Research highlights: nanopore protein detection and analysis

Abstract

In this article we highlight recent work using nanopores to detect and study proteins. Nanopores are excellent single molecule sensors, capable of rapidly characterizing small molecules with relatively modest instrumentation requirements. Although the vast majority of recent effort and attention surrounding nanopores has centered on detection and sequencing of nucleic acids, proteins represent a more difficult and diverse analyte population, with a wide range of sizes, structures, charges, among other characteristics. Nanopores can be used to detect the presence of proteins of interest as well as to study their enzymatic activity, binding to ligands, and secondary structure. We highlight new work describing detection of specific protein species in solution by coupling them to a strand of carrier DNA that is used to electrophoretically transport the proteins through conical glass nanopores. Additionally, we spotlight another approach for nanopore detection of protein and other analytes through detection of their binding to aptamers—measurements which were quantitative to pM concentrations. Finally, we highlight studies in which protein secondary structure and folding energetics were studied through the use of an unfoldase enzyme coupled to a protein nanopore, a technique capable of detecting the effects of single amino acid mutations on the stability of the folded protein.

Nanopore detection of specific protein species using carrier DNA

Biomolecules such as DNA and protein have been previously analyzed with sub-nanometer resolution using inorganic or protein nanopore sensors through the measurement of pA–nA currents that are modulated by the presence of an analyte of interest at the pore entrance or interior.¹ The small size of biological nanopores, typically 1–2 nm, limits the range of analytes able to pass through the pore, requiring most proteins to unfold. The large size of solid-state nanopores, which can be fabricated at controllable and predictable sizes from two to hundreds of nanometers, allow proteins to traverse the nanopore intact.² Although single molecules can be detected, resolution remains a problem in differentiating similarly-sized molecules, as the measured signal is strongly related to molecular volume.³ Another issue in protein measurement is transport of the analyte to and through the pore, as electrophoresis, which is very useful in nanopore DNA measurement, is less effective for neutral or weakly charged species.⁴ Bell and Keyser report a technique using unmodified solid-state nanopores that facilitates globular protein translocation and enables specific protein identification.⁵ Their approach uses designed double-stranded DNA carrier molecules capable of binding one or more analytes. The highly charged DNA can electrophoretically transport the bound protein through the nanopore for sensing, and the binding site selectivity enables highly specific detection.

Bell and Keyser designed their DNA carrier by using the 7 kb-long M13mp18 ssDNA genome and designed 190 complementary oligonucleotides 38 bases long that hybridize along the ssDNA genome with even spacing. Some of these complementary oligomers at 1, 3, or 5 sites in the middle of the 7 kb DNA strand were functionalized with biotin, which binds to the globular protein streptavidin with very high affinity. These hybridized DNA strands, and strands with protein bound to them, were detected using glass nanopores 15 nm in diameter fabricated from capillaries using a laser-based pipette puller (Fig. 1). Translocation of the hybridized DNA through the nanopore was electrically measured by a ∼1 ms long, ∼100 pA reduction in measured current. Binding of the streptavidin to the biotinylated oligomers was indicated by the presence of a 50 pA, 100 μs spike reduction in current during the DNA translocation. The magnitude of this spike increased if there were additional streptavidin-binding oligomers present in the middle of the DNA strand. This increased signal allowed detection of bound protein in 99% of the measured translocations of DNA with 5 binding sites, and a 5% false positive rate. If the biotin binding sites were spaced further apart (∼1700 bp), the passage of the individual bound proteins through the pore could be resolved.

Fig. 1 Nanopore protein detection using carrier DNA. (A) A 7228 base long nucleotide is hybridized with complementary oligonucleotides 38 bases long. Nanopore detection of this DNA is indicated by a constant reduction in current of duration ∼1 ms. (B) If one oligonucleotide is conjugated to biotin and bound to streptavidin, the presence of the protein can be seen as a brief decrease in current during the nanopore measurement. (C and D) This reduction in current can be increased by having 3 and 5 consecutive molecules of streptavidin bound to the carrier DNA. (E) If the spacing between bound streptavidin is increased, they may be resolved individually. Adapted from Bell and Keyser.⁵

This approach was also able to identify a single analyte from a mixture. Using DNA with 3 biotin binding sites, two mixtures were compared: one mixture consisted of 15 nM each of streptavidin, β-lactoglobulin, β-galactosidase, and lysozyme and a second of 15 nM each of bovine serum albumin, β-lactoglobulin, β-galactosidase, and lysozyme without streptavidin. Measurements of the first mixture resulted in 85% of translocation events exceeding a defined threshold detection current, while the second mixture resulted in 15% of the events exceeding threshold current (same as the control with no biotinylated DNA). To further decrease the concentration limit of detection, the number of threshold events in the control must be further reduced, possibly with improved data filtering. To demonstrate the adaptability of DNA carriers, the authors used a different chemical tag for selectively detecting antidigoxigenin antibodies. Similar specificity results were obtained, indicating that this technique is generally adaptable to many protein sensing applications.

Synthesis of DNA strands capable of binding to specific biomolecules is able to give a binary indication of their presence in a solution containing the target and a background mixture. With further improvement in data analysis or binding site design, quantitative concentration determination may also be possible. This technique is highly generalizable—a DNA carrier can be easily designed and created for any combination of specific ligand receptor binding pairs. The adaptability of this platform opens many possibilities including detection of antibodies and single molecules inaccessible with other techniques.

Quantitative protein concentration determination using aptamers

Protein nanopores are also able to detect the presence of protein and identify specific protein species; however, because of the small size of biological pores, this detection is usually indirect. Previous studies have used aptamers for analyte detection in nanopores, either requiring conjugation of aptamers to the pore for sensing^6,7 or direct detection of the aptamer–analyte conjugate.⁸ Extending this work, Li et al. describe use of aptamers for single molecule sensing for detection of VEGF, thrombin, and cocaine using DNA aptamers and an α-hemolysin nanopore.⁹

Specifically, the authors indirectly detect the target molecule via competitive binding to an aptamer initially bound to complementary ssDNA functionalized with ferrocene⊂cucurbit[7]uril (Fc⊂CB[7]) for readout (Fig. 2). Although the target-specific aptamer is initially hybridized to the readout DNA, in the presence of the target the aptamer preferentially binds to it, leaving the readout DNA molecule single-stranded and able to electrophoretically enter the α-hemolysin pore for detection. With this mechanism, the readout ssDNA will only enter the nanopore in the presence of the target analyte, and therefore detection of the readout ssDNA translates to detection of the analyte.

Fig. 2 Nanopore protein detection using aptamers and “readout” DNA. (A) Aptamer specific for analyte A is combined with a complementary DNA probe functionalized with ferrocene⊂cucurbit[7]uril (Fc⊂CB[7]) for readout. Analyte A binds to the aptamer, leaving the DNA probe single-stranded and able to enter the protein nanopore (B). (C) The presence of the Fc⊂CB[7] on the probe DNA results in large extended blockades allowing the indirect detection of A. Adapted with permission from Li et al.⁹ Copyright 2015 Wiley-VCH.

With Fc⊂CB[7] conjugated to the readout DNA, large, long-lasting signals (400 pA, 0.1–1 s) were obtained following capture of the readout DNA in the α-hemolysin pore. These signals had a reproducible signature related to the presence of the Fc⊂CB[7] and its dissociation from the readout DNA. This signature allows differentiation of measured nanopore blockades caused by capture of the readout DNA versus interfering off-target analytes. The binding strength between the aptamer and the complementary ssDNA was tuned by introducing specific base mismatches to ensure that the aptamer will preferentially bind to the target molecule. Aptamer binding was highly specific, with measurements in the presence of several off-target analytes at high concentration indistinguishable from control.

As with previous nanopore detection work,¹⁰ the concentration of these analytes could be quantitatively determined by the rate of measured blockades, with the detection limit set by experimentally accessible times; the lowest concentrations measurable corresponded to several blockades measured in an hour. The authors demonstrated quantitative detection of three analytes: VEGF₁₂₁, thrombin, and cocaine, with detection limits of 500 pM, 5 nM, and 5 μM, respectively.

To further decrease the concentration limit of detection, the authors added biotin to the 3′ ends of the aptamers, enabling them to be bound by streptavidin-coated magnetic beads. In the presence of the analyte, the readout DNA separated from the aptamer as the analyte bound to the aptamer, leaving analyte–aptamer bead conjugates and readout DNA molecules free in solution. Application of a magnetic field enabled removal of all beads, leaving behind only the readout DNA. This protocol decreased non-signal background events and, following sample concentration, improved the detection limit 100-fold to 5 pM when detecting VEGF.

This approach is potentially very useful because the sensor specificity for detection of a wide range of different analytes can be easily changed through the use of the same pore but with different aptamers. The large signal and long duration is amenable for use with low cost amplifiers and data acquisition systems, enabling production of practical instrumentation. Although not demonstrated here, with readout ssDNA functionalized with groups in addition to Fc⊂CB[7], simultaneous detection of multiple analytes may be possible with multiple aptamers paired to readout DNA functionalized with these unique functional groups to simultaneously determine concentration profiles for each analyte.

Protein folding studied with unfoldase-coupled nanopores

Although nanopores have been actively explored for DNA sequencing, analyzing proteins through amino acid sequencing is less illuminating as a result of the importance of secondary and tertiary structure on protein morphology and function. Large nanopores can detect whole native protein, but the measurements coarsely reflect protein volume and can be insensitive to point mutations or other functional changes in the protein that do not significantly affect structure. Small nanopores require protein denaturation for translocation and currents during this can assist in identifying and characterizing protein domains,^11,12 conformational structure and dynamics,^13–15 and bulk characteristics of the protein.¹⁶

In a follow-up to their 2013 study,¹¹ Nivala et al. used an α-hemolysin nanopore and the ClpXP complex to study stepwise denaturation of a model protein, S2-GT, consisting of four separate domains, two ubiquitin-like protein (Smt-3) domains on either end, a titin fragment (I27), and green fluorescent protein (GFP) each joined in series by short peptide linkers. The carboxy terminus was modified with a polyanion tail (to assist in translocation) and an ssrA tag, which is specifically recognized by ClpXP. S2-GT was placed in solution on the cis side of α-hemolysin and ClpXP was placed on the trans side. The ClpXP complex consists of ClpX, a motor protein which pulls the peptide chain, and ClpP, a ClpX associated peptidase that degrades the unfolded protein fed from ClpX. The addition of ClpP as well as an ATP regeneration mixture in the trans compartment increases the number of complete protein translocation events observed. Measurements were performed by applying a cis–trans electrical potential to drive the polyanion tail through α-hemolysin for binding and processing by ClpXP. Electrical current was measured as ClpXP pulled the peptide chain through the pore, which necessitated denaturation of the protein domains of S2-GT (Fig. 3).¹²

Fig. 3 Protein unfolding studied with a nanopore. (A) S2-GT, a protein construct consisting of Smt3, titin I27, GFP, and Smt3 is fed into an α-hemolysin nanopore using an anionic polyGSD tag terminated by ssrA, which is recognized by ClpXP, a motor protein protease complex which pulls polypeptide chains and degrades them. (B) Current through α-hemolysin measured with S2-GT and ClpXP shows (ii) insertion of polyGSD into the pore, (iii) pulling Smt3 against the pore entrance, (iv) unfolding of Smt3, and similar processes for I27 (v and vi), GFP (vii, viii, and ix), and the terminal Smt3 (x and xi). Protein modification to affect stability of the folded domains was observable in the recorded current (see text). Adapted with permission from J. Nivala, L. Mulroney, G. Li, J. Schreiber and M. Akeson, ACS Nano, 2014, 8, 12365–12375. Copyright 2014 American Chemical Society.

Analysis of the current showed that its temporal sequence was repeatable and able to be associated with pre-unfolding and translocation events of the four domains of the protein. The authors identified features of the measured conductance at the beginning and end of the translocation (sections ii–iv and x–xi) to be related to the two Smt-3 domains, with the remaining events associated with the I27 and GFP domains. To confirm this and to investigate the effect of a mutation in the protein, a destabilizing point mutation was added to the I27 domain. The measured current of S2-GT with the mutant I27 domain showed that sections v and vi were altered, with all other events remaining unchanged, indicating that v and vi were associated with pre-unfolding and translocation of I27 and vii–ix were similarly associated with GFP.

Similarly, a “superfolding” mutant of GFP was tested, which is more resistant to chemical denaturants as a result of eleven point mutations. All sections of the measured current associated with Smt-3 and I27 were unaltered, but section ix was replaced with three unique features, representing intermediate unfolding steps prior to translocation of the superfolding GFP through the pore. The authors were able to further investigate these sub-events by cleaving the protein halfway through the superfolding GFP, measuring a current signature that terminated after just the first of the three “sub-events”, as well as testing additional GFP variants containing rearranged domains.

In order to quantify the analysis of the current signatures, the dwell time, average amplitude, and standard deviation of the different sections were used as numerical descriptors. Sections ii through ix, previously associated with the first Smt-3, I27, and GFP domains, were confirmed by random forest analysis to enable differentiation between the different S2-GT mutants. These statistics were further aggregated with a Naïve Bayes classifier to produce a confusion matrix, with >86% successful identification of the protein variant. The Naïve Bayes classifier and confusion matrix are an interesting addition to the study, potentially allowing reliable identification of a protein variant through quantification of current patterns.

The results here allow identification of protein domains and their variants given a pre-assembled library of current signatures for all relevant domains. Through statistical analysis and comparison with the assembled library, protein domains could be reliably identified. Quantitative analysis of the kinetics of unfolding (for example as a function of transmembrane voltage) could provide insights into the energetics and unfolding pathways of enzymes and structurally important proteins.

References

1. B. M. Venkatesan, D. Estrada, S. Banerjee, X. Jin, V. E. Dorgan, M.-H. Bae, N. R. Aluru, E. Pop and R. Bashir, ACS Nano, 2012, 6, 441–450.
2. W. Li, N. A. W. Bell, S. Hernández-Ainsa, V. V. Thacker, A. M. Thackray, R. Bujdoso and U. F. Keyser, ACS Nano, 2013, 7, 4129–4134.
3. C. Plesa, S. W. Kowalczyk, R. Zinsmeester, A. Y. Grosberg, Y. Rabin and C. Dekker, Nano Lett., 2013, 13, 658–663.
4. M. Firnkes, D. Pedone, J. Knezevic, M. Döblinger and U. Rant, Nano Lett., 2010, 10, 2162–2167.
5. N. A. W. Bell and U. F. Keyser, J. Am. Chem. Soc., 2015, 137, 2035–2041.
6. S. Ding, C. Gao and L.-Q. Gu, Anal. Chem., 2009, 81, 6649–6655.
7. D. Rotem, L. Jayasinghe, M. Salichou and H. Bayley, J. Am. Chem. Soc., 2012, 134, 2781–2787.
8. R. Kawano, T. Osaki, H. Sasaki, M. Takinoue, S. Yoshizawa and S. Takeuchi, J. Am. Chem. Soc., 2011, 133, 8474–8477.
9. T. Li, L. Liu, Y. Li, J. Xie and H.-C. Wu, Angew. Chem., 2015, 127, 7678–7681.
10. Y. Wang, D. Zheng, Q. Tan, M. X. Wang and L.-Q. Gu, Nat. Nanotechnol., 2011, 6, 668–674.
11. J. Nivala, D. B. Marks and M. Akeson, Nat. Biotechnol., 2013, 31, 247–250.
12. J. Nivala, L. Mulroney, G. Li, J. Schreiber and M. Akeson, ACS Nano, 2014, 8, 12365–12375.
13. A. Oukhaled, B. Cressiot, L. Bacri, M. Pastoriza-Gallego, J.-M. Betton, E. Bourhis, R. Jede, J. Gierak, L. Auvray and J. Pelta, ACS Nano, 2011, 5, 3628–3638.
14. G. Oukhaled, J. Mathé, A. L. Biance, L. Bacri, J. M. Betton, D. Lairez, J. Pelta and L. Auvray, Phys. Rev. Lett., 2007, 98, 158101.
15. L. Payet, M. N. Martinho, M. Pastoriza-Gallego, J.-M. Betton, L. Auvray, J. Pelta and J. R. M. Mathé, Anal. Chem., 2012, 84, 4071–4076.
16. D. S. Talaga and J. Li, J. Am. Chem. Soc., 2009, 131, 9287–9297.

Footnote

† Edited by Dino Di Carlo.

---