The case of the missing allosteric ribozymes

Abstract

The RNA World theory encompasses the hypothesis that sophisticated ribozymes and riboswitches were the primary drivers of metabolic processes in ancient organisms. Several types of catalytic RNAs and many classes of ligand-sensing RNA switches still exist in modern cells. Curiously, allosteric ribozyme formed by the merger of RNA enzyme and RNA switch components are largely absent in today’s biological systems. This is true despite the striking abundances of various classes of both self-cleaving ribozymes and riboswitch aptamers. Herein we present the known types of ligand-controlled ribozymes and riboswitches and discuss the possible reasons why fused ribozyme-aptamer constructs have been disfavored through evolution.

Graphical Abstract

Many proteins are allosterically regulated by ligands that bind at a location distinct from the active site of an enzyme or the channel of a trans-membrane protein, which provides an efficient mechanism for regulating the production or transport of various biochemicals^1–3. Proteins are clearly adept at forming allosteric devices where a ligand binding event in one part of the polypeptide structure creates a structural change at a distal site to affect another function such as catalysis or transport. It is also known that RNA molecules can fold to form active sites that promote chemical reactions (Fig. 1A) or to form selective binding sites for metabolites or other biochemical targets (Fig. 1B). In recent decades, nucleic acid engineers have fused these two functions to create allosteric RNA constructs wherein ligand binding regulates the function of a ribozyme^4–10. Surprisingly, there have been very few examples of natural allosteric RNAs reported in the literature to date. While nature clearly is fully embracing protein-based allostery, there appears to be a strong bias against the use of allosteric RNAs by modern biological systems.

Fig. 1. The known classes of natural ribozymes and riboswitches.

(A) A total of 14 distinct ribozyme classes have been experimentally validated. The pie chart reflects the various general types of ribozymes and includes 13 classes that manipulate phosphate esters (gray shading) and one that synthesizes peptide bonds (ribosomes). (B) At least 45 distinct riboswitch classes have been experimentally validated, and many of these sense molecules that are derived from RNA nucleotides or their precursors. The pie charts are depicted in proportion to the number of classes of each type of RNA.

Allosteric ribozymes might have not been so scarce in ancient forms of life. In contemporary organisms, noncoding RNAs that mimic the functions of proteins are relatively rare but are believed to have predominated before ribosomes emerged in evolution to make the first genetically encoded polypeptides. At the height of this era called the RNA World^11,12, enzymes made of RNA are proposed to have run complex metabolic networks^13,14. Included in this collection of ribozymes are presumed to be all catalytic activities needed to make the common purine and pyrimidine ribonucleotides, capture energy and matter for the production of other valuable biomolecules, and synthesize polynucleotides that serve as genomic molecules or biocatalysts. Presumably, the ribozyme representatives we have today (Fig. 1A) are just a tiny sampling of the diversity of classes that existed before the appearance of protein enzymes.

Also, RNA-based receptors and switches would have been needed to regulate the production and action of ancient ribozymes in response to the concentrations of various metabolites or other signals. Among the essential biomolecules of the RNA World would have been many of the ubiquitous enzyme cofactors and second messengers we see in modern cells^13,14, which are replete with nucleotide-like chemical moieties¹⁵. Interestingly, most of the ~50 known classes of riboswitches found in modern bacteria selectively sense and respond to RNA building blocks or their precursors, nucleotide-like coenzymes, and signaling molecules derived from RNA nucleotides^16,17 (Fig. 1B). Given their ligand specificities, abundances, and phylogenetic distributions, it has been proposed that some of these riboswitch classes might have originated during the RNA World^18,19. Indeed, some riboswitches might be direct descendants from analogous ribozymes that used these ligands as coenzymes or substrates in ancient chemical reactions²⁰, or as allosteric effectors.

Importantly, during the RNA World, it seems possible (or inevitable) that aptamers and ribozymes would have worked in tandem, with ligand binding to the aptamer domain resulting in allosteric control over the action of its ribozyme partner. To assess the need for allosteric ribozymes in the RNA World, it is important to consider the level of complexity that metabolic processes would have required for ancient life forms to have thrived. One way to address this issue is to consider the metabolic state required to give rise to primitive ribosomes that fused amino acids to make the first polypeptides. It seems likely that the metabolic pathways needed to reach this era were remarkably complex. In addition to promoting the metabolic processes to make nucleic acids and many other supportive processes, advanced cells that began the transition from the RNA World to one that used proteins would have needed to produce at least a few amino acids as well as the ribozymes needed to activate these molecules and selectively join them via amide (peptide) bonds.

A near complete collection of the ubiquitous coenzymes present in today’s cells might also have been present in RNA World species, to be used by ribozymes to help make many of these essential metabolites^13,14. Furthermore, there might have been a large collection of nucleotide-based signaling molecules like those common in modern bacteria, such as cAMP, ppGpp, c-di-AMP and c-di-GMP, which would have been employed to help regulate these sophisticated metabolic and physiological pathways²¹. Instead of binding to protein receptors as they usually do currently, these signaling molecules would have been sensed by ancient riboswitches to modulate the production and activity of ribozymes associated with pathways and processes in need of regulation.

Some of the riboswitch classes in modern cells that sense these RNA-like compounds are exceedingly common in the bacterial domain of life¹⁷. Furthermore, there are 14 classes of well-validated ribozymes present among extant species (Fig. 1A), including nine classes of self-cleaving ribozymes²². Some self-cleaving RNA classes such as the hammerhead^23,24, HDV²⁵ and twister²⁶ ribozymes have up to several thousand representatives among the collection of sequenced bacterial genomes. Unexpectedly, in a recent, near-comprehensive search for riboswitches and ribozymes that reside in tandem arrangements in bacterial species (Breaker Laboratory, unpublished findings), we failed to find any previously undiscovered examples where a riboswitch aptamer and a ribozyme appear adjacent to each other as domains within a single transcript (Fig. 2). Therefore, the only allosteric ribozyme with this arrangement that has been experimentally validated to date is formed between an aptamer for c-di-GMP and a group I self-splicing ribozyme²⁷, as discussed in greater detail below.

Fig. 2. General architectures of engineered and natural ligand-controlled ribozymes.

(A) Allosteric ribozyme engineers have exploited arrangements wherein preexisting aptamer and ribozyme domains are arranged in serial (left) or wherein an aptamer is embedded within a structurally sensitive substructure in the interior of the ribozyme (right). Ligand binding either disrupts or stabilizes folding of the active ribozyme structure. (B) Riboswitch regulation of translation involves the regulation of peptidyltransferase ribozymes (ribosomes) by controlling intermolecular interactions between the mRNA and ribosomes. (C) The architecture of glmS ribozymes includes docking of the ligand GlcN6P at the active site, where it functions as a cofactor to promote the internal RNA phosphoester transfer reaction. (D) Group I self-splicing ribozymes bind to the small molecule substrate guanosine (or any of its 5ˊ-phosphorylated derivatives) to trigger the first of two phosphoester transfer reactions. (E) Eukaryotic riboswitches exploit aptamers to permit ligand-dependent (TPP) regulation of splice site access by spliceosomes. SS designates splice site. (F) True allosteric regulation is exhibited by group I self-splicing ribozymes whose 5ˊ splice site access is regulated by ligand (c-di-GMP) binding to an adjacent aptamer. In each model, red is the binding site for the small-molecule ligand, black is the ribozyme, and blue is the expression platform.

One partial answer for this deficit could be that all the self-cleaving ribozyme classes promote single turn-over reactions, and therefore they simply do not need allosteric control. However, these RNA-cleaving ribozymes could have easily been adapted by nature to serve as expression platforms for riboswitches. An allosteric single turn-over self-cleaving ribozyme could regulate gene expression much like an intrinsic terminator stem, which also works only once to terminate mRNA transcription (see discussion below). The presence or absence of the allosteric effector could thereby regulate mRNA stability via an allosteric self-cleaving RNA. Likewise, such an allosteric ribozyme could respond could cleave off the ribosome binding site (RBS) of an mRNA in a manner dependent on the allosteric effector, which again needs to happen only once to stop translation initiation.

Strangely, not a single allosteric ribozyme is known that functions as an expression platform by associating with a riboswitch aptamer. Thus, even though there are many representatives of known ribozyme and riboswitch classes, there are surprisingly few natural examples of allosteric ribozymes that integrate both the functions of a sensor/switch and an RNA enzyme. In the sections below, we describe the previously discovered examples of allosteric ribozyme systems and present some of the possible reasons why this very powerful and ancient RNA technology might have been largely abandoned by today’s organisms.

Architectures of engineered and natural allosteric RNAs

The ease of engineering allosteric ribozymes was demonstrated in the laboratory more than 20 years ago by researchers who exploited the modular nature of aptamers and ribozymes^4–10. RNA components, drawn from the known molecular “parts bin”, could be readily fused in a manner that retained the functions of the separate domains to yield constructs that function as designer allosteric ribozymes (Fig. 2A). For example, simple judicious grafting of aptamers onto structurally sensitive parts of hammerhead ribozymes gave rise to the first examples of allosteric self-cleaving RNAs that are activated or deactivated by binding to ATP⁴. More elaborate versions were made using two aptamers whose ligands bind cooperatively to maximally promote ribozyme action⁷. Subsequent efforts have been directed in pursuit of allosteric ribozymes for biosensing and gene control applications^9,10. Simply put, allosteric ribozymes might have initially been a speculative output of the RNA World theory, but RNA switch engineers are producing functional devices that are being developed for practical applications^28,29. This remains an active research area within the field of synthetic biology, and new demonstrations and improvements in the technology are continuing to be reported in the scientific literature.

Given the abundances of some natural self-cleaving ribozyme classes, it seems likely that genomic rearrangements that play out over evolutionary time would occasionally cause a riboswitch aptamer to be located adjacent to a self-cleaving ribozyme. By chance, some of these genomic rearrangements should result in architectures wherein ligand binding to the aptamer domain results in allosteric control of RNA strand scission by the ribozyme. Nature could achieve by chance what ribozyme engineers have done by rational design or by using directed evolution methods. Over eons, allosteric ribozymes that are beneficial to cells should have been retained and enhanced if they are competitive with other forms of biochemical and genetic regulation mechanisms, even in today’s cells.

As discussed further below, there indeed are a few natural examples of ligand-controlled ribozymes wherein the process of natural selection might have exploited the modular nature of RNAs. However, we believe that there are far fewer examples of natural allosteric ribozymes in modern cells than would be expected given the modular nature of RNA, the likely RNA-rich evolutionary history of life, and the relative ease of creating synthetic RNA switches. Thus, it is interesting to consider possible explanations for the dearth of allosteric ribozymes. Perhaps these RNAs are less effective than other gene control systems such as those made of protein, or perhaps they are more prone to loss because they are larger genetic targets for disruption than are other mechanisms of equal utility.

Most known riboswitch classes use ligand binding in a single aptamer domain to regulate gene expression via the action of an ‘expression platform’ (Fig. 3). These expression platforms primarily exploit one of two exceedingly simple mechanisms^19,30. The first expression platform type employs an intrinsic terminator stem (a strong base-paired hairpin followed by a run of U nucleotides). When formed, this architecture signals RNA polymerase to pause and then terminate transcription of the nascent mRNA^31,32. The second common expression platform type employs a simple base-paired region involving nucleotides of the ribosome binding site³³. Formation of this ‘anti-RBS’ stem occludes access by 16S ribosomal RNA and prevents translation initiation of the downstream open reading frame of the mRNA (Fig. 3).

Fig. 3. Two predominant expression platform mechanisms for riboswitches.

Both function by ligand-triggered folding pathway changes involving alternative hairpin formation. Red regions are portions of the aptamer that can base-pair with nucleotides that form structures involved in promoting or preventing gene expression. Transcription control involves the ligand-regulated folding of an intrinsic terminator stem (strong stem followed by a run of U nucleotides), whose formation terminates RNA polymerase progression on the DNA template. Translation control involves the ligand-regulated folding of an anti-terminator stem, whose formation prevents ribosome binding and subsequent translation of the ORF. Likewise, simple ligand-dependent alternative folding of overlapping aptamer and ribozyme domains can yield allosteric ribozyme constructs.

For both riboswitch expression platform systems described above, the nucleotide sequence and structural requirements for the switch device appear to be fewer than the requirements for forming the active site for a high-speed ribozyme, even when only considering the small self-cleaving RNA classes. The simple base-paired region of a terminator hairpin or an anti-RBS stem of riboswitch expression platforms usually can tolerate one or more mutations without completely losing its function. In contrast, ribozymes require the formation of multiple conserved stems, and require the precise three-dimensional positioning of numerous strictly conserved nucleotides. Thus, blending the functions of an aptamer with a ribozyme in a manner that adds new functionality (allosteric regulation) probably requires higher ‘information content’³⁴ compared to the common riboswitch systems that regulate the formation of terminator or anti-RBS stems. In other words, allosteric ribozyme systems are probably larger genetic targets that are more likely to be disrupted by single mutations, which are common events in evolution.

It is important to note for this discussion that ribosomes are actually ribozymes that have an active site for peptide bond formation formed entirely by RNA^35,36. Therefore, a riboswitch that regulates RBS availability is regulating the function of a ribozyme by allosterically controlling ribosome access to its mRNA template. Although this riboswitch mechanism is exceedingly common, it does not quite represent the single-RNA allosteric ribozyme architecture that we believe is underrepresented in nature today. However, riboswitches that regulate RBS availability reduce the risk of mutational disruption by exploiting a separate ribozyme complex, a ribosome, such that each regulatory device has no need to preserve its own tethered ribozyme domain.

The additional evolutionary demands placed on unimolecular allosteric ribozymes do not preclude their emergence or use in modern cells, just that they should be expected to be far rarer than systems that achieve the same biochemical objectives with a lower risk of mutational disruption (reduced genetic liability). Indeed, extant cells are not entirely devoid of ribozymes whose actions are controlled by the binding of small molecules. In the following three sections, we describe ligand-dependent ribozyme reactions that are abundant in modern cells, but that have characteristics that cause us to classify them as distinct from the allosteric ribozymes we are seeking.

glmS self-cleaving ribozymes

An exceptionally common class of ligand-dependent RNAs is comprised of glmS self-cleaving ribozymes³⁷. These gene control elements are almost exclusively located in the 5ˊ untranslated regions (UTR) of glmS genes³⁸, which code for a protein enzyme that produces the modified sugar molecule glucosamine-6-phosphate (GlcN6P). Each glmS ribozyme is activated by GlcN6P binding, which triggers mRNA cleavage and the eventual degradation of the coding region by RNase J1³⁹. Thus, each glmS ribozyme also serves as a riboswitch that represses expression of the enzyme that otherwise produces more of the metabolite ligand.

However, glmS ribozyme-riboswitch RNAs are not allosteric (Fig. 2C and Fig. 4, Type I). Rather, they use GlcN6P as a cofactor to promote RNA strand scission by an internal phosphoester transfer reaction. Specifically, biochemical^40–43 and structural^43,44 evidence indicates that the amine group of GlcN6P is important for protonating the 5ˊ-leaving group to accelerate the RNA cleaving reaction. This ribozyme function requires a relatively sophisticated RNA structure, and the subsequent control of gene expression relies on the action of a second, protein-based nuclease (RNase J1) to degrade the ORF. This brings up a critical question relevant to the present discussion. Why do the vast majority of riboswitches use allosteric control of simple, non-ribozyme expression platforms such as intrinsic terminator stems (transcription termination) or use anti-RBS stems to regulate a ribozyme in trans (translation control) (Fig. 3), whereas glmS riboswitches use the direct action of a cofactor-dependent ribozyme?

Fig. 4. Types of ribozymes naturally regulated by the binding of small molecules.

Parentheses identify the compounds (cofactor, riboswitch ligand, or substrate) that regulate the function of the ribozyme. Filled circles identify which domain of life is known to employ each ribozyme system.

The following fact might help explain the unusual choice nature has made to use a cofactor-dependent ribozyme to regulate gene expression. Unlike the cognate ligands of most other riboswitch classes, the target ligand for glmS ribozymes (GlcN6P) is highly similar in chemical structure to glucose-6-phosphate and to various other hexose phosphate isomers that are exceedingly abundant in cells. Indeed, the hexose-phosphate pool in E. coli cells has been measured at nearly 1 mM⁴⁵, which is well in excess of the concentration of GlcN6P needed to fully activate glmS ribozymes^37,40. Cells must therefore employ a genetic switch that precisely discriminates against hexose phosphate molecules that lack the 2-amino modification.

Perhaps glmS ribozymes use two biochemical steps to strongly reject compounds that are both closely related to GlcN6P and high in cellular concentration. The first specificity check involves simple ligand binding, which is like that used by almost any other riboswitch aptamer. The danger of using only ligand affinity to establish selectivity is apparent when considering the high concentrations of competing hexose sugars in cells such as glucose, fructose and their phosphorylated derivatives, which create a particularly difficult molecular recognition challenge.

Specifically, it is known that glucose-6-phosphate can be bound by glmS ribozymes^40,43,46, and that high concentrations of such competitors can at least modestly inhibit glmS ribozyme function even in cells⁴⁶. If a glmS ribozyme used its ligand only as an allosteric trigger, it would occasionally be improperly triggered by other hexose sugars. This is likely true for all conventional riboswitches which might be falsely triggered by a high concentration of a close analog. However, glmS ribozymes inherently employ a second specificity check by the requirement of a properly positioned amine group in their active sites, which must be present to promote mRNA strand scission and subsequent downregulation of translation. Thus, cells might require a cofactor-dependent ribozyme to exploit a catalytic mechanism that ensures GlcN6P ligand specificity⁴³. This dual specificity requirement (binding and cofactor function) cannot be achieved by a typical riboswitch aptamer or by an allosteric ribozyme because both would be triggered only by ligand binding. Therefore, the special features offered by a cofactor-dependent ribozyme might explain why the unusual function of glmS ribozymes is so widely used by bacterial cells.

Group I self-splicing RNAs

Another ligand-dependent ribozyme class is abundantly represented by group I self-splicing RNAs⁴⁷. These ribozymes are well-known for initiating their first phosphoester transfer reaction upon the binding of guanosine or any of its 5ˊ-phosphorylated derivatives such as GTP. For this ribozyme class, a compound such as guanosine or GTP serves as a substrate by providing its 3ˊ-oxygen atom as the attacking nucleophile (Fig. 2D and Fig. 4, Type II), and therefore the ligand is neither a cofactor nor an allosteric effector.

Note that certain common forms of guanosine derivatives, such as cGMP, c-di-GMP, and ppGpp cannot function as substrates to trigger splicing by the ribozyme. These compounds have been proposed to be signaling molecules used by RNA World organisms^21,48, but they lack a free 3ˊ-hydroxyl group that would permit ribozyme-mediated nucleophilic attack at the 5ˊ splice site. Two distinct riboswitch classes have been discovered that respond to c-di-GMP^49,50 and another class is controlled by ppGpp⁵¹. Both of these signaling molecules are synthesized from GTP under certain physiological conditions. Perhaps the depletion of GTP (and thus guanosine and other 5ˊ-phosphorylated guanosine derivatives) causes some group I ribozymes to splice less frequently, which would suppress gene expression.

Indeed, it has been proposed that group I ribozymes in some genetic settings do function as a form of metabolite-triggered riboswitch⁵⁰. For the vast majority of these ribozymes, they are unlikely to be regulated in an allosteric fashion. It is known that arginine can inhibit group I ribozyme function by competing with guanosine or GTP binding by the ribozyme⁵². However, splicing inhibition does not proceed via an allosteric process and is not known to occur in cells as a natural form of gene control. Therefore, neither the typical configuration for group I ribozymes, nor the observed inhibitory effect of arginine, represents the missing allosteric ribozymes we are seeking.

Riboswitch Control of Nuclear mRNA Splicing

Intriguingly, a different type of riboswitch-mediated regulation of RNA splicing is observed in many species of eukaryotic organisms. Specifically, there are several well validated examples of riboswitches for the enzyme cofactor thiamin pyrophosphate (TPP) embedded in the introns of various algae, plants and fungi^53–59. TPP binding regulates alternative splicing of introns located in the 5ˊ UTRs⁵⁵, the 3ˊ UTRs^56,57, or within the coding regions^58,59 of mRNAs related to the transport or biosynthesis of this ubiquitous enzyme cofactor. These intron-based riboswitches regulate spliceosome action by controlling access either to splice sites^55–58 or to the binding sites of splicing factor proteins⁵⁹ within certain pre-mRNA substrates. Ligand binding to the intron either directly or indirectly regulates the function of spliceosomes. It is notable for this discussion that each spliceosome carries a catalytic core made of RNA, which is believed to be a direct descendant of ribozymes based on the catalytic architecture of group II self-splicing RNAs⁶⁰.

Therefore, the eukaryotic TPP riboswitches in introns also regulate the action of ribozymes by controlling alternative structure folding of an RNA substrate, much like some more conventional riboswitches regulate the ribozymes that make peptide bonds (Fig. 2E and Fig. 3, Type III). As noted above, many bacterial riboswitches regulate the action of the peptidyl transferase (amide synthase) ribozyme at the heart of each ribosome - not by direct allosteric control of the ribozyme, but by regulating which mRNAs are bound by the ribozyme. Given that the catalytically active structures of the active sites of ribosomes and spliceosomes require docking with their RNA targets, the ligand-dependent regulation of intron or mRNA folding also could be considered a type of allosteric regulation. Ligand binding to the aptamer of a eukaryotic TPP riboswitch can alter access to the splice site and thus alter the active site structure of the ribozyme, although the ligand binding site is in a separate RNA strand, not covalently linked to the ribozyme. Yet again, however, these sophisticated multi-molecular RNA devices do not quite fit the description of the missing allosteric ribozymes formed by covalently linking aptamers and ribozymes.

True Allosteric Group I Ribozymes

Although exceedingly rare, examples of true, single-RNA allosteric ribozymes do exist in modern cells (Fig. 2F and Fig. 4, Type IV). The bacterial species Clostridioides difficile (previously known as Clostridium difficile) carries a remarkably intricate riboswitch formed by the fusion of an aptamer and a group I self-splicing ribozyme that exhibits allosteric control within a single RNA transcript⁵⁰ (Fig. 5). The ligand-binding domain is a c-di-GMP aptamer (class II⁵⁰) that senses a bacterial second messenger comprising a circular RNA dimer formed with guanosine monophosphates joined via two 3ˊ,5ˊ-phosphodiester linkages. When c-di-GMP is bound by the aptamer, the adjacent ribozyme allosterically adopts the structure necessary to promote the nucleophilic attack by GTP at the proper 5ˊ splice site⁵⁰. Thus, when both c-di-GMP and GTP are present, the favored splicing reactions yield a processed RNA transcript that can be translated to yield the desired protein product⁶¹.

Fig. 5. A true allosteric ribozyme formed by a single RNA transcript.

The colocalization of a c-di-GMP-II riboswitch aptamer and a group I self-splicing ribozyme permits allosteric control of RNA splicing⁴⁸. When c-di-GMP is bound, the aptamer folds as depicted, which permits the formation of the ribozyme base-paired substructure called P1. The ribozyme, a ~600 nucleotide domain immediately following the aptamer, then promotes attack by GTP (here called GTP₁) at the normal 5′ splice site (5′ SS). The splicing process continues via attack by the newly liberated 3′ oxygen atom of nucleotide 101 on the 3′ SS located more than 500 nucleotides downstream. This yields an intact mRNA that is ready for translation by ribosomes⁵⁹. In the absence of c-di-GMP binding, the precursor RNA folds differently. Nucleotides shaded blue base-pair to preclude the formation of P1, thus permitting alternative base-pairing of the nucleotides shaded in orange (P1* stem). This causes GTP (called GTP₂) to attack at a distal position, thereby generating a ribozyme splicing product that lacks an RBS and therefore cannot be translated⁵⁹. Thus, each precursor RNA transcript functions as a two-input Boolean AND logic gate (see the truth table) wherein both c-di-GMP and GTP are required for gene expression to be ‘ON’⁴⁸. Figure adapted from a previous publication⁴⁸.

Numerous other tandem riboswitch systems are known that stack aptamers from different classes to create Boolean logic gates^51,62–65. The allosteric aptamer-ribozyme system from C. difficile is naturally arranged to form a genetic “AND” gate, wherein maximal gene product formation is expected when both c-di-GMP and GTP are abundant. Presumably, this bacterial species monitors the concentrations of both c-di-GMP and its immediate metabolic precursor GTP to determine the level of expression of the associated gene^50,61. This allosteric ribozyme is particularly intriguing because it uses only RNA components to build an intricate gene control system. Specifically, it uses (i) an RNA aptamer to sense (ii) an RNA-derived signaling molecule, which activates (iii) an RNA enzyme that uses (iv) an RNA nucleotide substrate to initiate the splicing of a precursor mRNA transcript. Furthermore, accurate riboswitch-regulated splicing by the group I ribozyme yields a processed mRNA that carries a functional RBS, which of course is recognized by yet another ribozyme (a ribosome). Currently, this highly interactive system represents the most sophisticated RNA-driven device known, which provides an opportunity for researchers to examine its function and to reflect on how RNA World organisms could have reached a high level of metabolic and genetic complexity without relying on protein factors.

Obviously, the conjoined c-di-GMP aptamer and group I ribozyme system is large, and its structure is far more complex than those of typical riboswitch classes. Thus, the genetic price for achieving this Boolean AND gate capability (Fig. 5) is high because cells presumably must preserve the collaborative function of the aptamer, the ribozyme, and their functional interplay, despite natural mutations that occur over evolutionary time. Earlier, it was noted that a high information content might be one reason why allosteric ribozymes are rare among extant species. This tandem arrangement between a c-di-GMP aptamer and a group I ribozyme is even more complex than a similar fusion imagined between an aptamer and any of the known self-cleaving ribozyme classes. So, why do we see rare examples of the former arrangement, and few (if any, see below) examples of the latter? Given the fact that self-splicing ribozymes are more abundant than self-cleaving ribozymes, and that self-splicing ribozymes are frequently components of mobile genetic elements^66,67, perhaps there are simply more chances for allosteric self-splicing ribozymes to emerge in evolution compared to allosteric self-cleaving ribozymes, even if these composite devices do not persist for long against constant assault from natural mutations.

Concluding Remarks

RNA engineers have long ago demonstrated that allosteric ribozymes can be created by joining preexisting aptamer and catalytic domains using either rational design or directed evolution methods^4–10. Given the relative ease with which researchers can combine otherwise independently functioning RNA domains to yield RNA switches, it seems certain that natural evolution also could harness these roles in a similar manner. Several types of natural ribozymes are much larger than self-cleaving ribozymes, and thus present even more opportunities for the evolution of allosteric function. Intriguingly, ribosomal RNAs are known to experience shape changes in response to the binding of small molecules such as antibiotics^68–70. Whether short and simple, or long and structurally complex, many types of allosteric ribozymes could have been exploited by RNA World organisms long ago. Any ribozyme function that could have benefitted from regulation in response to the concentrations of substrates, products, or signaling molecules would be a candidate for allosteric control.

These speculations, and the other RNA switch examples described above, help us refocus on our original question: where are the modern examples of natural allosteric ribozymes made by fusing aptamers and self-cleaving RNAs? It has been proposed⁷¹ that HDV-like self-cleaving ribozymes associated with the glmM (phosphoglucosamine mutase) gene of Faecalibacterium prausnitzii are regulated by the presence of GlcN6P. We already know that GlcN6P is a powerful cofactor for promoting the action of glmS ribozymes, which serve as riboswitches for controlling genes related to this modified sugar^37–44. Therefore, a precedent exists for this type of genetic control. Albeit, the effects on HDV-like self-cleaving ribozymes are small (~2 fold), requiring high concentrations of GlcN6P (20 to 30 mM), and no distinct aptamer domain exists to sense the allosteric effector. If this phenomenon is biologically relevant, it does not representative of the robust, multidomain allosteric RNA devices that once might have predominated in the RNA World.

These HDV-like ribozyme examples, as well as the natural riboswitch systems described above, again only provide a glimpse of the types of RNA-based molecular devices that presumably could be deployed by modern cells. It is now abundantly clear that cells can exploit the modular nature of functional RNAs to build more complex devices. Tandem riboswitches^62–65 and allosteric group I ribozymes^50,61 help illustrate these capabilities. However, what we now know also serves to highlight the glaring absence of allosteric self-cleaving ribozymes in nature. There should be plenty of opportunities for ligand-responsive ribozymes to serve as riboswitches in modern cells.

There are still opportunities to discover examples of aptamers in productive configurations with self-cleaving ribozymes. Many thousands of riboswitch classes remain to be discovered¹⁷ and directed evolution experiments⁷² suggest that there are also many different RNA architectures that can operate as self-cleaving ribozymes. Regardless, it seems unlikely that researchers have missed a large collection of allosteric ribozymes in the genomes of organisms that have already been sequenced. Therefore, we might need to form an intellectual basis for coming to terms with the likelihood that allosteric ribozymes are rarer in modern cells than random chance would suggest.

Despite the biases against allosteric ribozymes evidently caused by the forces of evolution, we remain very optimistic that synthetic RNA devices that exploit fused aptamer and ribozyme constructs can be made to work well in cells. The joining of aptamers and ribozymes in a manner that yields precise sensors and switches is relatively straightforward for nucleic acids engineers, and proofs of principle abound in the literature^4–10,28,29. To successfully harness this likely ancient RNA technology, the goal is to create fused RNA constructs that retain the high affinity and specificity of the aptamer domain with the robust activity of the ribozyme domain. Several challenges have been identified and must be overcome to create allosteric ribozymes that have useful applications in cells⁷³. However, by reverse engineering natural ribozymes, riboswitches and their rare natural allosteric combinations, researchers can exploit the same mechanisms used by nature to build entirely new complex RNA devices.