest living relative.
For decades, experimental studies in chimpanzees have provided unique insight into human HCV infection (Bukh 2004) but chimpanzees do not present with identical clinical outcomes. For example, chimpanzees may clear infection more efficiently (Bassett et al., 1998), rarely develop hepatic diseases similar to humans (Walker 1997) and are refractory to IFNα therapy (Lanford et al. 2007). Moreover, HCV evolves more slowly in chimpanzees, possibly due to stronger immune pressure that reduces replication (Ray et al. 2000). We propose that the enhanced activity observed with PtIFNλ4 contributes to the distinct chimpanzee response to HCV infection.
Acute HCV infection in humans and chimpanzees (and human hepatocytes in vitro) selectively stimulates type III over type I IFN production (Park et al. 2012, Thomas et al. 2012). A heightening of the IFN response to HCV has been postulated to explain the capacity to control HCV infection (Sheahan et al. 2014, Boldanova et al. 2017). Enhanced expression of ISGs in chimpanzees due to higher IFNλ4 activity could lead to greater inhibition of viral infection by coordinating a more efficient adaptive immune response, which is critical for clearance and disease (Thimme et al. 2002). We observed enhanced expression of genes involved in antigen presentation and T cell mediated immunity alongside HCV restriction factors in chimpanzees compared to humans and in IFNλ4 K154E stimulation in vitro. Additionally, IFNλ4 can inhibit type I IFN signalling (Fan et al. 2016) and inflammation (Blazek et al. 2015). Therefore, IFNλ4 with enhanced activity may act as a core co-ordinator of both protective innate and adaptive immunity.
Based on available IFNλ1 and IFNλ3 crystal structures (Miknis et al. 2010, Mendoza et al. 2017), the equivalent position to aa154 in IFNλ4 is located on the IFNλ receptor (R) 1-binding helix F, although the glutamic acid side chain faces inward towards the opposing IL10R2-binding helices (A and D; Figure 2I). This position forms non-covalent intramolecular interactions with two regions (residues K64/K67 and T108 [IFNλ3 only]) mediated by the free carboxyl group of glutamic acid. In IFNλ4, the E154-interacting positions are not conserved with IFNλ1/3 but biochemically homologous positions exist (e.g R60 and R98). We propose that E154K prohibits these critical interactions, reducing HsIFNλ4 activity via affecting receptor binding. Our data support a direct role on protein activity rather than altered production or secretion. Additionally, modelling of L79F showed that leucine sits internally and that replacement with phenylalanine would likely disrupt packing of the helices (Figure S2I).
In humans, the E154 variant was found only in African rainforest ‘Pygmy’ hunter-gatherers from west central Africa (Lachance et al. 2012). A recent study in Pygmies from Cameroon, including the Baka and Bakola groups, showed low seroprevalence of 0.6% and no evidence of chronic HCV infection (Foupouapouognigni et al. 2011). By contrast, infection in other groups in Cameroon has a seroprevalence of ~17% (Njouom et al. 2003). One explanation for this difference is higher IFNλ4 activity in populations with the K154E variant, which would enhance HCV clearance and lower endemic transmission. As San and Neanderthal and Denisovan lacked E154, Pygmy populations likely re-acquired it rather than retaining it following divergence of chimpanzee and humans. The factors driving divergent evolution of IFNλ4 within and between species are not known but we speculate that different microbial burdens might play a role, such as exposure to highly-pathogenic zoonotic infections in the Congo rainforest (Mulangu et al. 2016), a habitat shared by Pygmies and chimpanzees.
Our data beg the question as to why the vast majority of humans do not encode the more active E154 variant. We propose that it is likely that the apparent selective disadvantage the less active IFNλ4 K154 allele confers in the face of HCV infection is counterbalanced by selective advantages in other contexts. For example, type III IFN signalling has been shown to enhance disease and impede bacterial clearance in mouse models of bacterial pneumonia (Cohen et al. 2013), suggesting that IFNλ4 with a lower activity could be beneficial during non-viral infections. To conclude, our study supports a significant and non-redundant role for IFNλ4 in controlling immunity whose activity has been repeatedly attenuated during human evolution, commencing with E154K.
AUTHOR CONTRIBUTIONS
CGGB, EAC and JMcL designed the experiments. CGGB, EAC, ICF, SS and DM conducted the experiments. CGGB, EAC, SS, AdSF, JLM, KCG, SF and ST provided and analyzed data. CGGB and JMcL composed the manuscript. All authors critically reviewed the manuscript.
EXPERIMENTAL PROCEDURES
Comparison of human and chimpanzee intrahepatic gene expression during acute HCV infection
Previously published datasets of intrahepatic differentially-expressed genes from liver biopsies were used to compare human and chimpanzee transcriptomic responses to early HCV infection. Studies focusing on acute HCV infection (0 to 26 weeks) in humans and chimpanzees were acquired through manual literature search using Pubmed and compiled. For chimpanzees, data was acquired from 4 studies (Bigger et al. 2001, Su et al. 2002, Nanda et al. 2008, Yu et al. 2010) and one report was employed for human data (Dill et al. 2012). The study by Dill et al. comprised single biopsy samples from each of six individuals, while in toto the chimpanzee studies combined data from ten animals with multiple, serial biopsies. All studies were carried out using similar Affymetrix microarray platforms except Nanda et al. who used IMAGE clone deposited arrays. Humans were infected with HCV genotype (gt)1 (n = 2), gt3 (n = 3) and gt4 (n = 1) while chimpanzees were experimentally infected with HCV gt1a (n = 6), gt1b (n = 3) and gt2a (n = 1). Gene names and fold-changes were manually converted to a single format (fold change rather than log2 fold change for example) to allow comparative analysis. Human biopsies were taken between two and five months after presumed infection following known needle-stick exposure, and serial chimpanzee biopsies were taken at different time points from between one week and one year following HCV infection. For comparative purposes, differentially-expressed genes in chimpanzees were included if they were detected during a time period overlapping with the human data. We identified a ‘core’ set of chimpanzee differentially-expressed genes (independently characterized in at least two studies) and compared them to the single human transcriptome study data at equivalent time points (between 8 and 20 weeks post-infection). This approach generated a set of core chimpanzee genes (genes found differentially-expressed in at least 2 studies, >2 fold change compared to controls and during the time frame compared to humans) for comparison with the human data. This is reflected in the ten-fold higher numbers of differentially-regulated genes found in the one human study compared to the ‘core’ (narrowed down) set assembled from four chimpanzee studies. These gene sets were compared to determine their degree of species-specificity or species-similarity using Venn diagram analysis (http://bioinfogp.cnb.csic.es/tools/venny/). The gene lists of humans and core genes for chimpanzees are shown in the Supplementary Data File 1. For the chimpanzee-specific genes, mean expression values were determined at each time point from individual animals.
IFNλ gene sequence analysis
Complete known human IFNL4 genetic variation along with associated frequency and ethnicity for the human population were collected from the 1000 Genomes database available at the time of study (June 2016) (http://browser.1000genomes.org/index.html). The reference sequence for the human genome contains the frameshift ‘TT’ allele and so potential effects of variants on the HsIFNλ4 predicted amino acid sequence were identified manually following correction for the frameshift mutation (TT to ΔG). The effect of all single nucleotide polymorphisms (SNPs) on the open reading frame (ORF) was thus assessed and re-annotated as synonymous or non-synonymous resulting in the selection of coding variants reported here. Inspection of whole genome sequence data from African hunter-gatherers was carried out using previously published datasets (Lachance et al. 2012). We remapped the raw reads of six San individuals (four Juǀʼhoan and two ‡Khomani San) in the Simon Genomic Diversity Project (Mallick et al. 2016) to human reference genome (hg19) and conducted variant calling using the haplotype caller module in GATK (v3). Two Juǀʼhoan individuals were heterozygous at rs368234815 (TT/ΔG genotype, Supplementary Data File 2). The genotypes of rs368234815 in Neanderthal and Denisovan were extracted from VCF files that were downloaded from http://cdna.eva.mpg.de/denisova/VCF/hg19_1000g/ and http://cdna.eva.mpg.de/neandertal/altai/AltaiNeandertal/VCF/. Neanderthal and Denisovan all contained only ΔG alleles (Supplementary Data File 2). Amino acid sequences for mammalian IFNλ genes were obtained from NCBI following protein BLAST of the wt HsIFNλ4 polypeptide sequence. Multiple alignments of IFNλ amino acid sequences were performed by MUSCLE using MEGA7. Accession numbers of specific IFNλs used in the experimental section of this study were as follows: HsIFNλ3: Q8IU54; HsIFNλ3, Q8IZI9.2; and for IFNλ4: Homo sapiens AFQ38559.1; Pan troglodytes AFY99109.1; Macaca mullata XP_014979310.1; Pongo abelii (orangutan) XP_009230852.1, Bos taurus (cow) XP_005219183.1, Felis catus (cat) XP_011288250.1.
Structural modelling
A homology model of the HsIFNλ4 structure was generated using the RaptorX online server (http://raptorx.uchicago.edu). The resultant HsIFNλ4 structural model was then structurally aligned with both HsIFNλ1 (PDB 3OG6) (Miknis et al. 2010) and HsIFNλ3 (PDB 5T5W) (Mendoza et al. 2017). Visualization, structural alignments, and figures were generated in Pymol (The PyMOL Molecular Graphics System, Version 1.8).
Recombinant DNA manipulation and generation of IFNλ expression plasmids
DNA sequences encoding the ORFs of HsIFNλ4, PtIFNλ4 and MmIFNλ4 (based on accession sequences above) were synthesized commercially with a carboxy-terminal DYKDDDDK/FLAG tag using GeneStrings or Gene Synthesis technology (GeneArt). As a positive control for functional assays, the HsIFNλ3 ORF was codon optimised (human) to ensure robust expression and antiviral activity and is termed ‘HsIFNλ3op’. All IFNλ4 coding region sequences were retained as the original nucleotide sequence without optimisation. This precluded a direct functional comparison of HsIFNλ3op and wt HsIFNλ4. Synthesized DNA was cloned into mammalian expression vectors (pCI, Promega) using standard molecular biology techniques. At each cloning step, the complete ORF was sequenced to ensure no spurious mutations had occurred during plasmid generation and manipulation. Single amino acid changes were incorporated using standard site-directed mutagenesis protocols (QuickChange site-directed mutagenesis kit [Agilent], or using overlapping oligonucleotides and Phusion PCR).
Cell lines
A549 (human lung adenocarcinoma), Huh7 (human hepatoma), HEK293T (human embryonic kidney), U2OS (human osteosarcoma), Vero (African Green Monkey kidney) and MDCK (Madin-Darby canine kidney) cells were grown in DMEM growth media supplemented with 10% FBS and 1% penicillin-streptomycin. Non-differentiated human hepatic progenitor HepaRG cells and derivatives were cultured in William’s E medium supplemented with 10% of FBS, 1% penicillin-streptomycin, hydrocortisone hemisuccinate (50 µM) and human insulin (4 µg/mL). All cells were grown at 37°C with 5% CO2.
Plasmid transfection and production of functional IFNλ
Plasmid DNA generated by midiprep of bacterial cultures (GeneJET plasmid midiprep kit, ThermoScientific) was introduced into cells by lipid-based transfection using Lipofectamine 2000 or Lipofectamine 3000 (ThermoFisher) following manufacturer’s instructions. To produce IFN-containing conditioned media (CM) or measure protein production, HEK293T ‘producer’ cells were grown to near-confluency in 12 (~4 × 105 cells per well) or 6-well (~1.2 × 106 cells per well) plates and transfected with plasmids (2 μg) in OptiMEM (1-2 mL) overnight. At approximately 16 hours (hrs) post transfection (hpt), OptiMEM was removed and replaced with complete growth media (1-2 mL). CM containing the extracellular IFNλs was harvested at 48 hpt and stored at -20ºC before use. Although antiviral activity was observed at 16 hpt, we chose 48 hpt to harvest CM to ensure robust production and secretion of each IFNλ. Intracellular IFNλs also were harvested from transfected cells at 48 hpt. CM was removed and replaced with fresh DMEM 10% FCS (2 mL) and then frozen at -70ºC. To prepare cell lysates with IFNλ activity, plates were thawed and the cell monolayer was scraped into the media and clarified by centrifugation (5 minutes [mins] x 300 g) before use. CM or lysates were diluted in the respective growth medium for each cell line before functional testing as described in the text.
Relative quantification of RNA by reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR)
Total cellular RNA was isolated by a column-based guanidine thiocyanate extraction using RNeasy Plus Mini kit (genomic DNA removal ‘plus’ kit, Qiagen) and following the supplier’s protocol. cDNA was synthesised by reverse transcribing RNA (1 μg) using random primers and the AccuScript High Fidelity Reverse Transcriptase kit (Agilent Technologies); the recommended protocol was followed. Relative expression of mRNA was quantified by qPCR (7500 Real-Time PCR System, Applied Biosystems) of amplified cDNA. Probes for ISG15 (Hs01921425), Mx1 (Hs00895608) and the control GAPDH (402869) were used with TaqMan Fast Universal PCR Master Mix (Applied Biosystems). The results were normalised to GAPDH and presented in 2−ΔΔCt values relative to controls as described in the text. HCV genomic RNA was quantified by RT-qPCR as described previously (Jones et al. 2010).
Global transcriptomic measurements and corresponding pathway analysis
IFN-competent cells (A549) were stimulated with IFN CM (1:4 dilution) in 6-well plates (~1.2 × 106 cells) for 24 hrs and global gene expression was assessed by RNA-Seq, using three biological replicates per condition. Sample RNA concentration was measured with a Qubit Fluorometer (Life Technologies) and RNA integrity was determined using an Agilent 4200 TapeStation. All samples had a RNA integrity number of 9 or above. 1.5 μg of total RNA from each sample was prepared for sequencing using an Illumina TruSeq Stranded mRNA HT kit according to the manufacturer’s instructions. Briefly, polyadenylated RNA molecules were captured, followed by fragmentation. RNA fragments were reverse transcribed and converted to dsDNA, end repaired, A-tailed, ligated to indexed adaptors and amplified by PCR. Libraries were pooled in equimolar concentrations and sequenced in an Illumina NextSeq 500 sequencer using a high output cartridge, generating approximately 25 million reads per sample, with a read length of 75 bp. 96.3% of the reads with Q score of 30 or above. Data was demultiplexed and fastq files were generated on a bio-linux server using bcl2fastq version v2.16. RNA-Seq analysis was performed using the Tuxedo protocol (Trapnell et al. 2012). Differential gene expression was considered significant when the observed fold change was ≥2.0 and FDR/q-value was <0.05 between comparisons. Pathway analysis was carried out using Ingenuity Pathway Analysis [IPA] (Ingenuity Systems, Redwood City, CA, USA).
Western blot analysis
Cell growth media was removed and monolayers were rinsed once with approximately 0.5mL PBS before lysis using RIPA buffer (ThermoFisher) containing protease inhibitor cocktail (1x Halt Protease inhibitor cocktail, ThermoFisher, or cOmplete™, Mini, EDTA-free Protease Inhibitor Cocktail, Sigma Aldrich) for 10 mins at 4ºC before being frozen at -20 °C overnight. Lysates were collected into a 1.5 mL sample tube and clarified by centrifugation (max speed for 15 mins). Samples (10 µl) from the soluble fraction were heated to 90ºC for 10 mins with 100 mM dithiothreitol (DTT)-containing reducing lane marker at 90ºC for 10 mins. Samples were run on home-made 12% SDS-PAGE gels alongside molecular weight markers (Pierce Lane marker, Thermofisher) before wet-transfer to a nitrocellulose membrane. Membranes were blocked using a solution of 50% PBS and 50% FBS for 1 hour at room temperature and then incubated overnight at 4°C with primary antibodies in 50% PBS, 50% FBS and 0.1% TWEEN 20. Secondary antibodies were incubated in 50% PBS, 50% FBS and 0.1% TWEEN for 1 hour at room temperature. Membranes were washed four times (5 mins each) following each antibody incubation with PBS containing 0.1% TWEEN 20. After the 4th wash following incubation with the secondary antibody, the membrane was washed once more in PBS (5 mins) and kept in ddH20 until imaging. Primary antibodies to the FLAG tag (1:1000) (rabbit, lot. 064M4757V) and α- tubulin (1:10000) (mouse, lot. GR252006-1) were used along with infra-red secondary antibodies (LI-COR) to anti-rabbit (donkey [1:10,000], 926-68073) and anti-mouse (donkey [1:10,000], C50422-05) to allow protein visualisation. Pre-stained, Pagerule Plus marker was used to determine molecular weights (ThermoFisher). Membranes were visualised using the LI-COR system on an Odyssey CLX and the relative expression level of proteins determined using LI-COR software (Image Studio).
Generation and use of IFN reporter cell lines
An IFN reporter HepaRG cell line was generated to measure the activity of IFNs by introducing the EGFP ORF fused to the ISG15 ORF separated by ribosome skipping sites by CRISPR/Cas9 genome editing. We chose to introduce EGFP in-frame to the N-terminus of the ISG15 ORF because it is a robustly-induced ISG. To facilitate this we also introduced the blasticidin resistance gene (BSD). BSD, EGFP and ISG15 were separated using ribosome skipping 2A sequences (P2A and T2A). Transgene DNA was flanked by homology arms with reference to the predicted target site. Homology donor plasmids for CRISPR/Cas9 knock-in were generated through a series of overlapping PCR amplifications using Phusion DNA polymerase followed by sub-cloning into pJET plasmid. Plasmids for CRISPR/Cas9 genome editing (wt SpCas9) were generated using established protocols (Ran et al. 2013) in order to generate plasmids that would direct genome editing at the 5’ terminus of the HsISG15 ORF (exon 2). pSpCas9(BB)-2A-Puro (PX459) V2.0 was a gift from Feng Zhang (Addgene plasmid # 62988). All sequences are available by request. HepaRG cells grown in 6 well dishes were co-transfected with CRISPR/Cas9 editing plasmids targeting the beginning of the ISG15 ORF in exon 2 (exon 1 contains only the ATG of the ORF), and homology donor plasmids described above (1 µg each) using Lipofectamine 2000 and the protocol described above. Transfected cells were selected using puromycin (Life Technologies) (1 µg/mL) and blastocidin (Invivogen) (10 µg/mL) until non-transfected cells were no longer viable. Selected cells were cloned by single cell dilution, expanded and tested for EGFP induction following IFN stimulation. Positioning of the introduced transgene was assessed by PCR amplification on isolated genomic DNA from individual clones (data not shown). Primers were designed to include one primer internal to the transgene and another external to the transgene and found in the target loci (sequences available on request). For use as an effective IFN reporter cell line, cells had to demonstrate robust induction of EGFP expression following stimulation with IFN and have evidence of specific introduction of the transgene. This study uses clone ‘G8’ of HepaRG.EGFP-BSD-ISG15 cells. However, we have not tested whether there is a single transgene integration site or multiple ones nor confirmed that the EGFP produced following stimulation by IFNs results from the expression of the specifically-introduced transgene rather than off-target integration, which is theoretically possible. However, we do not predict this would affect the cells’ ability to act as a reporter cell line. For use in IFN reporter assays, stimulated cells (in 96 well plates stimulated for 24 hrs; ~5 × 104 cells per well) were washed, trypsinised and fixed in formalin (1% in PBS) at room temperature for 10 mins in the dark before being transferred to a round-bottomed plate and stored at 4°C in the dark until measurement. Non-stimulated cells were used as negative controls and the change in % EGFP-positive cells was assessed by flow cytometry using a Guava easyCyte HT (Merck Millipore).
Production of virus stocks for antiviral assays
Antiviral activity of IFNλs was determined using encephalomyocarditis virus (EMCV), influenza A virus (IAV; A/WSN/1933(H1N1)), Zika virus (ZIKV; Brazilian strain PE243) (Donald et al. 2016) and HCV (HCVcc chimeric clone Jc1) (Pietschmann et al. 2006). EMCV was obtained from and amplified on Vero cells and titrated on U2OS cells by plaque assay. IAV stocks were generated on MDCK cells and titrated by plaque assay on MDCK cells with protease (TPCK-treated trypsin, Sigma Aldrich). ZIKV was titrated on Vero cells by plaque assay. For all plaque assays, cells were grown in 12 or 6-well plates to ~90% confluency before inoculation with serial ten-fold dilutions of virus stocks in serum-free Optimem. Inoculum remained on the cells for two hrs before being removed and the monolayers were rinsed with PBS (1 x) and semi-solid Avicell overlay (Sigma Aldrich) was added. For EMCV and IAV, 1.2% avicell was used, diluted in 1X DMEM 10% FCS, 1% penicillin-streptomycin. For IAV titration, TPCK-treated trypsin was added (1 μg/mL). For ZIKV plaque assay, 2X MEM was used instead of 1X DMEM. HCVcc Jc1 was generated as described previously by electroporation of in vitro transcribed RNA into Huh7 cells and harvested at 72 hrs post electroporation. After filtration of the supernatant, HCVcc Jc1 stocks were titrated by TCID50 on Huh7 cells and stored at 4°C before use. HCVcc Jc1 TCID50 assays were performed using anti-NS5A antibody (Lindenbach et al. 2005). Infected cells at 72 hrs post infection were fixed and permeabilised with ice-cold methanol. Cells were rinsed in PBS, blocked with FCS at room temperature, incubated overnight with mouse monoclonal anti-NS5A antibody (9E10) at 4°C. After removal of the antisera, cells were rinsed 3 times with PBS containing 0.1% TWEEN 20, and then incubated in the dark at room temperature for 1 hour with secondary antibody [Alexa-fluor 488nm anti-mouse (donkey)]. Cells were finally washed with PBS containing 0.1% TWEEN 20 and NS5A-expressing cells were visualized with a fluorescent microscope.
Antiviral assays
Cells stimulated with IFNs were infected with viruses at the following multiplicities of infection (MOI): EMCV (MOI = 0.3; added directly to the media); IAV (MOI = 0.01); ZIKV (MOI = 0.01); HCVcc (MOI = 0.05). For IAV, ZIKV and HCVcc, the inoculum was incubated with cells for at two (IAV/ZIKV) or three hrs (HCVcc) in 0.5–1.0 mL serum-free Opti-MEM/DMEM at 37°C before removal. Cells were rinsed with PBS and then incubated with fresh growth media for the allotted time (24 hrs for EMCV, 48 hrs for IAV and 72 hrs for ZIKV and HCVcc). At the times stated for individual experiments, infected-cell supernatants were harvested and infectivity was titrated by plaque assay. IAV, ZIKV and HCVcc antiviral assays were all carried out in 12 well plates except for measurement of HCVcc infectivity by indirect immunofluorescence, which was measured in a 96 well plate. In the case of EMCV, a cytopathic effect (CPE) protection assay was employed to assess infectivity (Mohamed et al. 2009). Here, HepaRG cells were plated in a 96-well plates (~5 × 104 cells per well) and, when confluent, were incubated with two-fold serial dilutions of CM or lysate for 24 hrs before the addition of EMCV. At 24 hrs post infection with EMCV media was removed; cell monolayers were rinsed in PBS and stained using crystal violet (1% in 20% ethanol in H20) for 10 mins. Crystal violet stain was then removed and stained plates were washed in water. The dilution of ~50% inhibition of EMCV-induced CPE was marked visually and the difference determined relative to wt HsIFNL4.
Luciferase-expressing MLV pseudoparticles with (JFH1 HCV E1E2) were generated as described (Cowton et al. 2016) along with their corresponding E1E2 deficient controls (particles generated only with MLV core) and used to challenge IFN-stimulated Huh7 cells. Huh7 cells grown in 96-well plates overnight (seeded at 4 × 103 cells per well) were stimulated with IFNs for 24 hrs and transduced with HCVpp. 72 hrs later, cell lysates were harvested and luciferase activity was measured (Luciferase assay system, Promega) on a plate reading luminometer.
For HCV RNA replication assays, RNA was transcribed in vitro from a sub-genomic replicon (HCV-SGR) expressing GLuc (wild-type and non-replicating GND) (Domingues et al. 2015). In vitro transcribed RNA (200 ng) was transfected using PEI (1:1) into monolayers of Huh7 cells in 96-well plates overnight (seeded at 4 × 103 cells per well) that had been stimulated with IFNs (24 hrs). At the specified time points, total supernatants (containing the secreted GLuc) from treated Huh7 cells were collected and replaced with fresh growth media. 20µL (~10% of total volume) was used to measure luciferase activity and mixed with GLuc substrate (1x) (50 µL) and luminescence (as relative light units, RLUs) was determined using a luminometer (Promega GloMax). Pierce Gaussia Luciferase Flash Assay Kit (ThermoFisher) was used and the manufacturer’s instructions were followed.
Statistical analysis
For non-transcriptomic analysis (outlined above), Graphpad Prism was used for statistical testing, which included Students’ T test and ANOVA as described in figure legends. *** = <0.001; ** = <0.01; * = <0.05, are used throughout to denote statistical significance.
ACKNOWLEDGEMENTS
We are grateful to: Chris Boutell and Prof Alain Kohl for IAV and ZIKV respectively; Takaji Wakita, Ralf Bartenschlager and Arvind Patel for HCV reagents; Sam Wilson and Carol McWilliam Leitch for critically reading the manuscript. SF and ST were supported by NIH grants 1R01DK104339-0 and 1R01GM113657-01. This work was funded by the UK Medical Research Council (MC_UU_12014/1).