A peer-reviewed directed-evolution platform, an aptamer-based ultra-high-throughput screen, and a lead enzyme variant now supplying active mRNA, saRNA, and circRNA clinical programs — and the enzymatic engine behind Hzymes Biotech's full lightweight mRNA manufacturing solution.

Key Facts (At a Glance)

•     What was engineered: M30, a directed-evolution variant of T7 RNA polymerase (T7 RNAP), the enzyme used to manufacture therapeutic mRNA.

•     Who did it: A team led by Prof. Guang-Yu Yang — founder of Hzymes Biotech and a professor at Shanghai Jiao Tong University's School of Life Sciences and Biotechnology — developed M30 in his university laboratory and carried it into GMP-scale production through the company he founded.

•     Where it was published: Research, the Science Partner Journal published with AAAS — Qin W, Nie T, et al. "Directed Evolution of T7 RNA Polymerase Minimizes dsRNA By-product and Enables High-Fidelity mRNA Synthesis for Demanding Therapeutic Applications." Research. 2026;9:1172. DOI: 10.34133/research.1172.

•     Headline result: M30 shows roughly a 10-fold increase in catalytic efficiency over wild-type T7 RNAP at 37 °C, dsRNA by-product reduced to about 10% of wild-type levels at 37 °C and to roughly 0.1% at 50 °C, and 3′-end consistency raised from 5.6% to 60%.

•     How it was found: Ab-FADS, an aptamer-based fluorescence-activated droplet-sorting platform capable of screening up to 10⁸ enzyme variants per day at 99.4% sorting accuracy.

•     Where it's used today: GMP-scale manufacturing supplying active mRNA, self-amplifying RNA (saRNA), and circular RNA (circRNA) clinical development pipelines — and, as detailed below, the enzymatic core of Hzymes Biotech's lightweight mRNA production package.

Why This Enzyme Matters Right Now

mRNA is only as clean as the polymerase that transcribes it. Every dose of an mRNA vaccine or therapeutic — from a pandemic booster to a fully individualized cancer therapy built around a single patient's tumor mutations — starts with T7 RNA polymerase copying a DNA template into RNA in a test tube, a process called in vitro transcription (IVT).

That reaction has a well-documented flaw. T7 RNAP does not produce pure single-stranded mRNA; it also generates double-stranded RNA (dsRNA) as a by-product, through self-templated 3′ extension and promoter-independent antisense transcription. dsRNA is recognized by the innate immune system's pattern-recognition receptors — RIG-I, MDA5, TLR3, and PKR — triggering type I interferon responses, translational shutdown, and reduced protein expression from the intended mRNA. Karikó and colleagues showed in 2011 that removing dsRNA by HPLC could increase protein translation 10- to 1,000-fold in primary cells, which is why dsRNA control has shaped mRNA process design for over a decade.

Historically, the industry has answered this problem downstream, with chromatography. This article is about the alternative answer: not removing dsRNA after the fact, but engineering the polymerase so it forms less of it in the first place.

This article is the third in a connected series on personalized mRNA cancer vaccine manufacturing. The first, "34 Neoantigens, One Patient, One Batch: What Moderna and Merck's Phase 3 Win Demands of mRNA Manufacturing", examined why per-patient mRNA manufacturing — the model required by individualized neoantigen therapies like intismeran autogene (mRNA-4157/V940) — makes upstream dsRNA control an economic necessity, not just a quality nicety. The second, "Personalized Cancer Vaccine Race 2026: How Lightweight mRNA Manufacturing Cuts Production Time from 8 Weeks to Under 3", covered how plasmid-free DNA template methods (RCA/USTAT and PCR/SDT) compress the upstream template-prep bottleneck — and flagged that RCA-derived templates currently show higher dsRNA levels than the plasmid method, an impurity issue reducible only in part by linearization temperature tuning. This article picks up exactly that thread: it goes inside the peer-reviewed enzyme engineering that Hzymes founder Prof. Guang-Yu Yang built specifically to suppress dsRNA formation at its source, regardless of which template method feeds the reaction.

The Ab-FADS Platform: Screening 100 Million Variants a Day

Directed evolution works by generating large libraries of enzyme variants, screening them under conditions that mimic real use, and carrying the best performers into the next round. The bottleneck is almost always the screen: most laboratory assays can test thousands of variants, not hundreds of millions.

Prof. Yang's answer is Ab-FADS — an aptamer-based fluorescence-activated droplet-sorting platform. Each enzyme variant is compartmentalized in a microdroplet together with its reaction components; an RNA aptamer reports on transcription output as a fluorescent signal; and droplets are sorted at ultra-high throughput based on that signal. Reported performance:

Because droplets can be screened under non-standard conditions — elevated temperature, modified nucleotide chemistries — Ab-FADS selects for enzymes that work under the conditions of actual GMP manufacturing, not just idealized bench assays. That distinction matters: an enzyme optimized only for a standard 37 °C benchtop reaction can behave very differently once temperature, salt, or nucleotide chemistry shift for a real production process — including the elevated-temperature linearization step used in RCA/USTAT template workflows, discussed below.

Ab-FADS ultra-high-throughput screening platform

The M30 Mutant: What Four Rounds of Evolution Produced

Using energy calculations and cavity-volume analysis to build a combinatorial mutant library, Prof. Yang's team ran four rounds of directed evolution through the Ab-FADS platform, selecting simultaneously for catalytic activity, thermostability, and reduced dsRNA formation. The lead variant to emerge, M30, is reported to outperform wild-type T7 RNAP across every axis measured:

In animal studies cited in the published research, mRNA produced by M30-mediated IVT and delivered via lipid nanoparticles (LNPs) produced IL-6 and IFN-α levels close to baseline in mice — consistent with substantially lower activation of the same innate-immune sensors that unmodified-dsRNA transcripts trigger.

Four rounds of directed evolution generating mutant M30

M30-derived IVT products show reduced immunostimulatory effects

M30 binding affinity — enhanced DNA binding, reduced RNA rebinding

The Mechanism: "Strong DNA Binding, Weak RNA Rebinding"

Biophysical experiments and molecular dynamics simulations described in the paper indicate that M30 works by shifting the enzyme's binding behavior: it holds the DNA template more tightly while releasing RNA product more readily, rather than rebinding it. Because self-templated 3′ extension and antisense fold-back — two of the dominant known routes to dsRNA formation in T7 IVT — both depend on the polymerase rebinding RNA rather than releasing it, weakening that rebinding suppresses dsRNA formation at its mechanistic source, rather than removing it after transcription is complete.

This places M30 in the same conceptual family as other published low-dsRNA T7 engineering efforts — including Moderna's 2023 double-mutant, which raised 3′-end homogeneity from 6–12% (wild-type) to above 90% (Dousis et al., Nature Biotechnology, 2023) — while pursuing a distinct set of mutations discovered through an independent, ultra-high-throughput screening platform.

Closing the dsRNA Gap Left Open by Faster Template Methods

Our companion article on lightweight manufacturing profiled two plasmid-free DNA template methods now compressing personalized vaccine production: RCA/USTAT (Pfizer, 2026), which produces mRNA from circular template to finished product in 48 hours, and PCR/SDT (BioNTech, 2022), which builds a linear template in 24 hours. Both methods matched plasmid-derived mRNA on sequence fidelity (99.93% RNA-seq alignment) and functional T-cell activation. But the USTAT data disclosed one open gap: dsRNA levels from RCA-derived templates ran higher than the plasmid method, an impurity only partly addressed by optimizing linearization temperature.

That gap is precisely what M30 is engineered to close — because M30's dsRNA suppression operates on the transcription step itself, at the level of how the polymerase binds DNA and releases RNA, it applies regardless of which upstream template method (plasmid, RCA/USTAT, or PCR/SDT) generated the DNA feeding the reaction. Pairing a fast, plasmid-free template method with a low-dsRNA polymerase is, in effect, how the two halves of the personalized-vaccine manufacturing bottleneck — template speed and transcription purity — get solved together rather than traded off against each other.

From Bench to GMP: The Hzymes Product Line

The research has been carried into commercial manufacturing by Hzymes Biotech, the company Prof. Guang-Yu Yang founded to bring his laboratory's enzyme-engineering work to GMP scale. Four product lines, all built on the same Ab-FADS directed-evolution platform, are reported to be supplying active clinical pipelines:

1. Low-dsRNA T7 RNAP mutants (the M30 line).

Manufactured at GMP scale under a documented quality system, positioned for personalized cancer vaccines, gene-editing therapies, and high-dose mRNA therapeutic programs — the exact class of per-patient manufacturing profiled in our first companion article, and directly relevant to teams adopting the RCA/USTAT template method described in our second.

Hzymes proprietary library of T7 RNA polymerase variants

2. High-integrity T7 RNAP mutants.

A separate set of targeted substitutions aimed at reducing premature transcription termination rather than dsRNA specifically, reported to increase specific activity (U/mL) more than 10-fold over wild-type and to improve full-length transcript yield on difficult-to-transcribe and long RNA sequences, including self-amplifying RNA (saRNA).

High-specific-activity mutants deliver higher yield and improved integrity, including strong saRNA performance

3. Thermostable T7 RNAP variants for circRNA.

Engineered through multi-site mutation and directed evolution for a ~270-fold longer half-life and ~58-fold higher catalytic activity at 52 °C, with stable transcription reported across 37–55 °C. That thermal window matters specifically for circular RNA (circRNA): it allows one-step circRNA synthesis, where transcription and circularization happen in the same elevated-temperature reaction, instead of requiring a separate high-temperature circularization step — with the higher temperature also improving circularization efficiency.

4. The enzymatic core of Hzymes' lightweight mRNA production package.

As detailed in our second companion article, Hzymes' full lightweight solution — cell-free DNA amplification, a degenerate/enzymatic IVT process, rapid affinity chromatography kits (HBP005016, HBP005017), and pre-encapsulated LNPs — compresses total production time from the traditional 6–8 weeks to under 3 weeks. The Ab-FADS-derived T7 RNAP variants covered in this article are the enzymatic engine behind that package's IVT step, which is what allows the "degenerate/enzymatic IVT" component to be tuned per project without sacrificing the purity gains described here.

Beyond these product lines, Hzymes maintains a broader proprietary library of T7 RNAP variants optimized for different manufacturing needs, and supports the rest of the mRNA workflow — plasmid preparation, IVT process development, mRNA purification, LNP encapsulation, and analytical QC — along with CRO services, proof-of-concept development, and third-party quality testing for early-stage programs. Teams evaluating either plasmid-free template methods or downstream purification bottlenecks can request Hzymes' M30, high-integrity, or thermostable T7 RNAP variants, along with the rapid affinity chromatography kits and pre-encapsulated LNPs from the lightweight package, as a combined solution.

Hzymes' lightweight mRNA production package

Why an Engineered Enzyme, Not Just Better Chromatography

Our first companion article laid out the economic case in detail: a chromatography-based purification train typically recovers only 50–70% of transcribed mRNA, and every step in that train — resin cost, column changeover, cross-contamination validation — is a cost that recurs on every single batch. In a pandemic-scale campaign, that cost is amortized across billions of doses. In a personalized therapy, where every batch is one patient's unique sequence, it is paid again, in full, for every patient. Our second companion article added the time dimension: even a perfectly pure process is not viable per-patient if template preparation alone takes 3–4 weeks.

The research reviewed here addresses both constraints from the same root cause: dsRNA is a property of the polymerase and the template geometry, not merely of the downstream process or the template amplification method. An engineered enzyme that forms less dsRNA at the point of transcription — while also transcribing faster and tolerating the elevated temperatures used in fast template workflows — reduces the burden the purification train has to carry and removes a variable that fast template methods have not yet fully solved on their own.

Frequently Asked Questions

Who founded Hzymes Biotech?

Hzymes Biotech was founded by Prof. Guang-Yu Yang, a professor at Shanghai Jiao Tong University's School of Life Sciences and Biotechnology, to commercialize his laboratory's directed-evolution enzyme engineering — including the Ab-FADS screening platform and the M30 T7 RNA polymerase variant — for GMP-scale mRNA manufacturing.

What is T7 RNA polymerase used for?

T7 RNA polymerase is the enzyme most widely used to manufacture therapeutic and vaccine mRNA through in vitro transcription (IVT), copying a linearized DNA template into RNA in a cell-free reaction.

What is dsRNA and why is it a problem in mRNA manufacturing?

Double-stranded RNA (dsRNA) is an unwanted by-product of T7-mediated in vitro transcription. It is recognized by innate immune sensors (RIG-I, MDA5, TLR3, PKR), which can trigger interferon responses, reduce protein expression from the intended mRNA, and cause reactogenicity — making dsRNA control a core quality and safety concern in mRNA drug manufacturing.

What is the M30 mutant?

M30 is a directed-evolution variant of T7 RNA polymerase developed by Prof. Guang-Yu Yang, founder of Hzymes Biotech, at Shanghai Jiao Tong University. Published in Research (2026;9:1172), M30 shows roughly 10-fold higher catalytic efficiency than wild-type T7 RNAP, dsRNA by-product reduced to about 10% of wild-type at 37 °C and roughly 0.1% at 50 °C, and 3′-end consistency raised from 5.6% to 60%.

What is the Ab-FADS platform?

Ab-FADS (aptamer-based fluorescence-activated droplet sorting) is an ultra-high-throughput directed-evolution screening platform that can test up to 10⁸ enzyme variants per day at 99.4% sorting accuracy, including under non-standard conditions such as elevated temperature and modified nucleotides.

Does a low-dsRNA T7 RNA polymerase like M30 solve the dsRNA issue seen in RCA/USTAT template methods?

Published USTAT data shows RCA-derived DNA templates currently produce mRNA with higher dsRNA levels than plasmid-derived templates, an issue only partly addressed by optimizing linearization temperature. Because M30 suppresses dsRNA formation at the transcription step itself — independent of how the upstream DNA template was generated — pairing a low-dsRNA T7 RNAP such as M30 with a fast template method such as RCA/USTAT or PCR/SDT is one way to address both the speed and purity constraints of personalized mRNA manufacturing at once.

How is M30 different from Moderna's published low-dsRNA T7 RNA polymerase?

Both are engineered T7 RNAP variants aimed at the same problem — Moderna's 2023 double-mutant (Dousis et al., Nature Biotechnology) raised 3′-end homogeneity from 6–12% to above 90%. M30 was discovered independently through the Ab-FADS platform and is reported to combine dsRNA reduction with a roughly 10-fold catalytic efficiency gain and improved thermostability. Both are evidence that upstream enzyme engineering, rather than downstream purification alone, is becoming a primary route to low-dsRNA mRNA manufacturing.

Why does thermostability matter for circRNA synthesis?

Circular RNA (circRNA) synthesis benefits from elevated reaction temperatures, which improve circularization efficiency. A thermostable T7 RNAP that remains active and stable at 50–55 °C — such as Hzymes' thermostable variant, with a reported ~270-fold longer half-life and ~58-fold higher activity at 52 °C — enables one-step circRNA synthesis, combining transcription and circularization in a single reaction instead of a separate high-temperature step.

Disclosure: Prof. Guangyu Yang is the Founder of Hzymes Biotech, which participated in the research and commercialization of the T7 RNA polymerase variants discussed in this article. The underlying research on M30 and the Ab-FADS platform has been peer-reviewed and published in Research (see Reference 1). Product performance and application data presented in this article are based on Hzymes Biotech's experimental and product data.

References

1. Qin W, Nie T, Hei M, Li L, Pan Y, Luo M, Yang G-Y, et al. Directed Evolution of T7 RNA Polymerase Minimizes dsRNA By-product and Enables High-Fidelity mRNA Synthesis for Demanding Therapeutic Applications. Research. 2026;9:1172. DOI: 10.34133/research.1172 — also available via PMC12932865.

2. 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;39(21):e142. DOI: 10.1093/nar/gkr695.

3. Dousis A, Ravichandran K, Hobert EM, Moore MJ, Rabideau AE. An engineered T7 RNA polymerase that produces mRNA free of immunostimulatory byproducts. Nat Biotechnol. 2023;41:560–568. DOI: 10.1038/s41587-022-01525-6.

4. Gholamalipour Y, Karunanayake Mudiyanselage A, Martin CT. 3′ end additions by T7 RNA polymerase are RNA self-templated, distributive and diverse in character. Nucleic Acids Res. 2018;46(18):9253–9263. DOI: 10.1093/nar/gky796.

5. Triana-Alonso FJ, Dabrowski M, Wadzack J, Nierhaus KH. Self-coded 3′-extension of run-off transcripts produces aberrant products during in vitro transcription with T7 RNA polymerase. J Biol Chem. 1995;270(11):6298–6307. DOI: 10.1074/jbc.270.11.6298.

6. World Health Organization. Evaluation of the quality, safety and efficacy of messenger RNA vaccines for the prevention of infectious diseases: regulatory considerations. WHO Technical Report Series No. 1039, Annex 3. Geneva: WHO; 2021.

7. Ghosh S, Simms CL, Rohrer SD, et al. USTAT: Unified Sequential Template Amplification and Transcription — a fully synthetic mRNA manufacturing platform. npj Vaccines. 2026;11:112. DOI: 10.1038/s41541-026-01434-8.

8. de Mey W, De Schrijver P, Autaers D, et al. A synthetic DNA template for fast manufacturing of versatile single epitope mRNA. Molecular Therapy: Nucleic Acids. 2022;29:943–954. DOI: 10.1016/j.omtn.2022.08.021.

9. Hzymes Biotech. Company product information and manufacturer-supplied performance data. hzymesbiotech.com.

10. "34 Neoantigens, One Patient, One Batch: What Moderna and Merck's Phase 3 Win Demands of mRNA Manufacturing." Companion article 1. https://www.hzymesbiotech.com/articles/mrna-manufacturing-neoantigen-dsrna/.

11. "Personalized Cancer Vaccine Race 2026: How Lightweight mRNA Manufacturing Cuts Production Time from 8 Weeks to Under 3." Companion article 2. https://www.hzymesbiotech.com/articles/cancer-vaccine-race-2026-mrna-manufacturing-under-3-weeks/.