A System for In Vitro Selection of Fully 2’-Modified RNA Aptamers

Abstract

SFM4-3, KOD DGLNK, and Therminator polymerase are investigated for their compatibility with SELection with Modified Aptamers (SELMA), an aptamer discovery method that enables incorporation of large nucleobase modifications such as glycans. We demonstrated that with suitable modifications to the primer design and protocol, these enzymes are compatible with SELMA, enabling 2’-fluoro or 2’-methoxy ribose modifications at all positions. In the case of 2’-fluoro modifications, Therminator exhibits cleaner incorporation of an alkyne-modified nucleobase for click chemistry.

Aptamers are genetically encodable polymers, usually oligonucleotides, that have been selected in vitro from a random library to bind to a protein or molecule of interest.^1-4 Oligonucleotide aptamers are typically 20-100 bases in length and can achieve high binding affinities for their targets due to their specific folding, shape complementarity and positioning of functional groups.⁵ Aptamers possess several advantages compared with antibodies, including straightforward synthesis and quality control, low immunogenicity^{6, 7}, and small size.^{5, 8} Aptamers are typically discovered using SELEX (Systematic Evolution of Ligands by EXponential Enrichment), in which libraries of random DNA or RNA sequences flanked by constant primer regions are incubated with a target of interest and the bound sequences are amplified by PCR. Two aptamer therapeutics have been approved for clinical use. Pegaptanib targets vascular endothelial growth factor (VEGF) for the treatment of age-related macular degeneration.^{9, 10} Very recently in 2023, avacincaptad pegol, an aptamer targeting complement protein C5, was approved for the treatment of geographic atrophy (GA) secondary to age-related macular degeneration.^{11, 12}

A barrier to the development of oligonucleotide aptamer therapeutics has been the instability of naked RNA or DNA to nucleases in serum.^{13, 14} To address this, aptamers have been extensively stabilized with modified nucleotides (XNA), such as fluoro- (2’-F-) or methoxy (2’-OMe-) ribose, which confer nuclease resistance.¹⁴ Pegaptanib and avacincaptad are illustrative, as they both bear 2’-F- or -OMe at nearly every position. However, these modifications typically must be introduced into natural DNA or RNA aptamers after selection by SELEX, and trial and error is necessary to find substitutions that do not compromise the binding activity of the selected molecule. Incorporation of XNA throughout the selection process would be more desirable and efficient but requires technical innovation, as XNA libraries may not be easily synthesized and/or PCR amplified by polymerases with limited tolerance for modified nucleoside triphosphates (NTPs).^15-42

As a solution to this problem, our group has reported SELection of Modified Aptamers (SELMA, Figure 1)^43-48, which allows for extensive modifications, not only of ribose, but of the nucleobase (e.g., oligosaccharide attachment) and includes a variant that can use 2’-F pyrimidine NTPs⁴⁸ throughout the selection. In SELMA, PCR and/or RT-PCR of modified oligonucleotides is circumvented by tethering an unmodified, PCR-amplifiable DNA copy to each XNA sequence in the library.

Figure 1.

The SELection with Modified Aptamers (SELMA) construct contains a PCR amplifiable DNA genotype that is covalently linked to a phenotype strand containing XNA and/or 5-ethynyluridine modified by click chemistry.

However, this technique has not yet been adapted to selections with libraries containing the 2’-F modification on all four nucleotides, which would require an RNA or DNA polymerase capable of incorporating them. Although several polymerases are known to accommodate ribose-modified NTPs, the activity and fidelity of these enzymes remains poorly understood and difficult to assess by traditional sequencing methods. Ultimately, a better understanding of the performance of these enzymes is needed before they can be widely adopted for aptamer selection applications.

To this end, we have investigated three different polymerases capable of incorporating 2’-modified XNA into the SELMA protocol. Therminator DNA polymerase is a commercially available mutant of 9°N DNA polymerase that is able to incorporate a variety of non-canonical nucleic acids including 2’-F-modified NTPs (2’-F-NTPs).⁴⁹ SFM4-3, which also incorporates 2’-F-NTPs, is a variant of the Stoffel fragment of TAQ DNA polymerase developed by Chen et al. using directed evolution.⁵⁰ KOD DGLNK is a mutant KOD polymerase developed by Hoshino et al., that accepts both LNA and 2’-OMe-modified nucleotides.⁵¹ We purchased commercially available Therminator and expressed SFM4-3 and KOD DGLNK for generation of the ssXNA “phenotype” strand (G, Figure 2). Library construction begins with a short ssDNA strand containing the random library template (A), which is annealed to a phosphorylated overhanging “regeneration” primer (B). Bidirectional DNA polymerase extension creates a longer duplex (C) and lambda exonuclease digestion of the phosphorylated strand results in a single stranded hairpin (D). The 3’ end of D is self-priming and can be extended with XNA using one of the three above enzymes to afford double stranded hairpin (E). The XNA strand is then displaced (F) by treatment with a “hairpin poison primer” (HPP) and its extension with BST 2.0 to generate Form G. Compared with the DNA/DNA duplex of our original SELMA method, this relatively long primer is required for the strand displacement of the stronger 2’-F-RNA-DNA interaction.^{35, 52, 53} The resulting Form G library contains modified ssXNA that can adopt a sequence-dependent fold and can undergo selection towards a target, while PCR amplification of the tethered dsDNA in the bound library fraction affords a new template in double stranded Form H. The use of the biotinylated primer in PCR allows for strand separation and a return to ssDNA library template (Form A).

Figure 2.

XNA SELMA Library build. Blue = constant primer sites; red = random region; black = constant hairpin region; orange = XNA.

For KOD DGLNK and SFM4-3, some optimization was done to characterize the behavior of each enzyme at various temperatures. We found that, for our applications, increasing the extension temperature from previously reported conditions resulted in more efficient extension and higher yields of Form E (Figures S3, S4).^{50, 51} After determining the ideal extension conditions, each enzyme was used to form double stranded hairpin and subsequently strand displaced as shown in Figure 3.

Figure 3.

A) Scheme of hairpin extension and strand displacement. B) Denaturing PAGE gel of extended (lanes 1, 3 and 5) and displaced (lanes 2, 4, and 6) hairpin made with KOD DGLNK and MeO-RNA (lanes 1 and 2), SFM4-3 and F-RNA (lanes 3 and 4) and Therminator and F-RNA (lanes 5 and 6). The band below 50 bp in lane 4 is the NEB WarmStart aptamer.

To demonstrate compatibility of each enzyme with the SELMA workflow, we performed mock selections in which a known target sequence of 2’-F- or 2’-OMe-RNA was selected from a random library by hybridization to a biotinylated complementary oligonucleotide (Figure 4A). The target sequence contained an NheI restriction site, which enabled easy detection of the target sequence by the presence of four fragment bands after digestion with NheI (Figure 4B, Group 4). This sequence was spiked at a ratio of 1:1000 into N₂₅ random libraries generated with 2’-F-RNA using Therminator or SFM4-3, or with 2’-OMe-RNA using KOD DGLNK. Prior to selection, the known sequence was not visible by gel analysis of the spiked libraries following NheI digest (e.g., Figure 4B, Group 2). Each library was then subjected to one round of selection. Species that bound to the biotinylated complementary oligonucleotide were recovered using streptavidin magnetic beads. Bound material was then PCR amplified and digested with NheI, resulting in the expected four small fragment bands in all three selections (Figure 4B, Groups 3, 5, 6). Thus, we observed clear enrichment of our target sequence after just one round of selection using libraries generated with each of the enzymes tested here. Control selections containing only random library (Figure 4B, Group 1) or only NheI (Figure 4B, Group 4) sequence were also performed in parallel. The successful mock selection demonstrates the capability of the SELMA method to evolve a fully fluorinated or fully methoxy library and generate enrichment without the need for a reverse transcription step.

Figure 4.

A) NheI selection construct design and enrichment test. B = biotin; SA = streptavidin B) Gel analysis of NheI-digested templates from selections. NheI-treated samples are indicated with a (+). Group 1 = random library post selection, Group 2 =1:1000 NheI template:random library template pre selection, Group 3 = 1:1000 NheI template:random library template post selection, Group 4 = NheI template only post selection. Groups 1-4 were prepared with SFM4-3. Groups 5 and 6 are post-selection libraries analogous to Group 3 but made with KOD DGLNK/MeO-NTPs Therminator polymerase/F-NTPs, respectively. Complete gels corresponding to Groups 5 and 6 can be found in Figure S5.

To investigate chemical functionalization of libraries, we focused on 2’-F-RNA and used either Therminator or SFM4-3 to generate an F-RNA strand containing three 5-ethynyluridine (fEdU)⁴⁸ bases by primer extension against a 56-mer DNA template containing three adenines (Figure 5). Following CuAAC attachment of oligomannose (Man₉) glycan,⁴⁴ the results were visualized by gel, in which the functionalized F-RNA strand migrates higher due to added glycans. When Therminator was used to generate the F-RNA strand, the corresponding glyco-F-RNA migrates primarily as one band, corresponding to addition of three glycans, with a small amount of a lower band corresponding to incomplete functionalization. However, when SFM4-3 was used to generate the F-RNA, the glyco-F-RNA is a series of bands, corresponding to different numbers of attached glycans in some cases exceeding three. These results suggested substantial misincorporation of fEdU by SFM4-3 and have led us to focus on using Therminator for future libraries.⁵⁴

Figure 5.

Generation of alkyne-containing F-RNA/DNA duplex, CuAAC modification with Man₉-azide glycan, and gel electrophoresis. Lanes 1 and 3 = product of primer extension with Therminator and SFM4-3 respectively, in the presence of f(A,G,C,EdU) triphosphates. Lanes 2 and 4 = Therminator and SFM4-3-derived oligos after click functionalization to attach Man₉ glycan.

Conclusions

In conclusion, we have described several modifications to the SELMA methodology that allow for the incorporation and in vitro selection of fully 2’-modified XNA without the need for reverse transcription or post-SELEX medicinal chemistry optimization to incorporate nuclease resistant modifications. Additionally, we have demonstrated that fully 2’-modified XNA SELMA constructs can be successfully enriched towards a known target. This expansion of SELMA capabilities will simplify discovery of aptamers that could be directly used in the presence of nucleases. We have shown that the 2’F-RNA can be chemically modified by CuAAC, and we have found in the process that Therminator is less prone to misincorporation of fEdU than SFM4-3. Selection applications of SELMA with XNA and CuAAC modifications are currently underway with carbohydrate binding targets. Using this approach, we will develop potent modulators of clinically relevant carbohydrate binding proteins that are directly ready for use in vivo.

Supplementary Material

^‡Supplementary Information available: [details of any supplementary information available should be included here]. See DOI: 10.1039/x0xx00000x

Data availability

The derived data supporting these findings can be found in the ESI^‡. The raw data generated by Azenta Life Sciences and the python scripts to analyze this data can be found at https://github.com/emilyziperman/Polymerase-Fidelity-Analysis