Since saRNA can induce cap-dependent translation shutdown and cell death, we developed a microplate assay to simultaneously monitor cap-independent translation, cap-dependent translation, and cell number over time. In this assay, primary mouse FLS (Fig. S1) were plated in 6-well plates and transfected with saRNA constructs that used the Venezuelan equine encephalitis virus (VEEV) RdRp to self-replicate.

These constructs featured a dual-fluorescence design, with EGFP expressed through encephalomyocarditis virus (EMCV) IRES-mediated, cap-independent translation, and mScarlet3 expressed through subgenomic cap-dependent translation. This dual-reporter system enabled assessment of the effects of saRNA on translational control (Fig. 1a). Spectral overlap between EGFP and mScarlet3 was corrected using linear unmixing (Figs. S2a-c).

To monitor cell number, FLS were stained with BioTracker NIR680, a lipophilic carbocyanine membrane dye that enables long-term fluorescent labeling of cells.³² Carbocyanine dyes like BioTracker are weakly fluorescent in aqueous solution but exhibit high fluorescence when integrated into lipid bilayers.

Consequently, mock transfection with Lipofectamine MessengerMAX—a lipid transfection agent— enhanced BioTracker fluorescence compared to non-transfected cells, likely due to an interaction with the lipid components of Lipofectamine (Figs. S3a, S3b). Despite this interaction, BioTracker remains effective for indicating cell number, as treatment of mock transfected cells with the apoptosis inducer staurosporine³³ reduced BioTracker fluorescence (Figs. S3a, S3b). The BioTracker results were consistent with those obtained using the viability dye calcein AM (Figs. S3c, S3d), with the added advantages of eliminating the need for additional staining and wash steps and avoiding spectral overlap with red and green fluorescent proteins. This assay allows for seamless longitudinal monitoring of cap-independent translation, cap-dependent translation, and cell number using microplate readers or imagers.

Cells detect cytosolic dsRNA through several sensing pathways, including RIG-I,³⁴ melanoma differentiation-associated protein 5 (MDA-5),³⁵ protein kinase R (PKR),³⁶ and the oligoadenylate synthase (OAS)/RNase L pathway.³⁷ To broadly inhibit these pathways, we designed an saRNA construct, referred to as ‘E3’, that utilizes EMCV IRES cap-independent translation to express the vaccinia virus E3 protein (Fig. 1a). E3 protein is a pleiotropic inhibitor that binds and sequesters dsRNA, effectively blocking multiple dsRNA sensing pathways.³⁸

Given the key roles of PKR in cap-dependent translation shutdown and OAS/RNase L in cellular mRNA degradation in response to dsRNA, we sought to further inhibit these pathways in a second construct. This construct includes Toscana virus NSs, a ubiquitin ligase that promotes PKR degradation,³⁹ and Theiler’s virus L*, which inhibits RNase L.⁴⁰ To enable co-expression of these three proteins—E3, NSs, and L*—we incorporated 2A “self-cleaving” peptide sequences between the proteins, allowing the polyprotein to be cleaved into separate proteins during translation. As prior experiments showed that using multiple identical 2A peptides can reduce protein expression,⁴¹ we selected non-identical 2A peptides for this construct. This construct, named ‘E3-NSs-L*’, is expected to provide a more comprehensive inhibition of dsRNA-sensing pathways than the E3 construct.

Finally, as a control, we designed a ‘native saRNA’ construct that expresses a blue fluorescent protein (moxBFP) as a visual marker of IRES functionality in place of inhibitors of dsRNA-sensing pathways. Importantly, moxBFP expression did not overlap spectrally with the EGFP or mScarlet3 channels (Figs. S2a-c).

Following transfection of saRNA constructs, we monitored BioTracker, EGFP, and mScarlet3 fluorescence over 3 weeks (Figs. 1b, 1c). Inhibiting dsRNA-sensing pathways with viral proteins significantly enhanced saRNA transgene expression (Figs. 1d, 1e). Interestingly, the E3 construct produced the highest levels of EGFP and mScarlet3 expression, surpassing both native saRNA and the E3-NSs-L* construct. Native saRNA yielded low levels of transgene expression, while E3-NSs-L* showed intermediate expression.

Achieving durable transgene expression depends equally on preserving cell viability and maximizing transgene expression, as constructs that induce cell death undermine the goal of sustained transgene expression. Transfection of FLS with native saRNA caused both immediate and long-term reductions in cell number, as indicated by BioTracker intensity (Fig. 1f). While the E3 construct initially maintained cell number, BioTracker signals gradually diminished over time. In contrast, the E3-NSs-L* construct provided sustained protection against the decline in cell number. To quantify the cumulative effects of different saRNA constructs on cell number over the experimental time course, we performed area under the curve (AUC) analysis, demonstrating that E3-NSs-L* significantly protected against a decline in cell number compared to the E3 construct (Fig. 1g). These findings were further supported by experiments using CellTag, a cell number normalization stain (Fig. 1h), which revealed that increasing dsRNA-sensing pathway inhibition protects against cell loss (Fig. 1i). Together, these results suggest that while the E3 construct effectively enhances saRNA transgene expression, it is associated with greater cell loss, underscoring the advantage of the E3-NSs-L* construct for applications requiring sustained, non-cytotoxic gene expression.

saRNA is known to induce programmed cell death,^6,10,12–16 which often leads to cell detachment.⁴² Thus reductions in BioTracker signal following saRNA treatment are likely indicative of this process. However, since BioTracker is a lipophilic membrane dye, the reduction in signal may not necessarily correlate with cell viability or cytotoxicity. In addition to long-term cell tracking, lipophilic membrane dyes like BioTracker are often used to stain extracellular vesicles, which bud off from the plasma membrane.⁴³ During cellular stress, increased production of extracellular vesicles⁴⁴ could deplete the membrane-associated BioTracker dye from cells, resulting in a lower detectable signal. Therefore, the observed reduction in BioTracker signal could reflect enhanced extracellular vesicle production rather than cell death.

To provide more definitive evidence that inhibiting dsRNA-sensing pathways protects against saRNA-induced cytotoxicity, we conducted a longitudinal assay with annexin V staining over 6 days (Fig. 2a), followed by calcein AM staining on day 7 (Fig. 2b). Annexin V, a membrane-impermeable protein, binds to phosphatidylserine, which translocates to the extracellular side of the cell membrane during early apoptosis, enabling the detection of apoptotic cells with intact membranes.⁴⁵ If the membrane is compromised, annexin V can enter cells and stain intracellular phosphatidylserine, indicating late apoptosis or necrosis.⁴⁶ Calcein AM, a viability dye, accumulates in metabolically active cells with intact membranes, serving both as a marker of live cell number and an indicator of membrane integrity.⁴⁷

The apoptosis inducing agent staurosporine caused a transient increase in annexin V staining that decreased after the first day (Fig. 2c), along with a marked reduction in calcein AM fluorescence compared to mock transfection (Fig. 2d), observations that are consistent with extracellular phosphatidylserine translocation and/or increased membrane permeability. Similarly, native saRNA induced a temporary increase in annexin V staining during the first two days post-transfection (Fig. 2c) accompanied by reduced calcein AM staining (Fig. 2d). Co-expression of E3 prevented the initial increase in annexin V staining, but elevated it from day 2 onward, and also resulted in significantly reduced calcein AM intensity (Figs. 2c, 2d). In contrast, E3-NSs-L* had no significant effect on annexin V or calcein AM staining compared to mock transfection, supporting the idea that broader inhibition of dsRNA-sensing pathways, achieved by combining E3, NSs and L*, is more effective than E3 alone in preventing saRNA-induced cell death and preserving cell viability.

After characterizing how moderate and strong inhibition of dsRNA-sensing pathways affects saRNA-induced cytotoxicity, we next investigated molecular mechanisms of translational control to understand how this inhibition influences transgene expression. Eukaryotic initiation factor 2 alpha (eIF2α) is a central regulator of translation initiation; its phosphorylation serves as a cellular stress response to limit cellular cap-dependent protein synthesis, freeing ribosomes for cap-independent translation initiation instead.⁴⁸ Using in-cell western assays, we found that native saRNA transfection increased eIF2α phosphorylation, but both E3 and E3-NSs-L* effectively blocked this phosphorylation (Fig. 3a), consistent with previous reports of vaccinia virus E3’s inhibitory effect on eIF2α phosphorylation.²² Interestingly, E3-NSs-L* also increased total eIF2α levels, an effect not observed with the E3 construct (Fig. 3b).

Given that PKR phosphorylates eIF2α upon activation by dsRNA,⁷ we next assessed its levels following saRNA transfection. Consistent with previous reports,⁶ native saRNA transfection increased PKR levels (Fig. 3c). The E3 construct similarly increased PKR levels; however, co-expression of E3-NSs-L* mitigated this increase, suggesting that E3-NSs-L* effectively counteracts saRNA-induced PKR upregulation. This is in line with the known role of the Toscana virus NSs protein as a ubiquitin ligase that targets PKR for degradation.³⁹

We next examined the impact of saRNA transfection on eIF4E, a key regulator of translation initiation.⁴⁹ All constructs reduced eIF4E phosphorylation levels (Fig 3d), while total eIF4E levels remained unchanged (Fig. 3e).

eIF2α-and eIF4E-mediated translational control do not account for the greater transgene expression observed with the E3 construct compared to the E3-NSs-L* construct. This is evidenced by the lack of significant differences in eIF2α and eIF4E phosphorylation status or levels between the two constructs (Figs. 3a-b, 3c-e), suggesting that additional mechanisms beyond eIF2α-and eIF4E-mediated control contribute to the enhanced transgene expression seen with the E3 construct.

When OAS binds dsRNA, it synthesizes 2′-5′-oligoadenylate, which activates RNase L to cleave single-stranded RNA, leading to cellular mRNA degradation⁵⁰ and inhibition of mRNA nuclear export,⁵¹ thereby massively reducing host protein synthesis.⁵² Many RNA viruses, including dengue and Zika, evade RNase L by localizing their transcripts within viral replication organelles, protecting them from degradation while allowing continued protein production.^51,53 Similarly, the VEEV-derived saRNA used in this study replicates within these organelles,⁵⁴ likely protecting its transcripts from RNase L activity. Consequently, RNase L activation by the E3 construct may selectively deplete cytoplasmic host mRNA while sparing saRNA. This depletion reduces competition for ribosomes, thereby enhancing the translation of saRNA transcripts. In contrast, the E3-NSs-L* construct encodes Theiler’s virus L*, an RNase L inhibitor,⁴⁰ thereby preventing host mRNA degradation and forcing saRNA transcripts to compete directly with cellular mRNA for ribosomal access.

To determine whether RNase L activation contributes to the greater transgene expression observed with the E3 construct compared to E3-NSs-L*, we assessed RNase L activity by measuring ribosomal RNA (rRNA) cleavage, an indicator of RNase L activity.⁵⁵ RNA extracted from FLS transfected with the E3 and E3-NSs-L* constructs was evaluated with the RNA integrity number (RIN) algorithm, an automated Bayesian learning method for assessing rRNA integrity.⁵⁶ Cells transfected with E3 exhibited significantly lower rRNA integrity than those transfected with E3-NSs-L* (Fig. 3f). As prior studies have shown that RNase L activation primarily depletes host mRNA levels before extensive rRNA degradation,^50,52,57 our findings support the hypothesis that RNase L activation and subsequent reductions in cellular mRNA occur with the E3 construct but not with E3-NSs-L*.

Having addressed key saRNA-induced toxicities—including cell death, translation shutdown, and mRNA degradation—with the E3-NSs-L* construct, we next sought to design new saRNA constructs aimed at mitigating cytokine production. In these new constructs, E3-NSs-L*-EGFP was expressed via a single IRES using non-identical 2A peptides (Fig. 4a). On a second IRES, varying amounts of cellular inhibitors of inflammatory signaling were expressed. The control construct, called ‘moxBFP’, lacks cellular inhibitors of inflammatory signaling and instead expresses a blue fluorescent protein.

Given that activation of nuclear factor-кB (NF-κB) is a hallmark of viral infection and a master regulator of cytokine and chemokine induction,^58,59 we prioritized its inhibition using super repressor inhibitor of κBα (srIκBα). This dominant active form of IκBα cannot be phosphorylated and thus resists ubiquitination and degradation, forming a stable cytoplasmic pool of IκBα that prevents NF-κB nuclear translocation and downstream signaling.^60,61 We generated one construct expressing srIκBα alone (named ‘srIκBα’) and another expressing srIκBα in combination with Smad7 and suppressor of cytokine signaling 1 (SOCS1) using non-identical 2A peptides (named ‘srIκBα-Smad7-SOCS1’). Smad7 negatively regulates both transforming growth factor-β (TGF-β) and NF-κB signaling pathways,⁶² while SOCS1 inhibits type I, II, and III interferon (IFN) pathways as well as NF-κB signaling.^63–65

Using a bead-based immunoassay on FLS culture supernatants, we observed that native saRNA transfection induced cytokine responses associated with viral infection (Figs. 4b, S4a-m). Cytokine secretion was still apparent when dsRNA-sensing pathways were inhibited, although a reduction in tumor necrosis factor (TNF)-α, IFN-α, and IFN-β levels were observed with E3 expression. Similarly, C-X-C motif chemokine ligand 10 (CXCL10) levels were reduced with E3-NSs-L* expression. On the other hand, srIκBα co-expression broadly suppressed cytokine secretion, reducing all measured cytokines to levels comparable to mock-transfected FLS. The srIκBα-Smad7-SOCS1 construct did not further reduce cytokine levels beyond the suppression achieved by srIκBα alone.

We then examined the impact of inhibiting inflammatory signaling on cell number and transgene expression (Figs. 5a, 5b). Interestingly, the srIκBα construct caused both immediate and sustained reductions in cell number, as quantified by BioTracker signal, whereas the srIκBα-Smad7-SOCS1 construct maintained cell number comparable to mock transfection throughout the experiment (Fig. 5c). AUC analysis of the BioTracker data confirmed that srIκBα-Smad7-SOCS1 significantly mitigated the srIκBα-induced reduction in integrated BioTracker signal (Fig. S5a). Additional experiments using CellTag and calcein AM staining corroborated these findings, showing that srIκBα negatively impacted both cell number (Fig. S5b) and viability (Fig. S5c), effects that were prevented by co-expression of Smad7 and SOCS1.

All three constructs produced low levels of EGFP expression (Fig. 5d), which was expected given EGFP’s position as the fourth protein in a quad-cistronic 2A element (Fig. 4a). Previous studies have demonstrated that protein expression decreases significantly for downstream genes in such configurations.⁴¹ Notably, while the moxBFP and srIκBα constructs generated intermediate mScarlet3 levels, the srIκBα-Smad7-SOCS1 construct produced 2-to 5-fold higher mScarlet3 fluorescence compared to the others (Fig. 5e).

We next explored how srIκBα and srIκBα-Smad7-SOCS1 modulate translational regulation. Interestingly, srIκBα expression led to reductions in eIF2α phosphorylation, a change that was prevented by co-expression of Smad7 and SOCS1 (Fig. 6a). While neither construct affected total eIF2α levels or eIF4E phosphorylation compared to the moxBFP construct (Figs. 6b, 6c), srIκBα caused a significant reduction in total eIF4E levels (Fig. 6d). Given the critical role of eIF4E in cap-dependent translation, its depletion likely contributes to the poor transgene expression observed with srIκBα (Fig. 5e). These effects, combined with srIκBα-induced declines in cell number and viability (Figs. 5c, S5a-c), highlight the dual impact of srIκBα on translational machinery and cellular health. Importantly, Smad7 and SOCS1 co-expression preserved eIF4E levels, counteracting the disruption of cap-dependent translation caused by srIκBα.

Fibroblast activation protein-α (FAP-α) is a serine protease expressed on the surface of FLS that contributes to extracellular matrix degradation and tissue remodeling.^66,67 Its expression is minimal in normal adult FLS but increases significantly following inflammatory activation, such as in rheumatoid and osteoarthritis.^68,69 Since viral infections can induce fibroblast activation,^70,71 we hypothesized that prolonged saRNA transfection might similarly trigger FLS activation. To test this, we measured FAP-α levels using an in-cell western assay after 11 days of transfection. FLS transfected with moxBFP showed FAP-α levels comparable to those of mock-transfected cells, whereas srIκBα and srIκBα-Smad7-SOCS1 constructs resulted in significantly lower FAP-α levels compared to mock transfection (Figs. 7a, 7b).

These findings indicate that long-term saRNA transfection does not induce FLS activation, as measured by FAP-α levels, potentially alleviating concerns that saRNA might drive pathological fibroblast responses.

The srIκBα-Smad7-SOCS1 construct, which includes E3-NSs-L*, provides broad suppression of dsRNA sensing and inflammatory signaling pathways. This construct mitigates saRNA-induced cytokine release (Figs. 4b, S4a-m) and counters saRNA-and srIκBα-induced reductions in cell viability (Figs. 5c, S5a–c), enabling sustained cap-dependent transgene expression (Fig. 5e) without inducing fibroblast activation (Fig. 7b) or causing srIκBα-associated reductions in eIF4E levels (Fig. 6d). Together, these attributes position the srIκBα-Smad7-SOCS1 construct as a promising candidate for transient gene therapy applications.

To further enhance the therapeutic potential of this construct, we investigated the possibility of using a small-molecule drug to halt transgene expression. Achieving temporal control of transgene expression would enhance safety and flexibility in therapeutic contexts. The saRNA constructs in this study rely on the VEEV RdRp for self-amplification. VEEV is a mosquito-borne alphavirus capable of causing flu-like illness in humans, with the potential to progress to fatal encephalitis.⁷² Due to its pathogenicity, significant efforts have focused on developing small-molecule inhibitors of VEEV. One such compound is ML336, a potent inhibitor of the VEEV RdRp.⁷³

To test whether ML336 could inhibit transgene expression of the srIκBα-Smad7-SOCS1 construct, ML336 or vehicle was added to the culture medium one day post-transfection. EGFP and mScarlet3 fluorescence were monitored over 13 days (Fig. 8a), followed by calcein AM staining to assess cell viability (Fig. 8b).

ML336 treatment significantly reduced EGFP and mScarlet3 expression over time (Figs. 8c, 8d), demonstrating its efficacy in controlling saRNA transgene expression. While vehicle-treated cultures exhibited lower viability compared to mock-transfected cells—possibly due to reduced proliferation over the experiment’s duration—prolonged ML336 treatment did not reduce viability relative to mock-transfected controls (Fig. 8e).

This study aimed to develop and evaluate saRNA constructs that modulate innate immune responses, specifically targeting dsRNA sensing and inflammatory signaling pathways, to reduce saRNA-induced cytotoxicity, cytokine expression, and achieve sustained transgene expression. To this end, we engineered saRNA constructs that co-express multiple viral innate immune inhibitors and cellular inhibitors of inflammatory signaling. The constructs were designed to enable monitoring of cap-dependent and cap-independent translation using distinct fluorescent proteins.

The experimental approach included cloning the saRNA constructs using molecular biology techniques, followed by in vitro transcription and transfection into mouse primary FLS. Transfected cells were analyzed using fluorescence-based microplate assays to assess cell number, cell viability, phosphatidylserine exposure, and fluorescent protein expression. Further evaluations of translational control, FLS activation markers, RNA integrity, and cytokine secretion were conducted using in-cell western assays, capillary electrophoresis, and bead-based immunoassays. Finally, the ability to control saRNA transgene expression was explored using a small-molecule inhibitor targeting the VEEV RdRp.

The commercially available TagGFP2 E3L Simplicon vector (SCR725, Merck) served as the saRNA backbone in this study. After removing the TagGFP2-IRES-E3 sequence, various constructs were generated by combining different elements using restriction digests (FastDigest, Thermo Scientific), overlap extension PCR (Phusion, Thermo Scientific), and HiFi assembly (NEBuilder, New England Biolabs).

Genetic elements were sourced as follows: mScarlet3 from pDx_mScarlet3 (a gift from Dorus Gadella [Addgene plasmid #189754]),¹³⁸ IRES-EGFP from pIRES2-EGFP (Clontech), and IRES-E3 and IRES-PuroR (puromycin resistance) from the original TagGFP2 E3L Simplicon vector.

Custom gene synthesis (GeneArt, Thermo Scientific) was used to produce IRES-moxBFP, IRES-E3-T2A-NSs-P2A-L*-E2A-mEGFP, and IRES-srIкBα-P2A-Smad7-T2A-SOCS1 sequences. The IRES sequence was derived from pIRES2-EGFP (Clontech). Other sequences were derived from public databases: moxBFP (FPbase ID: SSTDU),¹³⁹ mEGFP (FPbase ID: QKFJN), E3 (UniProt: P21605-1), NSs (UniProt: P21699), L* (UniProt: P0DJX4), Smad7 (UniProt: O35253-1), and SOCS1 (UniProt: O35716). srIкBα was engineered from IкBα (UniProt: Q9Z1E3) by substituting serines 32 and 36 with alanines. All synthesized sequences were codon-optimized for mouse expression (GenSmart, GenScript).

Plasmids were cloned in DH5α competent Escherichia coli (High Efficiency, New England Biolabs) and purified by maxiprep (PureLink HiPure, Invitrogen). All plasmid sequences were verified using nanopore whole plasmid sequencing (Plasmidsaurus).

Plasmids were linearized using XbaI (FastDigest, Thermo Scientific) at 37°C for 3 hours. The linear plasmids were purified by phenol-chloroform extraction followed by sodium acetate ethanol precipitation. Uncapped RNA was synthesized in vitro using the T7 RiboMAX Large Scale RNA Production kit (Promega) at 37°C for 2 hours. After purification by ammonium acetate precipitation, the RNA was denatured at 65°C for 5 minutes. Cap-1 structures were then generated using the Vaccinia Capping System in conjunction with mRNA cap 2’-O-methyltransferase (both from New England Biolabs) for 45 minutes at 37°C. Following another ammonium acetate purification, the RNA was treated with Antarctic phosphatase (New England Biolabs) for 30 minutes at 37°C. After a final ammonium acetate purification step, the RNA was resuspended in THE RNA Storage Solution (Invitrogen), aliquoted, and stored at - 80°C until use.

All mice used in this study were 5- to 12-week-old wildtype C57BL/6J mice (Bicester, UK). Both male and female mice were included, though sex differences in response to saRNA transfection were not analyzed. Mice were housed in groups of up to five in a temperature-controlled room (21°C), with bedding, a red shelter, and enrichment materials provided. They were maintained on a 12-hour light/dark cycle, with food and water available ad libitum. Mice were euthanized by exposure to CO₂ gas in a rising concentration, and death confirmed by dislocation of the atlanto-occipital joint by applying pressure to the side of the neck using thumb and forefinger, procedures in line with the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012.

We cultured FLS from mouse knee joints following established protocols,¹⁴⁰ with slight modifications. In brief, knee joints were exposed by cutting through the overlying skin with dissecting scissors. The patellar tendon was then grasped near the patella with Dumont tweezers, and spring scissors were used to sever the patellar tendon along with surrounding tissues. Finally, the patella was excised by cutting the quadriceps tendon, and excess muscle and tendons were carefully trimmed away.

Both patellae from each animal were collected into a microcentrifuge tube containing ice-cold sterile phosphate buffered saline (PBS), then transferred to a 24-well tissue culture plate (Costar) with FLS culture media. The media consisted of DMEM/F-12 (Invitrogen) supplemented with 25% fetal bovine serum (Sigma), 1X GlutaMAX (Gibco), and 100 μg/ml Primocin (InvivoGen).

The patellae were maintained in a humidified incubator at 37°C with 5% CO2, with media changes every 1-2 days. FLS outgrowth was observed, and after approximately 10 days, when cells reached 70% confluency, the patellae were transferred to new wells for further FLS collection. Cells attached to the original wells were passaged using 0.1% trypsin-EDTA (Sigma) into a single well of a 6-well plate (Costar). Cells were later expanded into T25 flasks (BioLite, Thermo Scientific or CELLSTAR, Greiner Bio-One) and passaged upon reaching confluency. After the second passage, Versene solution (Gibco) was used for cell dissociation.

FLS were used between passages 3 and 8. In most experiments, cells from the same animal were divided and plated onto 6-well or 24-well plates, allowing for parallel manipulations with matched conditions. Occasionally, when cell number was low, cells from different animals were pooled before division. Once plated for experiments, media was changed every 2-3 days.

Cells were plated on poly-D-lysine-coated glass-bottom 35 mm dishes (P35GC-1.5-14-C, MatTek) and fixed with 4% paraformaldehyde for 10 minutes at room temperature without permeabilization. After fixation, cells were washed twice with PBS, followed by blocking with 10% normal goat serum in PBS for 1 hour at room temperature.

Cells were then incubated overnight at 4°C in blocking buffer, either without primary antibody (no primary control) or with cadherin-11 (extracellular) rabbit polyclonal antibody (1:200, DF3523, Affinity Biosciences). After three washes with PBS, cells were incubated for 1 hour at room temperature in blocking buffer with 1 μg/ml Hoechst 33342, BioTracker NIR680 (1:2000, Merck), and goat anti-rabbit Alexa Fluor 488 secondary antibody (1:500, Invitrogen).

After four additional PBS washes, cells were imaged in PBS using a Leica Stellaris 5 confocal microscope equipped with a 40X oil immersion objective. Tile-scanned images were stitched using Las X microscopy software (Leica) to generate high-resolution panoramic images.

BioTracker NIR680 was diluted 1:2000 in unsupplemented DMEM/F-12. T25 flasks containing FLS were washed once with HBSS (with calcium and magnesium, Gibco), then incubated with the diluted dye for 30 minutes at 37°C. After incubation, the flasks were washed three times with FLS media, with each wash lasting 10 minutes at 37°C. The FLS were then dissociated using Versene and plated onto 6-or 24-well plates.

For transfection in 6-well plates, the medium was removed and replaced with 1 ml of Opti-MEM I (Gibco). In a microcentrifuge tube, 500 ng of saRNA was diluted in 200 μl of Opti-MEM I. In a separate tube, 3 μl of Lipofectamine MessengerMAX (Invitrogen) was diluted in 100 μl of Opti-MEM I. After gently mixing, the complexes were incubated at room temperature for 5 minutes, then added dropwise to the cells. The cells were incubated with the complexes for 2 hours at 37°C, after which the media was removed and replaced with fresh FLS media. For transfection in 24-well plates, all volumes and amounts of saRNA were reduced fivefold.

Unless otherwise specified, black glass-bottom 6-well plates (P06-1.5H-N, Cellvis) were used for microplate imaging. Prior to imaging, the FLS media was replaced with Live Cell Imaging Solution (Invitrogen). Imaging was performed using the Odyssey M laser scanner (LI-COR), using LI-COR acquisition software with a plate offset of +1.45 mm and 100 μm resolution.

EGFP, mScarlet3, and BioTracker NIR680 were imaged using the 488, 520, and 700 channels, respectively. Spectral cross-excitation between EGFP and mScarlet3 in the 488 and 520 channels was corrected through linear unmixing analysis, as described below. Fluorescence intensity was quantified in ImageJ by measuring the integrated density within equal-area regions of interest for each well.

Fluorescence values (EGFP, mScarlet3, and BioTracker) were normalized by the day 0 BioTracker signal measured before transfection to account for variations in starting cell number.

saRNA constructs encoding individual fluorescent proteins (moxBFP, EGFP, or mScarlet3) were designed to assess cross-excitation among the fluorescent proteins used in this study. tSA201 cells (96121229, ECACC) were seeded in black glass-bottom 6-well plates coated with poly-D-lysine (Sigma) and transfected with the saRNA constructs. Cells were imaged the following day using an Odyssey M laser scanner, with fluorescence captured through the 488 and 520 channels. Expression of moxBFP was undetectable in both the 488 and 520 channels, and therefore it was excluded from further analysis.

Bleed-through of EGFP into the 520 channel and mScarlet3 into the 488 channel was quantified using ImageJ, with cross-excitation determined to be 11.32% and 0.94%, respectively. The emission signals were assumed to be linearly proportional to the sum of the intensities of each fluorophore, and the unmixed EGFP and mScarlet3 signals were calculated using Python.

FLS were stained with BioTracker NIR680 and plated on black glass-bottom 24-well plates (Sensoplate, Greiner Bio-One). Cells were transfected with saRNA or treated with 0.5 μM staurosporine (Cayman Chemical) as a positive control.

Annexin V-CF800 conjugate (Biotium) was diluted to 250 ng/ml in Live Cell Imaging Solution (Invitrogen). On each day of the assay, wells were washed with Live Cell Imaging Solution and replaced with diluted Annexin V-CF800 solution. Plates were incubated at 37°C for 15 minutes. Following incubation, plates were washed 3 times with Live Cell Imaging Solution before imaging the 700 and 800 channels with an Odyssey M imager set to +1.45 mm image offset and 100 μm resolution.

Image analysis was performed using ImageJ. To correct for unidirectional spectral bleed-through of BioTracker NIR680 into the 800-channel, subtractive compensation was applied by dividing the 700- channel image by a factor of 800 and then subtracting it from the 800-channel image. Due to the presence of a small number of high-intensity speckles in the 800-channel image, area rather than fluorescence intensity was quantified. The display range of the 800-channel image was set between 0.25 and 2.5 and the image was thresholded using the method of Li.¹⁴¹ The area of thresholded pixels within each well was then quantified and adjusted based on the BioTracker signal before transfection to account for variations in cell number. Data was then normalized to the average of the mock transfection condition.

Calcein AM (Invitrogen) staining was performed according to the manufacturer’s protocol. Briefly, cells were washed with HBSS and incubated with 2 μM calcein AM (diluted in HBSS) at 37°C for 30 minutes. Following three washes, cells were imaged in Live Cell Imaging Solution, and the 488-channel image was captured using an Odyssey M imager.

Although calcein AM and EGFP share overlapping spectra, calcein AM fluorescence was typically much higher than EGFP, making EGFP’s contribution to the measured Calcein AM signal negligible in most cases. However, when the E3 construct was used, EGFP expression was sufficient to interfere with the signal. To address this, in experiments that included use of the E3 construct, a 488-channel image was captured prior to calcein AM application and subtracted from the post-application image to accurately measure calcein AM fluorescence.

Fluorescence intensity was quantified using ImageJ. In experiments where BioTracker was used, fluorescence values were corrected by the pre-transfection BioTracker signal to account for differences in starting cell number.

Cells were plated on black, glass-bottom 24-well plates. Alongside mock and saRNA transfections, one well was reserved for background subtraction, which received no treatment. Unless specified otherwise, experiments were conducted 2 days post-transfection. Cells were fixed with 4% paraformaldehyde for 10 minutes at room temperature, followed by two washes with tris-buffered saline (TBS). Permeabilization was then performed using 0.1% Triton X-100 in TBS for 10 minutes, followed by two additional TBS washes. After permeabilization, cells were blocked with Intercept TBS Blocking Buffer (LI-COR) for 1 hour at room temperature with gentle agitation.

Primary antibodies were diluted in Intercept Blocking Buffer and incubated with the cells overnight at 4°C. The background subtraction well was incubated with blocking buffer alone, without primary antibodies.

The following primary antibodies were used: Phospho-eIF2α (Ser51) (D9G8) XP rabbit monoclonal (1:200, #3398, Cell Signaling Technology), eIF2α (L57A5) mouse monoclonal (1:200, #2103, Cell Signaling Technology), PKR rabbit polyclonal (1:500, #18244-1-AP, Proteintech), Phospho-eIF4E (S209) rabbit monoclonal (1:200, #ab76256, Abcam), eIF4E mouse monoclonal (5D11) (1:200, #MA1-089, Invitrogen), and fibroblast activation protein (73.3) mouse monoclonal (20 μg/ml, #BE0374, InVivoMAb). Except for the fibroblast activation protein antibody, which was conjugated to DyLight 800 using the manufacturer’s DyLight Antibody Labeling Kit (#53062, Thermo Scientific), all primary antibodies were unconjugated.

After washing the wells three times with TBS, cells were incubated for 1 hour at room temperature with the appropriate secondary antibodies or normalization stain, with gentle agitation. The secondary antibodies used were goat anti-mouse IRDye 800CW (1:800), donkey anti-rabbit IRDye 800CW (1:800), donkey anti-mouse IRDye 680RD (1:800), and CellTag 700 (1:500), all from LI-COR. The background subtraction well received secondary antibody but no CellTag. Following four additional washes with TBS, the plate was inverted and gently tapped on absorbent paper to remove excess liquid. The plate was then imaged using the Odyssey M imager with LI-COR Acquisition software using a plate offset of +1.45 and 100 μm resolution. Signal quantification was carried out using Empiria Studio software (LI-COR).

FLS were plated in 6-well plates and either transfected with saRNA or subjected to mock transfection. After 26-30 hours (allowing for saRNA replication but before substantial cell loss), cells were harvested by scraping in 350 μl RLT buffer (QIAGEN) supplemented with 10 μl/ml β-mercaptoethanol (Sigma). Cell lysates were homogenized using QIAshredder columns (QIAGEN), and total RNA was extracted using the RNeasy Mini Kit (QIAGEN) according to the manufacturer’s instructions. RNA quantity and quality were initially assessed using a N60 nanophotometer (Implen). Samples were aliquoted and stored at - 80°C until further analysis. For detailed quality assessment, samples were transported on dry ice and analyzed using the RNA Pico Kit on the 2100 Bioanalyzer system (Agilent), performed by Cambridge Genomic Services. RNA Integrity Number (RIN) values were determined for each sample.

FLS were stained with BioTracker and seeded onto 6-well plates. After transfection, the saRNA- lipofectamine complexes were replaced with 2 ml of fresh FLS media. Two days later, supernatants were collected and stored at-80°C until analysis. On the day of analysis, the supernatant was thawed on ice, and levels of 13 innate immune-related cytokines (IFN-γ, CXCL1, TNF-α, MCP-1, IL-12p70, CCL5, IL-1β, CXCL10, GM-CSF, IL-10, IFN-β, IFN-α, and IL-6) were measured using the LEGENDplex Mouse Anti-Virus Response Panel (740621, BioLegend) following the manufacturer’s protocol. Samples were run in duplicate. A custom 3D-printed PETG vacuum manifold (designed with Fusion 360) was used for vacuum filtration, and beads were analyzed using a CytoPLEX LX flow cytometer (Beckman Coulter). Data analysis was performed with the LEGENDplex Qognit Data Analysis Suite (BioLegend). To account for differences in cell number between wells, cytokine levels were corrected based on the BioTracker signal before transfection.

ML336 (Cayman Chemical) was dissolved in DMSO (Sigma) at a stock concentration of 10 mM, and aliquots were stored at-20°C. For treatments, ML336 was added to cell cultures at a final concentration of 1 μM, while vehicle control cultures received 0.01% DMSO.

All statistical analyses were conducted using GraphPad Prism 9, with specific tests detailed in the corresponding figure legends. All statistical tests were two-sided tests.