Selection of trans-acting ribozymes for site-specific RNA modification

4.1.1 Analogies with deoxyribozyme selection methods

Our partially structured pool and selection strategy enabled the selection of efficient ribozymes for site-specific modification of target RNA in trans. The modification occurs in most cases at a predetermined position. The pool was designed to include a hypothetical substrate sequence, which was paired to left and right recognition arms, leaving one bulged nucleotide as the desired modification site. The hypothetical substrate sequence was physically connected to the pool via a 14-nucleotide connecting loop. The ribozymes were therefore selected essentially as “self-biotinylating” catalysts. Removal of the connecting loop, however, was enough to convert these ribozymes into efficient trans-acting tools.

The design of our selection library was a simplified variation of the 10DM24 deoxyribozyme selection pool. The design of the 10DM24 selection library was rather complex with four binding arms, recognizing two RNA substrates. Pairing of the arms to the substrate sequences led to formation of a three-helix-junction (Zelin et al., 2006). The deoxyribozyme was selected for 2'-5' ligation of two RNA oligonucleotides and was further engineered for 2'-5' branching of an RNA target using a single nucleotide (Höbartner and Silverman, 2007).

The engineering process was done by removing the 5'-triphophorylated guanine nucleotide from the 5'-end of the right hand oligonucleotide substrate and supplying it as a free GTP.

The deoxyribozyme recognizes the GTP analog thorough Watson-Crick base-pairing to a cytosine in one of its binding arms (Höbartner and Silverman, 2007). By mutating the cytosine in this position, the specificity of 10DM24 can be switched towards other complementary NTPs (Höbartner and Silverman, 2007).

Our selection was performed directly for the single nucleotide branching activity using natural and non-natural ATP analogs. Hence, a “preorganized three-helix junction” setting was not necessary. The number of binding arms was therefore reduced to two recognizing a single substrate RNA. Through this selection strategy, ribozymes have been evolved that could be easily engineered for targeting various RNA sequences. The mode of NTP analog recognition of these ribozymes, however, may be more complex than 10DM24. In-depth investigation should thus be performed to elucidate the mechanism behind substrate specificity of these ribozymes, in order to engineer NTP analog specificity.

Analogies can also be made between our selection strategy and other deoxyribozyme selection methods. The constant regions of the deoxyribozyme selection pools are in most cases designed as binding arms complementary to the substrate sequence(s).

Deoxyribozymes acting on non-oligonucleotide substrates have also been selected following similar design principles. In these instances of DNA catalysts, oligonucleotides complementary to the constant regions of the pool are conjugated to the reaction substrates.

This pairing facilitates the localization of reactants in the vicinity of the random region, improving the likelihood of the isolation of the desired catalyst. The oligonucleotide (tethered) substrate is required to be ligated to the single-stranded DNA pool, before each selection round, as it is lost during the PCR amplification cycles. The evolved variants are also generally compatible with activity in trans when the target oligonucleotide or the oligonucleotide tethered substrate is not covalently attached to the catalyst (Silverman, 2009).

For selection of RNA catalysts acting on unmodified RNA substrates the physical linkage between the substrate sequence and the pool was made transcriptionally. The extra ligation step, necessary in case of 10DM24 deoxyribozyme was therefore not needed.

4.1.2 Basis of the modification site selectivity of the evolved ribozymes

Out of four ribozymes probed for modification site determination, three modified the predetermined adenosine. The other ribozyme labeled the target sequence at one nucleotide 5' to the bulged A. This proves that our selection strategy facilitates rapid evolution of ribozymes with the ability to modify this particular position. Two reasons may contribute to the site-selectivity of our selection process:

First, adenine nucleotides in a bulged structural context have been shown to have higher tendency to participate in branching reactions compared to other nucleotides. This conclusion was made by Silverman et al based on their experience of selecting 2'-5' ligase deoxyribozymes. The 7S11 deoxyribozyme developed by this group is the earliest example of 2'-5' RNA ligase deoxyribozymes (Coppins and Silverman, 2004a). The selection process of 7S11 was initially aimed at development of a linear 3'-5' ligase deoxyribozyme. 7S11, however, evolved in a way that it formed a triple-helix junction surrounding a bulged adenine nucleotide that served as the branch-site (Coppins and Silverman, 2005). This structural context was the inspiration for the selection of the 10DM24 ribozyme. The deoxyribozyme was initially selected for branching a bulged uridine, however, replacement of the uridine

with an adenine remarkably improved the ligation rate. These findings along with strong prevalence of bulged A-nucleotides at branch-sites of class II and spliceosomal introns, led to the conclusion that the bulged adenines are “inherently favored” for 2'-5' branching reactions. This conclusion was the basis of our choice for the modification-site adenine nucleotide (Zelin et al., 2006). Furthermore, Büttner et al discovered that the 5'-GAG-3' is the optimal modification context for 10DM24 (Büttner et al., 2014). For that reason, the bulged A for our ribozyme selection was also placed within a GAG context.

Additionally, the desired reactions for which the ribozymes were selected, involve the introduction of bulky groups to the ribose backbone of the RNA. These types of modifications are known to stall reverse transcriptase. Thus, the variants that self-modify at positions outside the priming range (i.e., in the random region) are eliminated from the pool due to failure in full-length cDNA synthesis.

4.1.3 Comparison with other trans-acting RNA modifying ribozymes

To our knowledge, no 2'-5' nucleotidyltransferase ribozymes had been developed prior to our work. However, several other artificial RNA-modifying ribozymes exist that possess trans-activity (Jadhav and Yarus, 2002; Kang and Suga, 2007; Lorsch and Szostak, 1995;

Poudyal et al., 2017; Saran et al., 2005). These ribozymes were mostly selected using unstructured RNA libraries. Except for some 5'- or 3'- modifying ribozymes, no predetermined modification site was considered during the selection process of most of these ribozymes. The modification sites were, therefore, randomly evolved. To turn these ribozymes into trans-acting variants, an in-depth investigation had to be performed to probe the modified position and predict or determine the secondary structure of the ribozyme.

Mutational studies are then often performed to minimize the ribozyme and determine the critical positions often resulting in loss of efficiency. The ribozyme is then split into the substrate segment and the catalyst segments. In some cases, the substrate and the catalyst segment interact via partially complementary regions. The complementary parts can sometimes be covaried to allow flexibility regarding the target sequence, however, in many cases these ribozymes have a narrow range of substrate sequences they can modify. The required modification context in those instances is complex and with a low probability of appearing in any given RNA sequence. For example, the kinase ribozyme 2PTmin3.2 is a trans-acting ribozyme that recognizes its target sequence via partial complementarity. This

ribozyme phosphorylates the 2'-OH of the first purine nucleotide within a 5'-RAAAANCG-3' context (Saran et al., 2005).

Our selection strategy using a partially structured pool allows the selection of ribozymes that can readily be adopted for trans-activity simply by removing the connecting loop and varying the binding arms based on the desired target. These changes do not cause any loss of efficiency. Moreover, the variants selected mostly demonstrated remarkable target sequence variability.