Advantages of mRNA display

Abstract

mRNA display is a robust in vitro selection technique allows the selection of protein libraries of trillions of variants. mRNA display relies upon a covalent linkage between a protein and its encoding mRNA molecule; the power of the technique stems from the stability of this link, and the large degree of control over experimental conditions afforded to the researcher. This chapter describes the major advantages that make the mRNA display technique superior to comparable in vivo and in vitro methods.

1. Introduction

mRNA display is an in vitro selection and evolution technique that enables screening of trillions of protein variants for desired functions in a single experiment. The key feature of this technique is the covalent linkage of an mRNA with its encoding protein which is established during in vitro translation [1, 2]. The stable genotype-phenotype linkage renders the protein directly amplifiable and enables the enrichment of mRNA-displayed protein variants with desired properties. This article will describe the advantages of the mRNA display technology in comparison to other in vivo and in vitro screening techniques (Fig. 1.). Specifically, we will discuss the value of interrogating more protein variants than is possible with most other techniques (1), the benefits of an in vitro method that frees protein expression from cellular constraints (2), allows the incorporation of unnatural amino acids in the protein variants (3), and permits protein screening under an unlimited range of desirable conditions (4). We will conclude by discussing the current limitations of the technique and how they are being overcome, as well as future prospects of mRNA display.

Fig. 1.

Advantages of mRNA display comprise the use of very large libraries of protein variants, the ability to readily incorporate unnatural amino acids and perform selections under a wide range of conditions. * Ribosome display uses libraries of similar size.

2. The numbers game – Advantages of large libraries

Directed evolution is a proven protein engineering method to generate proteins with modified properties by mimicking the natural process of mutagenesis and selection [3, 4]. The chance of identifying a protein variant with a desired function from a library of mutants is proportional to the number of protein variants that can efficiently be tested. Numerous methods have been developed to readily create large numbers of protein variants including techniques such as random mutagenesis by error-prone PCR, DNA shuffling, or the use of random synthetic DNA [5–10]. The generation of genetic diversity is therefore not the limiting aspect of a directed evolution approach, but instead the throughput of the subsequent screening step represents the bottleneck.

The following thought experiment will demonstrate the importance of library size in directed evolution approaches. Let us assume that in a typical binding pocket of a protein or active site of an enzyme there are about a dozen amino acids which line the cavity and form the first shell of contact to the ligand. Many of these residues (and more throughout the protein) will strongly influence the binding affinity or enzymatic activity. A common engineering approach to modify these properties is therefore to use site-saturation mutagenesis of these first shell residues to generate diversity for a directed evolution experiment [9–11]. A simple calculation shown in Table 1 exemplifies the dilemma the researcher quickly faces. The numbers demonstrate that classic cell-based screening methods with a throughput of 10²-10⁴ allow exhaustive sampling for only about up to 3 positions, and higher-throughput FACS-based screening or in vivo selection methods for up to 6 positions (Table 2). In reality, this problem can be further exacerbated several fold due to the redundancy of the genetic code. Several engineering teams have attempted to somewhat lessen this numbers problem by synthesizing so called “smart libraries”, which sample a limited number of codons to reduce library size while mostly maintaining the chemical diversity [10–12]. Unfortunately, these approaches only slightly delay the combinatorial “explosion” of possible variants. In any case, for smart libraries or not so smart ones, the chances of finding a protein variant with a potentially rare new or improved property increases proportionally with the number of variants interrogated.

Table 1.

Library size required to sample all possible variants in site-saturation mutagenesis

Positions randomized	Number of possible variants
1	20
3	8.0 × 103
6	6.4 × 107
10	1.0 × 1013

Table 2.

Differences in throughput for protein screening and selection techniques

Technique	Typical library size	References
Classic cell-based screen	102–104	[3]
Robotic screen	105	[98, 99]
FACS-based screen	107	[100]
In vivo selection	106–108	[17–19, 101]
Cell-surface display		[16]
Eukaryote	106–107
Prokaryote	108–1010
Phage display	~109	[20]
Ribosome display	1012–1014	[15]
mRNA display	1012–1014	[23]

mRNA display allows the use of protein libraries exceeding 10¹³ variants, which is several orders of magnitude larger than all in vivo methods and most in vitro techniques (Table 2) [13]. In contrast, cell-surface display and other selection methods that require in vivo expression of the protein library [9, 14–16] are limited by the transformation efficiency, which can be as high as 10¹⁰ for Escherichia coli [17] and 10⁷ for Saccharomyces cerevisiae [18] (Table 2). However, more typical in vivo libraries contain only about 10⁶-10⁸ unique sequences [19]. The widely used phage display technology has been reported to screen up to 10⁹-10¹² sequences [20–22], but again 10⁹ is more common because the technique is also limited by in vivo transformation into E. coli for propagation [20]. mRNA display, on the other hand, is performed entirely in vitro and therefore the number of sequences is not limited by a transformation step [13, 23, 24].

mRNA display selections with trillions of variant sequences have been successfully performed to isolate proteins with de novo functions, thereby demonstrating that large libraries can yield potentially rare protein variants [8, 25–27]. The generation of de novo functions, e.g. to instill binding or catalytic capability into a protein originally lacking those properties entirely, is still considered a great challenge in protein engineering [28]. The first example of mRNA display producing de novo proteins was the isolation of artificial ATP-binding proteins from a completely random polypeptide library [8]. In addition to yielding specific and tight binders, this work also determined for the first time the distribution of functional proteins in protein sequence space by identifying one ATP binder for each 10¹¹ random protein sequences. Subsequently, mRNA display also yielded artificial ATP-binding proteins from a library based on an unrelated protein scaffold [29]. Furthermore, de novo enzymes that catalyze a reaction unobserved in nature were produced by mRNA display from a library based on a non-catalytic protein scaffold [13, 26]. During the selection process for these de novo enzymes, surprisingly, the original scaffold was abandoned and instead a new fold was formed that lacks classic secondary structure motifs and, like the catalyzed reaction, has not been found in nature [30]. The large library size in mRNA display also enables the screening of an entire natural proteome, for example to identify protein-protein [31–33] or protein-drug interactions [34, 35], and to isolate kinase [36] or protease substrates [37, 38]. The length of polypeptides that have been successfully screened ranges from small peptides of <10 residues [1, 39] to proteins of ~650 residues [33, 40]. The display efficiency and thereby the size of library produced per volume of translation reaction slightly increases with decreasing size of the displayed peptide [31]. To date, mRNA display has produced a wide range of engineered polypeptides including peptide inhibitors, antibody mimics, and antibody fragments (Fab) [40–44].

Methods for screening large libraries for function are an integral part of the modern protein engineer’s toolbox and in vitro techniques such as mRNA display allow the highest variety of sequences to be tested. Besides large library size, further benefits of mRNA display include a high level of control over protein translation and the conditions under which the selections can be performed. These advantages will be described in detail in the following sections.

3. What’s in a cell? - Benefits of in vitro protein expression

All steps of the mRNA display procedure are performed entirely in vitro including protein expression. This fact alleviates many of the potential issues associated with in vivo protein production. The heterologous expression of proteins in E. coli or other cell-based systems can lead to the accumulation of protein aggregates that are detrimental to cell viability [45, 46]. Foreign, high copy number proteins can disrupt normal metabolism or lead to toxic products [47, 48]. Alternatively, the heterologous proteins themselves can in some cases be toxic to the cell [46, 49]. Cell-free protein expression removes both the need to maintain the viability of the cell and the danger of other cell components confounding a selection by acting as competing enzymes, potential inhibitors, or alternative substrates.

Alternative display methods, cell-surface and phage display, include an in vivo translation step and are therefore additionally limited by the requirement for transport of a folded protein across a cell membrane [16] or the release of a phage from a host cell [14, 21]. This effectively constitutes an additional undesired selection bias for proteins that can be successfully transported for display. The mRNA display technique has no such requirement.

Cell-based methods will likely degrade poorly folded protein library variants, but the experimenter has no real means to influence this process. In contrast, in vitro selections provide the option to fine-tune the removal of poorly folded proteins by selective protease treatment [50].

Cell-free protein translation in mRNA display is typically performed using eukaryotic cell lysates from rabbit or wheat germ [13] allowing a wider range of post-translational modification than would be possible with bacterial in vivo expression. Protein expression conditions can be controlled even more tightly with the use of the fully chemically-defined PURE translation system (Protein synthesis Using Recombinant Elements) [51, 52]. The PURE translation system is reconstituted from the purified components necessary for E. coli translation such as tRNAs, aminoacyl-tRNA synthetases, amino acids, ribosomes and release factors [53].

4. Expanding the protein alphabet with non-standard amino acids

The use of the PURE cell-free translation system also enables the facile incorporation of diverse non-standard amino acids (NSAA) into mRNA-displayed proteins, thereby vastly expanding the chemical properties of the selected proteins. NSAAs are either rare, naturally occurring amino acids different from the standard 20 residues, or completely synthetic amino acids that introduce novel chemistries to engineered proteins [54, 55]. NSAAs have been used to enhance protein stability, catalysis or detection [54–59]; enable subsequent protein modification via bio-orthogonal click chemistry [54, 59]; and generate biologically active peptides for pharmaceutical use. To date, more than 150 different NSAAs have been genetically encoded in proteins [60]. While NSAAs can be incorporated by different approaches both in vitro and in some cases in vivo [56], the PURE cell-free system is the most versatile method as it can use multiple orthogonal incorporation techniques in parallel by altering the composition of a user-defined translation system [54, 61]. mRNA display in combination with PURE can therefore readily realize the full potential of the ever-expanding repertoire of amino acid building blocks [54–56, 62].

Protein libraries containing NSAAs have been selected by mRNA display for a wide range of activities. The first such proof-of-principle mRNA display selection incorporated biocytin, a biotin derivative of lysine. The biotin-labeled peptides were then enriched by selection against a streptavidin agarose matrix [39]. In separate work, four-base codons were used to incorporate biocytin and other NSAAs to select for novel streptavidin-binding peptides [63]. Other groups have screened for inhibitors against Breast Cancer Associated protein 1 (BRCA1) [64], or improved the protease stability of peptides through unnatural methylation [65].

The PURE system has also been used in conjunction with mRNA display to introduce synthetic amino acids that allow further modification by click chemistry. Methionine was replaced with an alkyne-modified glycine, which enabled the post-translational modification with azide-modified high mannose glycans. These mRNA-displayed glycopeptides were then selected to tightly bind to a HIV broadly neutralizing antibody [66]. Although this is the only published combination of mRNA display and click chemistry to date, this principle could be used to immobilize an mRNA-displayed peptide to a surface for purification, trap a reaction substrate or product, or attach to a binding partner in solution.

NSAA incorporation has been harnessed to select macrocyclic peptide binders from mRNA-displayed libraries of up to 10¹³ variants [63, 67–78]. Circularized peptides are favored in drug screening due to their increased proteolytic stability, improved membrane permeability, and reduced conformational flexibility that can lead to tighter target binding [35, 79–81]. An overview of peptide cyclization strategies can be found in several reviews [35, 79–82]. One of these mRNA-displayed macrocyclic peptide libraries contained twelve NSAAs and only eight standard amino acids, yielding binders to thrombin with a low nanomolar dissociation constant [74]. NSAAs can also be used to perform the peptide cyclization itself. This has been achieved by a number of different means, including the formation of a thioether bond between a chloroacetyl-analog of tyrosine and a cysteine side chain [75, 83]; or by using 4-selenalysine that was subsequently converted to dehydroalanine and cyclized with a cysteine residue to form a linkage similar to natural lantipeptides [73]. In a further study, an altered mRNA display protocol, RaPID (Random nonstandard Peptide Integrated Discovery) [71, 83], produced macrocyclic peptides that functioned as isoform-selective inhibitors of Protein Kinase B [75]. These examples demonstrate that the in vitro translation feature of mRNA display enables the experimenter to harness the continuing progress in non-standard amino acid incorporation and promises to greatly expand the potential for protein engineering in both fundamental research and for practical applications.

5. Complete control over selection conditions due to robust in vitro format

Cell-based selections are by necessity limited to conditions that are compatible with cell growth and replication, while in vitro selections by mRNA display enable the researcher to control selection conditions freely. Numerous parameters such as pH, temperature, and ionic strength can be precisely controlled. The format also allows easy inclusion of additional components such as reaction substrates, binding targets, solvents or other chemicals. Unlike within the crowded cytosol of a cell where millimolar concentrations of potentially competing proteins, solutes and other factors could confound selection outcomes, all components of an in vitro selection system are entirely controlled by the experimenter.

In order to identify “winning” sequences by any protein screening or selection technique, each protein variant must be linked to its genetic information. For example, each gene variant is contained inside the cell or viral capsid for cell-based screening and phage display, respectively. Compared to the size of the protein variant molecule, a cell or a phage is a very large entity and will therefore alter the properties of the displayed protein. In mRNA display, this essential gene-protein link is dramatically miniaturized to a simple covalent bond between the protein and its encoding mRNA via the small molecule puromycin. Therefore, the bias that an attached cell or phage might exert on a protein selection is reduced to the bare necessity – the mRNA itself. In addition, this stable covalent linkage between the protein and mRNA further facilitates the use of stringent selection conditions. The ribosome display technology, which otherwise has many favorable features in common with mRNA display such as large library size and the in vitro format, uses the whole ribosome to keep mRNA and protein together in a non-covalent and therefore less stable manner [15]. The small size of the linkage used in mRNA display – relative to cell, phage, or ribosome display constructs – likely increases the chances of a displayed peptide maintaining the same physicochemical properties as the respective unfused peptide. For example, a library of mRNA-displayed fusions was selected for the ability to infiltrate cancer cells. It was subsequently demonstrated that the identified peptides alone had the same cell-penetrating properties as the mRNA-displayed fusions [84, 85]. The functionality of isolated peptides is therefore not significantly hindered by the presence of the fused mRNA, and peptide variants selected by mRNA display are accurately represented.

The stability of the covalent linkage between mRNA and protein also allows a wide variety of conditions to be used during a selection. For example, in order to evolve artificial ATP binders towards greater structural stability, a selection was carried out in the presence of increasing concentrations of the chemical denaturant guanidine hydrochloride [86, 87]. Furthermore, an artificial RNA ligase enzyme, originally selected at room temperature, was evolved for increased thermostability by performing the selection at elevated temperature. The resulting thermostable ligase variant had a melting temperature of 72°C, which was 35 degrees more stable than the most closely related variant selected at ambient temperature [88].

Cell-based screening and selection systems can yield false positive results due to the unintentional up-regulation of multicopy suppressors [89] with promiscuous activity that is similar to the desired activity. Nearly 40% of all E. coli proteins are estimated to be promiscuous [90]. Therefore, selections for enzymatic function through complementation of auxotrophic strains are complicated by the risk of chromosomal mutations leading to the overexpression of existing weakly promiscuous enzymes [91, 92] or by the library variant itself acting as a transcription factor to the same effect. The tight control over selection conditions afforded by the mRNA display method completely avoids such issues.

A typical round of an mRNA display selection takes approximately 2–3 days depending on selection conditions and purification steps [13, 93]. Recently however, this time has been drastically reduced to about two hours per round. An optimized version of the mRNA display protocol successfully accomplished six rounds of selection in just fourteen hours thanks to a one-pot transcription, translation, and puromycin coupling reaction [94]. This mRNA display version was termed TRAP display: transcription−translation coupled with association of puromycin linker. Furthermore, coupling mRNA display with continuous magnetic flow separation and subsequent high throughput sequencing has enabled the selection of IgG binders with nanomolar affinities in a single round [95]. These examples demonstrate that the in vitro format and the small, covalent gene-protein linkage in mRNA display provide great freedom to choose selection conditions and selection stringency.

6. Addressing potential concerns with mRNA display

mRNA display is a robust method for the selection of functional proteins from large libraries. While this technique compares favorably with other selection methods in many aspects, we will here also address potential concerns associated with the technique and ways to overcome them.

RNA stability:

A common question about mRNA display involves the perceived sensitivity of RNA to degradation. This question likely originates from people’s experiences isolating long RNA from lysed cells. However, due to the in vitro nature of the entire mRNA display procedure, RNA nuclease degradation is simply not an issue. Simple precautions such as using nuclease-free chemicals, wearing gloves during experiments and protecting against dust contamination are usually sufficient to avoid degradation [13]. Only when mRNA-displayed proteins are to be incubated with biological specimens does mRNA degradation become a concern due to the inherent presence of nucleases. Nevertheless, recent work on selecting cell-penetrating peptides showed that up to 60% of the mRNA-displayed library could be recovered intact after the incubation with cells [85]. Moreover, through a modification of the mRNA display procedure, cDNA-displayed protein libraries can be produced which avoid mRNA stability concerns altogether [96, 97].

Monomeric vs. dimeric proteins:

The design of the mRNA display approach results in a single polypeptide chain covalently linked to its encoding mRNA. Therefore, mRNA display has been limited in principle to the selection of monomeric proteins. But, recently, a new strategy was developed to address this limitation. The polypeptides of the heavy and light chains of an antibody Fab fragment were separately mRNA-displayed and then mixed to form heterodimers. After affinity selection, the heterodimers were individually encapsulated in water-in-oil emulsions and the winning genes were recovered by overlap-extension PCR [44].

7. Conclusions

In vitro selection of polypeptides by mRNA display has proven to be an excellent tool for peptide design, protein engineering, and the investigation of protein interactions. The ability to search very large libraries of candidate sequences dramatically outperforms all in vivo methods and most other in vitro selection techniques. The convenient option of incorporating unnatural amino acids in the protein libraries expands the possibilities of obtaining new structures and activities that have no natural precedent and could not be selected by other means. This avenue is being heavily pursued to generate macrocyclic peptide therapeutics with unnatural modifications, opening the way to more potent protein inhibitors. The covalent genotype-phenotype linkage in mRNA display not only allows easy recovery and identification of the selected candidates but also, due to its robustness, enables the use of a wide variety of selection conditions, many of them incompatible with in vivo methods. This can be especially valuable in enzyme engineering in order to evolve soluble, folded, thermostable proteins that are also compatible with organic solvents, extreme pH, or high salt concentrations.

In summary, if you want to search the largest possible libraries of protein variants under the widest range of selection conditions – the mRNA display technology should be your method of choice.