Removal of dsRNA byproducts using affinity chromatography

Abstract

Double-stranded RNA (dsRNA) molecules are immunogenic byproducts of in vitro transcription of single-stranded RNA (ssRNA). Removal of dsRNA from ssRNA is difficult because the byproducts have similar sizes, sequences, and charges to the desired ssRNA. Here, we describe a dsRNA-specific affinity resin that selectively removes dsRNA from ssRNA. Affinity purification reduced dsRNA levels by >100-fold, to as low as ∼0.00007% w/w of total mRNA, with no negative impact on RNA integrity. The purified RNA, synthesized with standard nucleotides, induced no inflammatory response in a reporter cell line assay designed to measure innate immune responses. Purified RNA induced greater protein expression and healthier cells. The immunogenicity of the affinity-purified RNA with standard nucleotides compares favorably to RNA synthesized with modified nucleotides and purified with cellulose or reverse-phase high-performance liquid chromatography (HPLC). dsRNA affinity purification provides a facile and scalable solution to the problem of immunogenic dsRNA byproducts in transcribed RNA. This approach will improve quality and safety of RNA vaccines and therapeutics.

Graphical abstract

Clark, Dillen, and colleagues demonstrate how a double-stranded RNA affinity resin can remove dsRNA byproducts from mRNA, eliminating all immunogenicity for mRNA prepared with unmodified nucleotides. This simple, effective dsRNA removal method is scalable for use in bioprocessing.

Introduction

Single-stranded RNA (ssRNA) is the constituent biomolecule in several novel vaccines and therapeutics. Immunostimulatory and therapeutic messenger RNA (mRNA), circular RNA (circRNA), and self-amplifying RNA (saRNA) are three types of ssRNAs used in biomedicines. The utility of mRNA vaccines was illustrated by the rapid development and deployment of mRNA vaccine in response to the COVID-19 pandemic. Currently, there are numerous clinical trials investigating the use of ssRNA for vaccines and therapeutic approaches,^1,2,3,4 which target a variety of diseases and inherited genetic conditions.

mRNA is a linear molecule with a 5′ cap that is required for immunotolerance and translation and a 3′ poly-A tail that increases stability and protects the mRNA from exonucleases. circRNAs are derived from ssRNA transcripts, and the circular topology eliminates the 5′ or 3′ termini that are susceptible to exonucleases. In place of the 5′ cap, circRNAs often use an internal ribosome entry site to initiate translation. saRNAs contain the 5′ cap and 3′ tail of mRNA, but are much longer (∼8–12 kb), as they encode viral RNA-dependent RNA polymerases that transcribe multiple copies of the target message.⁵ This self-amplification can improve and sustain target protein expression. ssRNA therapeutics are mostly formulated in lipid-nanoparticles, which enable the RNA payloads to disperse through an organism, and deliver the messages to the cytosol, the site of protein translation.

The presence of double-stranded RNA (dsRNA) byproducts in transcribed RNA is a major challenge for mRNA-, saRNA-, and circRNA-based therapeutics. Detection of dsRNA by endosomal Toll-like receptors (TLRs) or cytoplasmic RIG-1-like receptors leads to the expression of proinflammatory cytokines and type 1 interferons, resulting in the induction of an inflammatory response and inhibition of protein expression.⁶ Thus, maximal efficacy in vivo requires control of dsRNA during the manufacturing process. dsRNA byproducts are close in size to the primary ssRNA transcript,⁷ and in the case of loop-back 3′ hairpin extensions, likely contain poly-A tails.^8,9 As a result, dsRNA is not removed by size exclusion chromatography, tangential flow filtration, or poly-dT affinity chromatography.^10,11,12,13

Cellulose/ethanol purification and ion-pair RP-HPLC chromatography are two common methods for removal of dsRNA at manufacturing scale.¹⁴ Cellulose-based purification was first described as a tool to purify viral dsRNA from plant and fungal tissues.^15,16,17 More recently, the method was adapted to dsRNA removal from transcribed RNA.¹¹ Large amounts of cellulose slurry are required to process limited amounts of RNA (1 L of cellulose slurry can purify the dsRNA from approximately 120 mg of ssRNA¹¹), and the fibers are not compatible with the caustic agents utilized for in-process sanitization,¹⁸ limiting the scalability of this method to processes aimed at producing large quantities of clinical or commercial grade ssRNA.

Ion-pair RP-HPLC separates dsRNA from mRNA by isolating a sharp peak of mRNA, with the dsRNA species eluting in the post-peak tails.^12,19 Though RP-HPLC offers excellent separation of dsRNA from ssRNA, the scalability and safety of this process is complicated by the requirement of toxic and flammable solvents, elevated temperatures, and a low binding capacity of approximately 0.017 mg mRNA/mL resin.¹⁹

Both purification methods reduce dsRNA levels, as judged by immuno-dot blots with dsRNA-specific antibodies. However, the complete silencing of the inflammatory response in cell or animal models requires the use of modified nucleotides such as N1-methylpseudouridine in addition to cellulose or RP-HPLC purification.^11,12,20,21

Recently, the first dsRNA-specific affinity resin (AVIPure dsRNA Clear, Repligen Corporation) was released commercially. The resin contains a small, stable, protein-based ligand that demonstrates high affinity for the dsRNA double--helix, and low affinity for ssRNA. Here, we tested if this dsRNA-specific affinity resin could reduce dsRNA levels in transcribed RNA. We evaluated the affinity resin with RNAs produced with either standard in vitro transcription (IVT) protocols or IVT protocols optimized for production of low levels of dsRNA byproducts. We compared the affinity resin to cellulose and RP-HPLC and analyzed the performance of the resin with column runs and biochemical assays. Finally, we transfected affinity-purified RNAs into a reporter cell line and measured activation of the interferon-signaling pathway, target protein expression, and cellular health. Throughout our studies, we used non-modified nucleotides, and we discuss how our results compare to studies that utilized cellulose and RP-HPLC purification in combination with modified nucleotides.

Results

Comparing the dsRNA-specific affinity resin to cellulose and RP-HPLC

To measure the selectivity of the dsRNA affinity resin for dsRNA versus ssRNA, we performed batch-binding experiments utilizing a filter plate. The affinity resin was incubated with either commercial GFP mRNA (DasherGFP, see materials and methods) or a mixture of dsRNAs consisting of a commercial dsRNA ladder and a firefly luciferase duplex synthesized for this study (see methods). In parallel, samples were incubated with cellulose fibers and 16% ethanol in the previously described buffer.¹¹ Controls included an empty filter plate well and the base chromatography beads. After 1-h incubation, the samples were collected by centrifugation and analyzed qualitatively with gel electrophoresis (Figure 1A). Recovery of mRNA was 94% for the affinity resin and 73% for cellulose, based on the UV260 absorbance of the purified mRNA. In contrast, both cellulose and affinity resin bound all dsRNA species, as judged by their absence in the flow-through samples (Figure 1B). This result was confirmed with an immuno-dot blot using the dsRNA-specific J2 antibody (Figure 1C).²¹

Figure 1.

Comparison of dsRNA affinity resin, cellulose, and RP-HPLC

(A) Agarose gel of the GFP mRNA input, or flow-through fraction from blank beads, affinity resin, or cellulose. (B) PAGE of the input dsRNA mixtures, flow-through fraction from blank beads, affinity resin, or cellulose. (C) Immuno-dot blot of samples in B) with dsRNA specific J2 mAb and 100 ng RNA/spot. (D–F) Purification of GFP mRNA with cellulose, RP-HPLC, or affinity resin. (D) J2-immuno dot blot of purified RNAs, with 1 μg RNA/spot. (E) Agarose gel of purified RNAs, visualized with total RNA stain and UV. (F) Dual-color immuno-northern blot with total RNA in red channel, and J2-chemiluminescent signal in green channel. (G) Quantitative dsRNA ELISA analysis of the samples in (D–F). The % dsRNA is plotted on a logarithmic axis, with error bars representing 1 standard deviation. (H) Tabulation of ELISA results.

The double-stranded IVT byproducts constitute a very minor fraction of the total GFP mRNA, approximately 0.1%. Unlike the dsRNA mixture assayed in Figure 1B, dsRNA byproducts are not easily discernable with gel electrophoresis.⁷ To determine if the dsRNA affinity resin could remove trace levels of dsRNA byproducts from purified mRNA, we analyzed the dsRNA levels of the mRNA shown in Figure 1A. For further comparison, a portion of the RNA was purified by ion-pair RP-HPLC.¹² The RP-HPLC lane was underloaded due to the limited amounts of material that could be purified on an analytical-scale column (see materials and methods). An immuno-dot blot revealed complete removal of dsRNA byproducts by the affinity resin (Figure 1D). Relative to the dot blot presented in Figure 1C, which contains pure 100% dsRNA, the dsRNA levels are 0.08% for GFP mRNA, leading to a longer image exposure time and visualization of the faint dsRNA signals in the cellulose and RP-HPLC samples. An agarose gel confirmed that the mRNA remained intact throughout the various processing steps (Figure 1E). To analyze the size distribution of dsRNAs before and after purification, we performed an immuno-northern blot of the agarose gel⁷ (Figure 1F). The majority of the dsRNAs migrated above the GFP mRNA band, suggesting a larger molecular mass. The dsRNAs signals were markedly reduced or absent after purification with cellulose, RP-HPLC, or the affinity resin. To quantify the degree of removal, we utilized a commercially available ELISA kit (see materials and methods) that offered equal or better sensitivity than the J2 dot blot (Figure S1). Affinity purification reduced the dsRNA levels from 0.08% to 0.0003%, representing a 270-fold (2.4 log) reduction (Figures 1G and 1H). In comparison, the cellulose treatment reduced dsRNA by 13-fold and RP-HPLC by 58-fold.

Column runs with dsRNA affinity resin

Next, we evaluated the performance of the affinity resin in column runs with a 0.34-mL bed-volume column. The commercial GFP mRNA was diluted to 0.7 mg/mL in the column buffer (see methods), and the column was loaded to a challenge of 3.4 g GFP mRNA/L_resin. During sample application and chase, a plateau of UV255 signal was observed from 1.25–2.5 mls (Figure 2A), which tailed off as the sample application phase (0–2 mls) ended and a buffer chase began (2–3.5 mls). Following mRNA application, the bound dsRNA was eluted with 6 M guanidine hydrochloride (GuHCl), resulting in a sharp peak from 3.75–4.0 mls (Figure 2A). To determine if the affinity resin maintained performance after exposure to GuHCl, 3 additional purification cycles were performed with the same column, and the resulting chromatograms were indistinguishable.

Figure 2.

dsRNA affinity resin column chromatography

(A) Overlay of 4 cycles of GFP purification with dsRNA specific affinity resin. The flow through and elution phases are indicated. (B) Agarose gel of the input RNA, 4 flow-through samples, 4 elution samples, and 4 eluates subjected to a heat annealing step. (C) Immuno-northern blot of (B) with total RNA in red channel and J2-chemiluminescent signal in green channel. (D) Quantitative dsRNA ELISA analysis of the purification samples, and % recovery calculated with UV260.

Both input and flow-through samples ran as a single sharp band on an agarose gel (Figure 2B) visualized with a total-RNA stain. In contrast, the elution samples displayed a smear of lower molecular weight RNA in addition to the GFP mRNA band. We hypothesized that the low molecular weight fragments might represent dsRNA duplexes that were denatured by the 6 M GuHCl elution buffer. The elution fractions were annealed by heating to 75°C and cooling in the presence of 0.1 M NaCl, but this did not lead to appreciable hybridization.

An immuno-northern blot of the agarose gel revealed a diffuse band of high molecular weight dsRNA in the input which was completely absent in the flow-through (Figure 2C). The elution samples were enriched for the high molecular weight dsRNA observed in the input, which likely results from loop-back transcripts. Additionally, a very intense smear of smaller dsRNA was observed from the 150 base pairs (bp) dsRNA marker to the 3 kb ssRNA marker, overlapping with the low-molecular weight smear observed in Figure 2B.

The residual dsRNA content of the purified RNAs was 0.0013%–0.0016% w/w, representing an ∼84-fold reduction from the input containing 0.11% w/w dsRNA (Figure 2D). The elution contained 0.63% w/w dsRNA, a 6-fold increase relative to the input. For all 4 runs, approximately 82% of the input mRNA was recovered in the flow-through, and ∼19% recovered in the elution, suggesting some degree of non-specific binding.

Purification of mRNA synthesized with optimized IVT methods

The production of dsRNA byproducts can be controlled at the transcription stage.^{8,22,23,24,25} To determine how the affinity resin performed with RNA transcribed with conditions optimized for low dsRNA production, we prepared GFP mRNA using standard and optimized IVT protocols. Three independent replicates were prepared using each method. After initial ion-exchange purification, the samples were purified with the dsRNA affinity resin (Figure S3A) We selected ion-exchange purification to capture all dsRNAs produced during transcription, as oligo-dT purification might deplete abortive-transcript hybrids and other fragments that lack a poly-A tail. All RNAs remained intact through purification (Figures 3A,3F). No differences pre- and post-affinity purification were apparent on a non-denaturing agarose gel (Figure 3B) until contrast was enhanced, at which point faint high-molecular weight (HMW) bands could be visualized in the inputs (∼2,000 to >9,000 nts), and these bands disappeared in the column flow-through (Figure 3C). The high-molecular weight bands were not observed in the optimized IVT products, but a faint ∼2,400 nt band was reduced in intensity post-affinity purification. Immuno-northern blotting with the J2 antibody confirmed that the HMW dsRNAs were removed by affinity purification (Figure 3D). To visualize the byproducts in the optimized IVT RNA, twice the mass of optimized RNA was loaded in each lane and the contrast was manually enhanced (Figure 3D, right). The uncropped immuno-northern blots are presented in the supplemental data (Figure S2).

Figure 3.

Purification of mRNA produced with standard or optimized IVT

(A) Normalized capillary electrophoresis traces of RNA produced with standard or optimized IVT. Input, flow-through, and elution samples are plotted. (B) Agarose gel of samples pre- and post-affinity purification, visualized with automatic exposure and automatic contrast adjustment. (C) Agarose gel with manually enhanced contrast to show minor species. (D) Immuno-northern blot of the samples pre- and post-affinity resin. Due to lower dsRNA levels in the optimized IVT products, a greater mass of RNA was loaded and the contrast was manually enhanced. (E) Plot of all individual ELISA measurements from two assays, run on different days. (F) Quantitative dsRNA ELISA analysis of the standard and optimized replicates, and % integrity measurement calculated from (A).

The input, flow-through, and elution samples were analyzed with denaturing capillary electrophoresis (CE) (Figure 3A). At full scale, the curves of the input and flow-through samples were very similar. The elution fractions show a clear secondary peak at ∼2,400 nts, approximately twice the size of the primary 1,200 nt GFP peak. Rescaling the y axis to 15% reveals that the standard IVT elution fractions contain additional small peaks in the 500–900 nts range, minor peaks at 2,400 and 3,600 nts, and a broad peak between 4,000 and 10,000 nts (Figure S3B). For optimized IVT elution, the 500–900 nt peaks were not observed. The minor 2,400 and 3,600 peaks were observed, and the broad peak from 4,000 to 10,000 nts was less intense than the standard IVT elution (Figure S3C)

The mRNA transcribed with standard conditions contained 0.09% w/w dsRNA content (Figure 3F), matching the commercial GFP mRNA (Figure 2D). The mRNA transcribed with optimized conditions measured ∼0.016% w/w dsRNA, ∼6-fold lower than standard IVT. Affinity purification reduced the dsRNA levels by over 100-fold (Figures 3E and 3F), to ∼0.0006% w/w (standard IVT process) and ∼0.00007% (optimized IVT process). After affinity purification, the optimized IVT RNA had 8.3-fold lower dsRNA content than standard IVT RNA, and the fold-reduction values were greater for optimized IVT RNA (Figure 3F). All sample quantifications fell within the standard curve (Figure S1), including the purified, optimized RNA with 0.00007% w/w dsRNA, which corresponds to ∼1 ng/mL dsRNA content. The ELISA was highly reproducible, with consistent results across two ELISAs performed on different days (Figure 3E); 4–6 individual dilutions that fell on the standard curve are plotted.

Analytical chromatography of purified RNAs

As analytical chromatography methods can detect dsRNA impurities,^13,26we used high-performance size-exclusion chromatography (HP-SEC) and RP-HPLC to interrogate the RNAs pre- and post-affinity purification. For both RP-HPLC and HP-SEC, there were no obvious differences in any of the chromatograms (Figures S4A, S4D, S6A, and S6D). To look for subtle differences, the peaks were normalized and aligned using a graphing program. For the HP-SEC chromatograms, we observed that mRNA produced with standard IVT displayed a pre-peak shoulder that was absent post-affinity purification (Figure S4B). The mRNA from optimized IVT did not display the pre-peak shoulder (Figure S4E). For both transcription methods, a small peak at ∼1.83 mL was diminished post affinity resin (Figures S4C and S4F). Based on size standards (Figures S5A and S5B), we estimate this peak corresponds a mass of 3.5 kDa, or a ∼11 bp dsRNA (Figure S5C). For RP-HPLC, the differences were even more subtle (Figures S6A and S6D). Only a small area in the pre-peak region (2.5–3 min) was decreased after affinity purification (Figures S6B and S6E). No changes in the post-peak region, where longer dsRNAs are expected, were observed (Figures S6C and S6F).

Transfection of purified RNAs in interferon reporter cell line

To determine if the decrease in dsRNA content measured by our immunoanalytical methods correlated with reduced innate-immunogenicity in cells, we performed transfections with standard and optimized IVT mRNAs, pre- and post-affinity purification. The mRNAs were complexed with lipid based transfection reagents to deliver the various RNAs to the cytosol of A549-Dual reporter cells. The cells express a luciferase gene under the control of interferon response factors, which are activated by a signaling cascade that releases interferon in response to engagement of pattern recognition receptors such as RIG-I like receptors and TLRs.^27,28 One of these TLRs, TLR-3, specifically recognizes intracellular dsRNA. Poly I:C is an established TLR-3 ligand and provides a positive control for the assay. Negative control wells contained media only, and 3 replicate wells were prepared for each RNA sample.

Cells were assayed for luciferase activity at 24 and 48 h (Figure 4A). Unpurified standard IVT mRNA and the poly-IC positive control produced equally high luciferase activity at both time points. mRNA produced with the optimized IVT produced 80% less luciferase activity than standard IVT at 24 h and ∼40% less at 48 h.

Figure 4.

Transfection of standard and optimized RNA, with and without dsRNA affinity resin purification

(A–D) Three wells are transfected with each replicate RNA, and the mean values from all 9 resulting wells are plotted with the error bars representing 1 standard deviation. Non-treated cells (cells) are the negative control and poly-I:C (pIC)-treated cells are the positive control. (A) Interferon expression measured by luciferase reporter present in A549 cells. (B) GFP expression as mean fluorescence intensity measured with a flow cytometer. (C) % cells/debris measured with flow-cytometer. (D) % viability, measured with SYTOX RED viability dye and flow cytometer. ns non-significant, ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001.

The AVIPure dsRNA affinity resin reduced the inflammatory response of transfected RNA to baseline levels. There was no significant difference in luciferase activity between cells treated with media only, and those transfected with purified mRNA produced using an optimized IVT reaction. There was a significant ∼50% increase between cells treated with media only and those transfected with purified mRNA produced using the standard IVT reaction (Figure 4A).

The purified RNAs produced the highest per-cell GFP activity at the 24 h time point (Figure 4B). By 48 h, the GFP signals had leveled off for all cells except those transfected with optimized and purified RNA, which maintained significantly higher GFP expression at 48 h. Transfection efficiencies were ∼93%–96% for all cells (Figure S7).

Cell health correlated with dsRNA content of the transfected mRNA. Optimized IVT products produced healthier cells than standard IVT products, as judged by less cell debris (Figure 4C) and higher viability (Figure 4D). Cells transfected with purified mRNAs were even healthier, and no differences between standard and optimized IVT products were observed (Figures 4C and 4D).

Discussion

Validation of the dsRNA affinity resin

Here, we demonstrate that a dsRNA-specific affinity resin can remove dsRNA byproducts from transcribed RNA. We compared the affinity resin to two established dsRNA removal techniques, RP-HPLC and cellulose-ethanol purification, and demonstrate favorable performance of the affinity resin in terms of dsRNA clearance (Figure 1). We performed column runs with the affinity resin and observed consistent dsRNA removal and mRNA recovery in 4 purification cycles (Figure 2). Analysis of the dsRNA-enriched affinity column eluates by agarose gel and immuno-northern blot revealed a population of dsRNAs from approximately 150 to 1,000 bps in the commercial GFP sample (Figure 2C). This was surprising as only high-molecular weight dsRNA, migrating above the GFP band, was detected in the input RNA. Presumably, the smaller dsRNAs were below the limit of detection prior to the enrichment provided by binding and eluting to the affinity resin.⁷

Analysis of affinity column fractions with capillary electrophoresis

Due to the relatively minor fraction of dsRNA contained in the samples (∼0.02%–0.01% w/w, Figure 3F), it is not surprising that the differences between the input and flow-through are difficult to measure. The CE traces pre- and post-affinity purification nearly indistinguishable (Figure 3A). Only the elution fractions showed clearly discernable dsRNA peaks (Figures S3B and S3C). In the affinity column elution of the standard IVT RNA, smaller fragments in the 500–900 nt range were observed by CE (Figure S3B), possibly derived from extension of hybridized abortive transcripts. Affinity column eluates from both standard and optimized IVT displayed peaks at 2,400 and 3,600, possibly representing single and double loop-back transcripts. Broad peaks from ∼4,000 to 10,000 nt are observed in both elution fractions, and these could arise from multiple loop-back transcription events occurring with a single transcript. CE analysis clearly showed that integrity of the RNA was not impacted by affinity purification (Figure 3F).

Analysis of affinity column fractions with analytical chromatography

Other researchers have observed discrete pre-peaks for short abortive transcripts and post-peaks for long loop-back transcripts with CE and RP-HPLC.^13,26 However in this study, the RP-HPLC traces were mostly uninformative (Figures S5A and S5D), with the only notable difference being a small, broad pre-peak in the standard IVT RNA that decreased after affinity purification (Figures S3B and S3E). The post-peak region where long, immunogenic dsRNA species are expected to elute showed no differences (Figures S5C and S5F).¹² Smaller RNAs are expected to elute earlier in the acetonitrile gradient, larger RNAs later,^29,30 suggesting the very minor pre-peak difference may arise from dsRNAs that are smaller than the primary GFP peak.

Surprisingly, HP-SEC revealed two differences between the pre- and post-affinity resin samples. For the standard IVT RNA, a pre-peak shoulder disappeared after affinity purification (Figure S4B). This shoulder was not observed for the optimized samples (Figure S4E), and we speculate it may represent the smear that was observed in a contrast-enhanced agarose gel (Figure 3C). A small peak at ∼1.82 min was present in both standard and optimized RNAs, and it decreased after affinity purification (Figures S4C and S4F). This ∼11 bp peak could correspond to dsRNAs derived from abortive transcripts (Figure S5).

These results highlight the value of immunoanalytical methods, as dsRNA byproducts are extremely difficult to detect, characterize, and quantify using CE (Figure 3A), agarose gels (Figures 2 and 3), HP-SEC (Figure S4), or RP-HPLC (Figure S6). An additional benefit of the dsRNA affinity resin is the ability to concentrate dsRNA byproducts through bind/elute chromatography. For example, using CE, only the concentrated dsRNA eluates provided any information on the nature of the dsRNA byproduct sizes and distributions (Figures S2B and S2C). Similarly, the smaller dsRNAs in the commercial GFP were only detected in the concentrated eluates (Figure 3C). Analysis of enriched dsRNA byproducts can provide insights into the production of dsRNA at the transcription stage, and aid in the development of analytical and quality control assays.

Affinity purification in combination with optimized IVT

Optimization of IVT parameters and reagents can reduce dsRNA levels, improving the in vivo efficacy of the RNA products (8, 20, 22–25, 29). To test if the affinity resin could also purify feeds with low dsRNA content, we produced GFP mRNAs using standard IVT and IVT conditions optimized for low dsRNA production. Three replicate samples were prepared by each method, and consistent dsRNA levels were achieved across replicates through the optimized method, producing ∼80% less dsRNA than the standard method (Figure 3F). Affinity purification provided a greater-than 2-log reduction of dsRNA for both standard and optimized IVT RNAs. The dsRNA levels in the optimized IVT RNA were ∼8.3-fold lower than the standard IVT RNA, and this minor 8.3-fold difference was clearly evident in cell-based assays (see section “Standard nucleotide mRNA purified with affinity chromatography compared to modified nucleotide mRNA purified with solvent-based methods”).

We transfected the various RNAs into A549-Dual cells, a reporter cell line that produces luciferase in response to activation of the interferon-signaling pathway. Affinity-purification completely eliminated any inflammatory potential of transfected RNA (optimized IVT, Figure 4A), with no significant difference in luciferase relative to cells treated with media only. Affinity-purified standard IVT RNA increased luciferase by ∼30% at 24 h and ∼50% at 48 h, which is consistent with the ∼8.3-fold higher residual dsRNA levels measured by ELISA (Figures 3E and 3F). Unpurified standard IVT RNA increased luciferase activity by ∼100-fold (relative to untreated cells). This result suggests that combining optimized transcription with dsRNA removal is effective for producing non-immunogenic RNA.

Protein expression was highest in the samples with the lowest interferon signal (24 h, Figure 4B). At 48 h, only the optimized and purified RNA maintained a significant increase in GFP activity. Lower in vivo immunogenicity correlated with increased efficacy (Figure 4B) and lower toxicity (Figures 4C and 4D).

Standard nucleotide mRNA purified with affinity chromatography compared to modified nucleotide mRNA purified with solvent-based methods

The results obtained here with standard ribonucleotides and affinity purification compare favorably to those obtained with modified ribonucleotides and RP-HPLC purification. For example, in previous studies, when RNA was synthesized with standard nucleotides, purified with RP-HPLC, and transfected into dendritic cells, very high tumor necrosis factor alpha and interferon alpha titers were measured, despite a reduction in J2 dot blot signal. Complete reduction of immunogenicity was only achieved by combining RP-HPLC and modified ribonucleotides.¹² Similarly, using mouse models, complete elimination of immunogenicity was achieved only when cellulose purification was combined with modified ribonucleotides.¹¹ In contrast, we achieved a complete elimination of innate immunogenicity without modified nucleotides by combining IVT optimization and dsRNA-specific affinity purification.

Our results suggest that dsRNA affinity resins may reduce or eliminate the need for modified nucleotides. Modified nucleotides increase the cost of mRNA therapeutics and may have unwanted consequences. A recent study reported that N1-methylpseudouridine nucleotides caused frame shifting and expression of immunogenic non-sense transcripts.³¹ Beyond immunogenicity concerns, specific combinations of modified nucleotides and lipid encapsulants have different effects in vivo that are not directly related to inflammation, so there may be additional considerations related to nucleotide choice for a given vaccine or therapeutic.³²

Conclusions

Affinity removal of dsRNA requires no heat or organic solvents, is highly scalable, and compatible with standard bioprocessing operations. dsRNA affinity chromatography represents a viable solution for removal of dsRNA byproducts from transcribed RNA. However, even with dsRNA affinity chromatography, optimization of transcription remains crucial to control immunogenicity. Only the combination of optimized IVT and affinity purification completely eliminated the inflammatory response in reporter cells (Figure 4A).

Materials and methods

Gel electrophoresis and immuno-northern blotting

Agarose gel and immuno-northern blot techniques were described previously in detail.⁷ Agarose gels (1.5%) were prepared in 1× Tris-borate-EDTA (TBE) buffer, with SybrSafe total nucleic acid stain added to gel matrix (Thermo Fisher Scientific, S33102). Samples were prepared in DNA loading dye (NEB B7025S) as shown in Figure 1 and in RNA loading dye with formamide (NEB B0363S) as shown in Figures 2 and 3. Samples were not heated prior to electrophoresis. dsRNA and ssRNA ladders (NEB N0362S and N0363S, respectively) were loaded on the gels. Precast 10% TBE gels were used in Figure 1B (Thermo Fisher Scientific, EC62755BOX) and stained post-separation with SybrSafe in 1× TBE.

After imaging agarose gels, they were transferred to positively charged nylon membranes (Cytivia, 10416296) by wet-tank transfer (Galileo Reflection Maxi Tank) in 0.5× TBE buffer. Blots were blocked with commercial buffers (Bio-Rad EveryBlot, 12010020), incubated with J2-HRP conjugated antibody (Novus, NBP3-11395H), washed with 1× tris-buffered sailine with Tween-20 (TBST), and incubated with SuperSignal West Atto substrate (Thermo Fisher Scientific, A38555). A dual-color image was obtained by capturing the chemiluminescent signal and total RNA signal (from residual SybrSafe stain) using a Bio-Rad ChemiDoc imaging system.

Small-scale testing with filter-plates

Dasher GFP mRNA was purchased (Aldevron, 3870) and diluted to 0.5 mg/mL in 1× STE buffer (10 mM Tris pH 8, 100 mM NaCl, and 1 mM EDTA) or cellulose binding buffer (10 mM HEPES pH 7.2, 0.1 mM EDTA, and 16% ethanol). Cellulose fibers (Sigma-Aldrich, C6288) were prepared as described.¹¹ AVIPure dsRNA Clear resin (10 μL) (Repligen, 100RNA-5), blank beads (Tantti, GP22121V03), or 20 μL of prepared cellulose fibers were added to each well of a filter plate (Millipore, MSGVS2210). After washing and equilibrating with the appropriate buffers, 100 μL of GFP mRNA was added, the plate was incubated for 60 min with shaking, and then the flow-through samples were collected through centrifugation. The dsRNA mixtures (Figure 1B) consisted of 0.02 mg/mL dsRNA ladder (NEB N0363S) and 0.02 mg/mL annealed firefly luciferase duplex,⁷ otherwise the experiment was performed as described for GFP mRNA.

dsRNA immunoanalytics

Immuno-dot blots were prepared by pipetting 2 μL samples of RNA directly onto a nylon membrane and processing as described for immuno-northern blots. ELISA measurements utilized a commercial kit (Vazyme EasyAna dsRNA kit, DD3509EN-01), used as described in the product manual, except for a shortened signal development step. Several dilutions of the analyte RNAs were prepared, and all dilutions that fell on the standard curve were used for quantification and calculation of coefficient of variation.

mRNA synthesis

eGFP mRNA was transcribed in vitro from a linearized plasmid DNA template also encoding 5′ and 3′ untranslated regions and polyadenosine tail. Reactions were performed using wild-type T7 RNA polymerase and an NTP mix consisting of equimolar amounts of ATP, CTP, GTP, and unmodified UTP. Standard and optimized reactions comprised linear pDNA template, T7 RNA polymerase, MgCl2, DTT, NTPs, inorganic pyrophosphatase, and cap analog at differing ratios. Co-transcriptional capping was performed using CleanCap AG (Trilink). After IVT, the mRNAs were purified by ion-exchange chromatography. The mRNA was buffer-exchanged to water, filtered using a 0.2-μm syringe filter, and stored at −20°C until further use.

dsRNA affinity chromatography

Polishing chromatography was performed on either a Bio-Rad NGC (Figure 2A), or Cytiva ÄKTA Pure (Figure S3A) chromatography system using a column packed with AVIPure dsRNA resin (Repligen 100RNA-5). The column used for commercial GFP purification was a 0.35-mL bed volume; 0.3 × 5 cm column; the equilibration buffer was 25 mM HEPES, pH 7.2, 0.625 M NaCl, and 2.5 mM EDTA. The RNA was prepared at 0.7 mg/mL in running buffer, and a 2 mL sample was applied with a dynamic loop for a total challenge of 3.4 mg mRNA/mL resin, with a 1 min residence time (0.35 mL/min).

For standard and optimized IVT RNA, the column was a 1 mL 0.5 × 5 cm prepacked column (Repligen, 23052206), and the equilibration/wash buffer was 50 mM HEPES, 2 mM EDTA, 700 mM NaCl, and pH 7.2. The RNA was prepared at 0.5 mg/mL in running buffer, and a 4 mL sample was loaded onto the column for a total challenge of 2 mg mRNA/mL resin, with a 1 min residence time (1 mL/min).

After sample loading, the columns were washed with equilibration buffer until absorbance at 260 nm had returned to a baseline level. dsRNA was eluted from the column with 6 M guanidine hydrochloride, followed by re-equilibration with buffer. Flow-through and elution fractions were collected and buffer-exchanged to water for injection using Amicon Ultra-15 (Merck) centrifugal filter units. The RNA concentration was adjusted to ±1 mg/mL using water for injection and stored at −20°C until further use.

mRNA integrity

mRNA integrity was assessed using capillary gel electrophoresis on Fragment Analyzer (Agilent) as per the manufacturer’s instructions. Percentage mRNA purity was determined by performing AUC-analysis on the obtained electropherograms.

Analytical chromatography

For HP-SEC, a bioZen 1.8 μm dSEC-2, 200A column, 150 × 4.6 mm (Phenomex, 00F-4787-E0) was used. The mobile phase was 100 mM sodium phosphate, 200 mM sodium chloride, pH 6.8, and the flow rate was 0.5 mL/min. The molecular weight standards were Advance Bio 130 (Agilent, 5190-9416) and BEH200 (Waters, 186006518). For ion-pair RP-HPLC, a bioZen 1.7 μm Oligo 100 × 2.1 (Phenomex, 00D-4791-AN) column was used. Mobile phases were A: 100 mM triethylamine acetate, pH 7.0, and B: acetonitrile. The gradient from 5% to 20% B occurred over 10 min, and the column was heated to 60°C.

The resulting UV260 versus time data were exported as a .csv file and plotted in GraphPad Prism. The chromatograms were normalized, and the peaks were manually aligned to aid in comparison.

For RP-HPLC purification of commercial GFP (Figure 1), 3 purification cycles were performed with 10 μL injections of 1 mg/mL RNA. The main peaks were collected, pooled, and desalted with a silica spin column (NEB, Monarch T2050L).

In vitro assays

A549-Dual reporter (Invivogen) cells were transfected with eGFP mRNA and evaluated at 24 and 48 h post transfection for interferon response factor activation and green fluorescence. mRNA samples were complexed with Lipofectamine MessengerMax (Thermo Fisher Scientific) reagent in a 1 μg mRNA: 3 μL MM ratio per sample (100 ng/well). The complex incubated for 5 min and 20μL was added to a flat bottom 96-well plate. As a positive control, 100 ng/mL Poly IC (high MW) was complexed with Lipofectamine MessengerMax and added to the relevant wells at 20 ng/well. To this plate, 50,000 cells/well in 180 μL complete medium were seeded on top of the MM-RNA complex and placed in an incubator at 37°C and 5% CO₂.

After 24 and 48 h, 150 μL of supernatant was collected and centrifuged. The supernatant sample (20 μL) was transferred to a white 96-well plate; 50μL of 1× QuantiLuc detection solution per condition was added and luminescence was readout immediately. The cellular material was trypsinized and transferred to a 96-well U-bottom plate, where it was resuspended in 200μL fluorescence-activated cell sorting buffer containing SYTOX Red viability dye (Thermo Fisher Scientific). The resuspended cells were then analyzed on the flow cytometer (CytoFLEX V5-B5-R3, Beckman Coulter) for determining cell viability and GFP expression.

Data availability

All experimental data are available upon request.

Acknowledgments

This work was funded by the author’s employers, Repligen and etherna.

Author contributions

N.E.C., M.K., S.D.A., J.V.d.H., M.R.S., R.A.W., and S.D. performed experiments and analyzed data. N.E.C. prepared figures and drafted the manuscript. All authors designed experiments and edited the manuscript.

Declaration of interests

All authors are employees of Repligen (N.E.C., M.R.S., R.A.W., K.K., and T.C.S.) or etherna (M.K., S.D.A., J.V.d.H., and S.D.). Repligen manufactures and sells a variety of bioprocessing products, and etherna is an industrial RNA producer.