A comparative survey of the influence of small self-cleaving ribozymes on gene expression in human cell culture

ABSTRACT

Self-cleaving ribozymes are versatile tools for synthetic biologists when it comes to controlling gene expression. Up to date, 12 different classes are known, and over the past decades more and more details about their structure, cleavage mechanisms and natural environments have been uncovered. However, when these motifs are applied to mammalian gene expression constructs, the outcome can often be unexpected. A variety of factors, such as surrounding sequences and positioning of the ribozyme influences the activity and hence performance of catalytic RNAs. While some information about the efficiency of individual ribozymes (each tested in specific contexts) is known, general trends obtained from standardized, comparable experiments are lacking, complicating decisions such as which ribozyme to choose and where to insert it into the target mRNA. In many cases, application-specific optimization is required, which can be very laborious. Here, we systematically compared different classes of ribozymes within the 3’-UTR of a given reporter gene. We then examined position-dependent effects of the best-performing ribozymes. Moreover, we tested additional variants of already widely used hammerhead ribozymes originating from various organisms. We were able to identify functional structures suited for aptazyme design and generated highly efficient hammerhead ribozyme variants originating from the human genome. The present dataset will aide decisions about how to apply ribozymes for affecting gene expression as well as for developing ribozyme-based switches for controlling gene expression in human cells.

KEYWORDS:

• Ribozyme
• hammerhead
• hairpin, HDV, pistol, twister, twister sister
• gene expression
• self-cleavage
• human cell culture

Introduction

The discovery of ribozymes, catalytic RNAs, was a breakthrough in many scientific fields that was awarded shortly after with the Nobel Prize in Chemistry in 1989 [1,2]. Since then, many fundamental studies have been published about the general mechanisms of ribozyme catalysis [3–7]. In synthetic biology, the family of small self-cleaving ribozymes is of particular interest. They catalyse RNA strand cleavage via intramolecular transesterification reactions of the phosphodiester backbone [8]. Until to date, 12 classes of natural self-cleaving ribozymes are known mostly with detailed information about their cleavage mechanisms, structure, and natural environments [8–11].

The family of self-cleaving ribozymes includes (in order of discovery) the hammerhead (HHR) (1986) [6,12], hairpin (1986) [13], hepatitis delta virus (HDV) (1988) [14], Varkud satellite (VS) (1990) [15], glucosamine-6-phosphate synthase (glmS) (2004) [16], Vg1 (2008) [17], twister (2014) [18], twister sister, hatchet and pistol (all 2015) [19–21], epigenetic (B2)n (2020) [22] and the most recently discovered hovlinc (2021) [11] ribozymes. For some, only a single or few instances of the motif were found in genomes, while for others, numerous natural variants were identified [8]. Considering the broad distribution of self-cleaving ribozymes and the recent and ongoing developments of in vitro selection methods and bioinformatic analysis approaches, it is likely that further ribozyme classes will be discovered [18]. In addition, several artificial self-cleaving ribozyme classes generated by in vitro selection strategies have been reported [8,23,24].

Although the primary, secondary, and tertiary structures of those natural small self-cleaving ribozymes vary, they share common features, such as helical segments (stems) connected by unpaired nucleotides (loops) that are often conserved (Figure 1). Many comprise tertiary interactions between distal parts of the ribozymes to support or enable the folding of the catalytically active conformation [8]. In case of the HHR the comparably late discovery of such tertiary interactions explained how fast-cleaving ribozymes are obtained even in physiologically low Mg²⁺ concentrations [25]. Solving the crystal structure of the full length Schistosoma mansoni HHR revealed non-canonical interactions that align stems I and II and thereby stabilize the overall conformation and proper folding of the catalytic core [26].

Figure 1. (A) Exemplary mRNA containing a ribozyme in the 3’-UTR of a Renilla luciferase (hRluc) gene. After cleavage, the unprotected RNA is prone to degradation. (b-i) schematic secondary structures of small self-cleaving ribozymes. Structures are adapted to represent the sequence formats used in this manuscript. (B) Hepatitis delta virus (HDV) ribozyme, (C) Hairpin ribozyme, (D) Hovlinc minimal structure, (E) Hatchet ribozyme, (F) Pistol ribozyme, (G) Twister ribozyme, (H) Twister sister (TS)-4 ribozyme, (i) Hammerhead ribozymes (HHR) type I (left) and type III (right). The arrowhead depicts the location of the cleavage site within the RNA motifs.

Despite their similarities, each class favours slightly different sequence-specific requirements for activity [10]. Besides, computational investigation of the active sites revealed details of respective catalytic strategies, exploiting ions or water molecules or using certain co-factors [10,27]. A theoretical examination of these different catalytic strategies enabled predictions for the speed limit of the respective ribozyme class. By introducing alterations to the catalytic strategy, the acceleration of the internal phosphoester transfer could be optimized [28].

In contrast to these comparably well-studied catalytic properties of natural self-cleaving ribozymes, their physiological functions remain largely unknown. However, their abundance and wide distribution in nature (ribozymes have been found in all kingdoms of life) imply important functions [8]. The best-characterized function of small self-cleaving ribozymes is the processing of concatemeric copies of RNA genomes of subviral particles such as viroids, satellites and HDVs by hairpin, VS, some HHRs, and HDV-like ribozymes [8,9]. In the case of certain circular RNAs, self-cleavage can be provided by either HDV-like ribozymes or HHR motifs [29], while others seem to exploit hairpin ribozymes in the sense strand and HHRs, twister or hairpin ribozymes in the antisense strand [30,31]. Some HHRs and HDV-like ribozymes were found to be associated with retrotransposons [8,9]. Additionally, isolated cases of intronic HDV-like ribozymes and HHRs in amniotes have been proposed to be involved in gene regulation or mRNA biogenesis [32,33]. Weinberg et al. suggested that the function of some genomic ribozymes may depend on their interaction with endogenous proteins, an aspect that is not yet well understood [8]. In contrast, the glmS ribozyme found in some bacteria is well known to facilitate a negative feedback regulation in Gram-positive bacteria using glucosamine-6-phosphate (GlcN6P) as co-factor, hence acting as a riboswitch to control gene expression [34–36].

Due to their small size (<200 nt), compact and often simple structures, self-cleaving ribozymes are appealing to be exploited for the construction of artificial systems in synthetic biology. In 2004, Yen et al. showed that self-cleaving ribozymes can be used to control mRNA stability in mammalian cells and mice [37]. Here, ribozyme cleavage provides permanent translation repression by removing translation-relevant motifs such as the 5’-cap or poly-A tail, thereby generating unprotected ends susceptible to exonuclease digestion [38]. One strategy to conditionally control gene expression using self-cleaving ribozymes is to combine them with ligand-binding RNA structures, namely aptamers, creating so-called aptazymes. Already in 1997, Tang and Breaker constructed a functional allosteric ribozyme in the test tube by fusing a HHR with an ATP binding aptamer [39].

However, as for their ribozyme platforms, the function of such aptazymes was not transferable from the test tube into cells. The primary reason for the lack of functionality was the use of slow-cleaving HHR versions that lacked stem-I/stem-II interactions. When these tertiary interactions were included, it enabled the construction of artificial riboswitches based on the HHR as expression platform [40]. However, even in these cases it is proved that often the developed sequences were context-dependent and could not be transferred easily from e.g. bacteria or yeast to mammalian cells [41–43]. Hence, context-dependent optimization is often required in order to utilize small self-cleaving ribozymes for constructing artificial switches of gene expression.

Until now, aptazymes for mammalian cells are mostly based on the HHR and are predominately derived from HHR sequences from S. mansoni [20,42–45] or the satellite tobacco ringspot virus (sTRSV) [46,47]. Lately, aptazymes based on the HDV [43,48], pistol [49,50] and twister [43,51,52] ribozyme were developed. Artificial ribozyme-based riboswitches do not rely on the expression of specific protein factors, do not interfere with cellular mechanisms, and are easy to implement by inserting them into untranslated regions (UTRs) of a gene of interest. Multiple proof-of-principle studies have demonstrated that ribozymes and aptazymes can be used to regulate therapeutically relevant biological outcomes such as T-cell proliferation [47], adeno- and AAV-vectored transgene expression in cell culture, as well as mice [43,44,50,53–56] and direct regulation of viral gene expression and replication [57]. In these studies, the transferability of aptazymes to different genetic contexts is substantiated. However, often high variability of the regulatory range and fold-change is observed. Optimization of aptazymes and generation of novel aptamer-ribozyme pairs therefore remains an important area of research. Remarkably, apart from the glmS riboswitch mentioned above, natural examples of aptamer-regulated ribozyme activity are restricted to a single case of a cdiGMP-dependent self-splicing group I ribozyme in Clostridioides difficile [58,59]. To expand the toolbox of ribozyme platforms for aptazyme development, a comprehensive and comparative survey of the effect of inserting several different ribozymes into mRNAs in human cell culture was conducted in this study.

Results and discussion

In synthetic biology, new biological systems are developed from existing natural components. Conditional control of gene expression can be achieved by using artificially constructed riboswitches. Such systems are especially interesting for future applications in gene therapy due to their small genetic footprints and the absence of potentially immunogenic regulatory proteins. In principle, the integration of an active self-cleaving ribozyme into an mRNA provides permanent translation repression by removing translation-relevant motifs such as the 5’m⁷G-cap structure or 3’poly(A)-tail, thereby generating unprotected ends susceptible to exonuclease digestion (Figure 1(A)) [38]. Although these principles were envisioned already two decades ago, still no systematic study exists that compares all suited ribozyme motifs and insertion sites in the same setup for achieving maximal effects on gene expression. To further expand the available toolbox of small self-cleaving ribozymes, we included all known ribozyme classes that show favourable properties for straightforward aptazyme design (Figure 1). For that reason, we focused on small (<100 nt) and independently self-cleaving ribozymes that excluded the epigenetic ribozyme B2 [22] as well as the VS ribozyme due to their rather large sizes [15]. We also excluded the glmS ribozyme [60] due to the necessity of glucosamine-6-phosphate as cofactor. As an endogenous metabolite, this sugar activates the ribozyme already at low concentrations. Moreover, multiple intracellular hexose and aminohexose derivatives interact with the glmS ribozyme [61]. The Mn²⁺-dependent Vg1 ribozyme was also not examined as it has already been shown to be non-functional in cells and proposed to be a molecular fossil [17].

The insertion of structured elements within the mRNA alone can interfere with gene expression. Hence, as a control, ribozymes inactivated by small, defined mutations were tested, respectively, to visualize structure-induced effects (Table S1). All constructs were inserted into the 3’-UTR (30 nt downstream of the stop codon) of a Renilla luciferase (hRLuc) with a flanking sequence of (CAAA)₃ to minimize interactions with the surrounding sequence context. This insertion site is especially relevant for a potential therapeutic use since the 3’-UTR has already been shown to be easily exchanged between target genes and naturally occurring regulatory elements are found to be enriched in the UTRs [62]. We used the psiCHECK2 vector that encodes a second luciferase (Firefly) that can be utilized for normalization in order to control for transfection efficiency. In a dual- luciferase assay (DLA) as used by Beilstein et al. [20], the luciferase activities are measured (Figure 2). Ribozyme cleavage-induced reduction of luciferase expression results in lower activity. Hence, the difference in luciferase activity between the active and inactive ribozyme reflects its regulatory potential and can be quantified as ON/OFF-ratios.

Figure 2. Dual-luciferase assay results examining different ribozyme classes on their self-cleaving activity in HeLa cells. Relative Renilla luciferase activity is shown for constructs containing the active and inactivated ribozyme version. Ribozymes were tested within the 3’-UTR 30 nt after the stop codon. Shown are mean and standard deviation of three independent measurements, each performed in technical triplicates. The respective ON/OFF-ratios between active and inactive constructs are displayed above.

One highly active ribozyme occurs in the hepatitis delta virus (HDV) (Figure 1(B)). It is stated to be the fastest naturally occurring self-cleaving RNA [63], and functions efficiently in the test tube and in cell culture using the wild-type sequence [64]. Moreover, its catalysis is independent from metal ions, and its structure is robust to denaturants [63]. A unique feature is the position of its terminal cleavage site directly at its 5’-end. Notably, within our context, the active ribozyme decreases luciferase activity to about 4% (Figure S1). In nature, the HDV ribozyme often occurs with additional sequences flanking the catalytic core. They are presumed to facilitate correct folding but are not critical for the catalytic activity [64]. Our results show that by further reducing the HDV sequence to only the catalytic core (Figure S1) its activity is not impaired but enhanced and it even outperforms the HHR. The shortened version effectively reduces hRLuc activity to 2% (Figure 2).

The hairpin ribozyme (Figure 1(C)) belongs to one of the earliest described ribozyme classes. Here, a sequence derived from the chicory yellow mottle virus (sCYMV1) was used [65]. Although every catalysed reaction is an equilibrium between cleavage and re-ligation, the equilibrium of most known ribozymes is predominantly found on the side of the cleavage products. In the case of the hairpin ribozyme, harsh conditions can favour the ligation reaction [66], and also stabilization of the tertiary structure can promote ligation over cleavage [67]. Although the hairpin ribozyme from sCYMV1 was engineered to cleave efficiently in the test tube [67,68], within the cells only minor effects contributing to stabilization of the tertiary structure could cause the equilibrium to shift and could be responsible for the poor reduction of luciferase activity observed here (Figure 2).

With the hovlinc ribozyme (Figure 1(D)) we also included a ribozyme originating from mammalian cell lines, which was only recently discovered as a new ribozyme class. It is stated to be a hominin-specific ribozyme comprised of about 168 nt with its pH optimum between 9–10 [11]. Since we were focusing on shorter constructs in this study, we used a proposed minimal version of 83 nt which has been demonstrated to still show cleavage activity [11]. However, it is stated to lose about 90% of efficiency compared to the full length construct because of slower or improper folding [11].

For the hatchet ribozyme (Figure 1(E)), a wild-type sequence from a metagenomics search was selected (Ht-2) [69]. For 90% of the environmental sequences, the ribozyme is associated with an additional P0 stem upstream of the P1 stem. Since it was shown to not contribute to the cleavage activity, it was removed in the construct used in this study. Derived from the annotated sequence [69], the P4 stem was closed with a tetraloop (GAAA). For this motif, the presence of Mg²⁺ is critical and metal ion concentrations different to those in the natural environment could impair its activity [69]. In addition, when examining the crystal structure an equilibrium of dimerized and monomeric ribozymes was observed [70]. Further studies revealed that dimerization is not crucial for ribozyme activity; however, it could impair correct folding in the cellular environment. The hatchet ribozyme was shown to be efficiently cleaving in the test tube [69], however in the context of mammalian cells studied here, it does not seem to exert a significant influence on gene expression (Figure 2). For all three ribozyme classes, hairpin, hovlinc and hatchet, only small changes in hRLuc activity were observed (0.9–1.8 ON/OFF-ratio).

Together with the hatchet, the pistol ribozyme (Figure 1(F)) was discovered. For the pistol ribozyme, multiple environmental sequences exist. The motif derived from Alistipes putredinis exerts fast and efficient cleavage in experiments in the test tube [21]. In the past, this class of ribozymes was rather inefficient when brought into the mammalian cell context [51,71] and also here a rather weak performance was observed (Figure 2). However, its compact structure and tertiary interactions provide several potential aptamer connection sites. Recently, the Yokobayashi group developed artificial sequences of pistol ribozymes cleaving with high efficiency within mammalian cells [71]. This has already led to the construction of well-performing aptazymes [49].

For the twister ribozyme (Figure 1(G)) many natural sequences have already been characterized. One well-studied example is the environmental sequence 9 (env9), which was already used for the design of functional aptazymes in different organisms [43,52,72]. Since the twister motif consists of multiple stem-loop structures, many permutations are possible depending with which stem the ribozyme is connected to the mRNA (P1 and P3 have been realized already for env9) [18]. A study in yeast showed better performance and a higher ON/OFF-ratio when connected via the P3 stem [52]. In our experiments, the twister env9-derived sequence shows a three-fold reduction in luciferase activity (Figure 2). However, the remaining hRLuc activity of about 40% is still rather high with respect to most applications. Nevertheless, it was already shown that inserting a twister ribozyme from Nematostella vectensis strongly reduced reporter activity in another context [51]. Thus, potentially, within this class other motifs could be identified that result in more pronounced reduction of reporter gene activity.

The rather newly discovered twister-sister (TS) ribozyme (Figure 1(H)) class is divided into four different subgroups. In which each motif is forming multiple stem-loop structures [19], resulting again in a variety of possible permutations. We focused here on the TS-4 P1 construct closing the P4 stem with a (GGAAA)-loop. For the TS-ribozyme numerous similarities with the twister motif were attributed [19]. Here, they show a comparable performance although the overall expression levels of the active and inactive TS-variant are slightly lower. With an ON/OFF-ratio of 3.8 it shows a clear change in hRLuc activity (Figure 2). Despite the high OFF-state, both classes, twister and TS ribozyme, offer a lot of structural options to be combined with an aptamer structure. Thus, for applications which do not require a low OFF-state but a significant absolute change in gene expression, these two motifs could be well suited.

The most prominent class of small self-cleaving ribozymes is the hammerhead ribozyme (HHR) (Figure 1(I)). HHRs are categorized into types that indicate which stem is connecting the ribozyme to the surrounding RNA. The most prevalent permutation encountered in natural environments is the type II HHR directly followed by type I [73]. However, optimization studies in cell culture showed that the highest activity can be obtained if the HHR is connected to the mRNA via its P3 stem (type III) [56]. Here, we examined two frequently used topologies of the S. mansoni HHR, the N107 HHR t1 [37,44,56] and the N79 HHR t3 [20,37]. These HHR sequences were already used for the design of various aptazyme constructs [20,42,44]. For both ON- and OFF-switches, the HHR provides a high dynamic range, which was successfully exploited to control reporter gene expression. Similarly, here in our context, active HHRs of both topologies reduced the hRLuc activity with high efficiency. For the type III variant a relative Renilla signal of only 3% was obtained. An ON/OFF-ratio about 66-fold represents a high dynamic range. Optimizing these sequences, the Farzan group was able to increase the ON/OFF-ratio to 1200 within a different context [56]. We included the optimized HHR sequence T3H48 when we tested additional variants of the HHR in the dual-luciferase context (Figure 4).

Atypical type III HHRs identified in the genome of Drosophila pseudoobscura and HDV-like circular RNAs of a toad and a termite, as well as a type I HHR from a mouse gut metagenome (MGM) were selected due to their different tertiary interactions [29,74,75]. In contrast to the S. mansoni HHR versions where non-canonical interactions between stem-loops I and II stabilize the overall fold of the HHR structure, the D. pseudoobscura, toad, termite and MGM HHRs were predicted to exploit a canonical pseudoknot interaction (Figure 4(A)) [29,74]. This structural feature renders these HHR attractive for the rational design of aptazymes. In order to further increase, the available toolbox of this highly active ribozyme, we included human HHR sequences into our comparative analysis (Figure 4(B)). The presence of two HHR sequences in the human genome was revealed in 2010, namely in the intronic regions of the RECK and C10orf118 genes [76]. Cleavage activity has been confirmed for both motifs [74,76], but to our knowledge, no experiments in cells have been reported to date. The use of human HHR sequences for regulation of therapeutic gene expression might avoid potential immune responses against foreign RNA structures and sequences [77].

By testing all these different ribozyme classes within the same context, we were able to identify motifs with high performance in human cells, being ideal starting points for developing novel aptazymes. To further support the obtained data, the relative activity of the ribozyme classes was evaluated in another reporter assay (Figure S2). Controlling the eGFP expression, the selected ribozymes show comparable trends to the DLA-based screen. The HDV as well as the HHR t1 and t3 efficiently reduce gene expression (<14%) followed by TS-4, pistol and hairpin ribozyme exerting moderate reduction of eGFP levels when being active. The hovlinc and the Ht-2 ribozyme again show no effect on gene expression. However, it has to be mentioned that the DLA-based assay is considered more sensitive and reliable since cells show auto-fluorescence and high expression levels of GFP can interfere with cellular fitness [78].

As mentioned before, the performance of a construct often depends on the context in which it is tested. However, it should be possible to deviate some general trends on how to adjust parameters like position or distance to the ORF to increase the performance in certain designs. Since it would be impossible to vary and test many different parameters, we started to examine the efficiency of two well-performing ribozymes with regard to their position within the 3’-UTR (Figure 3). For this purpose, we designed (short)-HDV and HHR (type I N107)-based constructs in positions ranging from 1 to 50 nt downstream of the ORF of the hRLuc reporter (Figure S3). We have also tested the + 19 position (19 nt downstream of the ORF) for comparison with the well-performing aptazymes that have been applied successfully at this location by the Suess group [20]. Again, the relative reduction of luciferase activity was compared for a series of different constructs and positions (Figure 3). Interestingly, the ON/OFF-ratio increases rather consistently with increasing proximity of the ribozyme to the ORF. For the HDV ribozyme inserted immediately after the stop codon, the hRLuc activity is abolished almost completely to 0.3% compared to the parental construct lacking a ribozyme. With a dynamic range of about 419, also the HDV ribozyme also seems to be an ideal candidate when a low OFF-state is required. The same trend is also observed for the type I HHR, although the general performance as well as the distance effect are both comparably less pronounced.

Figure 3. Dual-luciferase assay of constructs containing the (shortened) hepatitis delta virus (HDV) ribozyme or the N107 type I hammerhead ribozyme (HHR t1) in different positions of the 3’-UTR of the Renilla luciferase (hRluc). The relative hRluc activity is shown for the active and inactive ribozyme, respectively. Shown are mean and standard deviation of three independent measurements, each performed in technical triplicates. The ON/OFF-ratios are indicated above the bars.

These findings are in accordance, with a previous study examining the effect on translation of varying lengths of 3’-UTRs in absence of a poly(A)-tail. It was found that the length of the 3’-UTR affects translation efficiency severely – the shorter the 3’-UTR the lower the translation efficiency observed [79]. This finding could also explain the low background reached by the HDV ribozyme, which cleaves at its 5’-end leaving a short 3’end. In contrast, the cleavage sites of HHRs are located in the centre of the structure and a less pronounced effect was observed. However, it cannot be excluded that other RNA motifs could differ in their behaviour. Observing this trend is important since it could allow for optimization of existing aptazymes, as well as informing future aptazyme designs to place the cleavage site as close as possible to the end of the ORF.

As previously stated, for the hammerhead ribozyme, mostly two motifs have been used to construct artificial switches of gene expression. However, besides the utilized motifs from S. mansoni and sTRSV numerous other variants with distinct structural characteristics are known (Figure 4). For the HHRs discovered in the human genome as well as in HDV-like circular RNAs from a toad and a termite activity in the test tube was reported [29,74,76]. However, for those motifs, no activity in cells has been reported so far. We examined those motifs and additionally the D. pseudoobscura HHR regarding their performance within the dual-luciferase reporter assay (Figure 4(D)). Expanding on a previous study on HHR variants, we tested the selected motifs alongside the MGM HHR and used the same sequence context as Wurmthaler et al. [75]. Here, the HHR sequences were inserted 16 nt downstream of the stop codon of the hRLuc with A-rich spacers (Figure S3). According to our data (Figure 3), this position is in a suitable range for HHR insertion without losing activity due to position-dependent effects. Despite their diverse natural origins, all of the selected HHRs proved to be active in human cells, but with varying efficiency. Namely, the D. pseudoobscura, termite and human RECK HHR were shown to abolish hRLuc expression efficiently (Figure 4(D)). In comparison to the type I and III HHR from S. mansoni (Figure 2) their active sequences show similar or in case of the termite HHR even stronger reduction of reporter activity. Both, the termite HHR and the rationally optimized type III HHR sequence T3H48 [56] reduce the relative luciferase activity to below 1% (Figure 4(D)).

Figure 4. Sequences and predicted secondary structures of hammerhead ribozyme (HHR) motifs from (A) a mouse gut metagenome (MGM) [74,75], the drosophila pseudoobscura genome (D. p.) [74], antigenomic HHR found in hepatitis delta virus (HDV)-like circular RNAs of a toad and a termite (toad, termite) [29] and (B) intronic regions of the human C10orf118 (hC10) and RECK (hRECK) gene [74,76]. (C) The optimized T3H48 HHR [56]. (D) Dual-luciferase assay evaluating the activity of the HHRs depicted in (a-c) in HeLa cells. Relative Renilla luciferase activity is shown for constructs containing the active and inactivated version. Shown are mean and standard deviation of three independent measurements, each performed in technical triplicates the respective ON/OFF-ratios between active and inactive constructs are displayed above.

In general, differences observed in HHR cleavage completion and rates in the test tube are not reflected in the efficiency in cells. For example, a much higher k_obs was observed for the MGM HHR (8.74 min⁻¹ at 6 mM Mg²⁺) [75] compared to the termite HHR (0.96 min⁻¹ at 10 mM Mg²⁺) [29]. However, the latter has proved to be 16 times more efficient in our gene expression context (9.5% vs 0.6%). This again highlights the context-dependency and the need to directly evaluate individual sequences in cellular applications. As a human-derived motif, the RECK HHR is attractive for biomedical applications since its conservation in the human genome could indicate an adaptation to the human cellular context and it could also lack a potential immune response directed against foreign RNA structures. Hence, in a next step, we aimed at improving the ON/OFF-ratio by sequence alterations (Figure 5). By shortening stem I of the ribozyme to generate a more compact structure, a more efficient version was generated (Figure 5(A,C)). Shortening the prolonged stem III of the hC10 HHR did not yield a significantly better version (Figure S4). For a potential application in the 5’-UTR, a start codon-free version of the short RECK HHR was generated by a single U to C mutation, which could be shown not to impair ribozyme activity (Figure S5). Furthermore, type III versions of the hRECK HHR were generated. In line with a study of Zhong et al. [56], in which they optimized the type I N107 HHR [37] by transforming it to a type III HHR, in our case the hRECK type III versions outperformed the type I versions exhibiting a remaining luciferase activity of less than 1%. Type III HHRs are thought to reduce mRNA stability more efficiently due to a facilitated disassembly of the stem structures after cleavage due to short leaving strand and fewer tertiary interactions, preventing translation or re-ligation [56].

Figure 5. (A) For optimization a shortened version (sRECK) of the human RECK HHR (hRECK) was generated by removal of the nucleotides shown in blue. (B) Two engineered type III HHR versions were generated with different capping sequences for stem I (RECK t3 v1 and v2). (C) Dual-luciferase assay evaluating the activity of the optimized RECK HHR versions depicted in (A) and (B) in HeLa cells. Relative Renilla luciferase activity is shown for constructs containing the active and inactivated version. Shown are mean and standard deviation of three independent measurements, each performed in technical triplicates. The respective ON/OFF-ratios between active and inactive constructs are displayed above.

In addition, we aimed at reprogramming the interaction of stem I and stem II of HHRs by elucidating the interchangeability of base pairs involved in the loop–loop interactions of the D. pseudoobscura and termite HHR. We replaced the sequences in the terminal loop of stem II and the internal loop of stem I in a complementary manner, speculating that the ribozymes maintain their activity if Watson–Crick base pairs could still be formed. However, the loop sequences were found not to be interchangeable, as high luciferase activities were observed in both states (active and inactive) (Figure S6).

In summary, in this study, we report the activity of various small self-cleaving ribozyme motifs with distinct structural characteristics and efficiencies in human cell culture. Many different ribozymes have been used in a variety of different contexts and hence the performance of the ribozymes is per se hard to compare. With this study, we aimed at comparing a series of ribozymes that should in principle be suited for the construction of synthetic RNA switches of gene expression for their potential to affect gene expression. By systematically comparing a variety of different ribozymes in different formats and inserted at varied locations, we provide an overview that should be helpful if a decision needs to be made regarding the type and format of the ribozyme and the optimum insertion position for the purpose of modulating gene expression or for constructing efficient ligand-dependent ribozyme switches. Based on a closer evaluation of two well-performing motifs, we demonstrate that ribozymes seem to work more efficiently the closer they are inserted downstream of the stop codon in the 3’-UTR. In addition, we add HHRs originating from the human genome to the repertoire of functionally active ribozyme sequences. The presented results provide a guide on which ribozymes to use and where to insert them into the 3’-UTR to efficiently regulate gene expression in mammalian cells.

Material and methods

Plasmids and cloning

The ribozymes were introduced via restriction sites or PCR with overhang primers containing a phosphorylated 5’-end into a psiCHECK-2 vector (Promega) carrying two luciferase genes, the Renilla (hRLuc) and Firefly luciferase (hLuc+). Flanked by a spacer sequence of (CAAA)₃ they were inserted in the 3'-UTR of the hRLuc, 30 nt downstream of the stop codon (if not stated otherwise). The PCR product was digested with DpnI (NEB) to get rid of the template plasmid and purified by gel electrophoresis. Recovery was done with the Zymoclean Gel DNA Recovery Kit. After ligation with Quick Ligase (NEB), the product was transformed into the XL10-Gold Escherichia coli strain (Stratagene). Individual colonies were verified to contain the correct sequence by Sanger sequencing at Eurofins Genomics/GATC Biotech or Azenta Genewiz. Sequences used are listed in Table S1.

Cell cultivation and transfection

HeLa cells (ATCC, cat. no. ATCC-CCL-2) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) + GlutaMAX-I + 10% foetal calf serum (Gibco/Thermo Fisher Scientific). The medium was supplemented with 1% Penicillin/Streptomycin. The cells were cultivated at 37°C and with 5% CO₂. Prior to transfection, within a 96-well plate the HeLa cells were seeded out to 8000 cells/well to achieve a confluency of ~ 70% after 24 h. Transfection was conducted using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. After 4 h at 37°C, the medium was exchanged, and after 20 h, the Dual-Luciferase Reporter Assay (Promega) was performed. Since the ribozymes only interfere with hRLuc expression and consequently its activity, the hLuc+ is used as an internal control. For each construct, the ratio of hLuc+ and hRLuc is calculated and normalized to the respective activity of the empty psiCHECK-2 plasmid (without any ribozyme) with full expression of both luciferases. Measurements were performed with Spark® multimode microplate reader (Tecan) with a settle time of 0 ms and an integration time of 2000 ms. All constructs were measured in technical triplicates in at least three independent experiments.

Concerning the eGFP-based screen, the cells were treated as described above, before and during transfection. After exchanging the medium the cells were kept at 37°C for 48 h prior to the measurement. After washing with 1× PBS the fluorescence was measured with Spark® multimode microplate reader (Tecan) using the Top reading. The eGFP (478 nm/524 nm) fluorescence was normalized on mCherry (568 nm/620 nm) fluorescence. Thereby the relative eGFP levels were obtained. All constructs were measured in technical triplicates in at least three independent experiments.

Acknowledgment

We would like to thank Monika Finke for initial work as well as Vera Hedwig for helpful discussions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data supporting the findings of this study are available within the article and its supplementary materials.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15476286.2023.2296203.

Additional information

Funding

We thank the Deutsche Forschungsgemeinschaft (DFG) for funding within project A5 of the CRC 969 ‘chemical and biological principles of cellular proteostasis’.