Biological macromolecules (DNA, RNA, and proteins) after synthesis undergo various chemical modifications in the form of chemical groups being added onto their residues. These modifications act as markers and carry extra information for precisely regulating gene expression. Protein post-translational modifications have been the focus for many years owing to their pharmacological roles in cellular modulation, thus making them attractive targets for drug discovery. However, the role of RNA modifications has gained a lot of momentum in recent years as modified non-coding RNAs (microRNA, lincRNA, etc.) have emerged as key players in cellular processes and disease progression1,2. Among all RNA species, the most heavily modified are rRNA and tRNA, and to date more than 170 such modifications have been identified3. Traditional methods to identify RNA modifications have been chromatography techniques and mass spectrophotometry4, and a lot more modifications have been newly discovered due to advancements in chemical methods combined with high-throughput sequencing. Nevertheless, due to limited availability of techniques to sensitively and accurately identify these modifications, the emerging field of RNA epitransciptomics remains difficult to explore; mainly given the challenge in discriminating between their similar chemical structures. In the latest issue of Nature Nanotechnology, Wang et. al. report a nanopore-based strategy that enables high-resolution detection of modifications on RNA5. Using this method, the authors were able to distinguish eleven different modified nucleoside monophosphates (NMPs) with their custom base caller that reports a 99.6% accuracy.

Biological nanopores offer a platform for tuneable sensing and real-time detection of single-molecules. In this proof-of-concept study, Wang and co-workers report the design of a Mycobacterium smegmatis porin A (MspA) protein where they engineered a single cysteine in its narrow pore constriction. The single cysteine residue then served as an adapter for phenylboronic acid (PBA), a compound that can covalently react with cis-diols, thus acting as an efficient sensor for reacting with the cis-diols present in the ribose of NMPs. Single-channel current recordings showed a ~100 pA drop in current as compared to the open channel current, immediately upon the addition of PBA. This suggested the binding of PBA with the cysteine, thereby occupying the pore. The authors first tested if the four canonical nucleotides could be separately detected with this MspA-PBA system. When a 200 mV voltage bias was continuously applied to the nanopore, representative current pulses were observed for the NMPs tested and simultaneous detection of all four nucleotides (AMP, CMP, UMP, GMP) produced distinct current blockade events for each NMP (Fig. 1).

Fig. 1: MspA-PBA based strategy for distinguishing canonical and modified NMPs.
figure 1

a, MspA porin with a single cysteine engineered in its pore constriction. b, Upon the addition of 3-maleimide PBA, it reacts with the thiol of the cysteine residue, which results in a ~100 pA drop in the current. This now represents the open channel current (Io), as shown in d. c, When NMPs are added to the cis side of the pore, it can reversibly bind with the PBA. d, This binding is observed as stochastic sensing events. e, Raw current blockage data is then fed into the custom machine learning algorithm and the current blockage levels (Ib) vs noise (as standard deviation) are plotted to obtain a scatter plot of all the NMP events.

Full size image

Apart from the most abundant N6-methyladenosine (M6A) modification found on mRNA, some other relevant RNA modifications include 5-methylcytosine (m5C), 7-methylguanosine (m7G), N1-methyladenosine(m1A), pseudouridine (Ψ), inosine (I) and dihydrouracil (D) (ref. 6). The authors next tested seven of these epigenetic RNA modified compounds (m5C, m6A, m7G, m1A, I, Ψ and D) along with the four canonical NMPs. Nanopore current blockade events showed clear differences among all the eleven types of NMPs, except for UMP and m5C, which showed some overlap. However, to further discriminate all populations, plotting the noise (as standard deviation) versus the percentage blockage levels resulted in complete resolution of all moieties. This points to the sensitivity of the MspA-PBA system in being able to distinguish all eleven types of NMPs tested and discriminate structural differences among the canonical counterparts.

The main challenge was to accurately detect and identify the subtle differences in the modifications. The authors built a custom machine learning algorithm which used representative data events that were generated for each NMP and used this data-set for training into a machine learning algorithm. The classifier algorithm was able to distinguish each of the NMP moieties with an impressive close to 100% accuracy.

The feasibility of this MspA-PBA system combined with the machine learning algorithm was put to test by detecting naturally occurring RNAs such as microRNA and also tRNA, which is known to have one of the most complex modifications6. In separate experiments, each of these RNAs were first enzymatically degraded and then added onto the cis side of the nanopore. NMP compositions obtained from the raw events from each of the RNA subsets were consistent with the actual sequence, suggesting that each type of event is identifiable and quantifiable by the algorithm.

This nanopore-based PBA sensor could serve as a starting point for exonuclease-based sequencing where cleaved nucleotides can be sequentially fed into the nanopore sensor for detection. This sensor coupled with a direct RNA sequencing approach, such as that of Oxford Nanopore Technologies sequencing method7,8, could enable full-length epitrancriptome profiling of RNAs.

The effects of viral RNA modifications have been in the limelight more recently9, given their significance in SARS-CoV-2 infection where the SARS-CoV-2 RNA contains several modifications such as m6A, Ψ and 2′-O-methylation. However, one of the limitations of this MspA-PBA system is that it cannot detect ribose modified NMPs, such as 2′-O-methylcytidine.

This MspA-PBA system presents a promising approach for nanopore based sequencing of epigenetic RNA modifications. Nevertheless, there are a few things to keep in mind, moving forward with this system: library preparation might require extra purification steps to remove certain reagents present in the mixture that could contribute to the noise during electrophysiological recordings. Extracting RNA from biological samples, for example, microRNA from tissues, is difficult because of their low abundance. Also, further model training of the algorithm will be required as newer types of modifications get identified.