Messenger RNAs (mRNAs) are intermediate molecules that communicate the genetic information encoded in DNA to the protein-synthesizing machinery of the cell. Due to their transient nature of expression and the advances made in mRNA technologies, including modifications to the mRNA and improvements to the delivery vehicle, mRNAs are being explored as a novel modality for therapeutics of various kinds (for example, immuno-oncology, vaccines, protein replacement therapies, and genome-editing drugs). The simplicity of the approach, that is, synthesizing an mRNA in vitro that resembles an endogenous mRNA and encodes a protein of interest, followed by its delivery and expression in vitro or in vivo, makes it appealing and has broad applicability.

The application of in vitro–transcribed RNA for use as a therapeutic requires large amounts of functional RNA with low immunogenicity. The technologies used for the synthesis of these in vitro–transcribed mRNAs—predominantly using T7 phage RNA polymerase (T7 RNAP)—are well established. T7 RNAP transcribes RNA with high fidelity from a DNA template that contains the phage-enzyme-specific promoter. Although the process of synthesizing RNA from a DNA template using T7 RNAP is robust, previous studies have identified the presence of certain byproducts of the in vitro synthesis process that trigger cellular immune responses, including double-stranded RNA (dsRNA), which has been shown to be a major trigger of the immune pathway.

Therefore, when synthesizing mRNAs for in vivo applications that seek to minimize cellular immune responses, it is critical to either eliminate these dsRNA contaminants from the mRNA preparations or reduce their formation.

What are these dsRNA byproducts?

The production of dsRNA byproducts has been suggested to occur predominantly through two distinct mechanisms. In the first, the RNA transcript that has been synthesized by the T7 RNAP (runoff transcript, Figure 1) serves as a template for the RNA-dependent RNA polymerase activity of the T7 RNAP in subsequent rounds of transcription.1 If the 3’-end of the runoff transcript has sufficient complementarity (in cis), it will fold back and result in extension of the runoff transcript. The resulting RNA will be extended at the 3’ end and can be distinguished from the main transcript under denaturing conditions on a gel (Figure 2A).

DDTut_NewEnglandBiolabs_Fig1 - Revised
Figure 1. Schematic representation of possible mechanisms of dsRNA byproduct formation during in vitro transcription

Short RNA fragments, such as abortive transcripts, can anneal to complementary sequences in the runoff transcript (in trans) and also result in formation of dsRNA byproducts (Figure 1B). The identity and nature of the dsRNA products will vary depending on whether the extension occurs in cis or trans, and a strong correlation between RNA accumulation and formation of the spurious products is expected because the RNA that is synthesized likely rebinds to the polymerase to initiate extension.

The second mechanism proposed for the formation of dsRNA byproducts posits that the RNAP might switch to the nontemplate strand, resulting in an RNA molecule that is complementary to the runoff product but synthesized in a promoter-independent manner (Figure 1C).2 Because the size of the antisense molecule will be similar to the size of the runoff product, it cannot be distinguished by denaturing gel electrophoresis. Rather, analysis of the dsRNA byproducts formed due to the presence of an antisense RNA molecule will require native conditions (Figure 2B).

How to detect dsRNA byproducts in the reaction?

One of the most prominent tools used to detect dsRNA byproducts in the in vitro transcription (IVT) reaction is an immunoblot-based assay that depends on recognition of the byproducts in the reaction with a dsRNA-specific antibody. This type of assay has been used for a number of different RNA sequences and is a good indicator for the presence/absence of dsRNA byproducts in the IVT reaction (Figure 2C). However, an accurate quantification of the amount of dsRNA in the reaction is difficult to achieve with this assay. There are other assays that might allow for quantification of the dsRNA byproducts in the IVT reaction.

Drug Discovery Tutorial, New England Biolabs Figure 2
Figure 2. Methods for detection of dsRNA byproducts. (A) Denaturing gel electrophoresis can be used to separate and detect 3’-end-extended dsRNA byproducts. (B) Native gel electrophoresis to detect sense-antisense dsRNA byproducts. (C) Anti-dsRNA (J2) antibody recognizes long stretches of dsRNA regions (>40 base pairs) and is therefore useful for long RNA transcripts. RNase III is used as a control as it preferentially degrades dsRNA. (D) Intact mass spectrometry can be used to quantify abundance and lengths of the different 3’-end-extended dsRNA species produced in an IVT reaction. (E) Controls for dsRNA.

Both qualitative and quantitative assays rely on the recognition of the dsRNA byproducts by the antibody, and distinguishing between a bona fide dsRNA byproduct and an intrinsic secondary structure in the RNA can be an issue for these antibody-based assays. Moreover, careful investigation needs to be performed when using modified mRNAs because the presence of modifications can also alter recognition of the dsRNA structure by the antibody.

dsRNA-mediated immune responses are perceived to be mediated, in part, by binding of the dsRNA to the cytosolic immune receptor MDA5. MDA5 is an ATP-dependent helicase, and binding of MDA5 to the dsRNA byproducts results in ATP hydrolysis, which can be measured by standard biochemical assays. Cryo-electron microscopy (cryo-EM) structures of MDA5-dsRNA have shown the presence of filaments, and analyses of these filament structures have shown that each MDA5 molecule spans 14 or 15 RNA base pairs. However, access to facilities that would allow for these kinds of cryo-EM analyses might be limited; therefore, in vitro assays for ATP hydrolysis in the presence of MDA5 can be a good surrogate method to understand the extent of the immune response that will be generated from an mRNA of interest.

To gain a greater understanding of the exact sequences and nature of the dsRNA byproducts, RNA-seq2 and intact mass spectrometry analyses (Figure 2D) can be performed. However, ligation bias in RNA-seq experiments against (or potentially for) structured RNAs could impact these comparisons. Intact mass spectrometry analyses of the RNA 3’-end will not give sequence information but will provide a distribution of the heterogeneity seen in the RNA of choice. However, these approaches might not be readily available to all and cannot be performed in a high-throughput fashion especially when a large set of sequences is being screened.

While setting up the assays, the right controls should be used to make sure that the sensitivity of detection is optimal. Synthetic RNA (with/without modifications) or other cost-effective controls, such as dsRNA ladders from New England Biolabs and synthetic poly(I:C), can be used to ensure that the assay recognizes the dsRNA species (Figure 2E). When using an antibody-based assay, one must keep in mind that detection by the antibody is dependent on the length of the dsRNA region, and the sensitivity of the assay may need to be adjusted to make sure that the assay can detect these dsRNA byproducts.

Since the immunoassay cannot distinguish whether the dsRNA product is formed due to 3’-end extension or synthesis of an antisense molecule, it is recommended to implement more than one method to characterize the nature of the byproducts, because understanding the nature of the molecule can guide which approach to take to prevent the formation of the dsRNA byproduct in the IVT reaction (see next section).

How to reduce formation of dsRNA byproducts in an IVT reaction?

Several methods have been described to circumvent the effects of the dsRNA byproducts that are generated in an IVT reaction. Many of these methods rely on the removal of the byproducts after completion of the IVT reaction, with analytical purification being the most predominant approach. Because the dsRNA byproducts cannot easily be distinguished from the main IVT RNA product, they cannot be separated by size-exclusion chromatography, which is used to remove other contaminants or byproducts such as abortive transcripts. However, extensive ion-pair reversed-phase high-performance liquid chromatography (HPLC)-based purification has been shown to separate dsRNA byproducts from the main IVT RNA products (Figure 3A).3

Drug Discovery Tutorial, New England Biolabs Figure 3
Figure 3. Reduction of dsRNA byproducts in IVT reactions. (A) High performance liquid chromatography (HPLC) purification to separate the runoff RNA product (II) from the dsRNA contamination (III). (B) Performing in vitro transcription at higher temperatures (>48ºC) using thermostable RNA polymerases reduces 3’-end-extended dsRNA byproduct formation as seen with multiple detection approaches.

The longer retention time of the dsRNA byproducts in the HPLC purification process suggests that this method can be used to separate 3’-extended dsRNA byproducts from the main IVT RNA. Although efficient, this approach results in an additional step in the mRNA synthesis workflow, involves specialized instrumentation, is not compatible with scaling up the reaction, and impedes the cost effectiveness of the approach.

Additionally, it is not clear whether antisense dsRNA products can efficiently be separated by this HPLC-based method, and their separation may depend on the experimental conditions used. A cellulose-based chromatographic method for removal of dsRNA contaminants, based on the selective binding of dsRNA to cellulose, has also been reported.4 Even though this purification strategy can be cost effective, and the single-use nature of this approach can prevent carryover from a previous purification, it might not be suitable for all mRNA sequences, because some mRNAs might have a higher propensity to form secondary structures that could lower the rate of recovery.

The alternative approach to post-synthesis purification is to prevent formation of the dsRNA byproducts in the IVT reaction by altering the IVT reaction conditions. Lowering the magnesium levels in the IVT reaction has been suggested to reduce the formation of dsRNA byproducts (formed by synthesis of antisense RNA) for a few specific templates2; however, lowering the magnesium concentration in the reaction also affects the total yield of RNA, which is undesirable for applications where large quantities of mRNA are desired.

Recently it has been shown that addition of a DNA oligonucleotide complementary to the 3’-end of the runoff transcript can also prevent the formation of the 3’-extended dsRNA byproducts,5 but the removal of such an oligonucleotide after synthesis of a therapeutic mRNA can be challenging and will require an additional enzymatic step, which is undesirable when trying to streamline the workflow and limit production costs.

The use of a thermostable RNA polymerase, such as HiT7® from New England Biolabs, has been shown to reduce the formation of the 3’-extended dsRNA byproducts (Figure 3B) in the IVT reactions without affecting the overall yield or the integrity of the RNA, and it does not require additional enzymatic treatments, which could provide an alternative to current IVT reaction workflows.6,7

The method best suited for the removal/prevention of the dsRNA contaminants will depend on the final application and the scale of RNA yield desired. For applications where scaling up is a prerequisite, a post-synthesis purification step can impede the final outcome. A better approach for those is to prevent the formation of the dsRNA byproducts during the synthesis process.

Future perspectives

The use of synthetic mRNA as a drug requires the mRNA to be devoid of any contaminating RNA and to be synthesized in large amounts. Therefore, while choosing a suitable method to get rid of dsRNA byproducts for synthesis of your favorite mRNA molecule, early consideration should be given to the final application and scale of the IVT reaction. The nature of the dsRNA byproducts synthesized from the sequence of interest (3’-extended vs. antisense) will also be a factor in deciding which approach to reduce the generation of dsRNA will work better for a given template. For example, template-encoded poly-A tailing can circumvent the formation of some of the dsRNA byproducts, but not all, and high-temperature transcription can reduce 3’-end-extended dsRNA byproducts, but not the synthesis of antisense RNA.7 Using high-yield reaction conditions to synthesize greater amounts of RNA is tempting for applications where scaleup is desired.  However, because formation of some dsRNA byproducts can be enhanced in the presence of excess RNA, the enzyme:template:NTP conditions should be taken into account and optimized for each and every sequence to minimize the formation of these dsRNA byproducts. The nature of the sequences in the DNA template or the RNA molecule that have a higher propensity to form dsRNA byproducts is not fully understood.

A better understanding of the sequence specificities can aid rationalized design of mRNA 3’-ends to prevent formation of these contaminants in the reaction. Even minimal changes in template sequence and/or reaction conditions can potentially affect the extent of dsRNA byproduct formation and should be taken into consideration when designing/modifying a template sequence. Finally, detection of the dsRNA byproducts can be difficult when making mRNAs with modified NTPs, and at least more than one method should be implemented to characterize and quantitate the dsRNA byproducts in such cases.

In conclusion, efficient and cost-effective production of synthetic mRNAs for therapeutic purposes will require a thorough understanding of the nature of the final mRNA product and will depend on a synthesis process that minimizes the production of contaminants that would otherwise require costly purification approaches.


Bijoyita Roy, PhD (, is a staff scientist and Monica Z. Wu, PhD, serves as a research scientist in the RNA and Genome Editing division at New England Biolabs.

1. Gholamalipour Y, Karunanayake Mudiyanselage A, Martin CT. 3’ end additions by T7 RNA polymerase are RNA self-templated, distributive and diverse in character-RNA-Seq analyses. Nucleic Acids Res. 2018 Oct 12;46(18):9253-9263.
2. Mu X, Greenwald E, Ahmad S, Hur S. An origin of the immunogenicity of in vitro transcribed RNA. Nucleic Acids Res. 2018 Jun 1;46(10):5239-5249.
3. Karikó K, Muramatsu H, Ludwig J, Weissman D. Generating the optimal mRNA for therapy: HPLC purification eliminates immune activation and improves translation of nucleoside-modified, protein-encoding mRNA. Nucleic Acids Res. 2011 Nov;39(21):e142.
4. Baiersdörfer M, Boros G, Muramatsu H, Mahiny A, Vlatkovic I, Sahin U, Karikó K. A Facile Method for the Removal of dsRNA Contaminant from In Vitro-Transcribed mRNA. Mol Ther Nucleic Acids. 2019 Apr 15;15:26-35.
5. Gholamalipour Y, Johnson WC, Martin CT. Efficient inhibition of RNA self-primed extension by addition of competing 3’-capture DNA-improved RNA synthesis by T7 RNA polymerase. Nucleic Acids Res. 2019 Aug 8. pii: gkz700. doi: 10.1093/nar/gkz700.
6. Roy B, Robb GB. Use of thermostable RNA polymerases to produce RNAs having reduced immunogenicity. US Patent 10,034,951.
7. Wu MZ, Asahara H, Tzertzinis G, Roy B. Synthesis of low immunogenicity RNA with high-temperature in vitro transcription.

This site uses Akismet to reduce spam. Learn how your comment data is processed.