The development of biologically contained Sudan virus as an alternative and safe tool for fundamental filovirus research

ABSTRACT

Sudan virus (SUDV; species Orthoebolavirus sudanense) remains a major public health threat, yet research is limited by restricted access to biosafety level 4 (BSL-4) facilities. To address this, we developed a biologically contained SUDV lacking the essential VP30 gene, restricting replication to VP30-expressing cells. We demonstrate efficient virus rescue, strict functional and genetic containment, and replication kinetics comparable to wild-type SUDV in permissive cells. Using heterologous VP30-expressing cell lines, we observe strong cross-functionality among orthoebolaviruses, whereas Marburg virus VP30 shows minimal activity, highlighting genus-specific constraints. The system supports high-throughput antiviral screening and confirms robust activity of remdesivir. In addition, resistance profiling identified substitutions at residue F549 of viral polymerase L as key determinants for remdesivir resistance, with additional mutations at M675. Together, this biologically contained SUDV system enables safe study of viral replication, antiviral discovery, and resistance evolution under lower biosafety conditions.

Introduction

Filoviruses, including Ebola virus (EBOV; species Orthoebolavirus zairense), Sudan virus (SUDV; species Orthoebolavirus sudanense), and Marburg virus (MARV; species Orthomarburgvirus marburgense), are among the most known lethal viruses, causing severe diseases with high case fatality rates [1]. While EBOV has historically been of high interest to the scientific community, SUDV remains less well characterized at the molecular level despite having caused nine documented outbreaks, with 970 confirmed cases and 507 deaths reported [2, 3].

Viral Protein 30 (VP30) functions as a transcriptional activator, facilitating the initiation of viral mRNA synthesis. It interacts directly with the nucleoprotein (NP) and the large polymerase protein (L), playing a key role in coordinating transcription initiation and potentially influencing the switch between transcription and replication phases [4, 5]. In Orthoebolaviruses, binding of VP30 is required at the gene start site of the NP gene, where it helps to overcome inhibitory RNA structure, specifically stem-loop formations that act as transcriptional roadblocks [6].

A major barrier to filovirus research is the requirement for biosafety level 4 (BSL-4) infrastructure, which restricts access to these pathogens [7]. While several BSL-2 compatible alternatives exist, including minigenomes, virus-like particles, and pseudotyped systems, these approaches have significant limitations, such as restricted replication cycles or incomplete viral life cycle representation [8, 9]. Biologically contained systems, where the essential VP30 gene is deleted from the viral genome and complemented in trans by stably transfected cells, offer a powerful solution enabling multicycle replication in lower biosafety conditions [10–12].

Previous studies have established functional cross-complementation between filoviral proteins, particularly VP30, across related species [13, 14]. However, these investigations have not included SUDV, leaving a critical gap in our understanding of genus-specific replication requirements and evolutionary conservation within the genus Orthoebolavirus.

Here, we present the first biologically contained SUDV system, completing the toolkit for major pathogenic filoviruses. We demonstrate robust viral rescue, strict containment, and replication kinetics comparable to wild-type SUDV. Using heterologous VP30 proteins, we reveal strong cross-functionality among orthoebolavirids while confirming genus-specific limitations with MARV VP30. Finally, we validate the system for high-throughput antiviral screening and resistance profiling, identifying novel mutations in RNA polymerase gene under remdesivir pressure.

Materials and methods

Cell lines

African Green monkey kidney cells (Vero E6; Vero C1008, ATCC) and Huh-7 cells (kindly provided by Ralf Bartenschlager, University of Heidelberg) were used [15]. Vero E6 cells were cultured in DMEM (Thermo Fisher Scientific, TFS) supplemented with 10% FBS (Biowest), 200 mM L-Glutamine, 1% sodium bicarbonate, 1% Penicillin–streptomycin, 0.2% Fungizone, and 0.02% Gentamicin (all TFS). Huh-7 medium included additional 1% Non-Essential Amino Acids (TFS).

Plasmids

A SUDVΔVP30 plasmid with eGFP reporter was assembled from three gene-containing fragments and T7 backbone using NEBuilder Hifi DNA assembly kit (NEB). Fragments were synthesized by GenScript based on SUDV strain Gulu (NCBI GenBank accession number NC_006432): Fragment 1 (NP, VP35, VP40), Fragment 2 (GP, eGFP, V24), and Fragment 3 (L gene). Fragments 1–2 were digested with NotI-HF and KpnI (NEB), Fragment 3 with SmaI (NEB). T7 backbone containing HdVRz was amplified using Q5 HotStart-Fidelity 2X master mix (NEB).

For remdesivir-resistant SUDVΔVP30, the region from L gene 3′ end through T7 promoter to GP gene upstream portion was excised using NdeI (NEB). The remaining antigenome was amplified in two overlapping fragments using site-directed mutagenesis to introduce L gene substitutions associated with remdesivir resistance, then reassembled using NEBuilder HiFi DNA assembly.

pCAGGs plasmids containing SUDV NP, L, VP35, and VP30 genes were created by amplifying genes from synthesized fragments using Q5 HotStart-Fidelity 2X master mix. SUDV and RESTV VP30 were derived from separately synthesized fragments in pUC57 vectors (Genscript). The pCAGGs backbone was obtained by digesting pCAGGS-L plasmid. Sequences were validated by Sanger sequencing (Macrogen Europe) and the SUDVΔVP30 antigenome by nanopore sequencing (Oxford Nanopore Technologies (ONT), Oxford, UK).

Lentiviral constructs

VP30-expressing cell lines were generated by lentiviral transduction. SUDV VP30 was inserted into pLenti6.3 vector (TFS) using NEBuilder HiFi assembly, with the Cytomegalovirus (CMV) promoter replaced by a Spleen Focus-Forming Virus (SFFV) promoter. An IRES cassette was placed between VP30 and blasticidin resistance gene by digesting pLenti6.3 with SpelI-HF and SalI-HF and pEF1a-IRES vector with NheI-HF and SalI-HF, then ligated using Quick Ligation kit (NEB). RESTV VP30 construct was generated using the same protocol.

For mCherry constructs, pLenti6.3 vector under SFFV promoter was used. NeoR from pcDNA3.1 backbone (TFS) conferred geneticin (G418) resistance. mCherry and P2A motif were derived from pEGFP-N1.eGFP-P2A-mCherry vector (kindly provided by Ramanujan Hegde, MRC, Cambridge), with c-Myc NLS added for nuclear localization. Fragment order: SFFV, NeoR, P2A, mCherry, c-Myc NLS.

Lentiviral production and transduction

HEK293FT cells were seeded in T25-flasks to reach 50% confluency and transfected with Lipofectamine LTX/PLUS Reagent (TFS). The transfection mixture contained 3 μg VP30-lentiviral construct, 5.83 μg psPAX2 vector, 12 μL PLUS reagent, 6 μL LTX, and 3.17 μg pMD2.G vector in serum-free Opti-MEM (TFS). After 5 min of incubation at room temperature, the mixture was added to the cells and incubated for 21 h at 37°C. Sodium butyrate (10 mM) was added for 3 h, then medium replaced. Supernatant was collected 21 h later and centrifuged at 2000g, 4°C for 15 min.

Vero E6 transduction used the Virapower hiPerform T-rex Gateway Expression System (TFS) with 6 μg/mL Polybrene (Sigma-Aldrich) according to the manufacturer's protocol. Following transduction of VP30 and mCherry-NLS, cells were maintained under continuous antibiotic selection with blasticidin (10 μg/mL) and geneticin (750 μg/mL) as appropriate. For all viral infection and replication experiments, antibiotic selection was omitted to prevent confounding effects on virus replication kinetics.

SUDVΔVP30 rescue

VeroE6-SUDV-VP30 cells (200,000 cells/well) were seeded in 12-well plates and transfected with 500 ng SUDVΔVP30 antigenome, 500 ng T7 polymerase, 500 ng NP, 250 ng VP35, and 1000 ng L using 6:1 TransIT-LT1 Transfection Reagent (Mirus Bio). Medium was replaced the next day. At 9 days post-transfection, the medium was passaged onto fresh VeroE6-SUDV-VP30 cells (90-100% confluency) in T75-flasks. A second passage was performed 10 days later, using the same conditions. After the final passage (10 days), the medium was centrifuged at 4500×g for 15 min, the supernatant was aliquoted, and stored at −80°C. New virus stocks were prepared by passaging the previous stock on fresh VeroE6-SUDV-VP30 cells and harvesting at 6 days post-infection.

RNA extraction

RNA was extracted from 100 μL virus stock using QIAamp 96 Viral RNA Kit (Qiagen) according to the manufacturer's protocol.

Nanopore sequencing and analysis

RNA was converted to cDNA and amplified using Sequence-Independent Single Primer Amplification (SISPA) method. To sequence the independent four isolates (A-D) directly after rescue, library was prepared using the ligation sequencing kit (SQK-LSK109 with EXP-NBD196 barcoding expansion, ONT, UK). The sequencing library was loaded on an R9.4.1 flow cell and run for 48 h on GridION. Raw sequencing data was basecalled with the super accuracy model, and subsequently demultiplexed and adapter trimmed with the GridION built-in software v4.2.3. Read mapping against the SUDV reference genome (NCBI NCBI GenBank accession number NC_006432) was performed with CLC workbench v22, with consensus polishing and variant calling by Medaka v1.0.1.

In addition, higher-passage isolates were sequenced on a PromethION device (ONT) using the ligation sequencing kit (SQK-LSK114 with EXP-NBD114 barcoding expansion). Sequencing library was loaded onto the R10.4.1 (FLO-PRO114) flow cell for deep sequencing. Basecalling and demultiplexing was done using Guppy v6.5.7 in super-accuracy mode. Reads were further trimmed with porechop v0.3.2 and mapped with minimap2 v2.29 to either the wildtype or SUDVΔVP30 genome, or against a SUDV-VP30 cassette only, depending on the experiment. The resulting BAM files were processed with Samtools v1.21 for sorting, indexing and enumerating the number of aligned reads. For resistance-selection experiments, variants were called using Bcftools v1.18. Variant call format (VCF) files were filtered to retain only high-confidence variants (QUAL ≥ 30), and per-site mutation frequencies were derived from the filtered VCFs.

Plaque forming unit assay

VeroE6-SUDV-VP30 cells (700,000 cells/well) were seeded in 6-well plates. The next day, 200 μL virus dilutions (1 × 10⁻¹ to 1 × 10⁻⁵, duplicate) were added and incubated for 1 h at 37°C, 5% CO₂ with gentle agitation every 10-15 min. Then 3 mL agarose-medium (17.6 mg/mL SeaKem ME agarose [Lonza] in 1:1 ratio with DMEM supplemented with 20% FBS, 200 mM L-glutamine, 1% Penicillin–Streptomycin, 2% NEAA, 1% Gentamicin, 0.2% Fungizone [all TFS]) was added at ∼40°C. After solidification, plates were incubated for 3 days and eGFP⁺ clusters were counted.

Monoclonal cell line selection

Polyclonal Vero E6 cells containing mCherry and SUDVVP30 were created by sequential lentiviral transductions. Vero E6 cells were transduced with mCherry lentivirus and selected with G418. Sufficient mCherry signal was confirmed before second transduction with SUDV-VP30 lentivirus.

For monoclonal selection, confluent polyclonal cells were trypsinized, counted (Countess 2, TFS), centrifuged at 820 rpm for 10 min, and diluted to 10 cells/mL in original medium. This was transferred to 96-well plates (100 μL/well, ∼1 cell/well). Wells were monitored daily by microscopy, and growing monoclonal lines were expanded and subjected to antibiotic pressure.

Antiviral compound screening assay

Monoclonal VeroE6-mCherry-SUDV-VP30 cells (30,000/well) were seeded in 96-well plates in DMEM + 2% FBS with 2 μM CP-100365 (P-gp efflux pump inhibitor) and incubated overnight at 37°C. Remdesivir (GS-5734) was added in 1:2 dilution series (10-0.4 μM) and incubated 1 h, including 6 negative (medium only) and positive (virus only) controls per plate. Then 600 PFU SUDVΔVP30 was added and incubated 8 days. High-content images were acquired with Arrayscan XTI (TFS) using mCherry nuclear signal for autofocusing (587 nm excitation) and cell counting. eGFP-positive cell percentages were calculated with HCS Studio (TFS). We used GraphPad Prism v9.5.1. for data analysis and Z-factor was calculated as robustness control. Values close to 1 were considered robust whereas values shifting close to 0 indicate very low robustness [16].

Heterologous functionality of VP30

Vero E6 cell lines stably expressing VP30 from EBOV, SUDV, MARV, or RESTV (30,000 cells/well in 96-well plates) were incubated overnight at 37°C in DMEM + 2% FBS, then infected in triplicate with 1000 PFU of EBOVΔVP30, SUDVΔVP30, or MARVΔVP30. High-content fluorescent imaging was performed after 6 days.

Experimental selection of de novo remdesivir-resistant variants

A four-step enrichment protocol adapted from Leyssen et al., generated remdesivir-resistant SUDV isolates [17]. After identifying the minimum inhibitory concentration of compounds and maximum viral dose, 216 wells were infected with 3.1 × 10² TCID₅₀/mL SUDV in presence of 6.25 or 12.5 μM GS-441524 (remdesivir parent nucleoside). After 7–10 days, supernatants from wells showing eGFP-plaque formation were collected and titrated to identify lowest infectious dose yielding detectable replication (>10% eGFP-positive cells). Selected isolates were expanded in T25 flasks under 12.5 μM GS-441524. Resistant isolates were harvested at 10 days post-infection (dpi) and controls at 6 dpi. Titrations compared all isolates without GS-441524. Dose–response assays used GS-441524 (50–0.39 μM) with infectious doses calibrated to yield 80–100% eGFP positivity in virus controls.

Results

Rescue of biologically contained SUDVΔVP30

To enable virus rescue and propagation, we first established a stable Vero E6 cell line expressing SUDV VP30 (VeroE6-SUDV-VP30). Cells were transduced with a lentiviral vector encoding SUDV VP30 under a constitutive promoter, yielding a polyclonal cell line. Using this custom cell line, we initiated the rescue of biologically contained SUDV. VeroE6-SUDV-VP30 cells were transfected with SUDV antigenome (containing the gene for eGFP that substitutes VP30) and support plasmids as described in methods. Cells were monitored daily for eGFP expression. Between days 4 and 5 post-transfection, eGFP-positive foci were observed in all four wells, indicating successful virus rescue. By day 8, widespread fluorescence confirmed robust viral replication, prompting passage of the supernatant onto freshly seeded cells. A second passage was performed ten days later to expand viral stocks. Supernatants from the original four wells (P0) were retained and propagated separately, yielding four distinct SUDVΔVP30 isolates (designated as isolates A, B, C, and D; Table S1).

To verify genetic stability and exclude isolates with coding-region mutations, all four isolates were sequenced. Each isolate displayed a single guanosine insertion at the 5′ genome terminus (position 0), a known and naturally occurring feature in filovirus replication [18]. Since this insertion does not affect coding regions or viral function, it was considered negligible for downstream applications. Three of the four isolates contained only non-coding insertions, whereas isolate D harboured a mutation in the GP2 coding region (full details are provided in Supplementary Table 1). Based on minimal genomic alterations and the absence of coding-region mutations, isolate A was selected for subsequent experiments. Its viral titre was then determined by plaque assay. For later experiments, new virus stocks were propagated, isolated, and sequence-verified using the same methods.

SUDVΔVP30 shows strict VP30-dependent functional containment and physiologically relevant growth kinetics

To demonstrate both functional containment and biological relevance of our SUDVΔVP30 system, we performed dual validation approaches: qualitative imaging to visualize infection specificity and quantitative longitudinal analysis comparing viral growth dynamics in VP30-positive and VP30-negative cell lines. Both cell lines were transduced with nuclear-localized mCherry (NLS-mCherry) reporters to enable concurrent monitoring of total cell numbers and normalized infection levels.

For both approaches, VP30-positieve and VP30-negative cell lines were infected with SUDVΔVP30 using a serial dilution in 96-well plates, beginning at 1:12 and followed by eight 4-fold dilution steps (20,000 cells/well). For the qualitative experiment, images were acquired 3 days post-infection from representative wells using widefield and confocal microscopy. The right images in Figure 1 showed ∼32% eGFP-positive cells, corresponding to an estimated MOI of 0.51 based on a standard endpoint titre of 2.4 × 10⁶ TCID₅₀/mL. In contrast, the left images, despite receiving the same infectious dose, showed no eGFP signal (Figure 1).

Figure 1.

Functional containment of SUDVΔVP30 demonstrated by imaging and longitudinal replication analysis. Top row: Widefield fluorescence microscopy of VeroE6-SUDV-VP30 (+VP30; right) and VeroE6-mCherry (−VP30; left) cells at 3 days post-infection with biologically contained SUDVΔVP30 at an MOI of 0.385. Bottom row: Confocal microscopy images of the same conditions. Orange signal corresponds to nuclear mCherry expression in host cells.

To quantify these observations, we performed a comprehensive 11-day time course analysis using our dual-reporter system (Figures 2(A and B)). The same plates were imaged every 24 h, and for each time point, we estimated an apparent infectious titre relative to the dilution series to track how the normalized eGFP signal increased over time. Time-dependent TCID₅₀ values were extrapolated from these data using two established methods: nonlinear regression analysis (nLR; 4-parameter logistic model) and classical Reed-Muench (RM) analysis [19]. The nonlinear regression approach, based on ≥50% eGFP positivity thresholds, showed continuous titre increases until day 10, followed by plateau formation. Reed-Muench analysis, which classifies any detectable eGFP signal as positive, yielded consistently higher apparent titres (>2 logs above regression estimates at early time points) but plateaued earlier at ∼10⁶−10⁷ TCID₅₀/mL from day 7 onward, highlighting this method's lower stringency.

Figure 2.

Functional containment of SUDVΔVP30 demonstrated by imaging and longitudinal replication analysis. (A) Viral titres over 11 days post-infection in VP30-expressing and VP30-absent cells calculated by extrapolating TCID₅₀/mL values via non-linear regression from daily eGFP-quantification using our dual-reporter system. (B) Statistical analysis of background eGFP signal in −VP30 cells. Percentage of eGFP-positive cells at the highest viral input was compared to cell control across 11 days using Mann–Whitney U test with Benjamini, Krieger and Yekutieli correction for multiple comparisons. No significant difference was detected on any day (q ≥ 0.74), confirming absence of replication in VP30-absent cells. All experiments were performed in three biological replicates.

Rigorous statistical analysis confirmed the complete absence of viral replication in VP30-deficient cells (Figure 2(B)). Mann–Whitney U tests comparing eGFP signals from the highest viral doses in VP30- cells versus uninfected controls revealed no significant differences on any day (q-values 0.75–0.99 after multiple comparison correction using the Benjamini-Krieger-Yekutieli FDR procedure). This demonstrates that background fluorescence was never exceeded, confirming absolute functional containment even under high viral input conditions over extended time periods.

These comprehensive analyses provide definitive evidence for both functional safety and biological relevance of the SUDVΔVP30 system. The complete absence of replication in VP30-deficient cells demonstrates stringent containment, while robust growth kinetics in permissive cells confirm physiologically relevant viral behaviour suitable for downstream applications.

Long-term genetic containment of the SUDVΔVP30 system

Functional containment analysis demonstrates single-passage safety but does not address long-term genetic stability. To evaluate the risk of VP30 re-acquisition through recombination, we performed extended passaging analysis with deep sequencing validation. The antigenome and VP30 expression plasmids were deliberately designed to exclude homologous overhangs, making recombination highly unlikely. To experimentally confirm genetic containment, SUDVΔVP30 was serially passaged ten times on VeroE6-SUDV-VP30 cells in four independent replicates, followed by comprehensive genomic analysis using SISPA and PromethION deep sequencing.

Genome-wide coverage analysis revealed consistent read depth across all viral genes in each replicate, with expected decreases in non-coding regions when mapped against the wild-type sequence (Figure 3). The VP30 locus showed the anticipated dramatic coverage reduction corresponding to the engineered deletion. To robustly quantify the actual presence of VP30-derived reads independent of genomic context, the datasets were additionally mapped against the VP30 cassette sequence alone. Across all isolates, comprising 222,850; 244,570; 195,073; and 310,316 total reads, respectively, no single read aligned to the VP30 cassette reference, even without applying quality filters. These results indicate a complete absence of detectable VP30-related sequences, confirming that no recombination events occurred.

Figure 3.

Whole genome sequencing of SUDVΔVP30 after 10 serial passages on VeroE6-SUDV-VP30 cells. Mapping coverage across the viral genome of 18891 bp for four independent replicates. Coverage was uniform across all coding regions, with a distinct drop at the VP30 locus, consistent with the engineered deletion, and indicating complete absence of recombination events.

Heterologous functionality of SUDV VP30

VP30 cross-functionality analysis provides insights into filovirus replication strategies and evolutionary relationships between genera. Previous studies demonstrated VP30 interchangeability in minigenome systems, particularly between EBOV and Lloviu virus, suggesting functional conservation across filovirus transcription machinery; however, SUDV has never been included in such studies [14]. To systematically evaluate VP30 cross-functionality, we established Vero E6 cell lines stably expressing VP30 from EBOV, SUDV, MARV, or RESTV, enabling full pairwise comparison across all available VP30 variants. The inclusion of RESTV VP30 was motivated by our aim to extend the intra-genus analysis of orthoebolaviruses, as we had already established such a cell line in parallel efforts aimed at developing a biologically contained RESTV system, though these rescue attempts were unsuccessful at the time of writing.

Each VP30-expressing cell line was infected with 1000 PFU of EBOVΔVP30, SUDVΔVP30, or MARVΔVP30 in 96-well format. At six days post-infection, high-content imaging quantified eGFP-positive cells as markers of productive viral replication (Figure 4(B)). Robust cross-functionality was observed among all orthoebolaviruses. EBOVΔVP30 and SUDVΔVP30 replicated efficiently with all orthoebolavirus VP30 proteins, achieving 85-99% eGFP-positive cells across EBOV, RESTV, and SUDV VP30-expressing lines (Figure 4(A and B)). Notably, SUDVΔVP30 showed superior replication in RESTV VP30 cells (99%) compared to EBOVΔVP30 (85%), despite RESTV VP30 sharing only 65-69% sequence identity, suggesting that structural compatibility extends beyond primary sequence similarity.

Figure 4.

Orthoebolavirids have a very strong overlap in the functionality of their VP30 protein within the genus but only show minor to no cross-reaction with Marburg virus. VeroE6 cells expressing either EBOV, RESTV, SUDV or MARV VP30 were infected with 1000PFU of either biologically contained EBOV, MARV or SUDV (A) Bar plots showing the average percentage of eGFP-positive cells for each condition. An ordinary one-way ANOVA was performed with Dunnett’s multiple comparison test between each condition and its respective negative control. (****: p <   0.0001). (B) Representative high-content fluorescence microscopy images (eGFP channel) at 6 dpi illustrating replication patterns across the matrix of VP30-expressing cell lines and corresponding viruses. Each column represents the virus; each row represents the VP30 source. NC = uninfected control. (C) Correlation between VP30 amino acid identity (relative to the native VP30 of each virus) and observed eGFP signal (% positive cells). A Spearman correlation coefficient of ρ =   0.72 (p =   0.0098) indicates a moderate correlation between VP30 sequence identity and its ability to support replication.

Cross-genus compatibility was minimal but consistently detectable. EBOVΔVP30 and SUDVΔVP30 showed negligible replication in MARV VP30 cells (0.69% and similar levels), while MARVΔVP30 achieved only 0.42% eGFP-positivity in EBOV VP30 cells. Interestingly, MARVΔVP30 demonstrated modestly higher replication in RESTV VP30 cells (2.02%), potentially reflecting partial structural compatibility or expression level differences across polyclonal cell lines. Statistical analysis using one-way ANOVA with Dunnett's multiple comparison tests validated these observations (Figure 4(A)). All orthoebolavirus-VP30 combinations yielded statistically significant eGFP signals compared to uninfected controls, confirming robust intra-genus compatibility.

Spearman correlation analysis between VP30 amino acid identity and eGFP-positivity across all virus-VP30 pairings revealed a correlation coefficient of 0.72 (p = 0.0098), indicating that sequence similarity generally aligns with functional cross-complementation (Figure 4(C)). However, outliers (e.g. moderate replication of MARV in RESTV VP30 cells despite lower sequence identity) demonstrate that additional molecular determinants beyond linear sequence similarity influence cross-functionality. These results confirm that VP30 functional compatibility is largely genus-restricted while revealing exceptions when sufficient structural conservation exists. The strong intra-genus compatibility supports the use of heterologous VP30 proteins for orthoebolavirus research, while limited cross-genus functionality suggests distinct evolutionary constraints on transcriptional machinery.

SUDVΔVP30 enables robust high-throughput antiviral screening

Beyond fundamental virology applications, our SUDVΔVP30 system provides an ideal platform for antiviral discovery and validation. The integrated eGFP reporter enables real-time visualization of viral replication dynamics, while stably expressed nuclear-localized mCherry (NLS-mCherry) facilitates accurate cell proliferation monitoring. Monoclonal cell lines co-expressing VP30 and mCherry were selected to ensure experimental reproducibility. To validate system utility for compound screening, we tested remdesivir (GS-5734), a broad-spectrum RNA polymerase inhibitor with established activity against SUDV and related viruses [20, 21]. VeroE6-SUDV-VP30 cells were treated with compound for one hour prior to SUDVΔVP30 infection, with antiviral efficacy assessed by eGFP quantification. Remdesivir exhibited robust dose-dependent antiviral activity against SUDVΔVP30, achieving an IC₅₀ of 2.33 μM with a selectivity index (SI) of 21.47 under standard conditions (Figure 5). The assay maintained excellent performance characteristics with a Z'-factor >0.9, confirming its suitability for high-throughput screening applications.

Figure 5.

Remdesivir (GS-5734) inhibits the replication of biologically contained SUDV in VeroE6-SUDV-VP30 cells in a dose-dependent manner, with and without Pgp inhibition. Cells were pre-treated with or without the P-glycoprotein (Pgp) inhibitor CP-100356 and incubated overnight. The next day, cells were exposed to a dilution series of remdesivir ranging from 50 μM to 0.39 μM and subsequently infected with 600 PFU of SUDVΔVP30. Each condition was tested in six replicates. eGFP expression was used as a proxy for viral replication (panel A), and mCherry signal was used to assess cell viability (panel B). eGFP values were background-corrected by subtracting the mean signal from negative control wells. Cell viability was normalized to uninfected, untreated control wells. Inclusion of CP-100356 resulted in increased antiviral potency but also increased cytotoxicity at higher remdesivir concentrations.

Given the high P-glycoprotein (Pgp) expression in VeroE6 cells, which can reduce intracellular drug concentrations [22], we evaluated whether Pgp inhibition would enhance remdesivir potency. Co-treatment with the selective Pgp inhibitor CP-100356 improved the IC₅₀ to 0.55 μM and increased the SI to 42.16, though with elevated cytotoxicity (CC₅₀ reduced from >50 μM to 23.02 μM). These results demonstrate that robust antiviral activity detection is achievable without efflux inhibitors, while co-treatment with CP-100356 can maximize assay sensitivity when desired. Both configurations maintain a Z'-factor >0.9, confirming the system's versatility and robustness for antiviral screening under varying sensitivity requirements.

SUDVΔVP30 enables comprehensive resistance profiling under lower biosafety conditions

RNA viruses, including filoviruses, mutate rapidly and develop resistance to direct-acting antivirals like remdesivir, as observed with SARS-CoV-2 [23, 24]. We investigated whether our SUDVΔVP30 system could serve as an authentic and safe platform for resistance profiling using GS-441524, the nucleoside precursor of remdesivir (wild-type IC₅₀ = 1.90 μM). This calculated IC₅₀ was used as the baseline reference prior to continuing with the study and is indicated as WT. We employed two complementary approaches: targeted mutagenesis introducing known resistance-associated mutations, and experimental selection using high-replicate screening under complete antiviral pressure to enrich naturally occurring resistant variants [17, 23].

To validate system capability for resistance analysis, we introduced polymerase mutations F549S and T563A into the SUDV genome, corresponding to established remdesivir resistance mutations F548S and T562A in EBOV [23]. The engineered SUDV/mut virus exhibited marked resistance with an IC₅₀ of 24.88 μM versus 1.90 μM for wild-type (∼13-fold increase), confirming that remdesivir engages SUDV polymerase similarly to EBOV and validating our system for functional resistance analysis (Figure 6 and Table 1).

Figure 6.

Resistance profiling of SUDV isolates exposed to GS-441524. Dose-response curves for virus controls (VC1 and VC2), engineered mutant (SUDV/mut), and twelve resistant isolates (R1–R12) selected under compound pressure based on read-out at 6 dpi. SUDV/mut, carrying two resistance-associated polymerase mutations (F549S and T563A), displayed a ∼13-fold IC₅₀ shift. Selected isolates showed IC₅₀ increases ranging from 1.5- to 7.2-fold.

Table 1.

Summary table of IC₅₀ values and fold-increases relative to the average IC₅₀ of virus controls. All data represent the mean ± SD of three replicates. Assays were performed without Pgp inhibition, and Z-factors exceeded 0.8 in all cases, confirming robustness.

Strain	IC50	Increase
WT	1.90	–
VC1	1.98	–
VC2	1.01	–
R1	10.79	7.2
R2	5.46	3.7
R3	4.66	3.1
R4	6.99	4.7
R5	5.06	3.4
R6	2.21	1.5
R7	5.22	3.5
R8	6.44	4.3
R9	4.88	3.3
R10	3.99	2.7
R11	7.46	5.0
R12	6.06	4.1
SUDV/mut	24.88	13.1

To determine whether our system could also model the natural emergence of resistant isolates, we adapted a high-replicate selection protocol. We infected 216 replicate wells with SUDVΔVP30 under fully inhibitory GS-441524 concentrations (6.25–12.5 μM). Approximately 6% of wells showed viral breakthrough after seven days, indicating resistant variant emergence. Supernatants from these wells were further enriched and expanded in the presence of 12.5 μM GS-441524. To ensure standardized conditions for subsequent dose–response analysis, each resistant isolate was titrated to determine an infectious dose that consistently yielded robust replication, defined as 80–100% eGFP-positive cells. Dose–response analysis of resistant isolates showed IC₅₀ shifts ranging from 1.5- to 7-fold compared to controls when readout was carried out at 6 dpi, with all resistant isolates maintaining excellent assay performance (Z'-factors >0.8) (Table 1). These controls were isolates passaged in parallel to the resistant isolates, though in the absence of GS-441524, to rule out any effect of extended passaging on resistance development.

Deep sequencing using SISPA and PromethION platforms analysed viral isolates at three time points: P0 (pre-selection, parental virus used for initial drug exposure), P1 (isolates recovered after first selection step), and P3 (isolates obtained after expansion in T25 flasks for 10 days following enrichment) (Figure 7(C)). Despite even mutation distribution genome-wide, a single dominant pattern emerged: F549 substitutions in the L protein occurred in 11 of 12 resistant isolates, confirming strong evolutionary pressure on this conserved site across ebolaviruses (Figure 7(A)). While F549S was previously reported in EBOV remdesivir resistance [23], we identified F549L as an equally prevalent novel variant. Among F549-substituted isolates, 5 carried F549S while 6 acquired F549L (7B). Statistical comparison revealed no significant difference in resistance levels between F549S and F549L groups (p = 0.57, unpaired t-test), indicating functional equivalence (Figure 7(D)).

Figure 7.

Genomic profiling of SUDV isolates reveals conserved and emerging mutations under GS-441524 pressure. (A) Genome-wide distribution of all synonymous or non-coding (black) and non-synonymous (red) mutations identified across resistant SUDVΔVP30 isolates. Each row represents a different viral isolate (R1–R12 and two virus controls, VC1 and VC2). (B) Isolate-level prevalence of the most frequently detected substitutions. (C) Temporal enrichment of specific substitutions across three timepoints (passage 0 (parenteral pre-selection isolate, light grey), 1 (breakthrough isolates from initial selection, dark grey), and 3 (clonal isolates from T25 flask expansion, black)), showing how mutations such as F549S/L, M675 V and P1880S were enriched during resistance selection. Values represent relative substitution prevalence based on sequencing read frequency. (D) IC₅₀ values from isolates carrying either F549S (n = 5) or F549L (n = 6) were compared to assess potential functional differences between the two substitutions. Prior to analysis, data were assessed for normality using the Shapiro–Wilk test, which confirmed a normal distribution in both groups (p = 0.33 and p = 0.78, respectively). An unpaired Welch’s t-test was subsequently performed, yielding p = 0.57, indicating no statistically significant difference between the two groups. Error bars represent standard deviation.

Analysis of the broader mutational landscape identified two additional L protein substitutions of potential interest: P1880S (detected in 3/12 isolates) and M675 V (unique to isolate R6). P1880S showed high prevalence at passage 0 (∼50% reads) and consistently co-occurred with F549 substitutions, suggesting passive enrichment rather than independent selection. In contrast, M675 V appeared in the sole resistant isolate lacking F549 substitutions and was enriched early, indicating a possible alternative resistance pathway. Although the IC₅₀ shift for this M675V-carrying isolate was only very limited compared with the more pronounced shifts observed in the other isolates, we considered it biologically relevant because this lineage consistently showed breakthrough replication during the enrichment steps under GS-441524 pressure, suggesting a weaker but reproducible phenotypic effect. Examination of the available EBOV L protein cryo-EM structure revealed that EBOV M674 (homologous to SUDV M675) localizes near F548 within the polymerase central channel at the RNA template entrance (Figure 8). This spatial proximity suggests M675 V may influence RNA positioning or binding pocket stabilization, supporting its potential role as an alternative resistance determinant.

Figure 8.

Structural proximity of F548 and M674 in the EBOV L polymerase. Cryo-EM structure of the EBOV L polymerase highlighting the side chains of Phe548 (red) and Met674 (blue), homologous to SUDV residues F549 and M675, respectively. The close spatial positioning of these residues near the RNA-binding groove supports the hypothesis that M675 may contribute to resistance. While F549S has a dominant effect on remdesivir resistance, the presence of M675 V in the absence of F549S in one slightly resistant SUDV isolate suggests a possible secondary role for this site.

These comprehensive analyses demonstrate that the SUDVΔVP30 system successfully recapitulates known resistance pathways (F549S), identifies novel variants (F549L, M675 V), and enables mechanistic insights with minimal biosafety requirements. The platform's ability to model natural resistance emergence under controlled conditions establishes its value for antiviral development and resistance monitoring.

Discussion

We successfully established the first biologically contained SUDV system, completing the lower biosafety toolkit for major pathogenic filoviruses and addressing a critical gap in accessible filovirus research. Our system demonstrates robust containment, authentic replication kinetics, and broad utility for both fundamental research in virology and antiviral development. The SUDVΔVP30 system exhibits strict functional and genetic containment, with no detectable replication in non-permissive cells and no recombination during extended passage. Notably, replication kinetics in permissive cells closely mirror wild-type SUDV, validating the biological relevance of this approach.

Crucially, viral titres were calculated using both classical Reed–Muench and nonlinear regression approaches, which differ in their underlying positivity criteria. The former method treats even minimally infected wells as positive, providing a conservative upper-bound estimate well suited for biosafety and risk-assessment purposes. The latter approach more accurately reflects the actual quantitative infection burden and is therefore more informative for interpreting viral amplification kinetics. Despite these different criteria, both methods yielded similar temporal patterns. This indicates that the discrepancy, in our study, did not affect the interpretation of viral amplification kinetics. Transparency regarding the method used remains essential for correct interpretation of absolute titre values, risk assessment and direct comparison with other studies.

Our heterologous complementation studies provide new insights into filovirus evolutionary relationships and functional constraints. Previous studies have systematically explored the interchangeability of filoviral nucleocapsid proteins using minigenome systems and infectious clone rescue, revealing important but sometimes conflicting conclusions. A pivotal study demonstrated that EBOV could not be rescued using MARV VP30, while RESTV VP30 fully supported EBOV replication, suggesting genus-specific limitations [13]. More recent findings extended this work to Lloviu virus (LLOV), where minigenome assays revealed that LLOV VP30 could substitute for EBOV VP30, and vice versa, although with reduced efficiency [14]. These findings challenged the idea of strictly genus-limited compatibility and suggested that sufficient structural conservation can permit functional interchangeability, even across genera.

Building on this foundation, we incorporated the highly pathogenic SUDV – previously excluded from such analyses – and used our biologically contained SUDVΔVP30 system to expand understanding of VP30 functionality. It should be noted that a very recent study reported rescue of SUDV using EBOV support plasmids, demonstrating that SUDV genomic RNA can be replicated and packaged by an EBOV nucleocapsid complex [25]. However, because the initial nucleocapsid proteins were isogenic, this approach does not allow conclusions about cross functionality.

In our study, we observed robust cross-functionality among orthoebolavirid VP30 proteins, confirming strong intra-genus conservation. In contrast, MARV VP30 failed to support replication of EBOVΔVP30 or SUDVΔVP30, consistent with earlier studies. Notably, we observed a modest increase in MARVΔVP30 replication when complemented with RESTV VP30. Whether this reflects subtle structural compatibility or differences in expression levels within our polyclonal lines requires further investigation.

Together, these findings and the discrepancies across systems underscore the importance of experimental context. For example, Mühlberger et al. [26] reported that MARV VP30 could modestly support EBOV-specific transcription in a minigenome system, while Theriault et al. [13] observed complete failure of EBOV rescue with MARV VP30 in an infectious clone system [13, 26]. The latter study proposed that full-length rescue systems may be more sensitive due to their reliance on the artificial encapsidation of naked RNA, multi-cycle replication, and virus spread, compared to the single-cycle readout of minigenome assays [13]. Our biologically contained system shares this latter advantage by enabling multicycle replication, while also preserving the native stoichiometry and temporal dynamics of all viral proteins, except for VP30. This hybrid design offers a biologically relevant and flexible platform for evaluating protein compatibility across filoviruses.

Our system successfully recapitulates known pharmacological profiles, with remdesivir IC₅₀ values matching those reported for wild-type SUDV in BSL-4 settings [27]. The identification of F549 as a primary resistance hotspot, emerging independently in 11/12 selected isolates, confirms evolutionary conservation of this critical site across orthoebolavirids, as it mirrors findings from prior EBOV studies [23]. Significantly, we identified F549L as a novel resistance variant alongside the previously described F549S, expanding the mutational landscape at this site. Both substitutions had comparable effects on resistance.

The emergence of M675 V in an F549-independent isolate suggests alternative resistance pathways that warrant further investigation. M675 V mapped structurally close to F549, supporting the hypothesis that it may exert a secondary or modulatory effect, influencing polymerase conformation or RNA interaction. Although the IC₅₀ shift of the M675 V containing isolate was only modest, we considered this lineage biologically meaningful because it was one of only 12 out of 216 independent lineages that consistently maintained replication under fully inhibitory GS-441524 concentrations, whereas wild type and passaged control lineages were uniformly suppressed under the same conditions.

Our pharmacological profiling in this experiment used a standardized 6-day readout to ensure the highest possible assay robustness, whereas the experimental selection allowed breakthrough over a longer window. In particular, the M675 V containing isolate displayed delayed breakthrough only beyond this timepoint under GS 441524 pressure, suggesting a resistance phenotype that manifests primarily as slower but persistent replication rather than a large early IC₅₀ shift. Extending the IC₅₀ readout beyond 6–8 days has, in our experience, a negative effect on assay robustness due to declining cell viability from that timepoint onwards, underscoring the need for complementary endpoints such as time to breakthrough as an additional parameter in future studies. Alternatively, studies incorporating targeted mutagenesis of M675 V, either alone or in defined combinations with established mutations like F549S or T563A, will be crucial to further delineate its functional role.

We acknowledge that dissemination of resistance data may raise category 1 dual-use research of concern (DURC) considerations [28]. However, the additional mutation identified here confers only a modest reduction in remdesivir susceptibility and does not result in complete drug escape. In contrast, detailed characterization of resistance pathways is critical for molecular surveillance and the rational design of next-generation antivirals, and these benefits are considered to outweigh the relatively limited incremental dual-use risk.

Several considerations should be noted when interpreting our results. First, biosafety requirements for ΔVP30 systems vary across jurisdictions. While our system is approved for lower biosafety use, regulatory frameworks may differ internationally. The genetic containment data presented here provide additional experimental support that may inform future risk assessments, while remaining subject to local regulatory frameworks. Second, the use of constitutive VP30 expression and heterologous complementation systems may not fully reflect native regulatory dynamics. In addition, differences in VP30 expression levels between cell lines were not quantitatively assessed and could contribute to variation in observed replication efficiencies. Finally, although our resistance selection approach rapidly identified key mutations, extended passaging may uncover additional adaptive pathways, as described for other filoviruses.

This work addresses the longstanding disparity in research focus between EBOV and other pathogenic filoviruses, particularly SUDV. By providing accessible research tools, our system should accelerate both fundamental understanding and therapeutic development for this neglected but significant pathogen. The demonstrated cross-species functionality studies open new avenues for investigating filovirus evolution, host adaptation, and the potential for broad-spectrum therapeutics.

Looking forward, this platform enables several critical research directions: systematic evaluation of natural SUDV variants, investigation of virus-host interactions in species-specific contexts, and development of combination therapies targeting conserved resistance hotspots. More broadly, the completion of such lower biosafety alternatives for all major filoviruses establishes a foundation for comparative studies that were previously impractical in BSL-4 settings.

Our biologically contained SUDV system represents a significant advance in filovirus research infrastructure, providing a safe, scalable platform for addressing fundamental questions while advancing therapeutic development for this important but understudied pathogen.

Funding Statement

This work was supported by KU Leuven: [Grant Number C1/3M170314].

Disclosure statement

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

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplemental Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/22221751.2026.2678654.