Antiviral Efficacy, Cytotoxicity, Transcriptomics, and Discriminatory Function of 3D8 scFv Against Dengue and Zika Viruses

TL;DR

This study shows that the antibody fragment 3D8 scFv effectively inhibits dengue and Zika viruses with minimal side effects.

Contribution

The study demonstrates 3D8 scFv's broad-spectrum antiviral activity and distinct host response signatures for dengue and Zika.

Findings

3D8 scFv reduces viral RNA, protein, and infectious particles in both dengue and Zika infections.

Machine learning classifiers accurately distinguish ZIKV and DENV2 infections based on transcriptomic signatures.

3D8 scFv induces a limited protective stress response without significant cytotoxicity.

Abstract

Flaviviruses such as dengue virus (DENV) and Zika virus (ZIKV) co-circulate widely and cause significant morbidity, yet effective broad-spectrum antivirals are limited. This study evaluated the antiviral efficacy, cytotoxicity, and host transcriptional responses to the nucleic acid–hydrolyzing antibody fragment 3D8 scFv in mono- and co-infection models. RNA sequencing of A549 cells treated with 3D8 scFv revealed a dose-dependent activation of the MAPK–HSP70 stress response, with minimal transcriptomic disruption at antiviral concentrations. Comparative transcriptomic analysis identified distinct host signatures for ZIKV and DENV2, and machine learning classifiers accurately distinguished infection states (AUC > 0.95). In Vero E6 cells, prophylactic treatment with 3D8 scFv significantly reduced viral RNA, protein expression, and infectious particle production for both viruses, including during co-infection. Optimized post-entry treatment also demonstrated antiviral activity. Cytotoxicity assays confirmed good tolerability at effective concentrations. These findings indicate that 3D8 scFv inhibits viral replication through early cleavage of viral nucleic acids while inducing a limited protective stress response, supporting its development as a broad-spectrum antiviral candidate.

Figures

Full text

1. Introduction

Flaviviruses, including dengue (DENV) and Zika (ZIKV), are expanding globally due to climate-driven shifts [1] in Aedes mosquito distribution, urbanization, and human mobility sustain transmission [2]. DENV causes ~390 million infections annually [3], while ZIKV is associated with neurological complications such as microcephaly and Guillain–Barré syndrome [4]. Despite antigenic similarities, their clinical outcomes differ [5], reflecting distinct host–virus interactions [6]. Current antivirals targeting lipid metabolism and replication pathways—such as triclosan, lapatinib, metformin, and JNJ-A07—show activity against both DENV and ZIKV [7]. However, viral manipulation of host signaling and immune evasion underscores the need for broad-spectrum agents effective regardless of viral sequence variation [8].

The 3D8 single-chain variable fragment (scFv) is a 27 kDa, antibody-derived catalytic fragment with non-sequence-specific nuclease activity, offering broad-spectrum antiviral potential [9]. It enters cells via caveolae-mediated endocytosis, accumulates in the cytoplasm, and degrades viral nucleic acids without entering endolysosomal or nuclear compartments [10]. This mechanism underlies efficacy against diverse DNA and RNA viruses, including classical swine fever virus [11], pseudorabies virus [12], herpes simplex virus [12], influenza A [13], severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) [14], and other coronaviruses [15]. It may also support oral delivery via engineered probiotics due to its stability [16]. Activity is effective at ≤10 µM, while higher concentrations (>20 µM) induce dose-dependent cytotoxicity, defining a narrow therapeutic window [12,17]. These features collectively position 3D8 scFv as a promising broad-spectrum antiviral candidate.

Despite the broad antiviral activity reported for catalytic antibody fragments, the cellular pathways engaged by 3D8 scFv remain unclear. Because 3D8 scFv hydrolyzes intracellular RNA in a sequence-independent manner, it may trigger RNA damage response pathways and alter transcriptional programs, thereby influencing both antiviral efficacy and cytotoxicity. Transcriptomics enables the unbiased detection of these gene expression changes, and recent work shows that flaviviruses co-opt MAPK—the integrated stress response—and TNF signaling to support replication and immune evasion [18]. Given the central role of MAPK pathways in coordinating cellular responses to viral infection and nucleic acid stress, they represent a plausible interface for 3D8 scFv-induced signaling. Integrating machine learning with transcriptomics can further resolve dose-dependent and mechanistically informative gene signatures, an approach increasingly applied in antiviral research [19] but not yet used to characterize catalytic antibody fragments.

We analyzed host responses and potential off-target effects of 3D8 scFv in A549 cells using transcriptomics and machine learning, revealing virus-specific signatures for ZIKV and DENV2 infection. At antiviral-effective concentrations (5–10 µM), 3D8 scFv, administered prophylactically or therapeutically, potently suppressed mono- and co-infections without cytotoxicity, supporting its potential as a safe, broad-spectrum antiviral.

2. Results

2.1. Dose-Dependent Transcriptomic Responses and Pathway Remodeling Induced by 3D8 scFv in A549 Cells

RNA-seq profiling of A549 cells treated with 3D8 scFv (0, 5, 10, and 40 μM) produced high-quality datasets, with >98% retained reads and >90% alignment (Table 1). Each condition included two independent biological replicates. At lower doses, expression of housekeeping genes remained strongly correlated with control (R^2^ = 0.9685 at 5 μM). Still, this linearity progressively declined at 10 μM (R^2^ = 0.9509) and 40 μM (R^2^ = 0.9311), reflecting gradual loss of basal transcriptome integrity (Figure 1a–c).

Differential expression analysis revealed the dose-dependent amplification of transcriptional reprogramming, with 68 DEGs (Differentially Expressed Genes) at 5 μM, 868 at 10 μM, and 2904 at 40 μM (|log_2_FC| ≥ 1, FDR < 0.05). At 5 µM, the transcriptional response was dominated by immediate-early/stress genes, with strong induction of EGR1 (log2FC = 6.33, FDR = 6.4 × 10^−8^), HSPA6 (5.74, 4.1 × 10^−5^), EGR2 (5.60, 1.5 × 10^−3^), FOSB (4.00, 1.5 × 10^−3^) and ARC (3.92, 2.5 × 10^−2^), whereas downregulated genes included MT-TR (−2.64, 3.3 × 10^−3^), MT-TI (−4.10, 2.5 × 10^−2^) and RYR3 (−2.90, 2.5 × 10^−2^). At 10 µM, the magnitude and breadth of induction increased, led by HSPA6 (9.86, 3.9 × 10^−65^), EGR1 (8.18, 1.2 × 10^−17^), EGR2 (8.13, 6.5 × 10^−15^), ARC (7.42, 2.1 × 10^−11^) and FOSB (7.01, 8.9 × 10^−19^), while representative suppressed genes included ASCL1 (−4.88, 2.8 × 10^−3^) and SCGN (−4.73, 1.3 × 10^−3^) (full ranked lists provided in Supplementary Excel File S1 and S2). At 40 µM, the strongest upregulation was observed for heat-shock/stress genes including HSPA6 (12.11, 4.2 × 10^−89^), HSPA7 (10.17, 2.7 × 10^−22^), EGR2 (9.44, 4.8 × 10^−55^), EGR1 (8.94, 1.5 × 10^−66^) and HSPA1B (8.85, 2.8 × 10^−23^), whereas marked suppression was observed for genes such as MT-TI (−7.06, 1.3 × 10^−4^), HMGCS2 (−6.43, 6.4 × 10^−4^) and FETUB (−6.19, 9.8 × 10^−4^). Complete DEG tables (including all up/down genes with log2FC and FDR for each dose) are provided in the attached Supplementary Excel File S2.

Volcano plots demonstrated proportional up- and downregulation across doses (Figure 1d–f). At the same time, multidimensional scaling showed clear segregation by concentration (Figure 1g). Venn diagram analysis highlighted both overlapping DEGs and unique gene sets at each dose, supporting conserved and dose-specific responses to 3D8 scFv (Figure 1h). These results indicate that 3D8 scFv drives concentration-dependent transcriptomic remodeling, with minor perturbations at low concentrations and extensive reprogramming at higher doses. To interpret biological significance of these dose-dependent DEGs, pathway enrichment and network-level analyses were performed.

Dose-Dependent Pathway Enrichment and MAPK-Centered Transcriptomic Reprogramming

Pathway enrichment was performed for each 3D8 scFv dose relative to the untreated control (0 µM). Over-representation analyses were conducted on DEGs (FDR < 0.05) using all expressed genes as the background, whereas GSEA used the full ranked gene list from each dose vs. control contrast; significance was defined at FDR q < 0.05. Transcriptomic pathway enrichment analysis revealed that 3D8 scFv induced progressive, concentration-dependent remodeling of host cellular pathways, with mitogen-activated protein kinase (MAPK) signaling consistently emerging as the most significantly enriched pathway across all treatment conditions (Figure 2a–c).

At 5 µM, pathway enrichment was broad and distributed, with MAPK signaling showing significant enrichment (normalized enrichment score (NES) = 1.58–2.31, FDR = 0.05–0.0001), alongside pathways related to synaptic signaling, endoplasmic reticulum (ER) protein processing, and estrogen signaling (Figure 2a and Figure S2a). These enrichment patterns were characterized by moderate effect sizes and relatively low false discovery rates, consistent with early adaptive transcriptional responses. At 10 µM, enrichment became more consolidated, with MAPK signaling displaying sustained statistical significance (NES = 1.67–2.07, FDR = 0.05–0.003) and increased network connectivity with immune- and stress-associated pathways, including tumor necrosis factor (TNF), p53, nuclear factor κB (NF-κB), and Hippo signaling (Figure 2b and Figure S2b). Compared with the 5 µM condition, enrichment at 10 µM was associated with higher coherence among stress-response pathways, indicating a shift toward coordinated transcriptional regulation.

In contrast, treatment with 40 µM 3D8 scFv resulted in a more restricted enrichment profile, dominated by antiviral and metabolic pathways. MAPK signaling remained significantly enriched at this concentration (NES = 1.74–1.90, FDR = 0.04–0.02), together with Janus kinase–signal transducer and activator of transcription (JAK–STAT), phosphoinositide 3-kinase–Akt (PI3K–Akt), and NOD-like receptor signaling pathways (Figure 2c and Figure S2c). Notably, the narrower NES range at 40 µM reflected reduced pathway diversity relative to lower concentrations.

Mapping differentially expressed genes onto KEGG pathway modules further confirmed dose-dependent engagement of ERK1/2, JNK, and p38 MAPK branches, together with enrichment of upstream regulatory kinases such as MAP2K1, MAP2K4, and transforming growth factor β–activated kinase 1 (TAK1) (Figure 2d). Among downstream effectors, members of the HSP70 family (HSPA6, HSPA1A, and HSPA1B) exhibited some of the highest fold changes and showed a clear, concentration-dependent increase in expression (Figure 2e).

Consistent with these findings, supplementary analyses (Figures S3 and S4) revealed coordinated upregulation of additional proteostasis-associated genes, including molecular chaperones and ubiquitin–proteasome system components. These results demonstrate that 3D8 scFv orchestrates a conserved MAPK–HSP70 stress response axis [20], shifting from broad adaptive signaling at low doses to highly focused antiviral and proteostasis pathways at higher concentrations.

2.2. Machine Learning Reveals Virus-Specific Transcriptomic Signatures Distinct from 3D8 scFv Responses

RNA-seq analysis of Zika virus (ZIKV, GSE132228) and dengue virus (DENV2, GSE111330) infections, benchmarked against 3D8 scFv-treated controls, revealed virus-specific transcriptional programs. ZIKV induced widespread upregulation (2354 upregulated and 307 downregulated; padj < 0.05), including stress regulators (HSPA6, GDF15), immediate-early transcription factors (EGR1, FOS, JUN), and immune mediators (Figure 3a–c). This comparison was performed to benchmark whether host transcriptional program induced by 3D8 scFv exposure (A549; 5–40 µM vs. 0 µM; no virus) aligns with or opposes virus-associated transcriptional signatures derived from public ZIKV/DENV2 infection datasets. Infection classifiers were trained on the public datasets and applied them to the 3D8 scFv-treated transcriptomes to evaluate ‘discriminatory’ behavior at the signature level, rather than conducting a direct matched infection experiment in the same cell system.

In MA plot, genes with negative log_2_ fold change (indicating downregulation) included NDUFC2, HNRNPM, and GTF3C6, which are associated with cellular stress responses and transcriptional regulation. In the volcano plot, genes with positive log_2_ fold change were prominently represented among the upregulated genes, including HSPA6, EGR1, EGR2, and GDF15 (Figure 3a,b). Figure S5a highlights several biologically relevant differentially expressed genes previously associated with Parkinson’s disease, including SNCA, LRRK2, PINK1, PARK7, and TH (tyrosine hydroxylase). Inflammatory mediators such as IL6 and TNF were also among the significant DEGs, reflecting transcriptomic alterations across neuronal, mitochondrial, and inflammatory pathways relevant to Parkinson’s disease. All ten top differentially expressed genes were significantly downregulated in Zika virus-infected samples compared with controls, with consistently higher log expression values observed in controls. The heatmap confirmed this pattern, showing uniform suppression across infected samples. Several downregulated genes, including ND1, ND5, and ND6, were associated with mitochondrial respiration, while others such as SRSF7 and RAVER1 were linked to RNA processing (Figure 3c).

In contrast, DENV2 imposed predominant suppression (1891 downregulated, 1095 upregulated), targeting stress- and immune-associated genes (JUNB, IER5L) and lncRNAs (CEBPB-AS1, PYCARD-AS1) (Figure 3d–f). DENV2-associated candidate markers were therefore summarized using the top-ranked DEGs and classifier features (Figure S5), rather than a separate meta-analysis. Figure S5B evaluated RNA-seq data quality for GSE111330. The dispersion estimate plot shows well-modeled gene-wise variance across expression levels, while the adjusted p-value histogram is enriched near zero, indicating a substantial number of significantly differentially expressed genes and robust biological signal.

The MA plot (Figure 3d) shows log_2_ fold change relative to mean normalized expression, with most genes centered near zero and a substantial subset significantly up- or downregulated following dengue infection. The volcano plot (Figure 3e) highlights these differences by separating significantly upregulated and downregulated genes based on effect size and adjusted p-value. Figure 3f presents expression profiles of the top 10 differentially expressed genes using boxplots and a hierarchically clustered heatmap. Boxplots show consistent downregulation across all top DEGs in Dengue samples, and the heatmap cleanly separates dengue and control groups. Notably downregulated genes include JUNB, IER5L, CEBPB-AS1, PYCARD-AS1, CAPN15, and ZNF219, which are associated with cellular stress and immune regulation.

Hierarchical clustering and UMAP confirmed infection-driven segregation from 3D8 controls. UMAP analysis of GSE132228 showed clear condition-dependent clustering, with Zika-infected samples separating from controls primarily along the UMAP1 axis. Infected samples clustered in a region associated with elevated expression of JUN, EGR1, ATF3, HSPA6, and DDIT3, while control samples occupied a distinct baseline cluster, with partial overlap reflecting biological variability (Figure S6a–c). Application of the ZIKV-trained model to 3D8 scFv-treated cells confirmed reversal of viral transcriptional profiles (accuracy = 0.98, AUC = 0.99). Figure S6d shows a UMAP projection of GSE111330, with clear separation between dengue-infected (green) and control (purple) samples, indicating a strong and consistent disease-specific transcriptional signature with low intra-group variability. Figure S6e,f present XGBoost classifier performance using 5-fold cross-validation. Training accuracy was 1.0 across all folds, while validation accuracy remained near-perfect except for a decrease in fold 3. Accuracy, precision, recall, F1-score, and AUC were consistently high across folds (AUC = 1.0), demonstrating robust discriminative performance despite minor variability. Similarly, applying DENV2-trained classifier showed that 3D8 scFv-treated samples were classified closer to uninfected controls, consistent with reversal/attenuation of the DENV2-associated transcriptional profile. Collectively, all the data demonstrated distinct ZIKV- and DENV2-driven host reprogramming, with 3D8 scFv restoring the transcriptome towards a healthy state (Figure 3). Standardized effect size analysis identified several biomarkers with moderate to large expression changes associated with Zika virus-related pathogenesis, including PIK3CA, NOS2, CASQ2, CXCL14, EZH2, and EXO1 (Figure S7a); these genes overlap with differentially expressed transcripts detected in the present dataset, indicating consistency across analytical approaches. Effect size estimates were generally accompanied by relatively narrow confidence intervals, suggesting stable estimation. Evaluation of potential publication bias using funnel plot analysis revealed a largely symmetric distribution of effect sizes relative to their standard errors across biomarker categories, with no strong evidence of pronounced small-study effects (Figure S7b), supporting the overall robustness of the observed associations while remaining consistent with expected inter-study variability. Figure S8 shows minimal overlap among features selected by the Target, Dengue_ML, and ZIKA_ML models, with only a single shared feature between each pair and noncommon to all three, indicating largely distinct virus-specific molecular signatures.

2.3. DENV2 and ZIKV Replication Varies with 3D8 scFv Delivery Time

Time-course assays revealed that the antiviral efficacy of 3D8 scFv was strongly dependent on its administration relative to viral exposure. Prophylactic treatment significantly suppressed viral RNA accumulation, whereas therapeutic treatment showed minimal effect, with mock samples serving as baseline controls and positive controls normalized to 1. In DENV2, intracellular viral RNA (E gene) transiently increased at 2–4 h post-infection before declining sharply, reaching ≤0.6-fold of untreated infected control levels by 6 h and remaining reduced by 12 h (Figure 4a). ZIKV exhibited a similar pattern, with modest early increases at 2 h followed by sustained suppression after 6 h (Figure 4b). In contrast, therapeutic treatment (administered after viral adsorption) showed limited inhibition of intracellular viral RNA for both DENV2 and ZIKV across 2–12 h, showing trends comparable to infected, untreated control (Figure 4c,d). These findings suggest that prophylaxis treatment enables 3D8 scFv to accumulate and target viral replication complexes during peak replication. In contrast, therapeutic administration is limited by the sequestration of viral RNA within membrane-bound replication compartments. The limited therapeutic efficacy likely reflects both rapid viral internalization and low intracellular stability of scFv proteins, which degrade after ~12 h, as well as their short plasma half-life [21,22]. However, improvement in the protein treatment regimen (5 µM, 2 exposures) resulted in measurable inhibition.

2.4. Prophylactic 3D8 scFv Inhibits DENV2 and ZIKV Replication in Mono- and Co-Infection Settings

A prophylactic treatment paradigm was employed to determine whether intracellular availability of 3D8 scFv before viral challenge alters subsequent flaviviral replication outcomes. Cells were exposed to 3D8 scFv under prophylactic conditions before infection with DENV2 or ZIKV at low multiplicity (MOI 0.1), and antiviral efficacy was evaluated at a fixed post-infection endpoint using viral RNA quantification, infectious progeny assays, and viral protein detection (Figure 5a). In DENV2 mono-infected cells (Figure 5b–d), prM and E transcript levels were reduced by approximately 80–85% relative to infected untreated controls (p < 0.001), coinciding with a ~75% reduction in infectious particle production (Figure S9a) and a pronounced decrease in E protein abundance (Supplementary WB 5d). ZIKV replication was even more strongly inhibited (Figure 5e–g), with prM and E transcript levels reduced to below 10% of control values (p < 0.0001), an 83% decrease in plaque-forming units (Figure S9b), and near-complete loss of detectable E protein (Supplementary WB 5g). Importantly, the antiviral effect of 3D8 scFv was preserved under prophylactic co-infection conditions (Figure 5h–j). Cells simultaneously challenged with DENV2 and ZIKV exhibited coordinated suppression of both viruses, with prM and E transcript levels reduced to approximately 20–25% of their respective controls, an overall ~78% decrease in infectious progeny release (Figure S9c), and substantial attenuation of dengue and Zika E protein expression (Supplementary WB 5j). Together, these data demonstrate that prophylaxis treatment with 3D8 scFv effectively blocks viral RNA synthesis, protein expression, and infectious particle production across both mono- and co-infection contexts.

2.5. 3D8 scFv Exhibits Potent Therapeutic Antiviral Activity Against DENV2 and ZIKV in Mono- and Co-Infection Settings

Therapeutic antiviral activity of 3D8 scFv was evaluated using a post-entry treatment paradigm in which Vero E6 cells were first infected with DENV2, ZIKV, or both viruses (MOI 0.1) and subsequently exposed to 3D8 scFv during ongoing replication (Figure 6a). Antiviral effects were quantified at 48 h post-infection using complementary measurements of viral RNA abundance, infectious progeny production, and viral protein expression. Under therapeutic conditions, 3D8 scFv significantly constrained DENV2 replication (Figure 6b–d). Viral prM and E transcript levels were reduced to approximately 30% and 20% of infected untreated controls, respectively (p < 0.001), and this reduction in intracellular viral RNA was accompanied by a ~73% decrease in plaque-forming units (Figure S10a). Consistent with these findings, immunoblot analysis revealed marked suppression of DENV2 E protein expression (Supplementary WB 6d), indicating effective inhibition of both viral genome amplification and downstream protein synthesis.

A comparable inhibitory profile was observed for ZIKV (Figure 6e–g). Therapeutic administration of 3D8 scFv reduced viral prM and E transcript levels to approximately 20–25% of control levels (p < 0.001), coinciding with a 76% reduction in infectious particle production (Figure S10b) and pronounced attenuation of E protein abundance (Supplementary WB 6j). The concordance between RNA-level suppression and reductions in infectivity supports a direct impact on productive viral replication. Importantly, antiviral efficacy was preserved under co-infection conditions (Figure 6h–j). Cells simultaneously infected with DENV2 and ZIKV and treated with 3D8 scFv exhibited coordinated reductions in prM and E transcript levels for both viruses to below 25% of their respective controls, together with an approximately 75% decrease in total infectious progeny (Figure S10c) and strong suppression of dengue and Zika E protein expression (Supplementary WB 6j). These results demonstrate that therapeutic administration of 3D8 scFv after viral entry remains sufficient to restrict flaviviral RNA accumulation, protein expression, and infectious particle production in both mono- and co-infection contexts. Although the magnitude of inhibition is reduced relative to prophylactic exposure, the sustained antiviral activity observed post-entry highlights the capacity of 3D8 scFv to interfere with established flaviviral replication.

3. Discussion

Current antiviral discovery for dengue virus (DENV2) and Zika virus (ZIKV) has primarily targeted viral proteins, including the NS2B–NS3 protease, NS5 polymerase, and E glycoprotein, with promising small molecules, peptides, and repurposed host-directed drugs [23,24,25,26,27,28]. However, protein-targeted inhibitors are prone to resistance, while host-directed strategies face toxicity challenges [29]. To circumvent these limitations, we evaluated the nucleic acid-hydrolyzing antibody fragment 3D8 scFv, which penetrates host cells and catalytically degrades viral genomes in a sequence-independent manner [14].

3D8 scFv was expressed and purified in E. coli and retained its catalytic activity toward both DNA and RNA substrates (Figure S1). Vero E6 cytotoxicity profiling (Figure S11) using CCK-8 assay confirmed that 3D8 scFv is well tolerated at concentrations ≤ 10 µM (>90% cell survival and a CC_50_ of approximately 14 µM). Prophylactic and Therapeutic assays in Vero E6 cells demonstrate potent antiviral activity against DENV2 and ZIKV under mono- and co-infection. Overall, prophylactic treatment produced stronger inhibition of viral RNA (particularly for ZIKV), whereas reductions in infectious progeny were broadly comparable between the pre- and post-infection treatment regimens [21,22]. These findings align with a “critical site” neutralization model, where catalytic RNA cleavage during early replication prevents productive infection [30,31,32].

Mechanistically, antiviral efficacy differs between pathogens. ZIKV infection repressed mitochondrial oxidative phosphorylation and RNA-processing pathways, whereas DENV2 suppressed immune and stress regulators, including inflammasome-associated lncRNAs (Figure 3). Importantly, pathogen-specific transcriptional signatures persisted under 3D8 scFv treatment. Machine learning classifiers trained on differentially expressed genes discriminated against infection states with near-perfect accuracy (AUC ≈ 1.0), identifying biomarkers such as PIK3CA, NOS2, and JUNB as key drivers—3D8 scFv-treated cells clustered with uninfected controls, indicating effective viral suppression without broad transcriptomic burden.

A key advantage of ML framework used here is that it enables data-driven separation of virus-specific host-response signatures from the transcriptional effects associated with 3D8 scFv exposure [33]. By training classifiers on infection-associated DEGs and examining feature importance, we identified compact gene sets that best discriminate ZIKV vs. DENV2 responses and can serve as candidate biomarker panels for infection state classification. In this context, ML complements conventional DEG/enrichment analysis by prioritizing the most informative features for diagnostic or mechanistic follow-up and by providing a quantitative metric of separability (AUC) that is robust to high-dimensional transcriptomic data. More broadly, this approach provides a framework to evaluate whether candidate antivirals shift host transcriptional profiles toward non-infected states and to nominate virus-specific pathways/genes for targeted validation [34].

At the host level, 3D8 scFv induced dose-dependent activation of MAPK signaling (ERK1/2, JNK, p38) and a strong induction of HSP70 chaperones (HSPA6, HSPA1A, HSPA1B), which are adaptive stress responses critical for proteostasis [20,35]. HSP70 family proteins have been reported to facilitate multiple stages of flavivirus replication, including DENV entry/replication complex formation and virion production, and pharmacological HSP70 inhibitors have shown antiviral activity against DENV [36,37]. Therefore, the observed induction of HSP70 transcripts after 3D8 scFv exposure may reflect a generalized proteostasis or stress response rather than a direct antiviral mechanism. Importantly, because HSP70 can support DENV replication, it cannot be ignored that HSP70 upregulation contributes to the comparatively smaller effect size observed for DENV2 in some readouts. A direct way to test this would be to combine 3D8 scFv with HSP70 inhibition (or knockdown) and assess whether antiviral effects are enhanced or diminished, while also monitoring cell viability. Such experiments would help clarify whether HSP70 induction is compensatory, proviral, or part of a protective host response that is independent of 3D8 scFv’s nuclease-mediated antiviral activity. These signatures were observed at antiviral concentrations without cytotoxicity, suggesting that 3D8 scFv promotes cellular resilience rather than causing damage. Its broad-spectrum nuclease activity has previously been shown to suppress influenza A [13] and coronaviruses, including SARS-CoV-2 [14], further underscoring its translational potential.

Despite these promising findings, key limitations still exist. First, in vivo validation is missing, and pharmacokinetic issues such as intracellular stability, tissue penetration, and delivery need to be addressed. 3D8 scFv showed antiviral activity in both prophylactic and therapeutic settings but the short intracellular persistence of scFv proteins and limited systemic half-life likely constrain the magnitude and duration of inhibition, particularly in post-entry regimens. Future studies could therefore focus on optimizing stability and delivery to prolong effective exposure. Second, although cytotoxicity assays and transcriptomic profiling in Vero E6 and A549 cells indicated minimal off-target effects, the potential for unintended nuclease activity in other cell types, or in vivo, remains to be evaluated. Lastly, although 3D8 scFv has a lower risk of resistance than protein-targeted antivirals, systematic benchmarking against existing drugs in clinically relevant settings is necessary.

We propose that intracellular 3D8 scFv binds to and cleaves viral genomes immediately after uncoating and during early replication, blocking protein synthesis and progeny release (Figure 7). Transcriptomic profiling combined with machine learning indicates that 3D8 scFv suppresses viral replication while maintaining host transcriptomic integrity. These findings underscore its potential as a broad-spectrum RNA-targeting antiviral, especially in regions where DENV2 and ZIKV co-circulate. Future efforts should focus on optimizing intracellular stability and delivery to improve efficacy during later stages of infection.

4. Materials and Methods

4.1. Cells and Virus Propagation

A549 cells were cultured in DMEM (GenDEPOT, Baker, TX 77412, USA) with 10% FBS and 1% penicillin-streptomycin at 37 °C, 5% CO_2_, and treated with 0–40 µM 3D8 scFv for cytotoxicity assays. Vero E6 cells (ATCC CRL-1586) were maintained similarly. DENV2 (KBPV-VR-29) and ZIKV (NCCP 43280) were propagated in Vero E6 cells, and then clarified by centrifugation and 0.45 µm filtration; titers were determined by plaque assay.

4.2. Production, Purification, and Nuclease Activity of 3D8 scFv Against Nucleic Acids

The 3D8 scFv gene was expressed in E. coli BL21 (DE3) pLysE, induced with 1 mM IPTG at mid-log phase, and incubated at 26 °C for 18 h. Following centrifugation and 0.22 μm filtration, the protein was purified via Capto-L affinity chromatography, eluted with pH 2.0 glycine–citrate buffer, neutralized, and verified by SDS-PAGE. Purified 3D8 scFv (0.5 µg) was incubated with plasmid DNA, rRNA, or in vitro transcribed viral RNA in Tris-buffered saline with MgCl_2_ at 37 °C, and degradation was monitored over time by agarose gel electrophoresis; viral RNA was synthesized using the MEGAscript T7 kit (Invitrogen^TM^ by Thermo Fisher Scientific Baltics UAB, V.A. Graiciuno8, Vilnius, Lithuania), and total RNA was extracted from mammalian cells with TRI reagent.

4.3. RNA Extraction, Sequencing, and DEG Analysis

Total RNA (RIN ≥ 8) was isolated with TRI reagent. For transcriptomic analysis, A549 cells were treated with 3D8 scFv at 0 (untreated control), 5, 10, or 40 µM. Each condition included two independent biological replicates. Libraries were prepared using TruSeq Stranded mRNA Kit (Illumina, Inc., San Diego, CA, USA) and sequenced on NovaSeq 6000 (Illumina, Inc., San Diego, CA, USA) (150 bp paired-end). Reads were quality-checked with FastQC v0.11.9, trimmed with Trimmomaticv0.3919, and aligned to GRCh38 (Ensembl GCA_000001405.28) with HISAT2 v2.1.0 [38]. Gene counts were obtained via featureCounts. Genes with ≥10 counts were excluded, and data were TMM-normalized in edgeR v3.28.1 (Bioconductor, R Foundation for Statistical Computing, Vienna, Austria) [39]. DEGs were defined as |log_2_FC| ≥ 1 with FDR < 0.05. Reference gene stability was assessed by R^2^ of log_2_(TMM) values, and sample clustering was evaluated by multidimensional scaling in limma [40] and visualized with ggplot2 [41].

4.4. Functional Annotation and Pathway Analysis

Differentially expressed genes (DEGs) were analyzed with DAVID v6.8 for GO and KEGG enrichment (FDR < 0.05) and visualized using REVIGO v1.8.1. GSEA v4.0.2 was performed on KEGG gene sets ranked by log_2_(TMM-normalized expression), with significant pathways (FDR < 0.05) [42] displayed as enrichment maps in Cytoscape v3.10.4, and core-enriched genes are shown in heatmaps. Pathway-specific modulation was further assessed using clusterProfiler v3.22 [43].

4.5. Machine Learning Classification of ZIKV and DENV2 Using DEGs from 3D8 scFv-Treated Cells

Public RNA-seq datasets for Zika (GSE132228) and dengue virus 2 (GSE111330) were processed in R (v4.x) with DESeq2 [44], using median-of-ratios normalization and variance-stabilizing transformation; differentially expressed genes (adjusted p < 0.05) were identified via empirical Bayes shrinkage, and the top 100 per pathogen were retained. Feature matrices and binary labels were exported to Python (v3.10) for XGBoost [45] (v1.7) model training with predefined hyperparameters (learning_rate 0.05, max_depth 6, n_estimators 500, subsample 0.8, colsample_bytree 0.8, gamma 0.1, reg_lambda 1.0) and scale_pos_weight to adjust class imbalance. Model performance was evaluated by stratified five-fold cross-validation. UMAP (v0.5) was applied independently for visualization. Reproducibility was ensured with seed 42.

4.6. Cytotoxicity of 3D8 scFv in Vero E6 Cells

Cell viability was assessed using the Cell Counting Kit-8 (CCK-8) assay, which is based on the reduction in the water-soluble tetrazolium salt WST-8 by cellular dehydrogenases in metabolically active cells to generate an orange formazan dye. The amount of formazan produced is proportional to the number of viable cells and was quantified by measuring absorbance. Vero E6 cells (4 × 10^4^/well) were seeded in 96-well plates and treated with 2–25 µM 3D8 scFv for 48 h at 37 °C and 5% CO_2_. After adding 10 µL CCK-8 and 90 µL PBS, cells were added and incubated for 1–3 h, and absorbance was measured at 595 nm. Data represents SD from three biological replicates (n = 3).

4.7. In Vitro Antiviral Capability of 3D8 scFv

The antiviral activity of 3D8 scFv was evaluated in Vero E6 cells (ATCC CRL-1586) cultured in DMEM containing 10% FBS and antibiotics at 37 °C and 5% CO_2_. ZIKV (NCCP 43280) and DENV2 (KBPV-VR-29) were propagated in Vero E6 cells, clarified, filtered (0.45 µm), and stored at −80 °C, with titers determined by plaque assay. Infection models included DENV2, ZIKV, and co-infection at MOI 0.1 in 24-well plates (1 × 10^5^ cells/well).

4.8. Time-Course Assays

Cells were pretreated with 5 µM 3D8 scFv for 2–12 h (prophylaxis) or treated post-infection (Therapeutic). After 1 h viral infection, cells were washed with DPBS and incubated with 3D8 scFv. At 48 h, RNA was extracted (TRIzol, Invitrogen™, Thermo Fisher Scientific, Cincinnati, OH, USA) and viral E gene expression quantified by RT-qPCR, normalized to GAPDH, and analyzed with the 2^−ΔΔCt^ method.

4.9. Prophylaxis Assays

Cells were exposed to 10 µM 3D8 scFv for 12 h (with an additional 5 µM dose at 6 h), then inoculated with DENV2 or ZIKV (MOI 0.1). After 1 h of adsorption and DPBS washes, complete medium was added and cells were maintained for 48 h. Supernatants were stored at –80 °C and RNA was extracted using TRI reagent from cell lysates.

4.10. Therapeutic Efficacy Assay

To assess therapeutic activity, Vero E6 cells were infected with ZIKV, DENV2 (MOI 0.1), or both for co-infection. Viral adsorption was carried out in serum-free medium for 1 h, followed by two DPBS washes. Cells were then incubated with 3D8 scFv (5 µM) in DMEM; a second 5 µM dose was administered 24 h post-infection to ensure sustained intracellular exposure. After 48 h, supernatants and cell pellets were collected and stored at −80 °C for molecular analysis.

4.11. RNA Isolation and RT-qPCR

Total RNA was isolated using TRI reagent, quantified spectrophotometrically, and 1 µg was used for cDNA synthesis. Viral and host transcripts were measured by one-step RT-qPCR (AccuPower® GreenStar™ qPCR PreMix (Bioneer Corporation, Daejeon, Republic of Korea)) on a Rotor-Gene Q system (v2.1.0). Each 20 µL reaction contained 50 ng RNA and gene-specific primers for ZIKV and DENV2 E and prM genes (Table S1).

4.12. Plaque and Plaque Reduction Assays

Viral infectivity and neutralization were evaluated in Vero E6 cells (5 × 10^5^/well) in 6-well plates. Tenfold-diluted antiviral supernatants were adsorbed for 1 h at 37 °C, after which cells were overlaid with 2× DMEM containing 1% SeaPlaque agarose and incubated for 3–4 days. Plaques were then visualized, counted, and quantified as percent reduction relative to untreated controls.

4.13. Immunoblotting

Cellular proteins were extracted with PRO-PREP (iNtRON Biotechnology, Gyeonggi do, Republic of Korea), and 20 µg per sample was resolved by SDS-PAGE and transferred to PVDF membranes. Blots were probed with rabbit anti-ZIKV envelope (GeneTex Inc. (Irvine, CA, USA), GTX133325) or mouse anti-DENV2 antigen (Santa Cruz BIOTECHNOLOGY, INC Dallas, TX, USA, clone D1-11, sc-65659), followed by HRP-conjugated goat anti-rabbit (Invitrogen, A21020) or goat anti-mouse (Invitrogen, G-21040) secondary antibodies. Signals were detected by ECL solution (GenDEPOT, USA, W3652-050) and imaged on an iBright 1500 system (Invitrogen™, Thermo Fisher Scientific). Co-infection lysates were processed in parallel with antibody combinations to distinguish ZIKV and DENV2 proteins.

4.14. Statistical Evaluation

Quantitative data is reported as mean ± standard deviation, unless otherwise indicated. GraphPad Prism v8.0 (GraphPad Software, La Jolla, CA, USA) was used for statistical analyses. Comparisons between experimental groups were performed using a two-tailed Student’s t-test. Differences were considered significant at p < 0.05.