TL;DR
This paper introduces RaptGen, a deep-learning tool that significantly accelerates the discovery and optimization of RNA/DNA hybrid aptamers targeting the SARS-CoV-2 spike protein. By combining a novel primer-less SELEX method with RaptGen's sequence profiling capabilities, the researchers efficiently identified and chemically modified a highly potent aptamer (SPA1-M13) that binds to SARS-CoV-2 and its variants, as well as other coronaviruses.
What Problem Does This Paper Address?
Developing effective aptamers for industrial applications requires not only identifying sequences that bind to a target but also optimizing them for length and chemical stability. Traditional SELEX methods can be lengthy and produce aptamers that are too long or not robust enough for practical use. Specifically for SARS-CoV-2, there's a need for highly specific and stable aptamers that can act as diagnostic tools or therapeutic agents, especially given the emergence of new variants. The challenge lies in efficiently navigating the vast sequence space to find and refine these optimal aptamers.
Methods
The core of this work involves a two-pronged approach: a modified SELEX process and the application of a deep-learning algorithm, RaptGen.
- Primer-less SELEX: This technique was employed to generate shorter aptamer candidates from the outset. Instead of using fixed primer sequences for PCR amplification, this method removes them from the library, allowing for the selection of shorter, functional sequences. An RNA/DNA hybrid library was used, incorporating 2'-OH-GTP, 2'-OH-ATP, 2'-deoxy-CTP, and 2'-deoxy-TTP, with GMP added to create a 5' monophosphate for subsequent ligation.
- RaptGen Analysis: After SELEX, the resulting sequences were fed into RaptGen, a deep-learning tool designed to analyze SELEX data. RaptGen embeds sequence data into a low-dimensional space, allowing for the identification of sequence motifs and the generation of profiles. This helps in selecting promising candidate sequences. The researchers iteratively applied RaptGen, first to the initial SELEX output and then to a refined dataset after removing non-binding sequences.
- Aptamer Truncation and Chemical Modification: Promising candidate aptamers were then truncated to find the shortest functional sequence. Following truncation, chemical modifications were introduced to enhance stability and binding affinity. This included exploring 2'-O-methyl modifications and phosphorothioate linkages at specific positions.
- Binding Assays: Surface Plasmon Resonance (SPR) was the primary method used to assess binding affinity (K<sub>D</sub>) and kinetics. Flow cytometry was also used to evaluate the aptamer's ability to bind to spike protein expressed on cell surfaces.
Datasets
The primary "dataset" used in this study was generated through the experimental SELEX process. This involved sequencing millions of oligonucleotide sequences from various rounds of selection against the SARS-CoV-2 receptor binding domain (RBD). This experimental data was then processed and analyzed by the RaptGen algorithm. No pre-existing public datasets were explicitly mentioned for training RaptGen in this paper, suggesting it was trained on the data generated within this study.
Key Insights and Ideas
- Deep Learning for Aptamer Optimization: RaptGen's ability to identify sequence motifs and profiles in a low-dimensional space proved crucial for selecting promising aptamer candidates from complex SELEX data, especially in a primer-less approach.
- Primer-less SELEX Efficiency: This strategy successfully generated shorter aptamer candidates, reducing the complexity of subsequent optimization steps.
- G-Quadruplex Hypothesis: The truncation experiments suggested that the initial aptamer (SPA1) might form a G-quadruplex structure, with specific guanosine deletions leading to a loss of binding. However, further analysis indicated it possessed more complex binding than a simple G-quadruplex.
- Chemical Modification is Key: Extensive chemical modifications, including 2'-O-methyl and phosphorothioate groups, were essential to achieve high affinity and stability in the final aptamer (SPA1-M13).
- Broad-Spectrum Binding: The optimized aptamer (SPA1-M13) demonstrated binding not only to wild-type SARS-CoV-2 RBD but also to various SARS-CoV-2 variants, SARS-CoV-1, and MERS-CoV, suggesting a broad therapeutic or diagnostic potential against coronaviruses.
Key Results
- Identification of SPA1: The initial SELEX and RaptGen analysis identified a promising RNA/DNA hybrid aptamer, SPA1, which bound to the SARS-CoV-2 RBD.
- Truncation to SPA1-T16: Through systematic deletion, a 26-nucleotide truncated aptamer (SPA1-T16) was identified, retaining significant binding activity.
- Development of SPA1-M13: Extensive chemical modification led to SPA1-M13, a highly optimized aptamer with nanomolar binding affinities (K<sub>D</sub> of 2.8 nM for RBD and 0.39 nM for spike protein trimer).
- Variant and Cross-Coronavirus Binding: SPA1-M13 showed binding to SARS-CoV-2 variants (E484K, N501Y, B.1.351, B.1.1.529) and also to RBDs of SARS-CoV-1 and MERS-CoV, with K<sub>D</sub> values ranging from 4.0 to 93.2 nM.
- Cellular Binding: SPA1-M13 demonstrated specific binding to SARS-CoV-2 spike protein expressed on HEK293FT cells, as confirmed by flow cytometry.
Limitations and Future Work
- Non-binding Sequence Exclusion: The initial RaptGen analysis of the primer-less SELEX data yielded less informative sequence profiles, necessitating an iterative approach of excluding non-binding sequences to refine the analysis.
- Mechanism of Action: While the aptamer binds to the RBD, the precise mechanism by which it inhibits viral entry or interaction is not fully elucidated, though it's suggested to be unrelated to the ACE2 binding surface.
- In Vivo Efficacy: The study focuses on in vitro binding and cellular assays. Further studies would be needed to evaluate the in vivo efficacy and therapeutic potential of SPA1-M13.
- Broad-Spectrum Inhibitor Potential: The authors suggest that the broad binding spectrum against different coronaviruses could be beneficial for developing a broad-spectrum inhibitor, implying future research directions in this area.
- RaptGen's Generalizability: While demonstrated for primer-less SELEX, further validation of RaptGen's applicability to other SELEX strategies and targets would be valuable.
Summary updated