mRNA Display: Evolving Peptides in a Test Tube
Peptide Discovery Technologies
10¹² unique sequences
mRNA display generates libraries exceeding one trillion peptide sequences, roughly 1,000 times larger than phage display libraries.
Newton et al., ACS Synthetic Biology, 2020
Newton et al., ACS Synthetic Biology, 2020
View as imageIn 1997, Richard Roberts and Jack Szostak demonstrated that attaching puromycin to the 3' end of an mRNA molecule could create a covalent bond between the mRNA and the peptide it encodes.[1] That single chemical trick created mRNA display, a technology that now screens libraries of over one trillion unique peptide sequences against biological targets in a single experiment. Where phage display relies on bacteria and viruses to present peptide libraries, mRNA display happens entirely in vitro. This freedom from cellular biology allows larger libraries, harsher selection conditions, and the incorporation of amino acids that cells cannot produce. The technology has become the primary discovery engine for macrocyclic peptides, a drug class positioned between small molecules and biologics. For the broader context of display-based peptide screening, see our pillar article on combinatorial peptide libraries.
Key Takeaways
- Roberts and Szostak invented mRNA display in 1997 by linking puromycin to mRNA, creating covalent fusions between peptides and their genetic code (PNAS, 1997)
- mRNA display creates libraries exceeding 10^12 unique peptide sequences, roughly 1,000 times the diversity ceiling of phage display (Newton et al., ACS Synthetic Biology, 2020)
- The RaPID system combines mRNA display with flexizyme-based genetic code reprogramming to produce macrocyclic peptides containing unnatural amino acids (Goto and Suga, Accounts of Chemical Research, 2021)
- Alleyne et al. used mRNA display to discover cyclic peptide PCSK9 inhibitors with single-digit nanomolar binding affinity, rivaling monoclonal antibodies (2020)
- Hurd et al. demonstrated mRNA display selection directly in mammalian cell lysates, eliminating the need for purified target protein (Angewandte Chemie, 2024)
- Zilucoplan, a macrocyclic peptide discovered via mRNA display, received FDA approval in 2023 for generalized myasthenia gravis
The invention of mRNA display
The concept behind mRNA display is straightforward: create a physical link between each peptide and the mRNA molecule that encodes it. Roberts and Szostak achieved this by attaching puromycin, an antibiotic that mimics aminoacyl-tRNA, to the 3' end of synthetic mRNA molecules.[1] During in vitro translation, the ribosome reads the mRNA normally until it reaches the puromycin at the end. At that point, puromycin enters the ribosomal A site and forms a covalent bond with the growing peptide chain. The result: a single molecule containing both the peptide (the functional domain) and the mRNA that encodes it (the informational domain).
This genotype-phenotype linkage is the foundation of the entire technology. Because the bond is covalent, the fusion molecule is stable enough to survive the stringent washing conditions used during affinity selection. That stability distinguishes mRNA display from ribosome display, where the linkage depends on the more fragile ribosome-mRNA-peptide ternary complex.
How a selection cycle works
A complete mRNA display experiment follows five steps, repeated across multiple rounds to enrich for the best binders.
Library construction. Researchers synthesize a DNA library with randomized codons at the positions encoding the peptide's variable region. A typical library encodes peptides of 6 to 15 amino acids with randomized sequences, flanked by constant primer regions. The DNA is transcribed into mRNA.
Puromycin ligation. A flexible DNA-puromycin linker is ligated to the 3' end of each mRNA. The linker must be long and flexible enough for the puromycin to reach into the ribosomal A site after the last codon is translated. Getting this linker chemistry right was one of the key technical achievements of the original 1997 work.[1]
In vitro translation. The mRNA-puromycin constructs enter a cell-free translation system, typically rabbit reticulocyte lysate or a reconstituted E. coli system (the PURE system). Translation proceeds normally until the ribosome stalls at the mRNA-puromycin junction. Puromycin then bonds to the nascent peptide, creating the mRNA-peptide fusion. Each molecule in the library now physically links its code to its function.
Affinity selection. The fusions are exposed to an immobilized target protein. Peptides that bind the target are retained; everything else washes away. Because there are no living cells involved, selection can occur under conditions that would be lethal to phage: high salt, detergents, extreme pH, elevated temperature, or in the presence of competitive binders.[2]
Amplification. Surviving fusions are reverse-transcribed to cDNA and amplified by PCR. The enriched pool is transcribed back into mRNA, and the cycle repeats. After 3 to 8 rounds, the library converges on high-affinity sequences. Individual clones are then sequenced and characterized.
Why library diversity changes the math
The most cited advantage of mRNA display is library size. Newton et al. (2020) summarized the case: mRNA display routinely generates libraries of 10^12 to 10^13 unique sequences, while phage display is limited to approximately 10^9 because the library must pass through bacterial transformation.[2]
That 1,000-fold difference has concrete consequences. A 10-amino-acid peptide has 20^10, or roughly 10^13, possible sequences. A phage library of 10^9 members samples less than 0.01% of this theoretical space. An mRNA display library of 10^12 members samples approximately 10%. The probability of finding a rare, high-affinity binder scales with the fraction of sequence space explored. Experiments comparing the two technologies side by side have confirmed this: mRNA display selections against the same target protein yield peptides with binding affinities three to four orders of magnitude tighter than those from phage display selections using shorter peptide libraries.[2]
Schumacher et al. (1996) illustrated the importance of library diversity through mirror-image phage display, where even modest increases in library size improved the identification of D-peptide ligands.[3] mRNA display removed the cell-based diversity ceiling entirely.
Incorporating unnatural amino acids
Natural proteins use only 20 amino acids. mRNA display operates in vitro, which means the translation machinery can be engineered to accept non-standard building blocks. Ma and Hartman (2012) demonstrated that the E. coli PURE reconstituted translation system could incorporate unnatural amino acids into mRNA-displayed peptide libraries for in vitro selection.[4]
This capability matters because unnatural amino acids can improve a peptide's drug-like properties: D-amino acids resist protease degradation, N-methylated residues enhance membrane permeability, and non-proteinogenic building blocks create conformational constraints. A peptide containing these modifications is more likely to survive in the bloodstream and cross cell membranes than one built entirely from natural amino acids.
The most advanced application of this concept is the RaPID system, described in detail below.
The RaPID system: genetic code reprogramming meets mRNA display
Hiroaki Suga's RaPID (Random nonstandard Peptides Integrated Discovery) system, developed at the University of Tokyo, represents the most commercially successful implementation of mRNA display. It combines the technology with flexizyme-based genetic code reprogramming.
Flexizymes are artificial ribozymes that charge tRNAs with virtually any amino acid, natural or not. The FIT (flexible in vitro translation) system uses reprogrammed flexizymes to assign unnatural amino acids to specific codons. By coupling FIT with mRNA display, the RaPID system produces libraries of macrocyclic peptides that contain non-proteinogenic building blocks, each linked to its encoding mRNA.[5]
Goto and Suga (2021) described the products as "pseudo-natural macrocyclic peptides" because they are synthesized by the ribosome but contain structural features absent from nature: D-amino acids, N-methylated residues, beta-amino acids, and cyclization through thioether bonds formed between N-terminal chloroacetyl groups and internal cysteine residues.[6]
Passioura and Suga (2017) reviewed the early portfolio of macrocyclic peptide modulators discovered through RaPID, including inhibitors of protein-protein interactions, enzyme inhibitors, and receptor modulators across a range of therapeutic targets.[7] Tsiamantas et al. (2019) published detailed protocols for RaPID-based selection, cataloging the expanding list of functional macrocyclic peptides discovered through the platform.[5]
A typical RaPID screen starts with a library of more than 10^12 macrocyclic peptides and, within two weeks, converges on leads with nanomolar or sub-nanomolar affinity for the target. Huang et al. (2019) provided a comprehensive review of how RNA display methods, including RaPID, have contributed to macrocyclic peptide drug discovery across dozens of challenging targets.[8]
What mRNA display has discovered
PCSK9 inhibitors for cholesterol reduction
Alleyne et al. (2020) used mRNA display to discover a series of cyclic peptide PCSK9 inhibitors with single-digit nanomolar binding affinity.[9] PCSK9 is a validated target for cholesterol reduction, currently addressed by expensive monoclonal antibodies (evolocumab, alirocumab). The mRNA display-derived peptides represent a potential alternative that is smaller, cheaper to manufacture, and potentially amenable to oral delivery. This work demonstrated that mRNA display can find binders competitive with antibodies against clinically validated targets.
Cell-permeable cyclic peptides
Bowen et al. (2020) used mRNA display to identify membrane-permeating cyclic peptides, addressing one of peptide drug development's most persistent challenges: getting peptides inside cells.[10] By selecting for cell permeability alongside target binding, the study showed that mRNA display can optimize multiple drug-like properties simultaneously rather than fixing one property at a time.
KRAS inhibitors
Kage et al. (2024) published structure-activity relationships for middle-size cyclic peptides that inhibit KRAS, derived from mRNA display screening.[11] KRAS was considered "undruggable" for decades because it lacks the deep binding pockets that small molecules typically require. The macrocyclic peptide format, with its larger surface area and conformational pre-organization, can grip flat protein surfaces that small molecules cannot. The fact that mRNA display-derived peptides can reach this target illustrates the technology's access to pharmacological space beyond the reach of conventional drug classes.
mRNA display without purified protein
One limitation of traditional mRNA display is the requirement for purified target protein, which can be expensive and technically difficult for membrane proteins or multisubunit complexes. Hurd et al. (2024) developed a method to perform mRNA display selections directly in mammalian cell lysates, identifying cyclic peptides targeting the BRD3 extraterminal domain with low to sub-nanomolar affinity without ever purifying BRD3.[12] This advance expands the range of accessible targets and removes a significant bottleneck from the discovery workflow.
Dendritic cell targeting for vaccines
Kawaguchi et al. (2026) used the RaPID platform to develop dectin-1-binding peptides that target dendritic cells for antigen delivery, opening applications in vaccine development and immunotherapy beyond the traditional small-molecule and enzyme-inhibitor space.[13]
mRNA display vs. phage display
The relationship between mRNA display and phage display is complementary, not purely competitive. For a detailed comparison, see the dedicated article. Liu et al. (2024) developed new cell-penetrating peptides through a novel phage display platform, confirming that both technologies continue to evolve.[14]
| Feature | mRNA Display | Phage Display |
|---|---|---|
| Library diversity | 10^12 to 10^13 | 10^8 to 10^9 |
| System | In vitro (cell-free) | In vivo (requires E. coli) |
| Unnatural amino acids | Yes (via genetic code reprogramming) | Very limited |
| Selection conditions | Any (including harsh detergents, extreme pH) | Limited to phage-compatible conditions |
| Cyclization | Routine (thioether, disulfide, lactam) | Possible but more constrained |
| Cost per screen | Higher (complex molecular biology) | Lower (established workflows) |
| Regulatory track record | Growing (zilucoplan approved 2023) | Decades of validated drug discovery |
The 2018 Nobel Prize in Chemistry recognized phage display (George Smith, shared with Frances Arnold and Gregory Winter) for its foundational contributions to directed evolution.[15] mRNA display is younger but gaining ground. The technologies increasingly serve different niches: phage display for antibody engineering and simple peptide identification, mRNA display for macrocyclic peptide drug discovery where chemical diversity and library size are paramount.
For the high-throughput screening approaches that complement display technologies, and for how AI is accelerating peptide drug discovery alongside experimental selection, see the dedicated articles.
The commercial landscape
You et al. (2024) reviewed the state of cyclic peptide drug discovery and noted that the first drugs identified through display screening approaches have reached the market, with many more in clinical trials.[16]
PeptiDream (Tokyo, Japan) licensed the RaPID technology from Suga's laboratory and has become the most commercially advanced mRNA display company. Their pipeline includes multiple macrocyclic peptide candidates in clinical development through partnerships with Bristol-Myers Squibb, Novartis, Genentech, Merck, and Janssen. PeptiDream's PDPS (Peptide Discovery Platform System) uses the RaPID system as its core screening engine.
Ra Pharmaceuticals used mRNA display to discover zilucoplan, a macrocyclic peptide complement C5 inhibitor. UCB acquired Ra Pharmaceuticals for $2.1 billion. Zilucoplan received FDA approval in 2023 for generalized myasthenia gravis, making it one of the first drugs derived from mRNA display to reach patients.
These commercial successes validate mRNA display as a production-ready drug discovery platform, not just an academic technique.
Limitations and honest gaps
Technical complexity. The puromycin ligation, cell-free translation, and fusion purification steps demand specialized molecular biology expertise. A typical mRNA display laboratory requires equipment and skills that are not part of standard medicinal chemistry or biology training. Phage display is simpler to set up and run.
Hit-to-lead attrition. Finding a binding peptide is the first step, not the last. Converting an mRNA display hit into a drug requires extensive optimization of potency, selectivity, metabolic stability, pharmacokinetics, and formulation. The macrocyclic format helps with stability but does not eliminate the optimization challenge.
Oral bioavailability remains elusive. Most mRNA display-derived macrocycles are still administered by injection. The macrocyclic format and unnatural amino acid incorporation improve oral exposure relative to linear peptides, but true oral bioavailability for most peptide drug candidates remains an unsolved problem across the field.
Limited long-term clinical data. Zilucoplan's 2023 approval provides the first real-world data on an mRNA display-derived drug. Most other candidates remain in preclinical or early clinical stages. The technology excels at generating high-affinity binders in vitro, but the path from binding hit to approved drug remains long, expensive, and uncertain.
Cost. A single mRNA display screening campaign costs more than a phage display campaign. The reagents for cell-free translation, the puromycin-linked oligonucleotides, and the specialized equipment add up. For targets where phage display libraries are sufficient, the added cost of mRNA display may not be justified.
The convergence with computational methods
mRNA display and machine learning are increasingly used together. Computational models trained on mRNA display screening data predict which sequences are likely to bind a target, reducing the number of experimental rounds from the typical 6 to 8 down to 2 or 3. Conversely, computationally designed peptide scaffolds can be validated through mRNA display selection.
This convergence is reshaping the economics of peptide discovery. Trillion-member experimental libraries combined with computational filtering represent the current state of the art. For more on how AI is changing this landscape, see the dedicated article.
The Bottom Line
mRNA display links each peptide to its encoding mRNA through puromycin, enabling in vitro selection from libraries exceeding one trillion sequences. The RaPID system combines this with genetic code reprogramming to produce macrocyclic peptides containing unnatural amino acids, yielding drug candidates against targets from PCSK9 to KRAS. Zilucoplan's 2023 FDA approval provides the first clinical validation of an mRNA display-derived drug. The technology's main constraints are cost, technical complexity, and the same hit-to-drug attrition that affects all peptide therapeutics.