RNA elements and their biotechnological applications in plants

Summary

Engineering of plants for improved traits and efficient heterologous protein production can be achieved by modifying or introducing cis‐ or trans‐acting RNA elements. The function of these elements depends not only on their nucleotide sequence but also on their highly dynamic higher order structures. In this review, we explore RNA regulatory elements with established or potential application in plant biotechnology. We discuss RNA elements involved in translational control, transcript stability, and protein coproduction, as well as RNA domains that mediate conditional expression, RNA decay, or cap‐independent initiation. While some of these elements can be used in transiently or stably transformed plants, others have proven valuable in plant‐based in vitro expression systems. Additionally, we highlight RNA elements important for plastid gene expression. Finally, we examine RNA elements that are yet to be applied in plant biotechnology but have been successfully used in other organisms or require further understanding before they can be effectively utilized.

	Contents
	Summary	2517
I.	Introduction	2517
II.	Translational control as a tool for fine‐tuning gene expression	2519
III.	Functional RNA domains	2521
IV.	Other RNA elements	2526
V.	RNA elements in engineered chloroplasts	2529
VI.	Perspectives	2532
	Acknowledgements	2532
	References	2532

I. Introduction

Genetic engineering techniques have allowed us to cultivate plants with improved properties or use them as expression systems for synthesis of heterologous proteins. To achieve desired plant qualities or ensure effective protein production, it is important to control the levels of expression both spatially and temporally. Key tools to achieve the desired gene expression levels in the modified plants are cis‐ or trans‐acting RNA regulatory elements, which can be modified in the endogenous genome or introduced with heterologous sequences.

Similarly to other eukaryotes, plants export messenger RNA (mRNA) from the nucleus to the cytosol, where translation occurs (Fig. 1). Before export, mRNA undergoes essential processing steps, including intron splicing, the addition of a 5′ 7‐methylguanosine cap, and polyadenylation at the 3′ end. In the cytosol, the 5′ cap facilitates recruitment of translation initiation factors, enabling the preinitiation complex (PIC) to attach to the mRNA and scan for the start codon. The PIC scans in a strictly 5′‐to‐3′ direction until it reaches the start codon, usually the 5′‐most AUG codon, where the 80S ribosome assembles and translation initiates. The poly(A) tail ensures mRNA stability and interacts with the poly(A)‐binding protein, which associates with a translation initiation factor at the 5′ cap. This association circularizes the mRNA, which can increase translational efficiency, enhance tethering of initiation factors to the mRNA, and expedite ribosome recycling after translation termination (Jackson et al., ²⁰¹⁰). Viral RNA without a 5′ cap or a poly(A) tail possesses alternative elements, such as internal ribosomal entry sites (IRESs), that stabilize the RNA and allow its efficient translation (Miras et al., ²⁰¹⁷). In addition to expression of plant nuclear and viral genes, gene expression in plastids, where the nature of transcription and translation differs, is also of interest in biotechnology.

Fig. 1.

Processing and translation of nucleus‐derived mRNA in eukaryotes. RNA elements are highlighted in bold. eIFs, eukaryotic translation initiation factors. Specific translation initiation factors are illustrated individually and numbered. PIC, preinitiation complex, which consists of the 40S ribosomal subunit, eIF1, eIF1A, eIF3, eIF5, and a ternary complex of methionyl initiator tRNA, eIF2, and GTP; the PIC scans for a start codon in association with the cap‐binding complex: eIF4A, eIF4E/eIFiso4E, and eIF4G/eIFiso4G (Browning et al., ¹⁹⁹²; Jackson et al., ²⁰¹⁰; Brito Querido et al., ²⁰²⁴). The initiation factors remain associated with the small ribosomal subunit until start codon recognition and recruitment of the 60S ribosomal subunit but are omitted from the illustration for simplicity. Our knowledge of the eukaryotic translation machinery mostly comes from yeast and mammalian cells. Study of plant translation has revealed and will continue to reveal that the machinery differs in multiple ways both in its composition and regulation, for example through the use of the plant‐specific alternative initiation factors eIFiso4E and eIFiso4G (Browning, 2014). AGO, Argonaute; CPMC, cleavage and polyadenylation molecular complex; PABP, poly(A)‐binding protein; uORF, upstream open reading frame. The illustrated RNA elements are not present on all mRNAs.

Controlling the amount of protein synthesized is crucial for efficient use of resources and response to environmental stimuli. This is, in part, regulated at the transcriptional level. However, regulation at the post‐transcriptional level is the faster and more direct response, which can then be sustained by transcriptional modulation. Post‐transcriptional regulation can occur through altered translational efficiency or mRNA stability. Regulatory elements on a specific RNA can exert control by themselves or through interaction with proteins or other RNAs, and this control can be exerted constitutively or in response to particular internal or external cues (Hershey et al., ²⁰¹²).

In the case of mRNA, characterization of RNA species mostly focusses on the primary structure – the nucleotide sequence – which determines the amino acid sequence of the encoded polypeptide chain and its interaction with regulatory RNAs and the translation machinery. However, similarly to polypeptide chains, RNA forms higher order structures. It adopts secondary structures through intramolecular base pairing, tertiary structures that define its three‐dimensional conformation, and even quaternary structures involving the assembly of multiple interacting RNA molecules. RNA structure is also highly dynamic, constantly shifting between different conformations. The likelihood of adopting a particular structure is influenced by various factors, including proteins, other RNA molecules, small molecules, and external conditions, such as metal ion concentration, pH, crowding, and temperature. Both the nucleotide sequence, which ensures interaction specificity, and the multi‐tiered RNA structure with capacity to change are fundamental for the function of RNA regulatory elements, whether it is interaction with other molecules, sensing of environmental cues, catalysis, regulation of gene expression, RNA splicing, or RNA decay (Mustoe et al., ²⁰¹⁴; Leamy et al., ²⁰¹⁶; Ganser et al., ²⁰¹⁹).

In this review, we explore RNA regulatory elements, highlighting their current biotechnological applications in plants and plant‐derived expression systems, while proposing potential avenues for their expanded use. RNA elements leading to replicon‐based systems, such as viral replicating expression systems, have been extensively discussed elsewhere (e.g. Peyret & Lomonossoff, ²⁰¹⁵) and are beyond the scope of this review.

II. Translational control as a tool for fine‐tuning gene expression

1. Start codon recognition

Besides the start codon itself, efficient translation initiation is also facilitated by the sequences that flank the start codon, known as the Kozak sequence (Kozak, 1981, 1986, ^1987a). Genetic, biochemical, and structural studies of yeast and animal PICs have revealed that recognition of the Kozak sequence involves initiation factors (eIF) 2, 1A, as well as the ribosomal protein uS7 and the 18S rRNA (Jackson et al., ²⁰¹⁰; Brito Querido et al., ²⁰²⁴). Although the purine at position −3 and the guanine at +4 relative to the first nucleotide of the coding sequence (CDS) were found to have the strongest impact on start codon recognition in eukaryotes, nucleotides at positions −2 and −1 also have a great influence in plants (Kozak, ^1987a; Lukaszewicz et al., ²⁰⁰⁰; Sugio et al., ²⁰¹⁰). Poor context can allow bypass of the AUG by the scanning PIC and initiation at a downstream AUG (Fig. 2a), a phenomenon known as leaky scanning (Kozak, ^1987a).

Fig. 2.

Fine‐tuning gene expression through translation initiation control. (a) Kozak sequence facilitates efficient translation, and it can vary in strength. Translation initiation can be further increased by translation enhancer sequences, such as the Tobacco mosaic virus Ω sequence, which interacts with HSP101. (b) Upstream open reading frames (uORFs) inhibit the translation of the main open reading frame (mORF). A stronger initiation context at the uORF, greater length, and shorter uORF–mORF distance increase its inhibitory effect. Some uORFs with highly conserved sequences, such as the FvebZIPs1.1 uORF, inhibit mORF expression in response to a metabolite by an unknown mechanism. (c) Strong secondary structural elements in the 5′ untranslated region (UTR) can impede the recruitment of the 43S preinitiation complex (PIC) to the mRNA, its scanning, or translation initiation. A structure a short distance downstream of a noncanonical start codon can facilitate translation initiation. In the 5′ UTR of the Arabidopsis thaliana SMXL5, the RNA‐binding protein JULGI can induce an inhibitory G‐quadruplex. H, not guanine (A/C/U); R, purine (A/G); Y, pyrimidine (C/U). The initiation factors remain associated with the small ribosomal subunit until the recruitment of the 60S ribosomal subunit but are omitted from the illustration for simplicity. Blunt‐ended arrows indicate inhibition, and sharp arrows indicate enhancement or production.

Design of protein‐producing expression systems as well as modification of endogenous genes should, hence, consider the translation initiation context. However, due to the divergence of the scanning complex, optimal Kozak sequences can vary between species: (GCCRCCAUGGC in monocots (and a similar sequence in nonflowering plants) and AAAAAAAUGGC in dicots (Joshi et al., ¹⁹⁹⁷; Nakagawa et al., ²⁰⁰⁸; Rangan et al., ²⁰⁰⁸; Gupta et al., ²⁰¹⁶)). Optimization of the Kozak sequence contributes towards high protein production. On the contrary, the choice of a suboptimal context, particularly at the positions −3 to −1 if the second codon has to be preserved, can be used to fine‐tune the efficiency of translational initiation, as demonstrated in mammalian cells (Xie et al., ²⁰²³). The strength of Kozak sequence variants has been assessed in dicot, monocot, and gymnosperm protoplasts and suspension cells (Lukaszewicz et al., ²⁰⁰⁰; Sugio et al., ²⁰¹⁰). Adjustment of expression at the translational level may be desirable in situations where modification of the promoter region, to affect the transcript abundance, is undesirable or insufficient. Suboptimal contexts are also found around AUGs upstream of the main start codon; these initiate regulatory upstream open reading frames (uORFs).

2. Upstream open reading frames

In contrast to prokaryotes, eukaryotic transcripts tend to be monocistronic; nevertheless, a significant proportion of transcripts (over 30% in 11 species of plants; Von Arnim et al., ²⁰¹⁴), have uORFs that precede the main ORFs (mORFs). The effect of uORFs on the mORF is generally inhibitory, as ribosomes are usually recycled after termination (Fig. 2b). However, small ribosomal subunits can remain associated with mRNA, resume scanning, and re‐initiate translation at a downstream ORF. The efficiency of re‐initiation is correlated with the following parameters: short distance between the uORF and the mORF (Kozak, ^1987b); short uORF length (Luukkonen et al., ¹⁹⁹⁵); and efficient elongation at the uORF (Kozak, 2001; Pöyry et al., ²⁰⁰⁴; Lin et al., ²⁰¹⁹), which can be influenced by codon usage, strong secondary structure, or nature of the nascent peptide. It has therefore been hypothesized that efficient re‐initiation is conferred by undissociated initiation factors (probably eIF4G and eIF3) and acquisition of an initiator‐methionyl‐tRNA‐GTP‐eIF2 complex by the ribosome that has resumed scanning. Nevertheless, re‐initiation efficiency rarely exceeds 50% (Jackson et al., ²⁰¹²). Translation of uORFs overlapping the mORFs is unlikely to allow re‐initiation at the mORF, although instances of backward scanning have been reported (Matsuda & Dreher, 2006; Wang et al., ²⁰²²; Dever et al., ²⁰²³). In addition, translation of uORFs can be bypassed through leaky scanning, which occurs more frequently if their start codon is non‐AUG or in a suboptimal context (Hinnebusch et al., ²⁰¹⁶; Dever et al., ²⁰²³).

Notably, translation of uORFs can regulate mORF expression in response to internal conditions, and leveraging this mechanism has valuable potential in plant engineering. The transcription‐factor‐encoding TBF1 is regulated by two conserved uORFs, which derepress the mORF translation in response to accumulation of uncharged tRNA^Phe during stress (Pajerowska‐Mukhtar et al., ²⁰¹²). Certain metabolic genes are also regulated by uORFs that respond to the products of the associated metabolic pathway, for example, the synthesis of S‐adenosylmethionine decarboxylase in plants is negatively regulated by polyamines that act on the translation of two conserved uORFs. Translation of the first, very short uORF happens more frequently at low polyamine levels, bypasses the strongly inhibitory second uORF, and allows re‐initiation at the mORF (Hanfrey et al., ²⁰⁰⁵). Other examples of genes regulated by uORFs affected by small molecules are the sucrose‐regulated transcription factor gene bZIP11 (Wiese et al., ²⁰⁰⁴; Xing et al., ²⁰²⁰), the phosphocholine‐regulated phosphocholine biosynthesis gene XIPOTL1 (Tabuchi et al., ²⁰⁰⁶; Alatorre‐Cobos et al., ²⁰¹²; Von Arnim et al., ²⁰¹⁴) and the well‐known ascorbate‐responsive ascorbate biosynthesis gene GGP with a noncanonical ACG start codon (Laing et al., ²⁰¹⁵).

For biotechnological purposes, translation can be further tuned by the introduction or modification of uORFs: those with a stronger Kozak sequence exert a stronger inhibitory effect on the mORF and vice versa (Xie et al., ²⁰²³). Manipulation of gene expression using uORFs is demonstrably feasible. Lettuce cultivars with an increased foliar ascorbic acid content and tolerance to oxidative stress were produced by disruption of the GGP uORF (Zhang et al., ²⁰¹⁸). Strawberries with varying sugar contents were generated by mutating the sucrose‐responsive FvebZIPs1.1 uORF and cross‐fertilizing the mutants with one another (Xing et al., ²⁰²⁰). Conversely, tunable inhibition of translation was achieved in rice by introduction of new uORFs or extension of existing uORFs (Xue et al., ²⁰²³).

3. Translation enhancer elements

In addition to Kozak sequence for optimal translation initiation, heterologous as well as endogenous gene expression can be increased by the introduction of translation enhancer elements, often derived from viruses. The most extensively used enhancer is the CA‐rich Tobacco mosaic virus Ω sequence, which binds heat shock protein 101 (HSP101) and increases the recruitment of eIF4F (Fig. 2a). This function overlaps with the function of 5′ cap and the poly(A) tail (Wells et al., ¹⁹⁹⁸; Gallie, ²⁰⁰²). Another biotechnologically utilizable element comes from the Alfalfa mosaic virus RNA4 and showed a greater effect in rice (a monocot) than the Ω sequence (Jobling & Gehrke, 1987; Datla et al., ¹⁹⁹³; Shen et al., ²⁰²³). Finally, synthetic enhancers have been developed, such as CA‐rich 5′ untranslated regions (UTRs) derived from random library screening (Kamura et al., ²⁰⁰⁵), synJ (Kanoria & Burma, 2012), ARC‐1, which is complementary to an internal 18S rRNA region (Akbergenov et al., ²⁰⁰⁴), and 5S0 (Peyret et al., ²⁰¹⁹).

4. RNA structures as regulators of translation initiation

Much of our understanding of how RNA secondary structure affects eukaryotic translation initiation has been gained from studies in animal and yeast systems; these findings are relevant to plants as they share the general mechanism of translation initiation and should therefore be taken into account when designing expression systems or modifying genes. In yeast and mammals, the efficiency of translation initiation is normally reduced by the presence of strong secondary structure in particular regions of the 5′ UTR. For example, RNA stem–loop or G‐quadruplex structures located close to the 5′ end can inhibit the recruitment of the PIC to the mRNA. The helicase activity of eIF4A (and associated helicases) unwinds mRNA stem–loop structures as the PIC scans for the start codon (Brito Querido et al., ²⁰²⁴). Nevertheless, scanning can be obstructed by highly stable 5′‐UTR stem–loop or G‐quadruplex structures, which has been demonstrated in plant systems (Kozak, 1989; Kwok et al., ²⁰¹⁵). Furthermore, stem–loop structures located just upstream of the start codon can reduce translation initiation efficiency. They may act as a physical barrier, obstructing the PIC, or alternatively, an RNA hairpin may displace the PIC from the start codon before GTP (guanosine triphosphate) hydrolysis by eIF2 and the subsequent joining of the 60S ribosomal subunit. Conversely, a stem–loop structure just downstream of a start site with a noncanonical start codon or poor context is known to enhance translational efficiency, likely by positioning or prolonging the positioning of the scanning PIC at the start site (Fig. 2c; Kozak, ¹⁹⁹⁰; Wang et al., ²⁰²²; Cao et al., ²⁰²⁴).

Structural elements in the 5′ UTR can play a role in conditional translational regulation: for example, in vascular plants, the translation of SMXL4/5, which inhibits phloem differentiation, can be suppressed through the formation of 5′‐UTR G‐quadruplex, facilitated by the sucrose‐inducible RNA‐binding protein JULGI (Fig. 2c). Editing this system or adapting it to regulate other genes holds significant potential in improving crop yields (Cho et al., ²⁰¹⁸; Cao et al., ²⁰²⁴).

A comparison of 5′ UTRs of highly expressed genes in plants has revealed that weak secondary structure of the 5′ UTR is linked to high‐expression levels (Peyret et al., ²⁰¹⁹).

III. Functional RNA domains

While the main role of mRNA is directing the synthesis of a polypeptide in a regulated manner, a number of mRNA species have additional sensory or catalytic functions that are instrumental in regulating their translation. A region of RNA can adopt a complex secondary and tertiary structure that forms a functional domain, analogous to a domain in proteins. An important fact to keep in mind when designing expression systems that rely on higher order structural elements is that the structure does not form in isolation and can be affected by other regions of the RNA as well as interacting proteins. A regulatory element may function effectively in its native context, but positioned next to a desired CDS or other regulatory elements in an artificial expression system, it may form an alternative structure, compromising its functionality. In silico predictions of local secondary structure may give indications of the compatibility of regulatory secondary structural elements with the designed expression system, but experimental verification ultimately provides the highest certainty.

1. Riboswitches

RNA can fold into riboswitches: higher order structural elements that change conformation (‘switch’) in response to binding ions or small molecules. This alters the expression of the gene on the riboswitch‐controlled mRNA. In biotechnology, riboswitches confer inducibility of gene expression on the post‐transcriptional level, which is valuable in situations when constitutive expression is undesirable, such as when the protein of interest negatively impacts the expression system. In stably transformed plants, an inducible system allows protein production once the plant reaches a suitable size and avoids interference of the transgene expression with development or growth. While numerous classes of riboswitches have been described in bacteria, bacterial riboswitches are ineffective in eukaryotes (except in plastids) due to the inherent differences in translational mechanisms. Currently, the only known eukaryotic riboswitches bind the cofactor thiamine pyrophosphate (TPP) and its chemical analogues. They are often located within introns and regulate alternative splicing of mRNA, which can produce splice variants with premature stop codons, an alternative 3′ UTR that destabilizes the mRNA, or uORFs that inhibit translation of the mORF (Fig. 3a,b; Breaker, ²⁰¹⁸; Croft et al., ²⁰⁰⁷).

Fig. 3.

RNA switches responsive to ligands and temperature. (a) Arabidopsis thaliana THIC 3′ untranslated region (UTR) alternative splicing is regulated by a thiamine pyrophosphate (TPP) riboswitch. Binding of TPP leads to an alternative 3′ UTR (blue), removal of a 3′ regulatory region (yellow), and unstable mRNA. CPMC, cleavage and polyadenylation molecular complex. (b) Chlamydomonas reinhardtii THI4 5′ UTR alternative splicing is regulated by a TPP riboswitch. Binding of TPP leads to retention of an upstream open reading frame (uORF; pink) and inhibition of THI4 translation. (c, d) Artificial riboswitches developed for expression systems based on wheat‐germ extract. Upon ligand binding, a riboswitch can inhibit translation by obstructing the recruitment of the 43S preinitiation complex (PIC: c). In a different construct, a riboswitch can activate translation by triggering ribosomal shunting (d). Following the translation of a uORF, the small ribosomal subunit restarts scanning. In the absence of a ligand, the scanning is hindered by an RNA hairpin or translation re‐initiates at one of two in‐frame decoy start codons, inhibiting the translation of the main ORF. In the presence of the ligand, the riboswitch forms a rigid stem, which allows scanning ribosomes to shunt over the structure, bypass decoy start codons, and initiate translation of the main ORF. The inhibitory uORF sequence with two in‐frame start codons is illustrated in pink. (e) G‐quadruplexes in the 3′ UTR stabilize the transcripts at low temperature. They do not form at higher temperatures, reducing the mRNA stability. (f) Regulation of translation by a ThermoSwitch. At low temperature, PIC scanning is obstructed. The ThermoSwitch adopts a relaxed conformation at a higher temperature, which is more easily unwound. The initiation factors remain associated with the small ribosomal subunit until the recruitment of the 60S ribosomal subunit but are omitted from the illustration for simplicity. Blunt‐ended arrows indicate inhibition.

TPP riboswitches have been used to control gene expression in Arabidopsis and tomato seedlings. However, their application in vascular plants appears limited so far, as ligand addition could reduce TPP‐riboswitch‐associated RNA abundance by up to fivefold (Wachter et al., ²⁰⁰⁷), while a significant reduction in THIC transcript levels was observed in a separate study only when thiamine auxotrophs were used (Bocobza et al., ²⁰⁰⁷). These findings suggest that the current biotechnological application of TPP riboswitches in vascular plants is mainly restricted to monitoring intracellular TPP levels (Bocobza & Aharoni, 2014).

On the contrary, TPP‐riboswitch‐based engineering has been more successful in the unicellular green alga Chlamydomonas reinhardtii. Chlamydomonas possesses two TPP‐responsive riboswitches, upstream of the genes THIC and THI4, which additionally also bind 4‐amino‐5‐hydroxymethyl‐2‐methylpyrimidine or 5‐hydroxyethyl‐4‐methylthiazole, respectively (Moulin et al., ²⁰¹³). The THI4 riboswitch controls alternative splicing that removes an inhibitory uORF in the absence of the ligand. In the engineered expression system, the THI4 riboswitch is replaceable with alternative TPP riboswitches that have different ligand specificities and affinities, and it can successfully control the expression of heterologous genes (Mehrshahi et al., ²⁰²⁰).

Apart from one aptazyme (see the Catalytic RNAs section for further detail), no other riboswitches have been engineered in plants to control nuclear‐encoded genes. However, several artificially designed riboswitches have been assessed in vitro using wheat‐germ extract and are potentially applicable as in vitro biosensors. These ab initio riboswitches alter gene expression by mechanisms listed below and respond to a variety of ligands, such as theophylline, tetracycline, and riboflavin‐5′‐phosphate (Tabuchi & Yokobayashi, 2021):

• A strong RNA secondary structural element to obstruct ribosome association (Fig. 3c; Ogawa et al., ²⁰¹⁸);

• Disruption of the proper folding of an IRES (Ogawa, 2011);

• Activation of ribosomal shunting (see Pooggin et al., ²⁰⁰⁰), allowing bypass of decoy start codons (Fig. 3d; Ogawa, ²⁰¹³); and

• Inclusion of a 3′‐cap‐independent translational enhancer (CITE; see the Cap‐independent translation initiation section; Ogawa et al., ²⁰¹⁷).

While artificial riboswitches in plants remain underexplored, many have been developed and tested in other eukaryotic systems, especially yeast and mammalian cells. In addition to alternative splicing similarly to native eukaryotic riboswitches (Culler et al., ²⁰¹⁰), these artificial riboswitches utilize some noteworthy mechanisms, such as ligand‐induced −1 programmed ribosomal frameshifting (Lin & Chang, 2016) or a protein‐binding aptamer upstream of a non‐AUG start codon that stimulates translation (Horie et al., ²⁰²⁰). Future research may show whether these riboswitches are functional and sufficiently effective in planta.

2. Plant RNA structures conferring temperature‐sensitive regulation

While the RNA thermometers of bacteria and bacteriophages have been recognized since the 1980s (Altuvia et al., ¹⁹⁸⁹), eukaryotic temperature‐sensitive RNA elements and their mechanism are less understood. Relatively recent studies have shown that plants possess RNA structures that confer responsiveness to temperature through effect on RNA stability or indeed translational efficiency (Chung et al., ²⁰²⁰; Thomas et al., ²⁰²²; Yang et al., ²⁰²²; Lastovka et al., ²⁰²⁴).

In plant mRNA, 3′ UTRs can respond to low temperatures by forming G‐quadruplexes that stabilize the transcript (Fig. 3e). This phenomenon was shown in A. thaliana seedlings subjected to a temperature of 4°C, along with its role in regulating root growth in cold temperatures. The RNA G‐quadruplexes regulated the mRNA abundance and had little effect on translational efficiency, independently of their location on the transcript. The effective stabilization of a reporter mRNA by this element reveals its capacity to regulate other mRNA species, whether endogenous or heterologous, and it therefore shows great promise for engineering of cold‐tolerant plants (Yang et al., ²⁰²²).

Control of plant translational efficiency by a temperature‐sensing RNA structure was also demonstrated in A. thaliana. A 5′‐UTR RNA hairpin, known as the ‘ThermoSwitch’, was identified in several transcripts that exhibited increased translational efficiency when the temperature rose by 10°C (17–27°C). This structure varied in sequence among different genes. Subsequent in vitro studies revealed that the shift range can be as narrow as 22–27°C to trigger enhanced protein synthesis of reporters under ThermoSwitch control (Chung et al., ²⁰²⁰). Eukaryotic ThermoSwitches differ from the bacterial RNA thermometers in multiple ways, in part due to the different nature of prokaryotic and eukaryotic translational initiation. Bacterial RNA thermometers change conformation or gradually melt at warmer temperatures and thus reveal (or obscure) the ribosome‐binding site. On the contrary, plant ThermoSwitches have a seemingly dual mechanistic function: below the threshold temperature, they impede the scanning of the PIC, which travels from the 5′ end of the mRNA, from accessing the start codon; above the threshold temperature, the ThermoSwitch adopts a relaxed conformation, which enables the PIC to reach the start codon (Fig. 3f). The partially molten hairpin additionally enhances translation, possibly by temporarily impeding the incoming PIC while the previous complex initiates translation (Chung et al., ²⁰²⁰; Thomas et al., ²⁰²²). The molecular mechanism is the subject of ongoing research. A recent study demonstrated that the PIF7 ThermoSwitch, when integrated into an Agrobacterium‐mediated transient expression system, enables temperature‐sensitive regulation of gene expression in Nicotiana benthamiana, indicating its potential as a biotechnological tool for homogeneous induction of gene expression, independent of chemical inducers or suppressors (Lastovka et al., ²⁰²⁴).

3. Catalytic RNAs

Analogously to enzymes, RNA can form domains – ribozymes – that catalyse chemical reactions. While the ribozyme active site consists of RNA, some ribozymes, such as ribosomes, contain protein components (Cech, 2000). The catalytic functions of most naturally occurring ribozymes are RNA cleavage or splicing in cis. Self‐splicing introns found in mitochondria and plastids rely on cis‐acting ribozymes that splice the pre‐mRNA sequence independently of spliceosomes (Mukhopadhyay & Hausner, 2021). Cis‐cleaving ribozymes are used by viruses, satellite viruses, and viroids, including the plant pathogens Avocado sunblotch viroid and Tobacco ringspot virus satellite RNA to cleave their RNA during replication (De la Peña et al., ²⁰¹⁷). Ribozyme motifs are widespread in genomes of all kingdoms of life, and their identification in plant genomes has led to the discovery of retrozymes, nonautonomous retrotransposons that encode circular RNA (Cervera et al., ²⁰¹⁶; De la Peña et al., ²⁰¹⁷). Of the 10 families of cis‐cleaving ribozymes, the hammerhead ribozyme is the most extensively characterized (De la Peña et al., ²⁰¹⁷).

Naturally occurring ribozymes can be converted to trans‐acting ribozymes by separating the catalytic region from the substrate region and modifying the substrate‐recognition sequences to target a specific RNA species (Fig. 4a). Trans‐acting ribozymes are true catalysts, as they retain their activity following cleavage of the substrate. A minimal hammerhead ribozyme sequence active in vitro has been identified, which consists of the catalytic core flanked by substrate‐recognition sequences that direct cleavage 3′ of NHH (N is any nucleotide and H is any nucleotide apart from guanine; Peng et al., ²⁰²¹). However, the cleavage activity of the minimal ribozyme is reduced in vivo at low concentrations of magnesium. The activity in these conditions is improved by the addition of tertiary stabilizing motifs, which bring about interaction between the ribozyme loops and increase the effectiveness both in vitro and in vivo (Carbonell et al., ²⁰¹¹). Further refinements include, for example, the construction of cis‐trans‐cis ribozyme cassettes, where self‐cleaving ribozymes liberate a trans‐acting ribozyme without flanking sequences that could interfere with folding or substrate specificity (Fig. 4b; Bussière et al., ²⁰⁰³).

Fig. 4.

Catalytic RNA and cap‐independent translation initiation. (a) Ribozymes acting in cis and in trans. A hammerhead ribozyme structure is illustrated. (b) A cis‐trans‐cis ribozyme transcribed from a suppression cassette. The trans‐acting ribozyme is excised from the construct and targets the gene of interest for cleavage. (c) An aptazyme cleaves plant mRNA in cis in response to theophylline. The cleaved mRNA is then degraded. (d) A suppression construct that contains two coding sequences separated by a ribozyme. The cleaved transcript is processed into small interfering RNA (siRNA) using an RNA‐dependent RNA polymerase (RDR) and a Dicer‐like endoribonuclease (DCL). The generated siRNA associates with Argonaute (AGO) and inhibits the translation or cleaves the mRNA of the target genes. (e) The Triticum mosaic virus (TriMV) internal ribosomal entry site (IRES) can bring about internal translation initiation, thereby allowing the translation of both coding sequences in an artificial bicistronic construct. The function is independent of initiation factor 4E, but requires initiation factor 4G (eIF4G), which interacts with the upper stem of the first IRES hairpin, and initiation factor 4A. (f) Cap‐independent translational enhancer (CITE) facilitates translation of uncapped mRNA in vitro. Initiation factors, and possibly also the 43S preinitiation complex (PIC), are recruited to the 3′ CITE and delivered to the 5′ end, from where the PIC scans. Pairing of the 5′ and 3′ structures circularizes the mRNA, similarly to the circularization of capped and polyadenylated mRNA (below). Blunt‐ended arrows indicate inhibition.

Trans‐acting ribozymes have been directed against viral or viroidal RNA to generate resistant transgenic plants and successfully reduced the severity of symptoms. Targeting the negative‐sense RNA of positive‐strand RNA viruses and viroids conferred stronger protection (Atkins et al., ¹⁹⁹⁵; Yang et al., ¹⁹⁹⁷; Huttner et al., ²⁰⁰¹). Although the use of a minimal ribozyme showed success against viruses and viroids, the effectiveness varied between plant species and the inheritance of the resistance phenotype was irregular due to RNA silencing interfering with ribozyme accumulation (Yang et al., ¹⁹⁹⁷; Han et al., ²⁰⁰⁰). Multimeric ribozymes, ribozymes fused with satellite RNA that exploit the viral replication machinery, and ribozymes with tertiary stabilizing motifs were more successful than minimal ribozymes (Kwon et al., ¹⁹⁹⁷; Carbonell et al., ²⁰¹¹).

Reduction of endogenous gene expression using trans‐acting ribozymes had varying levels of success. An approximately fourfold reduction in target mRNA abundance was achieved in transgenic maize with a multimeric ribozyme (Merlo et al., ¹⁹⁹⁸). A cis‐trans‐cis stabilized ribozyme cassette in transgenic potato attained a reduction of targeted mRNA abundance to 36–50% in a third of the transgenic plants (Bussière et al., ²⁰⁰³). While trans‐acting ribozymes offer precision in cleavage and lower risk of off‐target effects, RNA silencing has a stronger inhibitory effect (Akashi et al., ²⁰⁰⁵).

Synthetic ligand‐dependent ribozymes – aptazymes – have been engineered by coupling an aptamer domain with a ribozyme domain. Similarly to riboswitches, binding of a ligand to the aptamer leads to a structural rearrangement of the RNA, which activates or inactivates the ribozyme domain. A cis aptazyme is usually placed in the 3′ UTR so that translation initiation is not inhibited by the higher order structures, and the conditional cleavage destabilizes the mRNA (Win & Smolke, 2007; Peng et al., ²⁰²¹). Aptazymes responsive to multiple signals can be integrated into logic gates (Win & Smolke, 2008). Trans‐acting hammerhead aptazymes have been developed as well (Zhou et al., ²⁰²³). While aptazymes have chiefly been implemented in bacteria, yeast, and mammalian cells, a theophylline‐sensing hammerhead aptazyme successfully controlled the expression of plant nuclear‐encoded genes (Fig. 4c; Shanidze et al., ²⁰²⁰). This offers a prospect of developing new inducible expression systems, for example for plant molecular farming, which may be informed by the aptazymes developed for use in other organisms. Alternative aptamers would be necessary for applications in theophylline‐producing plant species. In addition to controlling gene expression, aptazymes on reporter genes can serve as noninvasive sensors of intracellular protein, metabolite concentration, or temperature (Win & Smolke, 2007; Park et al., ²⁰¹⁹). In yeast, a theophylline aptazyme on GFP (green fluorescent protein gene) was used to monitor the intracellular xanthine concentration (Win & Smolke, 2007), and it is conceivable that aptamers sensing alternative ligands could be applied for this purpose in plants.

Self‐cleaving ribozymes have been combined with RNA silencing (see the RNA silencing section) to reduce plant gene expression: truncation of sense and antisense transcripts to retain them in the nucleus induced RNA silencing (Fig. 4d; Buhr et al., ²⁰⁰²). Considering this application, it could be interesting to fuse an aptazyme sequence with a target plant gene sequence, even an artificial trans‐acting small interfering RNA gene, to induce RNA silencing that is not reliant on conditionally active promoters, similarly to a system constructed in mammalian cells by fusion of an aptazyme with a primary microRNA sequence (Kumar et al., ²⁰⁰⁹). In another application combining two mechanisms, ribozymes were used in the production of guide RNAs (gRNAs) for CRISPR‐Cas9 plant genome editing: They allow the placement of multiple gRNA sequences on a single, potentially conditionally regulated transcript, which is cleaved by cis‐acting ribozymes into the individual parts (He et al., ²⁰¹⁷).

4. Cap‐independent translation initiation

RNA viruses that replicate in the cytoplasm are excluded from nuclear capping machinery. Viruses that do not encode their own capping enzymes produce uncapped RNAs that initiate translation via cap‐independent mechanisms. Moreover, some RNA viruses that possess viral capping machinery still produce subgenomic RNAs that remain uncapped and are translated through alternative initiation strategies (Kneller et al., ²⁰⁰⁶). Functional RNA structures can enable translation initiation independently of the 5′ cap. An IRES is an RNA structure that serves as a platform for ribosome recruitment and assembly at an adjacent start codon. This recruitment avoids the canonical scanning of the PIC for start codons from the 5′ end. IRESs have been found in plant viruses, such as Triticum mosaic virus, the IRES of which is over 700 nt long. It forms two bulged stem–loop structures in the 5′ leader and has features similar to picornaviral IRESs (Roberts et al., ^{2015, 2017}; Miras et al., ²⁰¹⁷; Jaramillo‐Mesa et al., ^{2019, 2022}). Even some cellular genes, such as the maize heat shock gene HSP101 or the tobacco HSF1, contain an IRES, which ensures that their translation is sustained when cap‐dependent translation is downregulated (Dorokhov et al., ²⁰⁰²; Son & Park, ²⁰²³). An IRES can direct the translation of an additional ORF on the same RNA, a phenomenon that is well characterized in animal viruses (e.g. Dicistroviridae; Martinez‐Salas et al., ²⁰¹⁸). However, the mechanisms of IRES‐dependent translation and stress‐induced inhibition of cap‐dependent translation in plants are not yet well understood, and the mechanistic nature of plant virus intergenic IRESs, such as the one present in crucifer‐infecting Tobamovirus (crTMV), which is short and has a simple secondary structure, has been called into question (Dorokhov et al., ²⁰⁰²; Miras et al., ²⁰¹⁷; Son & Park, ²⁰²³).

While virus‐derived IRESs have been successfully deployed in animal cells and yeast to produce bicistronic transcripts, there has been mixed success in plants (Wang & Marchisio, 2021). IRES sequences derived from both animal and plant viruses led to co‐expression of genes from bicistronic transcripts in vitro, in transfected protoplasts, as well as in transgenic plants (Urwin et al., ²⁰⁰⁰; Dorokhov et al., ²⁰⁰²). The Triticum mosaic virus IRES facilitated bicistronic reporter gene translation in wheat‐germ extract independently of the 5′ cap, the cap‐binding eIF4E, or the poly(A) tail (Fig. 4e; Roberts et al., ^{2015, 2017}). Although a crTMV IRES could direct bicistronic gene co‐expression in transgenic rice, a StopGo sequence (see StopGo sequence) was markedly more effective (Ha et al., ²⁰¹⁰). This IRES was also successfully used to coproduce coronaviral proteins in Agrobacterium‐mediated transient expression experiments (Moon et al., ²⁰²²). However, in attempts to construct bicistronic expression vectors for Chlamydomonas, IRESs derived from an array of plant and animal viruses were found to be ineffective (Onishi & Pringle, 2016). The IRES from Encephalomyocarditis virus could not bring about bicistronic expression in transgenic rice (Jung et al., ²⁰¹¹).

Other elements used by RNA viruses to avoid the requirement for a 5′ cap or a poly(A) tail are CITEs, which include the Barley yellow dwarf virus translation element or a variety of CITEs in Tombusviridae (Meulewaeter et al., ¹⁹⁹⁸; Guo et al., ²⁰⁰¹; Fabian & White, ²⁰⁰⁴; Wang et al., ²⁰⁰⁹; Zuo et al., ²⁰¹⁰; Simon & Miller, ²⁰¹³; Truniger et al., ²⁰¹⁷). These structurally diverse elements bind translation initiation factors, particularly the cap‐interacting eIF4G or eIF4E, and some interact directly with ribosomal subunits. The effect of most 3′ CITEs depends on a long‐distance base‐pairing with a 5′‐end structure on the same RNA to deliver the recruited translation machinery to the 5′ end (Fig. 4f). In contrast to the IRES mechanism, the ribosomes recruited by a CITE scan for a start codon from the 5′ end rather than entering internally (Rakotondrafara et al., ²⁰⁰⁶; Nicholson et al., ²⁰¹⁰). The fact that CITEs are functional when placed in the genomes of other viruses demonstrates their modularity and presents a prospect to use them in artificial constructs (Nicholson et al., ^{2010, 2013}; Miras et al., ²⁰¹⁴).

CITEs enhance the existing capacity of the translational machinery to initiate at uncapped mRNA, which is particularly useful in in vitro translation systems: CITEs have been used to achieve high translational efficiency of uncapped mRNA with short UTRs in wheat‐germ extract (Ogawa et al., ²⁰¹⁴). CITEs increase the translation of uncapped mRNA delivered to protoplasts, although capping the mRNA led to higher expression levels (Guo et al., ²⁰⁰¹; Chattopadhyay et al., ²⁰¹¹; Fan et al., ²⁰¹²; Gao et al., ²⁰¹²). On the contrary, in Agrobacterium‐mediated transient transformation experiments, inclusion of synthetic CITEs did not increase protein yield compared with CITE‐free constructs, possibly due to unsuitable sequence context or because the element did not further enhance the translation of 5′‐capped mRNA (Peyret et al., ²⁰¹⁹). It would be interesting to find out if CITEs or IRESs reinforce gene expression under stress conditions, and if so, if this capacity could be exploited in engineering stress‐tolerant crops.

IV. Other RNA elements

1. StopGo sequence

In addition to IRESs, an RNA element that ensures simultaneous synthesis of two or more proteins encoded on a single transcript is the StopGo sequence, first identified in the Foot‐and‐mouth disease virus, where it is known as the 2A sequence. Originally believed to be an autocatalytic cleavage site in the viral polyprotein (Ryan et al., ¹⁹⁹¹), the StopGo element encodes a polypeptide sequence that causes a skip in the formation of a glycine‐proline peptide bond (‘ribosomal stutter’), release of the nascent chain by release factors, and continuation of translation, which generates a second polypeptide with an N‐terminal proline (Fig. 5a; Atkins et al., ²⁰⁰⁷; Donnelly et al., ²⁰⁰¹; Doronina et al., ²⁰⁰⁸; Sharma et al., ²⁰¹²). A 30‐aa‐encoding StopGo sequence is the shortest length that produces efficient stutter independently of the flanking sequences; reducing its length can tune the ratio of joined and separated products (Minskaia et al., ²⁰¹³). The major appeal of this element is its capacity to achieve co‐ordinated protein production, which is difficult to achieve otherwise, even when the genes have identical promoters (Halpin et al., ¹⁹⁹⁹). StopGo sequences are effective when introduced to a variety of eukaryotic cells or in vitro translation systems, including plant‐derived ones (Halpin et al., ¹⁹⁹⁹; Donnelly et al., ²⁰⁰¹; Ha et al., ²⁰¹⁰; Burén et al., ²⁰¹²; Zhang et al., ²⁰¹⁷; Lee et al., ²⁰²⁰; Larsen et al., ²⁰²³), which makes them very useful biotechnological tools. The disadvantage is that the StopGo‐derived C terminus is linked to the upstream‐encoded protein, which can disrupt its folding, interactions, localization, or activity. This problem is alleviated by encoding a modified intein domain upstream of the StopGo sequence that cleaves off the StopGo‐derived C terminus (Fig. 5b; Zhang et al., ²⁰¹⁷).

Fig. 5.

StopGo element, stop codon readthrough, and intron‐mediated enhancement. (a) StopGo element, such as the Foot‐and‐mouth disease virus 2A, allows coproduction of two polypeptide chains encoded in the same open reading frame. As the StopGo element is translated, the ribosome releases the polypeptide and resumes translation. The C terminus of the released polypeptide contains the 2A sequence. (b) Encoding a modified intein domain (intein*) upstream of the StopGo element allows the synthesized polypeptide to self‐cleave, removing the intein‐ and 2A‐derived C terminus. (c) ‘Leaky’ stop codons allow a small proportion of ribosomes to read through and continue synthesizing a protein isoform with an extended C terminus. (d) Enhancing introns can augment expression by multiple mechanisms that increase transcript accumulation and translation efficiency. CPMC, cleavage and polyadenylation molecular complex.

2. Stop codon readthrough

Specific stop codons can be redefined to encode an amino acid instead of termination, which allows a proportion of ribosomes to translate beyond the ‘leaky’ stop codon, yielding protein variants with alternative C termini (Fig. 5c). These events occur in multiple virus families, and hundreds of potential sites exist in plant transcriptomes (Firth et al., ²⁰¹¹; Sahoo et al., ²⁰²²; Zhang et al., ²⁰²⁴). Readthrough is facilitated in cis by the stop codon context or an adjacent secondary structure, but the stimulatory mechanisms are not clearly understood (Firth et al., ²⁰¹¹; Palma & Lejeune, ²⁰²¹). The more efficient use of nucleotide sequence allows viral genome compression, while in biotechnology, sequence length constraints are a lot more relaxed. Yet, stop codon readthrough has found multiple biotechnological applications, where a readthrough rate of 2–5% was induced by a UAGCAAUUA stop codon context. It was used to produce Tobamovirus particles with heterologous polypeptides on the surface, which have potential pharmaceutical use or confer heavy metal resistance (Hamamoto et al., ¹⁹⁹³; Fujiyama et al., ²⁰⁰⁶; Shingu et al., ²⁰⁰⁶). A reporter system was developed in Chlamydomonas where the CDSs of the investigated gene and a fluorescence gene were fused at a leaky stop codon, producing both a functional protein and a nonfunctional but trackable fluorescent fusion protein (Caspari, 2020).

3. Expression‐enhancing introns

Once transcribed, pre‐mRNA undergoes further processing to become mature mRNA before being exported out of the nucleus. These processing steps include the splicing of introns by the spliceosome, a ribonucleoprotein complex. In addition to alternative splicing, which produces mRNA with different combinations of exons and can affect expression by adding or removing regulatory RNA elements, the very presence of an intron can increase expression on the transcriptional and post‐transcriptional levels (Fig. 5d). Introns of certain plant genes were shown to exert effects specific to tissue, developmental stage, or stress. The phenomenon is called intron‐mediated enhancement (IME) and has been used in biotechnology of plants ranging from algae to angiosperms, even though its precise mechanisms, of which there are multiple, so far remain unclear. The proposed mechanisms include recruitment of transcription factors, chromatin remodelling, reduction in RdRP‐dependent RNA silencing, and inhibition of premature RNA cleavage and polyadenylation. So far, the clearest mechanism is association with proteinaceous trans‐factors, including exon junction complexes and serine‐arginine‐rich proteins, deposited during intron splicing, which can enhance nucleus export and translational efficiency (Pydiura & Blume, 2023; Zhong et al., ²⁰²³).

An expression‐enhancing intron has to be located in the proximity of the transcription start site and possess efficient splice sites, the nature of which can differ among plant clades (Zhong et al., ²⁰²³). On the contrary, 3′‐UTR introns are known to stimulate mRNA degradation by nonsense‐mediated decay, a pathway that removes mRNA with premature stop codons (Kertesz et al., ²⁰⁰⁶). The strength of IME varies with a range of intronic properties, such as length, position, and orientation, as well as intronic and flanking sequence composition, including the presence of certain motifs (Chung et al., ²⁰⁰⁶; Pydiura & Blume, ²⁰²³). Clarification of the molecular mechanisms and how they relate to specific intron properties will aid in the design of effectively enhanced or conditionally expressed constructs.

4. 3′ UTR elements

Most genes that are efficiently expressed require 3′ end polyadenylation of the mRNA, which promotes both its transcription and translation, ensures mRNA stability and export from the nucleus, and protects the transcript from RdRP‐mediated RNA silencing in plants. Cleavage and polyadenylation are signalled by cis‐acting elements, which are located chiefly downstream of the CDS and influence the gene expression level. In both stably integrated and transiently expressed plant expression constructs, the cleavage and polyadenylation signals are present in ‘terminators’ (distinct from prokaryotic terminators), also known as 3′ regulatory regions, the use of which in plant biotechnology was reviewed by Bernardes & Menossi (2020).

The cleavage and polyadenylation site determines the 3′ UTR of the transcript, which can contain a range of elements with stabilizing, destabilizing, or translation‐modulating effect. The elements can form higher order structures or constitute binding sites for silencing RNAs or proteins, most of which are adaptors for effector proteins or mediate co‐translational protein–protein interaction. Study of the protein‐binding sites, the RNA‐binding proteins, and the sequence properties can guide the engineering of 3′ UTRs, but it is complicated by the markedly different behaviour of the 3′ UTR elements in isolation and in a wider sequence context (Mayr, 2017). Identification of 3′ UTR inhibitory elements has enabled their deletion from the genome, which increased the expression of the modified endogenous genes (Wang et al., ²⁰²⁴). Also important for 3′ UTR design is the finding that, despite the stabilization of mRNA by G‐quadruplex structures, the strength of plant 3′ UTR secondary structure is inversely correlated with mRNA stability (Yang et al., ²⁰²²; Zhang et al., ²⁰²⁴). Another intriguing finding was the suppression of translation caused by a 3′‐UTR G‐quadruplex structure in the A. thaliana gene HIRD11 (Yang et al., ²⁰²⁰; Cao et al., ²⁰²⁴).

The 3′ UTRs in gene expression constructs can contain viral elements, such as the previously discussed CITEs, to increase the protein yield. A noteworthy element is the Y‐shaped structure in the Cowpea mosaic virus RNA‐2. This element enhances mRNA accumulation likely by protecting the mRNA from 3′‐exonuclease‐mediated degradation and has been effectively utilized in high‐expression transient expression systems (Meshcheriakova et al., ²⁰¹⁴; Peyret et al., ²⁰¹⁹).

5. RNA silencing

RNA silencing, the suppression of expression directed in a sequence‐specific manner by trans‐acting regulatory RNAs, is an important means of gene expression regulation involved in a wide range of processes from development, to stress responses, to antiviral defence. The process is set in motion by double‐stranded RNA (dsRNA), which is processed into fragments, associates with a multiprotein complex, and brings about cleavage or translational inhibition of complementary RNA. RNA silencing constitutes a useful alternative to gene deletion in cases where a deletion compromises plant growth or reproduction, and it has been extensively used to study gene function and to improve crop or decorative plants. RNA silencing in biotechnology was reviewed in detail by Ossowski et al. (2008); Tiwari et al. (2014); Verma & Modgil (2024).

6. RNA modification

RNA base modifications, such as to N⁶‐methyladenosine, can change the stability and translation of mRNA. The exerted effect on gene expression is important for a variety of processes, including stem cell fate determination, fruit ripening, and stress responses. Therefore, manipulation of RNA modification holds great promise in crop improvement, as reviewed by Tang et al. (2023) and Tang & Wang (2024).

V. RNA elements in engineered chloroplasts

An alternative to editing the nuclear genome is the modification of the chromosome in plastids. Benefits of this strategy, such as high‐expression levels (in part due to the high plastome copy number per cell) and the absence of mechanisms for silencing heterologous genes in plastids, may for certain purposes, outweigh its greater technical challenges (Bock, 2015). The wide variety of metabolic pathways occurring in plastids presents the option to boost the biosynthesis of useful endogenous metabolites, such as carotenoids, or introduce enzymes that use plastidial metabolic substrates to synthesize foreign metabolic products, such as bioplastics (Wurbs et al., ²⁰⁰⁷; Bohmert‐Tatarev et al., ²⁰¹¹; Bock, ²⁰¹⁵). Heterologous genes can also be expressed in other types of plastids, particularly amyloplasts and chromoplasts, to produce proteins of interest in nongreen tissues, such as roots and fruit, but the overall gene expression is markedly lower (Valkov et al., ²⁰¹¹; Zhang et al., ²⁰¹²; Caroca et al., ²⁰¹³; Bock, ²⁰¹⁵).

1. Start codon recognition

Plastids are believed to have arisen from ancient Cyanobacteria (Martin & Kowallik, 1999); as these are prokaryotes, their translational regulation differs substantially from the one in the eukaryotic cytosol. In prokaryotes, the small ribosomal subunit interacts with a ribosome‐binding site a short distance upstream of the CDS (Fig. 6a). An efficient ribosome‐binding site often contains the purine‐rich Shine–Dalgarno (SD) sequence, which forms base pairs with the 3′ end of the 16S rRNA (the anti‐SD sequence), and has a weak secondary structure (Wen et al., ²⁰²¹; Bryant et al., ²⁰²³). While the SD interaction with the anti‐SD in the 16S rRNA is typically important for efficient translation in plastids, Cyanobacteria and chloroplasts contain a large proportion of genes that lack an SD sequence (Scharff et al., ²⁰¹¹; Lomsadze et al., ²⁰¹⁸). While their 16S rRNA contains the conserved prokaryotic anti‐SD sequence, the 3′ end tends to form a secondary structure that obscures it (Weiner et al., ²⁰¹⁹). Successfully expressed heterologous genes inserted into plastid genomes are generally designed to include an SD sequence. Importantly for biotechnology, introduction of a modified 16S rRNA gene with an accessible anti‐SD sequence can enhance the translation of a plastidial gene of interest that has an SD sequence (Weiner et al., ²⁰¹⁹). Moreover, bacteriophage T7 gene 10 5′ UTR is used in plastids to enhance translation of the downstream CDS of interest (Olins et al., ¹⁹⁸⁸; Bock, ²⁰¹⁵).

Fig. 6.

Gene expression in plastids. (a) In plastids, multiple genes can be cotranscribed into a polycistronic mRNA, which is processed by nucleases (red scissors). 30S ribosomal subunits are recruited to ribosome‐binding sites (RBS), often by binding the Shine–Dalgarno sequence with their 16S rRNA, and the 50S subunit associates once the start codon is reached. Intercistronic expression elements (IEE) are the binding sites for pentatricopeptide/tetratricopeptide repeat (PPR/TPR) proteins, which are encoded in the nucleus and imported into the plastid. Binding of PPRs/TPRs increases mRNA stability and remodels its secondary structure, which may reveal ribosome‐binding sites or nuclease sites. mRNA is additionally protected from degradation by secondary structural elements. ORF, open reading frame. RNA pol., RNA polymerase. (b) In an artificial theophylline‐responsive expression system, a riboswitch controls plastidial gene expression by suppressing/enabling the synthesis of T7 RNA polymerase (T7 RNAP). The gene of interest is under the control of a T7 promoter (P_T7) and transcribed only by T7 RNAP. (c) An RNA thermometer controls gene expression by obscuring the RBS at low temperatures and revealing it at higher temperatures. Black blunt‐ended arrows show inhibition that aids gene expression, and red blunt‐ended arrows indicate inhibition that obstructs gene expression.

2. Polycistronic transcripts and mRNA processing

Another distinctive feature of prokaryotic and plastidial gene expression is the cotranscription of multiple genes onto a single polycistronic mRNA (Fig. 6a). This is particularly beneficial for engineering of metabolic pathways or producing protein complexes, as multiple genes can be joined into an operon for cotranscription (Bohmert‐Tatarev et al., ²⁰¹¹; Lu et al., ²⁰¹³; Lin et al., ²⁰¹⁴; Fuentes et al., ²⁰¹⁶; Long et al., ²⁰¹⁸). In contrast to bacteria, plastids often process mRNA: removing introns and cleaving polycistronic transcripts with nucleases into monocistronic fragments (Barkan, 2011; Bock, ^{2022, 2015}). The intercistronic nuclease cleavage is evidently not site‐specific, and monocistronic fragments probably are stabilized degradation intermediates (Barkan, 2011). The expression of some genes relies on or is improved by the nucleolytic processing into monocistronic fragments. This is partly because the unprocessed mRNA contains inhibitory secondary structures that hinder expression (Barkan et al., ¹⁹⁹⁴; Hirose & Sugiura, ¹⁹⁹⁷; Felder et al., ²⁰⁰¹). Other genes, even heterologous operons, are expressed even without the accumulation of monocistronic fragments (Staub & Maliga, 1995; Barkan, ^{2011, 1988}; Yu et al., ²⁰²⁰). Because the dependence of translational efficiency on nucleolytic processing is difficult to predict, constructed operons usually contain intercistronic expression elements (IEEs) between genes to increase the reliability of expression (Zhou et al., ²⁰⁰⁷; Bock, ²⁰²²). These elements can recruit nuclear‐encoded RNA‐binding pentatricopeptide/tetratricopeptide repeat (PPR/TPR) proteins, which protect the bound mRNA fragment from exonucleases, and remodel the local secondary structure, thus derepressing translation and possibly promoting intercistronic cleavage by exposing endonuclease sites (Felder et al., ²⁰⁰¹; Pfalz et al., ²⁰⁰⁹; Prikryl et al., ²⁰¹¹; Hammani et al., ²⁰¹²; Legen et al., ²⁰¹⁸; Macedo‐Osorio et al., ²⁰²¹). It is important to consider compatibility of IEE with native PPR/TPR proteins when introducing a heterologous sequence (Bock, 2022). Additionally, it is also important to account for the fact that an excessive presence of IEEs can sequester PPR/TPR proteins, potentially destabilizing endogenous transcripts (Legen et al., ²⁰¹⁸). This issue can be addressed by using alternative IEEs in the inserted or modified sequences and encoding a PPR/TPR protein in the nuclear genome that specifically recognizes these alternative IEEs (Rojas et al., ²⁰¹⁹; Yu et al., ^{2019, 2020}).

To achieve high protein synthesis levels, the engineered plastidial expression systems incorporate 3′ UTR sequences downstream of CDSs. These sequences protect the transcript or processed transcript fragment by forming a strong mRNA stem–loop structure (Stern & Gruissem, 1987; Zhou et al., ²⁰⁰⁷). An IEE also protects the 3′ end even of fragments lacking a stabilizing secondary structure (Legen et al., ²⁰¹⁸). The 3′ UTRs used in such systems can be derived from existing plastidial genes, such as rbcL (RuBisCO large chain gene), and may originate from other species (Lu et al., ²⁰¹³).

3. Conditional expression

Conditional expression of plastidial genes may be desirable in scenarios where high levels of their expression could be harmful to the plant, when gene expression is needed in specific tissues, or when suppressing the expression of a gene essential for survival or development is necessary. Conditional regulation can be achieved using RNA elements for post‐transcriptional control. For example, a modified PPR‐protein‐binding site that recruits an artificial PPR protein encoded in the nucleus under the control of an inducible or tissue‐specific promoter (Rojas et al., ²⁰¹⁹; Yu et al., ²⁰¹⁹). This method has demonstrated success in conditionally stimulating the expression of plastidial genes, with a 20‐fold increase in reporter accumulation upon induction and a 15‐fold reduction in nontarget tissues. In a similar system in Chlamydomonas, the THI4 riboswitch was used to control the production of a nuclear‐encoded TPR protein (Ramundo & Rochaix, 2015).

The introduction of a synthetic theophylline‐responsive riboswitch achieved tightly regulated inducible expression without the need to modify the nuclear genome, but with c. 10 times lower maximal expression level than a riboswitch‐free construct (Verhounig et al., ²⁰¹⁰). The expression was enhanced in an improved system utilizing the bacteriophage T7 RNA polymerase promoter, which is not recognized by the plastidial RNA polymerases. The T7 promoter controlled the target gene, and its transcription was activated by induction of a riboswitch‐controlled T7 RNA polymerase gene in the plastidial chromosome (Fig. 6b; Agrawal et al., ²⁰²²; Emadpour et al., ²⁰¹⁵). Despite this improvement, the system achieved a narrower expression range, with only sevenfold difference between induced and baseline expression, compared with the broader range achieved using the modified PPR protein system. Computational tools for design of prokaryotic riboswitches built from diverse RNA aptamers could guide the development of plastid‐based phytosensors (Espah Borujeni et al., ²⁰¹⁶), but their design will be complicated by the differences in riboswitch effectiveness between bacteria and chloroplasts, as was the case with glycine‐, adenine‐, and thiamine‐pyrophosphate‐responsive riboswitches (Verhounig et al., ²⁰¹⁰).

The Chlamydomonas chloroplast contains an RNA thermometer that obscures the SD sequence of psaA at 25°C and derepresses translation at higher temperatures. The thermometer was successfully used to control heterologous protein synthesis, and an engineered extension of the stem suppressed expression at the lower temperature while enabling inducibility at 35–40°C (Fig. 6c; Chung et al., ²⁰²³). As with riboswitches, a collection of thermometer elements that were functional in E. coli did not confer temperature‐sensitive expression in chloroplasts, although a possible reason is the lack of 5′ UTR stabilization rather than the ineffectiveness of the thermometers.

In bacteria, small noncoding RNAs (sRNAs) regulate gene expression by forming duplexes with mRNA, which can inhibit the mRNA translation or target it for degradation (Ahmed et al., ²⁰¹⁸). A variety of sRNA species have been detected in chloroplasts, some antisense to chloroplastidial genes, but their effect on gene expression requires further investigation (Anand & Pandi, 2021). Exploring the inducible production of sRNA complementary to the ribosome‐binding site of target genes could provide a potential strategy for conditionally suppressing gene expression in plastids.

It is noteworthy that despite the prokaryotic origin of plastids, their gene expression mechanisms have marked differences from the ones in bacteria, and this ought to be considered when adapting bacterial RNA elements for use in plastids.

VI. Perspectives

The discovery and study of the variety of RNA regulatory elements have opened up new avenues for their application as powerful tools in plant biotechnology, with benefits spanning from basic research to agricultural innovation. Primarily, but not exclusively, acting at the post‐transcriptional level, these elements enable modulation of heterologous and endogenous gene expression both in cis and in trans, facilitate co‐ordinated gene expression, and enable conditional regulation. Plant molecular genetics research, particularly of differential gene expression and RNA structure, is poised to expand this repertoire further, perhaps with the aid of artificial intelligence to predict RNA structure and interaction with proteins. Beyond plant genomes, plant viruses and subviral pathogens serve as a rich resource of functional RNA elements, optimized through evolution to manipulate host machinery and drive efficient gene expression. Plant viruses have additionally been adapted into autonomously replicating vectors, which we have not discussed in this review (Peyret & Lomonossoff, 2015).

Among the diversity of identified RNA elements (Table 1), many have untapped potential in plant biotechnology. First, the feasibility of employing certain elements, such as the newly characterized thermosensitive RNA elements or aptazymes, in plant biotechnology has already been demonstrated (Shanidze et al., ²⁰²⁰; Lastovka et al., ²⁰²⁴). Further optimization will enhance their practicality for real‐world application. Second, some RNA elements that have been successfully tested in animal cells or yeast are yet to be explored in plants. These include aptazyme‐based logic gates and a broader range of aptamers for translation‐regulating riboswitches which could expand the toolkit for gene regulation in plant systems (Win & Smolke, 2008).

Table 1.

Biotechnologically useful RNA regulatory elements.

Element	Location	Effect	References
Translational control
Kozak sequence	Translation initiation site	Crucial for translational efficiency. Tunable. Efficient sequences vary among plant clades	Lukaszewicz et al. (2000); Sugio et al. (2010)
uORF	5′ UTR	Inhibitory, but tunable. Its translational efficiency inversely correlates with the main ORF expression	Von Arnim et al. (2014); Zhang et al. (2018); Xing et al. (2020); Xue et al. (2023)
Translational enhancer	5′ UTR	Increased translational efficiency	Gallie (2002); Akbergenov et al. (2004); Kamura et al. (2005); Kanoria & Burma (2012); Peyret et al. (2019); Shen et al. (2023)
Secondary structural elements	5′ UTR, translation initiation site, 3′ UTR	A strong 5′‐UTR structure inhibits translation. The effect can be conditional. Possibly enhanced translation initiation at weak sites by hairpins just downstream	Wang et al. (2022); Cao et al. (2024)
Functional RNA domains
Riboswitch	Splice sites in vivo, UTRs in vitro	Inducible/suppressible gene expression by a variety of ligands	Mehrshahi et al. (2020); Tabuchi & Yokobayashi (2021)
Temperature‐sensitive elements	UTRs	Inducible gene expression by temperature	Thomas et al. (2022); Yang et al. (2022); Lastovka et al. (2024)
Ribozyme	3′ UTR or trans‐acting	RNA decay, potentially amplified by RNA silencing	Peng et al. (2021)
Aptazyme	3′ UTR	RNA decay inducible by a ligand	Shanidze et al. (2020); Peng et al. (2021)
Internal ribosomal entry site	5′ UTR, intercistronic region	Expression of genes from polycistronic transcripts	Urwin et al. (2000); Dorokhov et al. (2002); Roberts et al. (2015); Moon et al. (2022)
Cap‐independent translational enhancer	5′ and 3′ UTR	Efficient translation of uncapped RNA in vitro	Ogawa et al. (2014)
Other RNA elements
StopGo sequence	Coding sequence	Coproduction of multiple proteins encoded on the same transcript	Halpin (2005); Zhang et al. (2017)
Stop codon readthrough	Translation termination site	Coproduction of protein variants with alternative C termini	Hamamoto et al. (1993); Fujiyama et al. (2006); Shingu et al. (2006); Caspari (2020)
Expression‐enhancing intron	Proximity to transcription initiation site	Increased transcript accumulation and possibly also translational efficiency	Pydiura & Blume (2023); Zhong et al. (2023)
3′ regulatory region	3′ end	Essential for transcript stability, nuclear export, and translation	Bernardes & Menossi (2020)
CPMV RNA‐2 Y‐loop	3′ UTR	Enhanced transient gene expression, probably by mRNA stabilization.	Meshcheriakova et al. (2014); Peyret et al. (2019)
miRNA and ta‐siRNA	Trans‐acting	RNA decay, translational inhibition, transcriptional inhibition	Ossowski et al. (2008); Tiwari et al. (2014); Verma & Modgil (2024)
RNA base modification	UTRs or CDS	Altered RNA stability or translational efficiency, alternative polyadenylation	Tang et al. (2023); Tang & Wang (2024)
RNA elements in plastids
Shine–Dalgarno sequence	Translation initiation site in plastidial mRNA	Efficient translation initiation at a specific site	Bock (2015, 2022); Weiner et al. (2019)
T7 gene 10 5′ UTR	5′ UTR in plastidial mRNA	Increased translational efficiency	Olins et al. (1988); Bock (2015)
Intercistronic expression elements	Intergenic regions in plastidial operons	Transcript stabilization and gene co‐expression. Can rely on inducible proteins	Bock (2015, 2022)
Riboswitches and thermometers	5′ UTR in plastidial mRNA	Inducible gene expression by ligands or temperature	Verhounig et al. (2010); Chung et al. (2023)

CDS, coding sequence; miRNA, microRNA; mRNA, messenger RNA; ta‐siRNA, trans‐acting small interfering RNA; UTR, untranslated region.

Furthermore, some RNA elements require further investigation but show strong potential as valuable biotechnological tools. RNA localization elements or ‘postcodes’, primarily characterized in cereal endosperm cells, interact with trans‐acting factors that guide them to a specific subcellular location. This targeting is mainly achieved through interaction with cytoskeletal motor proteins (Tian et al., ²⁰²⁰). In addition to temporal control of expression conferred by inducible elements, RNA postcodes could offer means of intracellular spatial control and improve the assembly of multiprotein complexes or import of the encoded proteins into organelles. Additionally, long‐distance mobility has been described of mRNA and other major RNA classes through phloem, even to distinct organs. RNA elements, such as tRNA‐like structures or polypyrimidine motifs, have been linked to long‐distance transport; however, the underlying mechanism remains unclear and requires further investigation (Kehr & Kragler, 2018). It has been suggested that mobile RNA could be used in phytosensors to transmit a signal from the sensing tissue to readily accessible parts of the plant (Mahmudul et al., ²⁰²²). In addition, 3′‐end tRNA‐like structures have also been linked to RNA stabilization and translational enhancement, likely in a conditional manner, and further insight into their function and mechanism will allow us to evaluate their biotechnological potential (Wu et al., ²⁰²²).

Adaptation of RNA elements for biotechnological purposes is not always straightforward: elements like viral IRESs are functional in their native setting, but less effective in artificial systems (see the Cap‐independent translation initiation section) and would benefit from optimization. Sequence context, plant tissue, and the plant species can influence the effectiveness of certain RNA elements, which has to be taken into account before their introduction into the expression system of interest. While gaps exist in our knowledge of many RNA elements as well as of plant molecular genetics and stress responses, filling them will pave way for the development of improved expression systems and their application in plant engineering.

Competing interests

None declared.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.