Characterisation of aptamer-target interactions by branched selection and high-throughput sequencing of SELEX pools

Abstract

Nucleic acid aptamer selection by systematic evolution of ligands by exponential enrichment (SELEX) has shown great promise for use in the development of research tools, therapeutics and diagnostics. Typically, aptamers are identified from libraries containing up to 10(16) different RNA or DNA sequences by 5-10 rounds of affinity selection towards a target of interest. Such library screenings can result in complex pools of many target-binding aptamers. New high-throughput sequencing techniques may potentially revolutionise aptamer selection by allowing quantitative assessment of the dynamic changes in the pool composition during the SELEX process and by facilitating large-scale post-SELEX characterisation. In the present study, we demonstrate how high-throughput sequencing of SELEX pools, before and after a single round of branched selection for binding to different target variants, can provide detailed information about aptamer binding sites, preferences for specific target conformations, and functional effects of the aptamers. The procedure was applied on a diverse pool of 2'-fluoropyrimidine-modified RNA enriched for aptamers specific for the serpin plasminogen activator inhibitor-1 (PAI-1) through five rounds of standard selection. The results demonstrate that it is possible to perform large-scale detailed characterisation of aptamer sequences directly in the complex pools obtained from library selection methods, thus without the need to produce individual aptamers.

Figure 1.

An overview of the method for parallelised aptamer characterisation. After five rounds of selection for binding to the serpin PAI-1, the enriched RNA pool was subjected to one round of parallel selections for binding to different PAI-1 variants (wild type and single-residue alanine mutants) and conformers (active and latent form) as well as PAI-1 presented by different antibodies or the natural extracellular matrix protein, vitronectin (VN). After a washing step, the retained RNA was reverse transcribed, PCR amplified with barcoded primers, and sequenced in one reaction by Illumina sequencing.

Figure 2.

Characterisation of the input pool. (A) 69 945 unique sequences were observed at least one time. The graph depicts how the number of unique sequences decreases when increasing the minimum abundance threshold. For example, 20 206 sequences of the 69 945 were detected in the pool at least 10 times. (B) Sequence abundance as the percentage of pool (%) for the top 1000 most prevalent unique sequences in the input pool. Samples for HTS analysis were prepared three times in similar ways. Either from the re-amplified original SELEX template which was used for transcribing the RNA used in this study (input pool; black dots) or directly from the original SELEX template (Original #1; dark grey dots, and Original #2; light grey dots).

Figure 3.

Binding sites of aptamers paionap-5 and -40 on PAI-1. Left panel: PAI-1 in a cartoon representation with residues important for binding of the aptamers shown as red (Arg78, Lys82, Phe116 and Arg120 recognised by both paionap-5 and -40) and yellow spheres (Lys124 recognised only by paionap-40) (14). Middle panel: Surface representation model of PAI-1 using same colour codes. Residues in blue (Lys71 and Tyr81) are unimportant for aptamer binding and were included in the branched selection-HTS analysis. Right panel: The position of the somatomedin B domain of vitronectin (green) when bound to PAI-1. The figure was prepared using PyMOL (the PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC) and PDB ID: 1OC0.

Figure 4.

Secondary structure prediction for paionap-5 variants. (A) The predicted secondary structure for paionap-5 (SEQ ID 551), generated by RNAshapes. Grey boxes denote helical regions, constant flanking regions are underlined, and bases with at least 90% conservation are shown with black spheres. (B) Ten paionap-5 related sequences with high structural and sequence similarity based on RNAshapes predictions were aligned using LocARNA. The average secondary structure is indicated in dot-bracket notation (top). Conserved bases in the random region are indicated by an asterisk (bottom). Bases in individual sequences predicted to undergo Watson–Crick base-pairing are highlighted in grey. Most of the constant regions were omitted for clarity as these regions were invariant. Remaining constant region bases have been marked by underlining. (C) Consensus structure and regions of high sequence conservation based on the LocARNA alignment. (D and E) Consensus structures of other paionap-5 related families with the same binding site.

Figure 5.

Secondary structure prediction for paionap-40 variants. (A) Twenty sequences with structural and sequence similarities to paionap-40 (SEQ ID 53156) were aligned using LocARNA. (B) The predicted average structure for paionap-40 variants based on the LocARNA alignment. See legend to Figure 4 for details. W denotes A or U, and D denotes G or T or A.

Figure 6.

Evaluation of the PAI-1 mutant specificity of the thousand most prevalent sequences of the input RNA pool. For each single-residue alanine mutant used in the HTS-microplate assay a plot of enrichment factors for wild type (EF_wt; x-axis) versus mutant (EF_variant; y-axis) is shown. The x-axis crosses the y-axis at EF_variant = 1 and the y-axis crosses the x-axis at EF_wt = 1.

Figure 7.

Secondary structure prediction for RNA variants binding to the PAI-1:vitronectin complex. LocARNA alignment of the ten variants of SEQ ID 58088 (A) and SEQ ID 415 (B) after one round of selection for binding to the PAI-1:vitronectin complex. (C and D) The predicted secondary structures for SEQ ID 58088 and SEQ ID 415, respectively, based on the LocARNA alignment. See the legend to Figure 4 for details.