Multi-pass, single-molecule nanopore reading of long protein strands with single-amino acid sensitivity

Background

Single-molecule protein sequencing technologies

Traditionally, proteins have been studied using Edman degradation (Edman et al., 1950), mass spectroscopy (MS) and enzyme-linked immunosorbent assay (ELISA) (Engvall and Perlmann, 1971). Edman degradation involves the selective removal of one amino acid at a time from the N-terminus of a protein and identifying it through a series of chemical reactions, while MS uses the specific mass-to-charge ratio of molecules for characterisation. ELISA relies on the specific binding between an antibody and its target antigen. Despite the age of these methods, we are experiencing a “renaissance” of these methods with recent improvements in both throughput and accuracy (e.g. Swaminathan et al., 2018).

Additionally, new methods relying on fluorescence and nanopores for single-molecule sensing have emerged in recent years. Yet, many of them are still under development and not widely applied. Single-molecule fluorescent protein fingerprinting allows the observation of individual protein molecules using DNA-PAINT imaging, offering insights into heterogeneity within samples. This method can track conformational changes, interactions, and structural dynamics at a granular level. Meanwhile, affinity-based approaches capitalise on specific molecular interactions to isolate and identify proteins or protein complexes. These newer techniques complement traditional methods, enriching our understanding of protein behaviour, structure-function relationships, and intricate cellular processes. An overview and description of existing and evolving methods can be found as in Figure 1 of Alfaro et al. (2021). For further details, we recommend Restrepo-Pérez et al. (2018) and Timp and Timp (2020).

Phosphoproteome Profiling

Protein phosphorylations are typically quantified through mass spectrometry-based methods (Yu & Veenstra, 2021) such as tandem mass spectroscopy. However, the readout of these methods does not achieve single-amino acid sensitivity, necessitating the use of computational algorithms and models to pinpoint the specific phosphorylation sites. To achieve high-throughput analysis, multiplexing methods have been developed recently (e.g. Mertins et al., 2018 using isobaric tags) to simultaneously measure multiple samples over time and various doses in a single assay minimising variability and noise between samples. More detailed information can be found for instance in Feng, Sanford et al. (2023), which we highlighted in a previous preLights article.

Key Findings

Overview

Keisuke Motone, Daphne Kontogiorgos-Heintz and colleagues have developed a new method to detect and sequence full-length proteins at single-amino acid resolution using a specialised protein unfoldase to guide proteins through a nanopore. By introducing a sequence that induces the unfoldase to slip, it is possible to re-read the protein molecules increasing sequencing accuracy. Moreover, the authors show that posttranslational modifications can be read out simultaneously to the amino acid composition. However, the authors also clearly state the limitations of their pioneering study and mention challenges to be solved in the future.

Key Finding I: Achieving single-amino acid sensitivity through the use of a ClpX unfoldase

There are similar challenges between the sequencing of nucleic acids and of peptides via nanopore  – while chains can be translocated through nanopores with the use of a voltage difference across the membrane, such translocation is typically much too fast. Oxford Nanopore has resolved this problem by utilising a motor protein to ratchet the nucleic acids through the pore in a slow, controlled manner. Similarly, the authors have included a blocking domain within their protein, such that the protein gets stuck in the pore, before the ClpX motor is used to translocate the protein in a trans to cis direction at a regulated speed (Figure 1). This ensures that translocation takes place slowly enough for a usable current signal to be recorded.

Fig. 1 Nanopore protein sequencing with the aid of an unfoldase. Figure adapted from Motone, Kontogiorgos-Heintz et al. (2023) Figure 1A, BioRxiv published under the CC-BY-NC-ND 4.0 International licence.

The authors then designed 59 amino acid-long protein constructs that contain repeating sequence blocks and introduced single amino acid mutations at the central position within each block, with the ends of the blocks marked with double tyrosine mutations (which result in a characteristic dip in current) (Figure 2A). They were able to observe distinct and reproducible ionic signatures that corresponded to the single amino acid mutation within each construct (Figure 2B), showing that their method is sensitive to single amino acid residues, albeit in an artificial context.

Fig. 2 Sequencing at single-amino acid level sensitivity . Figure adapted from Motone, Kontogiorgos-Heintz et al. (2023) Figure 2A and 2B, BioRxiv published under the CC-BY-NC-ND 4.0 International licence.

The authors tested various machine learning models and were able to achieve an accuracy of 28% for all 20-way amino acid classifications (compared to 5.5% for a dummy model). They were also able to detect deamidation in asparagine, highlighting the potential of their technology to detect post-translational modifications.

Key Finding 2: Ability to re-read individual protein molecules

An important means of increasing the accuracy of sequencing methods is by resequencing the same molecule multiple times, and finding the consensus sequence. Such a method has been frequently utilised in long-read DNA sequencing, for example in UMIC-seq (Zurek et al., 2020) as utilised by Oxford Nanopore and in PacBio Circular Consensus Sequencing  (Wenger et al., 2019).

The authors demonstrate that such a strategy is also possible with their nanopore-based peptide sequencing by including a slip sequence at the N-terminal part of the peptide, together with the blocking domain at the C-terminal part of the peptide. The blocking domain (as shown above) traps the protein in the pore to enable motor-binding and trans to cis motor-driven translocation, while the slip sequence causes the motor protein to slip, resulting in cis to trans (voltage-driven) translocation (Figure 3). This makes multiple rounds of trans to cis (motor driven) and cis to trans (voltage-driven) translocation possible, resulting in the sequencing of the same sequence multiple times.

Fig. 3 Re-reading individual protein molecules. Figure adapted from Motone, Kontogiorgos-Heintz et al. (2023) Figure 5, BioRxiv published under the CC-BY-NC-ND 4.0 International licence.

Re-reading the same molecule improves the accuracy of the 20-way amino acid classification task from 28% (without re-reads) to 61% (with 10 re-reads). However, a shortcoming of this method is that the location at which the motor regains grip is not defined, such that some landmarks may be required to ascertain how the current traces align.

Key Finding 3: Ability to examine intact, folded protein domains

The end goal of a nanopore sequencing method would be to sequence natural proteins, many of which are folded or contain multiple folded domains. The authors tested their method on four different folded protein domains and could illustrate that the unfolding of folded domains is possible, although unsuccessful folding events can be seen, and that these are correlated with structural stability (Figure 4). This suggests that their method provides information on both structural (unfolding) and sequence (translocation) states of the proteins examined.

Fig. 4 Examining intact, folded protein domains. Figure adapted from Motone, Kontogiorgos-Heintz et al. (2023) Figure 6A, BioRxiv published under the CC-BY-NC-ND 4.0 International licence.

Conclusion and Perspective

We believe that the compatibility of the presented protein sequencing technology with existing ONT pores, which are already widely used for genome and transcriptome sequencing, makes this method a lot more accessible to scientists as their systems can be directly adapted for protein sequencing. However, given the early development stage of  protein sequencing, it seems to us that the only real-world application right now is peptide barcoding, given that the single aa-level sensitivity can only be achieved in very synthetic and artificial contexts. Nonetheless, peptide barcoding could have exciting applications in protein-based screens, and the speed at which this field is developing continues to excite us.

Still, we found it quite striking that around 20 amino acids can occupy the CsgG sensing region when in a stretched conformation, which leads to 20^20 = 1.048576e+26 possible states. That’s a lot of different possibilities to be read out in a single signal, especially given that the change in this signal is in the order of 10−12 Ampere. The complexity of uncoding must really be quite high and should be close to/exceeding the physical and practical limits. We are keen to see how groups within the field will deal with this challenge as they work towards single amino-acid resolution de novo protein nanopore sequencing.

Questions to the Authors

Q1: Could the method be modified to simultaneously sequence proteins alongside other omics modalities? If so, what are the challenges that need to be tackled?

Q2: Can this method be combined with BOSS-RUNS (Weilguny, L., De Maio, N., Munro, R. et al., 2023) to disregard previously sequenced/uninteresting protein species in real-time and refocus on species that haven’t been covered (in as much depth) before?

Q3: How uniform is the motion through the pore using unfoldase-mediated translocation? And how important is it for protein sequencing that the motion is uniform?

Q4: Based on intuition, I would have guessed that a much smaller pore volume and a much thinner membrane would have been needed for protein sequencing given that amino acids are in the order of a few Ångströms (Å) in size compared to DNA (⌀ 2nm). Why is protein sequencing working with a pore optimised for DNA sequencing? And would a smaller pore and a thinner membrane improve readout accuracy?

Q5: Do you think the ClpX unfoldase will be able to unfold all protein domains or only a subset of domains? Do you have some insights regarding methods to aid protein unfolding that would be compatible with your set-up?

Q6: What are the next steps to get towards true de novo peptide sequencing with nanopores? And how long do you think it will take the field to get there?

Q7: Do you think that mass spectrometry will still be relevant in 20 years (once protein sequencing becomes reliable and widely-used)?

References

Alfaro, J.A., Bohländer, P., Dai, M. et al. (2021) The emerging landscape of single-molecule protein sequencing technologies. Nat Methods 18, 604–617. https://doi.org/10.1038/s41592-021-01143-1

Edman, P., Högfeldt, E., Sillén, L. G., & Kinell, P.-O. (1950). Method for Determination of the Amino Acid Sequence in Peptides. In Acta Chemica Scandinavica (Vol. 4, pp. 283–293). Danish Chemical Society. https://doi.org/10.3891/acta.chem.scand.04-0283

Engvall E, Perlmann P. (1971) Enzyme-linked immunosorbent assay (ELISA). Quantitative assay of immunoglobulin G. Immunochemistry. https://doi.org/10.1016/0019-2791(71)90454-x

Feng, S., Sanford, J. A., Weber, T., Hutchinson-Bunch, C. M., Dakup, P. P., Paurus, V. L., … Wiley, H. S. (2023). A Phosphoproteomics Data Resource for Systems-level Modeling of Kinase Signaling Networks. bioRxiv. https://doi.org/10.1101/2023.08.03.551714

Mertins, P., Tang, L.C., Krug, K. et al. (2018) Reproducible workflow for multiplexed deep-scale proteome and phosphoproteome analysis of tumor tissues by liquid chromatography–mass spectrometry. Nat Protoc 13, 1632–1661. https://doi.org/10.1038/s41596-018-0006-9

Restrepo-Pérez, L., Joo, C. & Dekker, C. (2018) Paving the way to single-molecule protein sequencing. Nature Nanotech 13, 786–796. https://doi.org/10.1038/s41565-018-0236-6

Swaminathan, J., Boulgakov, A., Hernandez, E., Bardo, A. et al. (2018) Highly parallel single-molecule identification of proteins in zeptomole-scale mixtures. Nat Biotechnol 36, 1076–1082. https://doi.org/10.1038/nbt.4278

Timp W, Timp G. (2020) Beyond mass spectrometry, the next step in proteomics. Sci Adv.;6(2):eaax8978. https://doi.org/10.1126/sciadv.aax8978

Weilguny, L., De Maio, N., Munro, R. et al. Dynamic, adaptive sampling during nanopore sequencing using Bayesian experimental design. Nat Biotechnol 41, 1018–1025. https://doi.org/10.1038/s41587-022-01580-z

Wenger, A.M., Peluso, P., Rowell, W.J. et al. (2019) Accurate circular consensus long-read sequencing improves variant detection and assembly of a human genome. Nat Biotechnol 37, 1155–1162. https://doi.org/10.1038/s41587-019-0217-9

Yu, L. R., & Veenstra, T. D. (2021). Characterization of Phosphorylated Proteins Using Mass Spectrometry. Current protein & peptide science, 22(2), 148–157. https://doi.org/10.2174/1389203721999201123200439

Zurek, P.J., Knyphausen, P., Neufeld, K. et al. (2020). UMI-linked consensus sequencing enables phylogenetic analysis of directed evolution. Nat Commun 11, 6023. https://doi.org/10.1038/s41467-020-19687-9

Tags: oxford nanopore, post-translational modifications, protein sequencing, proteomics, sequencing

Posted on: 4 December 2023 , updated on: 15 December 2023

doi: https://doi.org/10.1242/prelights.36105

Read preprint (No Ratings Yet)