Chemically and enzymatically synthesized RNAs

The chemically synthesized leaderRNA mimics (65- and 73-mer) and the duplex RNA were obtained from AXOlabs (Kulmbach, Germany; see S2 Table). The 157-meric T7 transcript, encoded in plasmid pUC19_2–158 and prepared as described [26], as well as the 170- and 181-meric RNAs (S2 Table) were transcribed by T7 runoff transcription from pUC19_2–158 (157-mer) or pUC18 (170- and 181-mer) derivatives linearized with EcoR1 and purified by denaturing PAGE as described [37]. Cloning of the pUC18 derivatives was achieved by purchasing DNA oligonucleotides (Metabion, Planegg/Steinkirchen, Germany) encoding the respective RNA sequence and several extra nucleotides with terminal BamH1 and EcoRI sites, respectively. Using complementary end primers, the DNA single strands were converted to double strands by PCR, followed by cleavage with BamH1 and EcoRI and ligation into pUC18 cleaved with the same restriction enzymes.

Cell culture

Human embryo kidney cells (HEK293; DMSZ ACC 305) and human hepatocellular carcinoma cells (HuH-7; JCRB Cell Bank, JCRB0403) were cultivated at 37°C and 5% CO₂ in a humidified atmosphere in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 50 U/ml penicillin, 50 μg/ml streptomycin, 2 mM L-glutamine and 10% fetal calf serum (FCS). All components were purchased from Thermo Fisher Scientific (TFS). Cloning and propagation of MG plasmids was conducted in the E. coli DH5α strain using standard microbiological procedures.

Plasmids

Plasmids used in MG assays, such as derivatives of plasmid pCAGGS encoding the Zaire EBOV nucleocapsid proteins VP30, NP, VP35 or L, as well as plasmids coding for the T7 RNA polymerase, the EBOV-specific wild-type MG (pANDY 3E5E) or a replication-deficient variant thereof, were described earlier [38,14]. For the construction of bicistronic minigenomes plasmids pA-3E5E-GLuc-GFP and pCAGGS_Luc2 were used. For details on cloning of pA-3E5E-GLuc-GFP see S1 Table. Construction of plasmid pCAGGS_Luc2 was described previously [39].

Cloning of replication-competent (RC), replication-deficient (RD) monocistronic and RC bicistronic MG variants

MG variants were constructed as described [18,40,41], using Dpn I-based site-directed mutagenesis techniques (see S1 Fig). Cloning primers are specified in S1 Table. 3’-leader monocistronic MG variants were derived from the replication-competent (RC) EBOV minigenome pANDY 3E5E. Corresponding replication-deficient (RD) MGs were based on a pANDY 3E5E derivative lacking the terminal 55 nucleotides of the antigenome 3’-trailer replication promoter [14] using the same primers and mutagenesis setup as described for the RC MG 3’-leader variants. RC bicistronic MGs used in this study are based on a wild-type RC bicistronic (bici) MG harboring the first two EBOV cistrons, but the NP and VP35 ORFs replaced with the Renilla (RLuc) and Firefly (FFLuc/Luc2) luciferase ORF, respectively. The two ORFs, separated by the authentic NP-VP35 intergenic region (IGR), are flanked by the terminal EBOV leader and trailer regions, the NP 5’- and 3’-UTRs as well as the VP35 5’-UTR and the L 3’-UTR. Cloning of the wt bici MG and mutant variants thereof was conducted by standard restriction cloning and site-directed mutagenesis, specified and detailed in S1 Text, S1 Fig and S1 Table.

Infection of HuH7 cells with EBOV

Work with EBOV variant Mayinga (Accession number AF086833) was conducted at the biosafety level 4 (BSL4) laboratory at the Philipps University Marburg. For EBOV RNA preparations, 8 x 10⁶ HuH7 cells in DMEM supplemented with 50 U/ml penicillin (P), 50 μg/ml streptomycin (S) and 2 mM L-glutamine (Q) (abbreviated as DMEM_P+S+Q) were subjected to EBOV at a multiplicity of infection (MOI) of 3 for 1 h at 37°C. The inoculum was removed and cells were incubated in DMEM_P+S+Q additionally containing 3% FCS at 37°C and 5% CO₂ in a humidified atmosphere. Viral RNA was extracted from EBOV-infected cells 1 day post infection.

EBOV-specific MG assay

EBOV-specific MG assays were conducted as described earlier [18]. Briefly, 8 x 10⁵ HEK293 cells per well were seeded in 6-well plates (Greiner) and cultivated at 37°C and 5% CO₂ in a humidified atmosphere in DMEM_P+S+Q supplemented with 10% FCS (3 mL/well) 18–24 h before transfection. At 60–80% cell confluency plasmids coding for the T7 promoter-driven EBOV-specific MG variants (250 ng), the EBOV nucleocapsid proteins NP (125 ng), VP35 (125 ng), VP30 (100 ng), L (1000 ng), as well as a plasmid encoding T7 RNA polymerase (250 ng) were transfected into HEK293 cells using TransIT (Mirus). In the case of monocistronic MGs, the plasmid pGL4.13 (Promega) encoding FFLuc was additionally transfected for normalization of transfection efficiencies. Cells were harvested 48 h post transfection and lysed for luciferase reporter gene assays or RNA extraction.

Luciferase assays

Luciferase assays were performed as described [18]. For details, see the S1 Text.

RNA-Seq: cDNA library preparation and Illumina Sequencing

Total RNA was prepared from EBOV-infected HuH7 cells or from MG-transfected HEK293 cells and either enriched for polyA+ RNA or small RNAs (< 200 nt). Preparation of cDNA libraries was performed at vertis Biotechnologie AG (Freising, Germany). For details, see S1 Text.

Extraction and purification of total RNA for qRT-PCR analysis and Northern blotting

Total RNA of EBOV-infected HuH7 cells was isolated and purified using the mirVana miRNA Isolation Kit (with phenol; TFS) according to the manufacturer’s protocol. Total RNA of MG-transfected HEK293 cells was isolated using the RNeasy mini kit (QIAGEN) following the manufacturer’s protocol. An on-column digestion step using the RNase-Free DNase Set (QIAGEN) was included during isolation. RNA was eluted in RNase-free water. A second DNase treatment in the presence of 20 U RiboLock RNase Inhibitor (TFS) was performed by incubation with Ambion DNase I (TFS) at 37°C for 1 h. Purification was conducted using Roti-Phenol/Chloroform/Isoamyl alcohol (Carl Roth). RNA was precipitated by addition of three volumes EtOH:NaOAc [30:1; 3 M NaOAc (pH 5)]. The RNA pellet was washed with 70% EtOH, air-dried and finally redissolved in RNase-free water.

Northern blotting

LeaderRNA detection by Northern Blotting was performed essentially as described [42], using a digoxigenin-labeled RNA probe and immunological RNA detection with the DIG northern starter Kit (Sigma Aldrich/Merck). For more details of the protocol, see S1 Text.

qRT-PCR

The strategies for the detection and quantification of the different viral RNAs are schematically illustrated in S2 Fig (strategies 1–4). RNAs used for the generation of qRT-PCR standard curves are summarized in S2 Table. For further experimental details, see S1 Text.

Quantification and statistical analysis

Statistical analysis was performed using GraphPad Prism (version 8.1.1). For details (such as p values), see figure legends. A statistical significance level of 0.05 was chosen for analysis by the unpaired parametric Welch’s t test or the non-parametric Mann Whitney test.

qRT-PCR assay for innate immune induction

Analysis of innate immune induction by qRT-PCR was done essentially as described [43]. For details, see S1 Text.

Transcriptome analysis of EBOV-infected HuH7 cells reveals synthesis of abortive leaderRNAs

We performed a transcriptome (RNA-Seq) analysis of EBOV-infected HuH7 cells either enriching for small RNAs <200 nt to detect potential leaderRNAs or enriching for poly(A) RNA to detect NP mRNAs. In the small RNA libraries, considerable amounts of short transcripts antisense to the genome 3’-leader sequence were identified. The majority (~70%) was 60–80 nt in length (Fig 2A), thus resulting from termination shortly after the transcription start site (TSS) and before the PE2 region (Fig 1B). A smaller fraction of reads had a length of ~35 ± 10 nt. Hence, the RNA-Seq data provide evidence for the synthesis of abortive leaderRNAs by the EBOV polymerase complex. The majority of termination sites coincide with the region of the NP hairpin (HP) structure. The NP HP is predicted to form on the genomic and/or antigenomic RNA level (nt 56–78, see Fig 1B) and is involved in regulation of transcription by VP30 [26,41,44,45]. In the mRNA-enriched library, most NP mRNA reads had their 5’-end at the expected position 56 and minor fractions at nt 57 and 58 (Fig 2B). This implies that the viral polymerase can also initiate transcription at positions 57/58, although at present we cannot exclude that all transcripts start with G₅₆, but lost this (and the following) nucleotide during library construction or sequencing.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

RNA-Seq analysis of (A) small RNAs (<200 nt) or (B) poly(A) RNAs derived from EBOV-infected HuH7 cells. (A) Length of EBOV leader transcripts in small RNA libraries. The vast majority of reads had their 5’-end at position 2, thus a 55-meric leaderRNA had its 3’-end at the TSS (position 56); leaderRNA reads with lengths between 15 and 100 nt were were defined as viral leader transcripts and used as read pool for leaderRNA length/3’-end analysis. Error bars represent standard errors of the mean (SEM) calculated for each transcript length based on three biological replicates (S3 Table). Dashed vertical lines demarcate arbitrary length windows, with percentages indicated; the antigenomic leader and part of the 5’-UTR of the NP mRNA are shown schematically below the graph. (B) Analysis of NP mRNA reads with 5’-ends between antigenome position 54 and 70 (% of reads at each position). The dashed vertical line marks the annotated EBOV transcription start site (TSS). Error bars indicate SEM based on three biological replicates (S3 Table). The sketch at the bottom shows the genomic 3’-leader with the expected transcription start site at nt 56 (light blue arrow). The RNA-Seq analysis is consistent with position 56 being the major TSS, but left the possibility open that some transcripts may be initiated at position 57 and 58 in EBOV-infected cells.

https://doi.org/10.1371/journal.ppat.1010002.g002

Northern Blot and qRT-PCR confirm the presence of leaderRNAs in EBOV-infected cells

To validate our finding of abortive leaderRNAs in EBOV-infected cells by RNA-Seq (Fig 2A), we performed Northern Blot analyses with a leader-specific T7-transcribed and digoxigenin-labeled probe that exhibits full complementarity to nt 1–80 of the expected leader transcript (Fig 3A, left). Indeed, we were able to detect leaderRNAs by using 20–40 μg of total RNA extracted from EBOV-infected HuH7 cells. In vitro-transcribed (ivt) RNAs identical to antigenome nt 2–78 and 56–158 as well as genomic nt 154–1 (Fig 1B) were used as positive and negative controls, respectively. The major fraction of leader transcripts migrated somewhat faster than the 2–78 nt reference RNA in native 20% PAA gels. Heterogeneous transcript lengths are consistent with the RNA-Seq data (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Northern Blot and qRT-PCR analysis of total RNA derived from EBOV-infected HuH7 cells.

(A) After electrophoresis on a native 20% PAA gel, leaderRNAs were detected by Northern blotting using a complementary, digoxigenin-labeled in vitro-transcribed RNA probe (nt 80–1 of the genomic RNA; schematically illustrated on the left). A T7-transcribed reference RNA (+) 2–78, corresponding to nt 2–78 of the antigenome, served as positive control and size marker; T7 transcripts (+) 56–158 and (-) 154–1 (representing the corresponding antigenome or genome sequence; see Fig 1B) were included as negative controls. Total RNA from Mock (non-infected) cells was used as additional negative control. Leader transcripts of ~ 60–70 nt were detected 15 min and 30 min post exposure. (B) Quantification of the NP mRNA:leaderRNA and leaderRNA:cRNA ratios for 3 biological replicates with 2 technical replicates each based on C_T values determined according to strategy 1 depicted in S2A Fig. The column graph illustrates the resulting mean ratio ± SEM for NP mRNA/leaderRNA (8.75 ± 1.02) and leaderRNA/cRNA (4.19 ± 0.34). An exemplary calculation for one of the replicates is shown in S3 Fig. (C) Molar NP mRNA/leaderRNA ratio based on C_T values obtained according to RT-PCR strategy 4 (S2D Fig) and including standard curves, either based on the synthetic 73- or 65-meric leaderRNA mimic as standard.

https://doi.org/10.1371/journal.ppat.1010002.g003

LeaderRNA amounts are lower than those of NP mRNA

In a third approach we used quantitative real-time PCR (qRT-PCR) to estimate the molar ratio of leaderRNA to NP mRNA. For this purpose, we designed three primer sets for amplification of NP mRNA, leaderRNA and longer antigenomic RNA as a measure of replicative RNA (cRNA; S2A Fig). Notably, only two of the three primer sets are specific for a single viral RNA species (NP mRNA and cRNA), whereas the leader primer set amplified leaderRNA as well as cRNA. Using this qRT-PCR setup, we again confirmed leader transcript synthesis since the primer set detecting both, leaderRNA and cRNA, resulted in lower C_T values of ~16 than the cRNA-specific primer set (C_T ~18) (exemplary experiment shown in S3B Fig). C_T values obtained with the NP mRNA-specific primer pair were clearly lower than those obtained with the leaderRNA+cRNA primer pair (C_T ~13 versus C_T ~16), indicating that NP mRNAs are more abundant than leaderRNAs. As the leaderRNA primer set simultaneously detected leaderRNA and cRNA, we indirectly calculated the ratio of NP mRNA:leaderRNA (see S3 Fig for details). A total of three biological replicates were analyzed in technical duplicates to determine an average ratio of ~9:1 for NP mRNA:leaderRNA amounts and ~4:1 for leaderRNA:cRNA amounts (Fig 3B). To consider possible differences in RT primer efficiencies, we also determined the relation of RNA copy number and C_T values by use of enzymatically or chemically synthesized RNAs (see S2 Table for sequences) that are sequence-identical to the amplified portions of the different viral RNA species. The used primer sets are illustrated in S2D Fig (strategy 4; for primer sequences, see S1 Text, paragraph qRT-PCR, Strategy 4) and the derived standard curves are shown in S4A Fig (see also S4B Fig for specificity of the primer pairs). In this setup, the molar NP mRNA:leaderRNA ratio increased to ~21 to 32-fold (see example calculations in S4C Fig), depending on the leaderRNA mimic (73- or 65-mer) used for generating the standard curve (Fig 3C). The leaderRNA levels may be somewhat underestimated, as the RT primer for leaderRNA (covering nt 64–39), may anneal inefficiently or not at all to leaderRNAs <60 nt. Based on leaderRNAs <60 nt contributing ≤30% to all leaderRNAs (Fig 2A), the mRNA:leaderRNA ratio could maximally reduce by a factor of ~1.5. This would still correspond to a more than 10-fold molar excess of mRNA over leaderRNA.

LeaderRNAs are also produced in MG systems

We also performed RNA-Seq using small RNA (<200 nt) preparations derived from cells transfected with monocistronic EBOV MGs. This revealed the presence of abortive leaderRNAs as well (Fig 4A), the majority (~63%) also 60–80 nt in length owing to termination in the region between PE1 and PE2 (see Fig 2A). This demonstrated that the synthesis of abortive leaderRNAs is not restricted to virus infections but can be investigated using EBOV-specific MG systems as well, thereby enabling mechanistic and biochemical studies outside BSL-4 laboratories. We initially quantified leader transcripts by qRT-PCR using the approach specified in S2B Fig. qRT-PCR of Firefly luciferase mRNA, synthesized from a cotransfected plasmid, served as internal standard to cancel out fluctuations in plasmid transfection efficiency [18]. As both primer sets used for Rluc mRNA or leaderRNA amplification (S2B Fig) also amplified cRNA and primer efficiencies were comparable (both had E values of ~2.00), we could directly calculate an mRNA:leaderRNA ratio of ~70:1 by dividing the 2^-ΔC_T values for mRNA+cRNA and leaderRNA+cRNA. Hence, in the MG context, mRNA levels exceeded those of leaderRNA to a larger extent than in the EBOV infection setup (Figs 3B vs. 4B). As for the EBOV infection model, we also validated this finding on the basis of qRT-PCR standard curves (S2D and S4A Figs). A ~100 to 190-fold molar excess of Rluc mRNA was derived from this approach, depending on the leaderRNA mimic (73- or 65-mer) used for generating the standard curve (Fig 4C). Again, the ratio reduces by a factor of ≤1.5 due to inefficient reverse transcription of leaderRNA reads <60 nt (see above).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4.

(A) RNA-Seq analysis of leaderRNA length/3’ termination sites in small RNA (<200 nt) libraries derived from cells transfected with the wt MG and coexpressing VP30. The shown mean values (± SEM) are based on 3 to 5 biological replicates each. For more details, see legend to Fig 2A and S3 Table. (B) Quantification of viral mRNA to leaderRNA ratio in MG-transfected HEK293 cells by a two-step qRT-PCR. The qRT-PCR setup is schematically depicted in S2B Fig (strategy 2). Relative levels (2^-ΔC_T values; normalized to Firefly luciferase mRNA) of mRNA+cRNA, leRNA+cRNA, and cRNA in the presence (+) or absence (-) of the viral polymerase L are shown (mean 2^-ΔC_T = ~314 for mRNA+cRNA; mean 2^-ΔC_T = ~4.6 for leaderRNA+cRNA, and mean 2^-ΔC_T = ~2 for cRNA alone). The horizontal line marks the cRNA level. Mean values (± SEM) were derived from 5 independent experiments (exp.) with 2 (1 of 5 exp.) or 3 (4 of 5 exp.) technical replicates each. (C) Molar Rluc mRNA/leaderRNA ratio based on C_T values obtained according to RT-PCR strategy 4 (S2D Fig) and including standard curves, either based on the synthetic 73- or 65-meric leaderRNA mimic as standard. (D) LeaderRNA 5’-ends (start sites) in RNA-Seq libraries derived from EBOV-infected HuH7 cells (upper bar), wt MG-transfected HEK293 cells (central bar), or a mock (non-infected/non-transfected) HEK293 control spiked with a synthetic 65-meric leaderRNA (lower bar; see S2 Table for sequence). The color code is specified on the right. The number of biological replicates (n) is indicated on the right (for details, see S3 Table).

https://doi.org/10.1371/journal.ppat.1010002.g004

Viral leaderRNA synthesis is specifically initiated opposite of genome nt 2

A recent study reported that antigenome synthesis is not initiated opposite to the 3’-terminal G, but opposite to C₂ [16]. This raised the question whether the viral polymerase initiates leaderRNA opposite to C₂ as well or at another site as shown for RSV (see Introduction). The vast majority of leaderRNAs in our RNA-Seq libraries from EBOV-infected and MG-transfected cells also lacked the nucleotide complementary to genome position 1 (Fig 4D, upper and central bar). To ascertain that the absence of the first nucleotide is not an artefact of the RNA-Seq procedure or data processing, we performed an RNA-Seq experiment under identical conditions using a small RNA (<200 nt) preparation from HEK293 cells spiked with a synthetic 65-meric leaderRNA mimic (carrying a 5’-triphosphate; for sequence information see S2 Table) that included nt C₁. Sequencing revealed that almost 100% of the spike-in RNA reads included the 5’-terminal C₁ nucleotide (Fig 4D, bottom bar), while it was predominantly lacking in RNA samples derived from either EBOV-infected or MG-transfected cells (top and central bar). This indicates that not only viral antigenomes (cRNAs) but also leaderRNAs are intiated at genome position 2.

First genome nucleotide is dispensable for replication but preferred in transcription

As the results by Deflubé et al. [16] and our data (Fig 4D) indicated genome position 2 as the start site for RNA synthesis by the viral polymerase, we addressed the question whether the genome’s first nucleotide is not only dispensable for replication but for mRNA transcription as well, considering that leaderRNAs might be obligatory pre-products of viral NP mRNA synthesis [32]. We therefore constructed MGs lacking the 3’-terminal first (G₁; variant Δ1) or the first two (G₁ and C₂; variant Δ2) nucleotides, as well as a variant carrying an additional 3’-nucleotide (variant +G, Fig 5A). We first measured reporter activity of these variants in the context of a replication-competent MG (RC MG) backbone. The Δ1 variant was even slightly more efficient than the wt construct (Fig 5B). Deletion of nucleotides 1 and 2 (variant Δ2), however, strongly decreased reporter activity. Remarkably, an additional G residue at the genome 3’-end (variant +G) substantially increased reporter activity relative to the wt MG (Fig 5B). We then analyzed the same genome variants as part of the replication-deficient MG (RD MG) backbone. In the RD MG, the viral polymerase can utilize the vRNA genomes synthesized by T7 RNAP for mRNA transcription and production of replicative cRNA, but is unable to synthesize new vRNA genomes to boost viral transcription and replication. Strikingly, all three mutant variants showed a clear defect in the RD MG setup (Fig 5C), indicating that the wt 3’-end is optimal for viral transcription; even the +G variant, most active in the RC MG system, showed decreased transcriptional activity in the RD MG context. Variant Δ2 was essentially reduced to background (–L) levels, supporting the key role of C₂ for initiation of EBOV transcription and replication [16]. The levels of mRNA, cRNA and vRNA, derived from the same RC MG-transfected cells as used for reporter gene assays, showed the same ranking of the tested variants as observed in the reporter assay (cf. Fig 5D–5F and 5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Analysis of replication-competent (RC) and replication-deficient (RD) MGs with mutated genome 3’-ends.

(A) Genome 3’-end of the wt MG and three mutant derivatives that either lack the 3’-terminal G residue (Δ1), the 3’-terminal two residues (Δ2) or carry an extra 3’-G residue (+G). (B-C) Corresponding reporter gene assays of lysates from cells transfected with MG variants illustrated in panel A, either as part of the (B) RC MG or (C) RD MG backbone. Mean activity values (± SEM) of the native 3’-leader MG (dark gray bars) and mutant MGs (light blue bars) were derived from 3 independent experiments with 3 technical replicates each; data for the wt MG were set to 100%. As negative control, the plasmid encoding the L gene was omitted during transfection (–L; black bars). (D-F) Corresponding two-step qRT-PCR of RC MG samples using the same cells as in panel B (for qRT-PCR setup, see S2C Fig); color code as in panel B and C. Mean 2^-ΔΔCT values (± SEM) of viral mRNA (D), cRNA (E) and vRNA (F) derived from 3 independent experiments with 3 technical replicates each. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 (unpaired Welch’s t test).

https://doi.org/10.1371/journal.ppat.1010002.g005

Abortive leaderRNA synthesis does not result from alternative GE signals

The RNA-Seq results indicated that leader transcripts are predominantly terminated immediately after the polymerase has passed the NP GS signal in EBOV-infected (Fig 2A) or MG-transfected cells (Fig 4A). For RSV, it was noticed that leader promoter nucleotides essential for both transcription and replication are also conserved at the corresponding positions in the GS of the L gene [33]. This suggested the possibility that leaderRNAs may be terminated by an unidentified termination (GE) signal, thus also employing a mechanism similar to the stop-start mechanism ubiquitously used by NNS viruses for mRNA transcription at internal genes. To address the possible presence of an alternative GE signal that could explain the production of abortive leaderRNAs in EBOV, we constructed a bicistronic MG mimicking the first two EBOV genes, which had the following features (Fig 6A, top): it comprised the authentic 3’-leader and 5’-trailer sequences, the native NP-VP35 gene border, but the protein-coding regions of the NP and VP35 genes replaced with two different luciferase genes (corresponding to mRNA 1 and mRNA 2); in addition, the L 3’-UTR substituted for the VP35 3’-UTR, and the VP35 5’-UTR hairpin was replaced with the NP 5’-UTR hairpin (construct termed “bici NP-NP”). To find out if the NP hairpin itself or if cis-acting sequences within the NP hairpin structure are sufficient for alternative transcript termination, we first inactivated the NP GE signal at the NP-VP35 gene border by mutation (Fig 6A, bici NPGEmut-NP, second from top). In the third construct, a derivative of the second, we further inserted the 3’-terminal 55 leader nt (PE1) to investigate if either sequences between leader end and NP hairpin are recognized as alternative GE signals or if nucleotides in PE1 and the NP hairpin may somehow interact to induce transcript termination (Fig 6A, bici NPGEmut-leNP). Note that we expected synthesis of mRNA 2 and thus Firefly luciferase (FF) reporter gene activity only if an alternative GE signal compensated the inactivated NP GE signal at the NP-VP35 gene border. As positive control we included an additional derivative construct in which we replaced the NP hairpin at the 2^nd cistron with the VP40 5’-UTR hairpin (see Figs 1A and 6A, bottom). The VP40 HP harbors an overlapping GS-GE signal, thus in our bicistronic MG context it introduces a functional GE signal that enables termination of mRNA 1 despite the defective GE signal at the NP-VP35 gene border (Fig 6A, bici NPGEmut-VP40). In luciferase reporter assays, we did not observe substantial FF reporter activity for MG constructs 2 (NPGEmut-NP) and 3 (NPGEmut-leNP) that carried the mutant NP GE signal, while our positive control resulted in FF reporter activity comparable and even higher than that of the bici NP-NP MG construct (Fig 6B and 6C). Hence, it is unlikely that an alternative GE signal within the 3’-terminal 78 nt of the leader mediates transcript termination. Residual reporter gene levels above the background (–L) control in the case of constructs 2 and 3 might be explained by residual induction of termination by the mutated GE signal. Alternatively, this residual firefly luciferase activity could originate from translation of readthrough mRNAs (cotranscripts of mRNA 1 and 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Testing for alternative gene end (GE) signals in the genomic 3’-leader promoter using a bicistronic (bici) MG system reflecting the EBOV NP-VP35 gene border.

(A) Bici MGs encode a Renilla (light blue box) and Firefly (light red box) luciferase (for further details on color coding, see Fig 1). In construct bici NP-NP, the native VP35 5’-UTR hairpin at the second cistron was exchanged with the NP 5’-UTR hairpin. Then, the MG variant bici NPGEmut-NP was constructed, in which the native NP GE sequence (GEmut, light green box) was mutated (marked by red vertical lines, mutations shown in the box at the bottom). For this construct, one would expect Firefly luciferase expression only if the NP HP harbored an alternative GE signal. In the third test construct (bici NPGEmut-leNP), a derivative of bici NPGEmut-NP, the entire 3’-terminal 78 nt of the 3’-leader promoter were inserted into the intergenic region. In construct bici NPGEmut-VP40, the NP HP at the second cistron was replaced with the VP40 5’-UTR HP that introduces a functional second GE signal (GE2) to replace the inactivated GE signal at the end of the NP gene; this variant served as positive control. (B-C) Corresponding Renilla (B) and Firefly (C) luciferase reporter gene assays were used as an indirect readout for viral transcription of mRNA 1 encoding Renilla and mRNA 2 encoding Firefly luciferase. Activity values of the wt NP-NP MG were set to 100%.–L, negative control in which the plasmid encoding L was omitted during transfection. Mean values ± SEM were derived from 3 independent experiments with at least 2 technical replicates each. *p < 0.05; ****p < 0.0001; n.s., not significant (unpaired Welch’s t test).

https://doi.org/10.1371/journal.ppat.1010002.g006

VP30 suppresses leader transcript synthesis

The NP hairpin structure regulates VP30-dependent transcription initiation at the TSS [41,44]. To examine if VP30 affects leaderRNA synthesis, we applied the qRT-PCR setup illustrated in S2B Fig. As expected, mRNA+cRNA levels, predominantly representing mRNA levels (Fig 4B), largely decreased in the absence of VP30, while cRNA alone was significantly increased (Fig 7A and 7B). This is in line with previous qRT-PCR data demonstrating that a decrease of transcription correlates with an increase of replication [23,25,41]. In addition, leaderRNA+cRNA levels were also increased in the absence of VP30 (Fig 7C). This increase was even more pronounced (~1.6-fold; Fig 7C) compared to the increase observed for cRNA alone (~1.2-fold; Fig 7B), emphasizing that leaderRNA levels increase as well in the absence of VP30. A direct comparison of 2^-ΔC_T values obtained by the leaderRNA+cRNA and the cRNA-specific primer sets illustrates the overall relations (Fig 7D): in the presence of VP30 (gray columns), leaderRNA+cRNA levels are 2.39 fold higher than the cRNA level; this ratio increases to 2.54 fold based on the mean 2^-ΔC_T values obtained in the absence of VP30 (pink columns), in line with enhanced leaderRNA and cRNA synthesis in the absence of VP30. As mentioned before, the measured leaderRNA levels are lower estimates due to inefficient RT-PCR amplification of leaderRNAs <60 nt.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. leaderRNA synthesis in the presence or absence of VP30.

(A-C) Two-step strand-specific qRT-PCR (see S2B Fig) quantification of (A) mRNA+cRNA, (B) cRNA or (C) leaderRNA+cRNA using the 2^-ΔΔC_T method as described [18]. HEK293 cells were transfected with the RC MG encoding the native 3’-leader, either with or without coexpression of VP30. Transfections without the plasmid encoding L served as negative controls. Mean 2^-ΔΔC_T values (± SEM) were derived from 5 independent experiments (exp.) with 3 (4 of 5 exp.) or 2 (1 of 5 exp.) technical replicates. (D) Mean 2^-ΔC_T values (± SEM) of leaderRNA+cRNA versus cRNA, corresponding to samples analyzed in panel C and B, respectively. 2^-ΔC_T values were determined according to strategy 2 (S1 Text, S2B Fig). *p < 0.05; ***p < 0.001; ****p < 0.0001 (unpaired Welch’s t test).

https://doi.org/10.1371/journal.ppat.1010002.g007

RNA-Seq, as introduced in Fig 2, was performed with RNA from cells transfected with the wt NP MG, with a MG construct harboring a destabilized hairpin (NheI NP) and with the MG variant Δ5’ spacer lacking the capacity to form a hairpin structure owing to a 12-nt deletion in the spacer between PE1 and PE2 (Fig 8A) [40,44]. In line with the qRT-PCR results, we observed an increase in leaderRNA amounts in the absence of VP30 for the wt (NP) MG and the two mutant MGs (Fig 8B, red columns), while mRNA levels largely decreased in the absence of VP30 (Fig 8C). The findings show that this reverse regulation of leaderRNA and mRNA 1 levels by VP30 is not dependent on the presence of the native NP hairpin in the 3’-leader.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. RNA-Seq analysis of relative leader transcript and Renilla mRNA levels in RC MG systems, with (+) and without (-) cotransfection of the plasmid encoding VP30.

(A) Illustration of the MG encoding the wt NP HP and mutant derivatives either encoding the NheI HP [44] destabilized particularly on the mRNA level or the Δ5’ spacer variant [40] devoid of any hairpin structure on the genomic and mRNA level. Minimum free energies (MFE; ΔG) of the secondary structures were predicted by RNAfold using the default parameters [80]. Black underlined residues in the NheI HP mark mutated residues. For more contextual details, see Fig 1B. (B) Fraction of leaderRNA reads (in %) in small RNA (< 200 nt) libraries derived from cells transfected with the MGs illustrated in panel A; the corresponding fraction (0.11 ± 0.06) is also shown for EBOV-infected cells. For the definition of leaderRNAs, see legend to Fig 2A; error bars are standard errors. The difference between the relative read numbers of leaderRNAs with and without VP30 cotransfection is significant if all MGs are pooled together (***p = 0.00025, Welch’s t test). The individual difference for the NheI NP construct is significant as well (**p = 0.0022, Welch’s t test) but not for the other MGs (p = 0.13 for wt NP, p = 0.054 for Δ5’ spacer, Welch’s t test). (C) Fraction of Rluc mRNA reads (in %) in corresponding poly(A) RNA-enriched libraries. For details (panel B and C) on sample and library preparation, Illumina Sequencing, sequencing analyses and biological replicates, see the S1 Text and S3 Table. (D, E) Analysis of Rluc mRNA reads with 5’-ends between antigenome position 54 and 70 (% of reads at each position) in poly(A) RNA-enriched libraries derived from cells transfected with the Δ5’ spacer MG in (D) the presence (+) or (E) absence (-) of VP30. Percent values were normalized to the sum of all mRNA reads with 5’-ends mapping to positions 54 to 70, based on two biological replicates (± SEM) in panel D and three biological replicates (± SEM) in panel E (see S3 Table for details). The data indicate position 56 as the major transcription initiation site.

https://doi.org/10.1371/journal.ppat.1010002.g008

The position of transcription initiation is neither affected by deviations from the native NP hairpin structure nor by the presence or absence of VP30

We wondered if deviations from the native NP hairpin structure and presence versus absence of VP30 might affect selection of the transcription initiation site. However, RNA-Seq revealed that, even in the absence of VP30, mRNA synthesis was predominantly initiated opposite to genome nt -56 in case of the Δ5’ spacer MG (Fig 8D and 8E), the NheI NP MG (S5 Fig) and the wt (NP) MG (S5 Fig). Thus, recognition of the TSS and site-specific initiation of mRNA synthesis is basically the same in the presence or absence of VP30, although the overall efficiency of viral transcription is very low in the absence of VP30 [41].

LeaderRNA termination is independent of RNA structure formation potential at the TSS

Overall, 3’-ends of leaderRNAs, as inferred from RNA-Seq, map to the same region for EBOV-infected cells as well as cells transfected with the wt (NP) MG and the two aforementioned mutant constructs (Fig 9A–9D; for more details, see Discussion). Of note, reads terminating in the nt 70 region contain an increased fraction of non-templated nucleotides, which seems even more pronounced in the MG-transfected versus EBOV-infected system (S6A–S6D Fig). As the leaderRNA length profiles for wt (NP) MG and the Δ5’ spacer MG devoid of any secondary structure are very similar, we can conclude that termination of leader transcripts is not influenced by the potential to form RNA structures in the spacer between PE1 and PE2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9.

Comparative RNA-Seq analysis of leaderRNA lengths in (A) EBOV-infected HuH7 cells or HEK293 cells transfected with (B) MG NP HP (wt), (C) MG NheI HP or (D) MG Δ5’ spacer. Panel A is identical to Fig 2A and included for comparison; the mean vales (± SEM) in panels B to D are based on four (each wt MG and MG NheI) or five (MG Δ5’ spacer) biological replicates. For more details, see legend to Fig 2A and S3 Table.

https://doi.org/10.1371/journal.ppat.1010002.g009

Initiation and termination sites for leader transcripts are independent of VP30

We then comparatively analyzed the small RNA libraries for leaderRNA initiation and termination sites in RNA samples derived from cells transfected with the wt (NP) MG versus the NheI NP and Δ5 ‘spacer MGs, either in the presence or absence of VP30. For all three MG variants and independent of VP30, leaderRNA 5’-ends mapped to position 2 (S7A Fig). Only ~0.2% of reads mapped to position 1, similar to leaderRNAs isolated from EBOV-infected cells (Fig 4D). In all cases, the RNA-Seq data provided no evidence for substantial changes in the respective 3’-end patterns of leaderRNAs upon omission of VP30 (cf. S6E–S6G and S6B–S6D Fig). We conclude that VP30 neither affects the position of leaderRNA initiation nor the pattern of termination sites.

Leader transcripts are not synthesized if hexamer spacing between PE1 and PE2 is violated

The NP hairpin represents the major part of the spacer region separating PE1 and PE2 of the leader promoter (Fig 1B). It was recently shown that hexamer phasing in the EBOV 3’-leader promoter is not only crucial for replication [44] but also for initiation of viral transcription [18]. This raised the question whether hexamer phasing affects leaderRNA synthesis as well. To address this issue, we used two MG variants deviating from hexamer phasing in the promoter, one carrying the NP hairpin with a 1-nt deletion near the stem and a variant in which the NP hairpin was replaced with the corresponding element of the VP35 gene (Fig 10A). In contrast to the wt (NP) hairpin conforming to hexamer phasing (with 30 nt between genome pos. -51 to -80), both mutant MGs comprised only 29 nt in this region. As shown by qRT-PCR, the two mutant MGs were defective in antigenome (cRNA), mRNA and leaderRNA synthesis (Fig 10B–10D). We conclude that hexamer phasing is generally essential for initiation of RNA synthesis in the leader promoter.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. qRT-PCR analysis of leaderRNA synthesis from MG templates that deviate from hexamer phasing.

(A) Illustration of RC MG variants, either encoding the wt NP HP, a derivative thereof with a single nt deletion, construct NP-1 (stem), or a variant in which the NP HP was replaced with the corresponding hairpin of the VP35 gene; in the latter two constructs, the distance between nt 51 and 80 is not a multiple of 6 (29 nt). For detailed information on color code see legend to Fig 1B; ΔA, deletion of a single A residue. (B-D) qRT-PCR analysis (performed as described in Fig 7 and S2B Fig, strategy 2) of MG variants illustrated in panel A; (B) mRNA+cRNA level, (C) leRNA+cRNA level, (D) cRNA level. Mean 2^-ΔΔC_T values derive from 3 independent experiments with 3 technical replicates each.

https://doi.org/10.1371/journal.ppat.1010002.g010

Abortive trailer transcripts

We also identified putative abortive trailer transcripts in our RNA-Seq libraries from EBOV-infected and MG-transfected cells, attributable to prematurely terminated vRNA synthesis initiated on antigenomic cRNA. TrailerRNAs can be differentiated from leaderRNAs by sequence differences at nt positions 15/16, 22/23 and 25/26 (Zaire EBOV genome, NC_002549.1). TrailerRNAs were less (~20-fold) abundant than leaderRNAs in RNA-Seq libraries (cf. Figs 8B and S7C). As for leaderRNA, the vast majority of trailerRNAs were initiated at position 2 of the template RNA (cf. S7A and S7B Fig). We observed a non-significant trend toward increased trailerRNA reads in libraries from MG-transfected cells without VP30 coexpression (S7C Fig), consistent with elevated cRNA template synthesis in the absence of VP30 (Fig 7B). Relative to leaderRNAs, the fraction of trailerRNA reads in the length range of ~20–45 nt was increased compared with the nt ~60 to 80 fraction (cf. S6 and S8 Figs). This change was more pronounced for RNA libraries derived from MG-transfected compared with those from EBOV-infected cells (cf. S8A and S8B–S8D Fig). For MG-transfected cells, the patterns were essentially identical in the presence and absence of VP30 (S8B–S8D Fig versus S8E–S8G Fig). As observed for the leaderRNAs, reads terminating in the nt 70 region showed an increased 3’-terminal deviation from the genome-encoded sequence (S8 Fig).

Leader transcripts do not affect innate immunity

T7-transcribed, 5’-triphosphorylated leader RNAs from NNS viruses, such as measles, rabies, vesicular stomatitis virus (VSV) or Newcastle disease virus (NDV) were reported to induce the antiviral type I interferon (IFN) system by activating the virus sensor RIG-I [46,47,48]. Likewise, an EBOV leader transcript mimic comprising nt 2–56, synthesized in vitro by T7 RNA polymerase as well, showed RIG-I-dependent activation of the IFN-β promoter [47]. It remained unclear, however, whether side products generated by T7 transcription were responsible for this observation. We revisited this issue by analyzing effects of chemically synthesized leaderRNA mimics with 5’-triphosphate ends on innate immunity. We first tested this with a chemically synthesized 65-meric leaderRNA mimic whose 5’-terminal nucleotide is complementary to the very 3’-end of the genomic RNA (S2 Table). The 65-mer was selected because its 3’-terminus was frequently represented in the small RNA-Seq libraries derived from EBOV-infected cells. After finding that leaderRNAs are initiated at position 2 (Fig 4D), we further included a 73-mer mimicking leaderRNA initiated at the 2nd 3’-terminal genome position, which represented the major size range of leaderRNAs (Fig 2A). A validated synthetic duplex RNA, previously demonstrated to elicit RIG-I-mediated immune responses ([49] see S2 Table), as well as VSV RNA served as positive controls and tRNA as a negative control. Upon transfection of HEK293 cells with VSV RNA and the duplex RNA control, we observed significant increases in the mRNA levels of IFN-β and the cytokine CXCL10. By contrast, immunostimulatory effects were neither seen with tRNA nor with the 65- and 73-meric leaderRNA mimics (Fig 11). Successful transfection of HEK293 cells with the leaderRNA mimics was confirmed by qRT-PCR (S9 Fig). Also, in experiments where we cotransfected increasing amounts of the 65-mer with a constant amount of VSV RNA, we found the intracellular IFN-β and CXCL10 mRNA amounts to be elevated at constant levels irrespective of the 65-mer concentration in the transfection mix (S10 Fig). The absence of a significant inhibitory effect exerted by increasing amounts of 65-mer on VSV RNA-induced immunity argues against the possibility that leaderRNAs might act as a bait to sequester RIG-I and shut down the protein’s activity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Analysis of innate immunity induction in HEK293 cells upon transfection with chemically synthesized, 5’-triphosphorylated EBOV leaderRNA mimics (65-mer, 73-mer), using mRNA levels for IFN-β and CXCL10 as readout; mRNA levels were determined by qRT-PCR.

Vesicular Stomatitis Virus (VSV) genomic RNA isolated as described [81] and a validated RIG-I-activating duplex RNA (3P-G/AS G24; [49]) were used as positive controls and bulk tRNA from Saccharomyces cerevisiae as negative control. In the mock control, transfection was performed with H₂O instead of RNA. The graph with mean values and standard deviations is based on six independent experiments each. Statistical significance was evaluated using the Wilcoxon test *, p < 0.05; ns, not significant. For sequence and structure of the 65-mer, 73-mer and duplex 3P-G/AS G24, see S2 Table; for methodological details, see S1 Text.

https://doi.org/10.1371/journal.ppat.1010002.g011