Article

This article is licensed under CC-BY 4.0

[Link]

Deep Learning-Assisted Single-Molecule

Detection of Protein Post-translational
Modifications with a Biological Nanopore
Chan Cao,* Pedro Magalhães,# Lucien F. Krapp,# Juan F. Bada Juarez, Simon Finn Mayer, Verena Rukes,
Anass Chiki, Hilal A. Lashuel,* and Matteo Dal Peraro*
Cite This: ACS Nano 2024, 18, 1504−1515 Read Online
See [Link] for options on how to legitimately share published articles.

ACCESS Metrics & More Article Recommendations *

sı Supporting Information
Downloaded via [Link] on March 21, 2026 at 06:28:03 (UTC).

ABSTRACT: Protein post-translational modifications (PTMs) play a crucial role in countless biological processes,
profoundly modulating protein properties on both spatial and temporal scales. Protein PTMs have also emerged as
reliable biomarkers for several diseases. However, only a handful of techniques are available to accurately measure
their levels, capture their complexity at a single molecule level, and characterize their multifaceted roles in health
and disease. Nanopore sensing provides high sensitivity for the detection of low-abundance proteins, holding the
potential to impact single-molecule proteomics and PTM detection, in particular. Here, we demonstrate the ability
of a biological nanopore, the pore-forming toxin aerolysin, to detect and distinguish α-synuclein-derived peptides
bearing single or multiple PTMs, namely, phosphorylation, nitration, and oxidation occurring at different positions
and in various combinations. The characteristic current signatures of the α-synuclein peptide and its PTM variants
could be confidently identified by using a deep learning model for signal processing. We further demonstrate that
this framework can quantify α-synuclein peptides at picomolar concentrations and detect the C-terminal peptides
generated by digestion of full-length α-synuclein. Collectively, our work highlights the advantage of using nanopores
as a tool for simultaneous detection of multiple PTMs and facilitates their use in biomarker discovery and
diagnostics.
KEYWORDS: biological nanopores, protein post-translational modifications, deep-learning, single-molecule sensing, α-synuclein

P roteins are the major molecular building blocks of life,

dictating the spatial and temporal occurrence of most
biological functions.1 Their functional properties are
mainly determined by their native three-dimensional structure,
subcellular localization, and interactome. Post-translational
sensitive methods to detect PTMs, generate site-specifically
modified proteins, and decipher the PTM code of proteins still
lags behind.
The standard methods currently available for the detection
modifications (PTMs) play an important role in regulating all of PTMs are mass spectrometry (MS), and enzyme-linked
these properties of proteins and thus represent molecular
switches for coordinating, diversifying, and regulating cellular Received: September 10, 2023
function.2 Moreover, the majority of protein-based biophar- Revised: November 16, 2023
maceuticals approved or in clinical trials bear some form of Accepted: December 12, 2023
PTMs in order to enhance their efficacy for therapeutic Published: December 19, 2023
applications.3 Despite the impressive recent advances in
biochemistry and structural biology, the development of
© 2023 The Authors. Published by
American Chemical Society [Link]
1504 ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

antibody-based assays,4,5 while immunoprecipitation coupled simultaneous occurrence of different types and combinations
to MS (IP/MS) has also been increasingly used to detect and of PTM species that are of clear clinical relevance. Moreover,
quantify low abundant protein species. Recently, great progress the detection of low-abundance protein PTMs from clinical
has been made toward developing and improving MS methods samples is still poorly investigated.
and immunoassays to map, detect, and quantify modified Here, we demonstrate that an engineered form of aerolysin,
proteins in biological samples. For instance, proximity a founding number of a major class of β-pore-forming toxins
extension assays6 and single molecule array (Simoa)7 have which has shown an excellent sensing capability for DNA and
shown high sensitivity for detecting PTMs. One of the major peptide analysis,34−37 can be used to detect peptides
drawbacks of these immunoassays is that they are based on containing single or multiple types of PTMs. As a model
antibodies that are developed against single PTMs. Single- system, we used peptides derived from the C-terminal domain
molecule techniques such as electron tunneling have also of the presynaptic protein α-synuclein, the primary constituent
achieved the detection of single PTM on peptides (i.e., of the pathological hallmarks of PD (i.e., Lewy bodies and
phosphorylation) with high sensitivity.5,8,9 However, the Lewy neurites). This region is characterized by the clustering
development and optimization of these methods did not of different types of PTMs (e.g., phosphorylation, nitration and
account for the presence of multiple and/or neighboring oxidation) that are associated with pathology formation and
PTMs.5 One example is demonstrated in our recent work, disease progression in PD and other neurodegenerative
which shows that neighboring PTMs interfere with the diseases.38 Recent studies also suggest that specific α-synuclein
detection of phosphorylated α-synuclein at S129 by most PTMs (i.e., pS129) are elevated in the cerebral spinal fluid
pS129 antibodies.10 Therefore, a method to accurately identify (CSF) of PD patients and could be developed into diagnostic
protein PTMs is essential as increasing evidence suggests the biomarkers for PD and potentially for other neurodegenerative
PTM code is combinatorial in nature and involves complex diseases.39,40
interplay and cross-talk among PTMs.11,12 We observe that phosphorylation, nitration, and oxidation
PTMs have been implicated in a wide variety of diseases, significantly modulate the current amplitude and dwell time of
such as characterizing the changes in protein PTM types and the signals in comparison with the wild-type (wt) α-synuclein.
levels is highly relevant for early diagnosis and monitoring the Additionally, we show that the α-synuclein peptides bearing
progression of diseases, as well as for evaluating the efficacy of multiple PTMs and different PTMs combinations induce
emerging therapies.4,13 For example, several neurodegenerative characteristic ionic current signatures, which provides an
diseases including Alzheimer’s disease (AD) and Parkinson’s opportunity to investigate PTMs crosstalk and monitor the
disease (PD) are caused by the accumulation of misfolded and PTMs dynamics in the future. A deep learning approach was
aggregated proteins in the brain regions that are affected by the specifically developed to process the signals, providing an
disease (e.g., amyloid-β in amyloid plaques, Tau in neuro- automatic and fast way to classify the current readouts
fibrillary tangles, and α-synuclein in Lewy bodies and Lewy associated with each PTM combination. We also demonstrate
neurites). One shared characteristic among these proteins is the ability of aerolysin pores to detect unmodified α-synuclein
that their aggregated forms are heavily modified. PTMs have peptides when mixed with red blood cells (RBCs) at picomolar
emerged as key signatures of disease pathologies and are concentration as well as the possibility of using cathepsin D to
commonly used as the primary (bio)markers of disease generate the specific C-terminal peptides that bear most
progression and pathology formation, spreading and clearance PTMs, providing promising ground for actual clinical
in response to therapies.14,15 applications. In summary, our work presents a single-molecule
Nanopore sensing is an approach based on ionic current approach able to detect clinically relevant low-abundance α-
readout that can detect a single molecule as it is passing synuclein PTMs. Compared to other analytical methods like
through a nanometer scale pore.16 Such pores are either MS and ELISA,41,42 this nanopore-based approach offers
biological assemblies of proteins embedded in a lipid detection at single-molecule resolution and provides promising
membrane or are fabricated by a solid-state material.17 When ground to decipher PTMs patterns relevant for disease
a molecule of interest passes through such a pore, the electric diagnostics.
current signal is modulated and exquisitely sensitive to the
molecule of interest and thus can provide information about its RESULTS AND DISCUSSION
size, mass, charge, composition, structure and conformation in Characterization of the C-Terminal Region of α-
real-time.18,19 Nanopore sensing has achieved sequencing of Synuclein with a Nanopore. The ability of aerolysin to
ultralong DNA,20,21 and this success has inspired its detect PTMs was examined by means of single-channel
application for peptide and protein analyses.22−25 Compared recording experiments, as shown in Figure 1a. A lipid
with MS, nanopore sensing is faster, cheaper, free of ionization, membrane separates the chamber into two compartments, cis
and has an ultrasensitive detection of biomolecules in solution and trans, and holds an aerolysin pore that connects them. An
at the single-molecule level. These distinguishing features, in engineered variant of aerolysin is used here, namely K238A,
particular its sensitivity, make nanopores perfectly suited for where lysine residues at position 238 are substituted by alanine
protein analysis since there is no biochemical method available residues, providing a significantly enhanced resolution for
for protein amplification, like polymerase chain reaction, that biomolecular sensing compared to wt aerolysin.43 As
has been widely used for DNA amplification. Recently, several mentioned above, we focused on α-synuclein as a model
groups have explored the potential of using nanopore system because of the strong links between its PTM profile and
approaches to detect protein PTMs, including phosphoryla- the brain pathology of several neurodegenerative diseases,
tion,26−28 acetylation,29 propionylation,30 glycosylation,31,32 most notably in PD.38,44
and ubiquitination.33 However, these studies focus on The C-terminal domain of α-synuclein, encompassing the
identifying either different types of modifications or the same residues 124−140 (α-syn124−140) (Figure 1b) harbors several
modification at different positions but do not consider the PTMs that are found in pathological α-synuclein aggregates in
1505 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Figure 1. Characterization of wild-type α-synuclein C-terminal peptides. (a) Illustration of the single-channel recording setup composed of
two chambers separated by a lipid membrane that is formed across an orifice; the chambers are named cis and trans. Voltage is applied across
the pore using two Ag/AgCl electrodes. (b) The structure of α-synuclein in full-length (PDB: 1xq8), including the amphipathic N-terminus
(yellow), nonamyloid-component (NAC) domain (green), and the disordered C-terminal fragment (blue). Phosphorylation (red circle),
nitration (blue circle), and oxidation (green circle) of residues 124−140 are highlighted. Nanopore single channel recording of wt α-
syn124−140 peptide in 1.0 M KCl (c) and in 1.0 M KCl, 8.0 M urea (d) solutions, respectively, both buffered with 10 mM Tris, and 1.0 mM
EDTA at pH 7.4. Left panels of (c) and (d) report raw current traces, middle panels typical events, while the right panels show scatter plots,
I/I0 percentage, and dwell time histograms. All data were obtained by applying a voltage of +100 mV.

the brain of patients with PD and other neurodegenerative observed (Figure 1c, right): one exhibited a relative current
diseases.44 Many of these PTMs, including phosphorylation percentage (I/I0, see Methods) of 9.0 ± 2.0, while the other
and nitration, have emerged as reliable markers of α-synuclein 17.1 ± 2.0 (named hereafter PI and PII, respectively). These
pathology formation in human brains and animal models of PD values correspond to the mean and standard deviation derived
and related synucleinopathies.38 The C-terminal domain of α- from a Gaussian fit of the relative current histogram.
synuclein is rich in proline residues, it is highly negatively To gain further insights into this peculiar two-level signal, we
charged and does not adopt a stable secondary structure in the considered the structure of this α-synuclein fragment. Based on
monomeric state of the protein,45 which in this context is previous studies,49 there is a highly hydrophobic segment
particularly convenient as these peptides can be easily driven to between residues 125−129 that can produce a structurally
translocate through the nanopore by the potential applied compact cluster likely responsible for the initial lower residual
across the lipid membrane. All peptides were prepared using current. To test this hypothesis, we reasoned that destabilizing
Fmoc-based solid-phase peptide synthesis and purified as this conformation by the addition of urea could produce
previously reported46−48 (see Methods). signals expected for a more extended peptide.49 We thus used
α-Syn124−140 was added into the cis compartment and when a 8.0 M urea (along with 1.0 M KCl, 10 mM Tris, 1.0 mM
positive voltage was applied to the trans compartment, clear EDTA, and pH 7.4) to induce complete unfolding of the
and reproducible blockades of ionic current were obtained peptide, while aerolysin remained folded and functional
(Figure 1c,d). For wt α-syn124−140, a peculiar and well- (Supplementary Figure S1). Under these conditions, the
recognizable 2-level blockade event was observed and the open pore current was lower (25 ± 1.8 pA at +100 mV)
signal often contained a higher residual current at the last compared to the same salt concentration without urea (72 ±
fraction (Figure 1c, middle). As a result, two populations were 1.5 pA at +100 mV). This is because the additional urea
1506 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

significantly decreased the mobility of the ions.50,51 However, double-phosphorylated α-syn124−140, pY125pS129, and triple
in these conditions, wt α-syn124−140 translocated with a simple nitrated α-syn124−140, nY125nY133nY136 (Supplementary
one-level signal as evidenced by one population in the scatter Figure S4c,d). Here, the nanopore results showed that
plot and the I/I0 histograms (Figure 1d). These one-level compared to the unmodified peptide (wt) and the singly
current signals suggest that the addition of urea results in a phosphorylated peptides (i.e., pY125 and pS129), the
more extended conformation of the polypeptide chain, which translocation speed of pY125pS129 (0.45 ± 0.02 ms) was
facilitates its translocation through the pore. In addition, due to even faster, likely due to its increased negative charge. While
the decrease of ionic mobility, the dwell time became much the dwell time of nY125nY133nY136 was around 2-fold longer
longer, 8.6 ± 0.24 ms, approximately 4 times longer than the (5.18 ± 0.23 ms) compared to wt and nY136, and slightly
condition without urea. These results demonstrate that the longer than nY125: its relative current was the lowest among
conformation of largely disordered peptides such as those all peptides (7.2 ± 1.0) since the three modifications
obtained from α-synuclein can be monitored at the single- contribute to increasing the overall volume of the peptide
molecule level using a nanopore. and therefore induce a deeper blockade of ionic current. In the
Deep Learning-Assisted Detection of Different PTMs case of the peptides containing two types of PTMs at different
and Their Combinations. Next, we assessed the feasibility of positions, i.e., nitration at Y125 and phosphorylation at S129
using nanopores for detecting PTMs of α-syn124−140 peptide. (nY125pS129, Supplementary Figure S4e), only one pop-
We investigated different types of modifications occurring at ulation was observed. The relative current of nY125pS129 was
the same amino acid (i.e., phosphorylation and nitration of between the values of the single modifications nY125 and
Y125, pY125 and nY125), or the same type of PTM located at pS129, as observed also for the width of the relative current
different positions (i.e., pS129 and pY125, nY125 and nY136). distribution. The dwell time of nY125pS129 was slightly longer
Moreover, we considered peptides bearing multiple PTMs of than that of pS129, but identical with nY125. For all of these
the same type (i.e., pY125pS129 and nY125nY133nY136) and peptides, the dwell time decreases as the voltage increases,
of different types (i.e., nY125pS129). indicating that the collected signals are indeed induced by
As shown in Figure 2a, the current signatures of the different peptides translocating through the nanopore. Altogether, these
PTM combinations appeared to all have distinct current observations demonstrate that the engineered K238A aerolysin
characteristics. Compared to the unmodified peptide, the can capture the diversity of PTMs.
pY125 and nY125 α-syn124−140 peptide presented only one To classify these PTMs in a more precise, unbiased, and
level and their relative current were lower than the wt peptide automatic way that would allow a translation to clinical
(Supplementary Figure S2). This decrease in the relative applications in the future, we developed a tailored deep
current value is consistent with a deeper blockade induced by learning approach for processing the nanopore current
the additional volume due to the presence of the PTMs. readouts. A long short-term memory (LSTM) recurrent neural
Moreover, the dwell time varied significantly with different network was integrated to read the local extrema of events
types of PTMs, the fitted values being 2.58 ± 0.4 ms for wt, followed by a multilayer perceptron (MLP) to classify the
0.55 ± 0.08 ms for pY125, and 4.51 ± 0.5 ms for nY125, peptides (Figure 2b), a pipeline which is similar to our
indicating that phosphorylation speeds up the translocation previous analysis of informational polymers53 (see details in
process, while nitration significantly slows it down. The faster Methods). We first detected the events with a cutoff threshold
translocation of pY125 could be induced by the additional at 3σ from the open pore current and then computed statistical
negatively charged phosphate group. In addition, we features (i.e., mean residual current, standard deviation, and
investigated the effect of oxidation of methionine at position dwell time) and extracted the local extrema of the events
127 of α-syn124−140 (oM127, Supplementary Figure S3). (Supplementary Figure S5 and Methods). For each single
Similar to pY125 and nY125, oM127 also showed only one peptide, we randomly used 75% of the recorded data to feed
current level. the deep learning algorithm and train the model; the remaining
When the same PTM occurs at different positions in the 25% was used as validation set to compute the accuracy of the
peptide sequence, the ionic current and dwell time were model for each single peptide. The events are filtered using a
modulated differently. For example, the dwell time of pS129 threshold on the predictions confidence of the model which is
(3.62 ± 0.2 ms) is 6.5 times longer than that of pY125 (0.55 ± translated into a selection percentage.
0.08 ms). We hypothesize that phosphorylation at Y125 As shown in Figure 2c, this approach allows for differ-
(pY125) disrupts the hydrophobic cluster of α-syn124−140,49 all entiating wt α-syn124−140 from peptides that contain different
amino acids become more easily exposed and translocate in a PTMs at the same position Y125 (i.e., pY125 and nY125) with
linear fashion. In this case, the additional phosphate group of an accuracy of 94% at a 50% selection percentage. Control
pY125 contributes to faster translocation under the applied experiments with a mixture of wt and pY125 at different
voltages. Similar results were obtained when we compared concentration ratios were performed to further test the deep
nY125 and nY136 α-syn124−140 (Supplementary Figures S2b learning model (Figure 2d). First, a mixture of an equimolar
and S4b). Unlike nY125, which only showed one population, ratio (1:1) was measured in nanopore experiments and the
the relative current of nY136 was less pronounced compared to percentage of assignment of wt and pY125 was 52.8 ± 3.5%
that of wt and showed two populations like wt α-syn124−140. and 43.2 ± 3.1%, respectively. This is in line with the
This suggests that disrupting the local structure facilitates the theoretical predictions. When the ratio of wt:pY125 was
detection of PTMs in proteins by the aerolysin pore. As changed to 1:3, the percentage of wt assignment decreased to
reported recently, this could be also achieved by using 19.0 ± 0.3% while pY125 increased to 78.2 ± 0.7%. Finally, a
chemical denaturants such as urea or guanidinium chloride.52 mixture ratio of 3:2 (wt:pY125) was tested. As shown in Figure
To explore the possibility of using the nanopore for the 2d, the assignments of wt and pY125 were as expected 61.5 ±
detection of multiple PTMs occurring simultaneously on the 0.1% and 35.0 ± 0.2%, respectively. Therefore, the expectation
same peptide, we measured the ionic current response of is that any sequence in the library of peptides (wt, pY125, or
1507 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Figure 2. Discrimination of α-synuclein peptide PTMs by deep learning. (a) Typical ionic current signals of different PTM types and their
spatial combinations. The scale shown in the left top panel applies to all of the events. (b) Deep learning approach to signal processing: an
LSTM recurrent neural network was used to read the events, followed by an MLP to predict the peptides. (c) The normalized confusion
matrix over the predictions shown as percentage of wt, pY125, and nY125 peptides classification using a deep learning approach. (d)
Assignment percentage of different mixture samples of wt and pY125 at a concentration ratio of 1:1 (left), 1:3 (middle), and 3:2 (right). The
theoretical accuracy is shown by the white columns while the predicted accuracy is represented by blue columns. (e) Selection percentage
versus averaged accuracy obtained from deep learning approach of peptides, including wt, pY125, nY125, pS129, pY125pS129,
nY125nY133nY136, and nY125pS129. (f) The normalized confusion matrix over the predictions shown as percentage of wt, pY125, nY125,
pS129, pY125pS129, nY125nY133nY136, and nY125pS129 classification. Columns represent actual peptides from the test set, while rows
represent the peptides that the deep learning algorithm assigned them to. All data were obtained using 1.0 M KCl, 10 mM Tris, and 1.0 mM
EDTA buffer at pH 7.4 by applying a voltage of +100 mV.

nY125) can be identified directly with a probability of 94%. Data from wt and the 7 tested α-syn124−140 PTM variants
Furthermore, mixture experiments of wt and pS129 were (i.e., pY125, nY125, pS129, nY136, pY125pS129, nY125-
performed (Supplementary Figure S6), for which the deep nY133nY136, and nY125pS129) have been used as a training
learning model could predict the different mixture ratios with a set for the same deep learning model, with a resulting accuracy
probability of 96%. of 78.2% using a selection percentage of 25% (i.e., accuracy
1508 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Figure 3. Detection of α-synuclein peptides at low abundance concentration. (a) Raw current traces after addition of serial dilutions of Hb-
depleted RBC samples. For 1:1 condition, 3 μL of red blood cell lysates devoid of hemoglobin were added to the 150 μL nanopore chamber
(the total protein content is 0.4−0.5 μg/μL). (b) Raw current traces of Hb-depleted RBCs sample in the presence of 6 μM wt α-syn124−140
peptides. The RBCs used here is 1:10 dilution. (c) Raw current traces showing signals of α-syn124−140 in symmetric salt (top, 1.0 M KCl) and
in a gradient salt concentration (bottom, 0.15 M/3.0 M (cis/trans) KCl). (d) The correlation between signal frequency fsig and the
concentration of wt α-syn124−140 in gradient salt conditions ranging from 100 pM to 200 nM peptide in the cis chamber under +100 mV
applied voltage. The error bars represent standard deviations from at least three independent experiments under the same conditions.

dependence on selection percentage as shown in Figure 2e). If on biological samples.54 Nevertheless, even though some
the event is too noisy or does not contain enough information nanopores have been used for the detection of biomolecules in
as in the case of short events, the prediction confidence would biological fluids samples (e.g., plasma,55,56 serum,54,57−61
be lower. The selection percentage allows for a trade-off sweat,62 urine,62 and saliva62), the task of measuring biomarker
between accuracy and filtering without requiring heavy filtering peptides is often challenging due to the intrinsic low sensitivity
during the preprocessing of the events. As illustrated in Figure and selectivity of the techniques applied to bodily fluids.
2f, 13% of pY125 was confused with pY25pS129, because the We previously showed that aerolysin pores possess high
signals of both pY125 and pY25pS129 showed very short dwell sensitivity for biomolecular sensing, especially its engineered
times. This could be further improved if higher-bandwidth variants.43,63 Therefore, we explored if K238A aerolysin was
instruments have been employed. Additionally, 6% of nY125 able to detect α-synuclein peptides in clinically relevant
was confused with nY125nY133nY136, while the other 16% of conditions, red blood cells (RBCs) devoid of hemoglobin
this peptide was confused with nY125pS129, 10% of nY136 (Hb) proteins. We first determined the dilution necessary to
was confused with pS129 and the other 9% was confused with ensure a stable baseline of the ionic current through serial
wt. We think this prediction confusion was caused by the dilutions of Hb-depleted RBC samples. In the literature, no
similar dwell times between these modified peptides. None- particular instabilities of the lipid bilayer were observed when
theless, the deep learning approach we developed here testing RBCs except when dilutions were lower than 1:50.62
provides high reading accuracy for all of the tested PTMs Based on these previous reports, we scanned a range of
peptides, particularly for pS129 which is one of the most dilutions up to 1:100 (Figure 3a) in our setting and observed
relevant biomarkers for synucleinopathies. This is encouraging that after the addition of 3 μL of Hb-depleted RBCs extracts
also because a similar deep learning model can be trained for into the cis chamber, no obvious effects were detected for the
other systems, thus finding general applicability for signal membrane stability. Interestingly, during the experiments with
processing in nanopore sensing experiments. Hb-depleted RBCs, we could continually collect data for up to
Detection of α-Synuclein Peptides at Low Abun- 3 h, indicating that the aerolysin pore remained invariantly
dance Concentration. To date, most nanopore sensing intact likely due to the ultrastability provided by its double β-
applications are performed using purified peptides or protein barrel structure.35 Moreover, when the concentration of Hb-
samples (as done hereinabove). There is however the need to depleted RBCs samples increased in the chamber, occasional
develop methodologies whereby detection can be performed signals were also observed (Supplementary Figure S7).
1509 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Figure 4. Detection of C-terminal peptides of α-synuclein after CtsD digestion. (a) Possible peptide fragments from CtsD digestion based on
previous studies66 and MS characterization of CtsD digestion of full-length α-synuclein. (b) Raw current trace from nanopore measurement
of CtsD-digested samples. (c) Deep learning result from 5 independent experiments aimed at classifying α-syn125−140 from the mixture
sample.

Notably, all raw current traces shown in Figure 3 were Detection of α-Synuclein Peptides in Complex
collected 30 min after samples were added. Mixtures. Next, we sought to determine if the presence of
Next, we sought to quantify the presence of α-synuclein wt multiple fragments, produced for instance by proteolytic
peptide within the Hb-depleted RBCs, with the aim of cleavage of α-synuclein, could influence the nanopore
mimicking the real complexity of pathological conditions detection of C-terminal peptides bearing the desired PTMs.
within a clinical setting. Based on the previous tests, dilutions This aims at mimicking and describing the actual α-synuclein
of 1:10 were chosen as the background for Hb-depleted RBCs mixtures that one could find in an actual clinical sample. For
extracts. As illustrated in Figure 3b, after the addition of wt α- this purpose, we choose to work with cathepsin D (CtsD)
syn124−140, typical two-level signals were recorded. protease as previous studies showed that CtsD-mediated
While these results represent the potential of aerolysin pores digestion of α-synuclein results in the generation of a C-
to detect C-terminal α-synuclein peptide, another prerequisite terminal fragment spanning residues 125−140 (α-syn125−140),
toward the development of aerolysin pores into an efficient in addition to four other α-synuclein fragments of various
single-molecule proteomic device for biomarker detection is length67,68 (Figure 4a). The C-terminal fragment contains the
the ability to detect molecules in low abundance, as often great majority of known disease-associated PTM sites (p/
found in clinical samples. In previous studies, the frequency of nY125, pS129, nY133, nY136), including all the ones we
signals has been used to quantify the detected molecule. investigate above for α-syn124−140. To the best of our
Therefore, we used the equation fsig = kon [PTMs]0 to explore knowledge, CtsD is the only enzyme that cleaves α-synuclein
the theoretical detection limit of aerolysin nanopores for and results in the generation of a peptide fragment that
sensing α-synuclein peptides. As illustrated in Supplementary contains all the most relevant PTM sites, thus making it the
Figure S8, we verified that the frequency of blockade events, ideal choice for future studies to detect α-synuclein C-terminal
fsig, was proportional to the concentration of α-synuclein peptides bearing single or multiple PTMs. Other enzymes
ranging from 120 nM to 24 μM, when +100 mV were applied which cleave within the C-terminal domain, e.g., Glu-C, give
using a symmetric buffer of 1.0 M KCl. The physiological rise to much shorter peptides, thus precluding the detection of
concentration of α-synuclein in RBCs is however much lower single peptides containing multiple PTMs.44
than these tested conditions, namely 26.2 ± 3.0 μg/mL,64,65 We incubated CtsD with full-length α-synuclein (see details
which corresponds to 35 nM if 3 μL of the original sample was in Methods) and detected the resulting products by nanopore
added into the nanopore system. Therefore, to enhance the experiments afterward. In parallel, the same sample was
detection limit of our setting, we measured peptides in a characterized by MS and, as shown in Figure 4a, CtsD
gradient of salt concentration, which was shown to be an digestion produced three main peptide fragments, α-syn5−38, α-
efficient way to increase the capture rate for ssDNA.56,66 An syn95−124, and α-syn125−140. After the CtsD-digested α-
asymmetrical KCl buffer solution consisting of 0.15 M in the synuclein sample was added into the nanopore system, diverse
cis chamber and 3.0 M in the trans was thus used (Figure 3c), current signals were obtained (Figure 4b). Under these
which enabled the detection of wt α-synuclein peptides at far conditions, the deep learning model was able to identify the
lower concentrations (i.e., ∼100 pM, Figure 3d) than when C-terminal peptides α-syn125−140 with an average accuracy of
using canonical symmetric buffer conditions. 81.4 ± 1.1% from five independent experiments (Figure 4c),
Altogether, these results demonstrate that K238A aerolysin implying that the other proteolytic fragments, while having an
nanopores can detect C-terminal peptides of α-synuclein at effect of the current readout do not compromise the ability to
∼nM concentration (factoring in the dilution conditions), a detect the α-syn125−140 fragment. It should be noted that there
condition which is overlapping with the concentration at which is one amino acid difference between peptides used to train the
they are found in a RBC clinical sample. Considering that an deep learning model and this α-syn125−140 fragment; a
even lower detection limit could be achieved by applying difference which however does not seem to affect significantly
higher voltages, minimizing the volume of the chambers, or the accuracy of predictions, likely because Ala124 does not
optimizing the pore variants, the possibility of developing an contribute in a significant way to modulate the current signal
effective nanopore-based tool for the diagnosis of synucleino- during translocation, while not holding any PTM. Moreover, as
pathies appears within reach. α-syn5−38 is positively charged, we do not expect it to be
1510 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

captured by the nanopore when using +100 mV voltage. (oM127 and wt shown in Supplementary Figure S2) were provided
Therefore, the only possible fragment that may interfere with by GenicBio Limited. All peptides were characterized by Liquid
the measurement is α-syn95−124 since it is negatively charged. chromatography−mass spectrometry (LC-MS) as previously de-
However, it should be noted that the length of α-syn95−124 is scribed.70 Their purity was also assessed by UPLC analysis, on a
Waters Acquity H-Class system using a C18 column (with UV
two times longer than α-syn125−140 which may lead to a low detection at 214 nm and run time of 4 min (gradient 10% to 90%
capture rate in the conditions used here. As shown for DNA, acetonitrile) with 0.6 mL/min flow rate.
the capture of long ssDNA is dramatically reduced in the RBCs were prepared similarly as previously described.75 Briefly,
aerolysin nanopore.69 RBCs were lysed and further treated using the Hemovoid kit (Biotech
Support Group), aiming to remove hemoglobin (the most abundant
CONCLUSIONS protein in RBCs) and to enrich low-abundance proteins such as α-
synuclein. Importantly, all biological fluids used in this study are
In this work, we have shown that an engineered aerolysin derived from healthy controls.
nanopore can detect and distinguish peptides carrying different CtsD digestion experiments were performed in a total volume of
types and numbers of PTMs occurring at different residues. In 100 μL. Specifically, 2 μL of 500 μM CtsD (Sigma-Aldrich Chemie
addition, the measurements can capture subtle structural GmbH, Buchs, SG Switzerland), 3 μL of 100 μM full-length α-
features that are challenging to be characterized by other synuclein and 95 μL of buffer (40 mM sodium acetate, 50 mM NaCl
biophysical methods without fluorescently labeling or modify- and 5 mM DTT, pH 5) incubated at 37 °C for 20 h at 300 rpm in a
ing the peptides/proteins. Using a deep learning model for ThermoMixer C.
signal processing, all investigated PTMs could be automatically Aerolysin Productions. The recombinant K238A aerolysin
identified in a supervised context, which means that this proteins were generated from the aerolysin gene in the pET22b
vector with a C-terminal hexa-histidine tag as described in our
approach can be scaled up to identify more PTMs or scaled
previous work,43,53 and then expressed and purified from BL21 DE3
down to fit a specific application. Importantly, this nanopore pLys E. coli cells. Cells were grown to an optical density of 0.6−0.7 in
approach can reach a detection limit as low as 100 picomolar Luria−Bertani (LB) media. Protein expression was induced by the
concentration and is amenable for high-throughput applica- addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and
tions, which are challenging for other techniques such as MS. subsequent growth overnight at 20 °C. Cell pellets were resuspended
Finally, one major advantage of this nanopore-based approach in lysis buffer (20 mM Sodium phosphate pH 7.4, 500 mM NaCl)
is that it enables simultaneous detection of several protein mixed with cOmplete Protease Inhibitor Cocktail (Roche) and then
PTMs which is difficult to achieve using immunoassay/ lysed by sonication. The resulting suspensions were centrifuged
antibody-based methods. This is because the presence of (12.000 rpm for 35 min at 4 °C), and the supernatants were purified
multiple PTMs alters the biochemical properties of the through a HisTrap HP column (GE Healthcare) previously
equilibrated with lysis buffer. The protein was eluted with a gradient
antibody-targeting epitopes. over 40 column volumes of elution buffer (20 mM Sodium phosphate
Altogether, our findings advance the development of pH 7.4, 500 mM NaCl, 500 mM Imidazole), and buffer was
biological nanopores into efficient single-molecule proteomic exchanged into a final buffer (20 mM Tris, pH 7.4, 500 mM NaCl)
devices and diagnostic tools. We envision that this could be using a HiPrep Desalting column (GE Healthcare). The purified
achieved by detecting the circulating peptides directly, or by protein was flash-frozen in liquid nitrogen and stored at −20 °C.
isolating target proteins from various biological fluids using Single-Channel Recording Experiments. Phospholipids of 1,2-
specific antibodies70 and then digesting them into smaller diphytanoyl-sn-glycero-3-phosphocholine (DPhPC) powder (Avanti
peptides and subsequently detecting them through a nanopore Polar Lipids, Alabaster, USA) were dissolved in octane (Sigma-
as reported in the recent studies.67,71 In the future, clinical Aldrich Chemie GmbH, Buchs, Switzerland) for a final concentration
of 1.0 mg per 100 μL. Purified protein was diluted to the
samples could be treated directly with the specific protease, concentration of 0.2 μg/mL and then incubated with Trypsin-agarose
rather than extracting the target proteins, and subsequently (Sigma-Aldrich Chemie GmbH, Buchs, SG Switzerland) for 2 h under
detected and classified using nanopores. Several very recent 4 °C temperature. The solution was finally centrifuged to remove
studies have shown that nanopore sensors can be used identify trypsin.
different types of proteins by assessing characteristic blockage Nanopore single-channel recording experiments were performed
event features linked to peptide fragments produced by on Orbit Mini equipment (Nanion, Munich, Germany) and an
digestion of the proteins with proteases.72,73 Such nanopore- Axopatch 200B Amplifier system (Molecular Devices, San Jose, USA).
based molecular diagnosis platform would hold also promise to In the Orbit Mini setup, DPhPC membranes were formed across a
detect multiple biomarkers simultaneously, broadly extending MECA 4 recording chip that contains a 2 × 2 array of cylindrical 50
μm diameter in a highly inert polymer. Each of the four cavities
the potential field of application, for instance for the detection
contains an individual integrated Ag/AgCl-microelectrode and
of biomarker signatures for cancer diagnosis.74 sustains one DPhPC bilayer. If not indicated otherwise, the
The ability of the aerolysin pore to detect and distinguish measurement chamber temperature was set to 20 °C. Data was
between long peptides bearing multiple PTMs also provides a collected at 10 kHz sampling rate with a 5 kHz low-pass filter.
more precise remapping of PTMs in proteins at the single- Ionic strength gradient experiments were carried out as follows.
molecule level, which remains a challenge for MS and other Teflon films with 50 μm apertures were mounted in Teflon chambers
methods. In conclusion, the nanopore-based technology using high-vacuum grease (Dow Corning Corporation, Midland, MI,
presented here, besides the natural advantages of being fast, USA). The films separated two compartments (cis/trans) only
cheap, label-free, and high-throughput, provides the possibility connected through the Teflon film aperture with one Ag/AgCl
of being developed into a portable diagnostic device with electrode in each compartment. Apertures were pretreated with 1 μL
of 2% (v/v) hexadecane in hexane on both sides using a standard
medical and commercial potential. pipet, and the chamber was mounted in the recording setup. DPhPC
bilayers were formed by folding as described previously.76,77 Briefly,
METHODS electrolyte solution was added to both sides taking care that the level
Synthesis of C-Terminal α-Synuclein Peptides and Their stayed below the aperture, and lipids (10 mg/mL in pentane) were
PTMs. The majority of peptides used in nanopore experiments were added onto the electrolyte surface in both compartments. After the
produced as described in previous works.44−46 A few peptides pentane evaporated, the electrolyte level was raised above the

1511 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

aperture, and a lipid bilayer was formed. The quality of the lipid matrix for wt and oM127 peptides, the mixture
bilayer was monitored through its capacitance, and its stability was experiments of wt and pS129 peptide, the red blood
verified through the application of 150 mV over the course of at least
5 min. After peptide addition, the cis chamber was carefully mixed by
cells lysate recording and the correlation between signal
pipetting up and down. Currents were sampled at 200 kHz and low- frequency and analyte concentration (Figures S1−S8)
pass filtered at 100 kHz with an Axopatch 200B (Molecular Devices, (PDF)
LLC., San Jose, CA, USA).
Peptides (lyophilized powder) were prediluted in 10 mM Tris and
1.0 mM EDTA solution (pH 7.4) to a stock concentration of 500 μM
AUTHOR INFORMATION
and added to the cis side of the chamber in 1.0 M KCl solution Corresponding Authors
buffered with 10 mM Tris and 1.0 mM EDTA (pH 7.4) to the final Chan Cao − Institute of Bioengineering, School of Life Sciences,
concentration indicated in the figure caption. All experiments shown Ecole Polytechnique Fédérale de Lausanne, EPFL, Lausanne
here were repeated with at least 4 different pores. 1015, Switzerland; Department of Inorganic and Analytical
Signals Processing and Classifications Using Deep Learn-
ing. The signal processing was done in the same way as previously Chemistry, Chemistry and Biochemistry, University of
reported.53 The open pore current distribution is measured by fitting Geneva, 1211 Geneva, Switzerland; [Link]/0000-
a Gaussian function on the peak distribution of the current with the 0003-2592-0690; Email: [Link]@[Link]
highest mean current. The events are extracted using a current Hilal A. Lashuel − Laboratory of Molecular and Chemical
threshold at 3σ from the open pore current distribution. The relative Biology of Neurodegeneration, Brain Mind Institute, School of
current percentage (I/I0) is computed from the mean open pore Life Sciences, Ecole Polytechnique Fédérale de Lausanne,
current (I0) and the mean residual current (I). The dwell time, EPFL, Lausanne 1015, Switzerland; [Link]/0000-
average relative current, relative standard deviation of the current 0001-7682-8320; Email: [Link]@[Link]
rel = (σ0 is the value of the open pore current standard deviation Matteo Dal Peraro − Institute of Bioengineering, School of Life
0
and σ is the residual current standard deviation) and local extrema are Sciences, Ecole Polytechnique Fédérale de Lausanne, EPFL,
computed. The events are selected on the basis of the dwell time (0.2 Lausanne 1015, Switzerland; Email: [Link]@
to 100.0 ms) and the average relative current (0 to 40%) discarding [Link]
the events that are too short, too long, or that do not block the
current sufficiently. Authors
The machine learning pipeline is composed of two steps. The first Pedro Magalhães − Laboratory of Molecular and Chemical
one is the classification of every event, and the second is the Biology of Neurodegeneration, Brain Mind Institute, School of
assessment of the quality of the prediction of the classifier. The neural
network architecture for both the classification and the assessment is a Life Sciences, Ecole Polytechnique Fédérale de Lausanne,
long, short-term memory (LSTM) neural network followed by a EPFL, Lausanne 1015, Switzerland
multilayer perceptron (MLP) using the position in time and relative Lucien F. Krapp − Institute of Bioengineering, School of Life
current of the local extrema for each event as input features. The Sciences, Ecole Polytechnique Fédérale de Lausanne, EPFL,
features are rescaled by a fixed factor to decrease the training time. Lausanne 1015, Switzerland
The classifier is composed of an LSTM with state size 128 without Juan F. Bada Juarez − Institute of Bioengineering, School of
any activation function followed by 6 fully connected hidden layers of Life Sciences, Ecole Polytechnique Fédérale de Lausanne,
size 256 with rectified linear unit (ReLU) as activation functions and EPFL, Lausanne 1015, Switzerland
finally an output layer of size 8 with softmax activation function. The Simon Finn Mayer − Institute of Bioengineering, School of Life
assessment is done with a scaled down version of the classifier with a
LSTM with a state of size 32, 3 fully connected hidden layers of size Sciences, Ecole Polytechnique Fédérale de Lausanne, EPFL,
64 with hyperbolic tangent activation functions, and an output layer of Lausanne 1015, Switzerland
size 1 with sigmoid activation function. The neural networks for the Verena Rukes − Institute of Bioengineering, School of Life
classification and assessment are trained together using a 3-parts loss Sciences, Ecole Polytechnique Fédérale de Lausanne, EPFL,
function. The first part is the full classification cross-entropy loss of Lausanne 1015, Switzerland; [Link]/0009-0008-
the predictions from the classifier and the peptides label. The second 2704-179X
part is the assessment of the cross-entropy loss between the predicted Anass Chiki − Laboratory of Molecular and Chemical Biology
and actual prediction validity from the classifier. The third part is the of Neurodegeneration, Brain Mind Institute, School of Life
reinforcement classification loss, which is the full classification cross-
Sciences, Ecole Polytechnique Fédérale de Lausanne, EPFL,
entropy loss scaled by the assessment prediction.
The deep learning classifier is benchmarked using cross-validation, Lausanne 1015, Switzerland
where the first (in time) 75% of the events of a measurement is used Complete contact information is available at:
for training and the last 25% for validation. In the case of the mixture [Link]
experiments, we use the holdout method as they are not present in the
training data set (or test/validation data sets). The data set composed Author Contributions
of nY125, nY125nY133nY136, nY125pS129, nY136, pS129, pY125,
pY125pS129, and wt contains 519,000 events. The data set for C.C., M.D.P., and H.L. conceived the project and designed the
oM127 and wt contains 83,000 events. The mixture experiments study. C.C. performed the experiments on Orbit Mini,
(WT:pY125) data set contains 88,999 events. supervised all single-channel experiments and did data analysis.
P.M. characterized all C-terminal α-synuclein peptides and
ASSOCIATED CONTENT prepared all clinical samples. L.F.K. performed signal
*
sı Supporting Information
processing and developed the deep learning pipeline. J.F.B.J.
conducted CtsD digestion and nanopore experiments. S.F.M.
The Supporting Information is available free of charge at
and V.R. carried out experiments on the asymmetric buffer.
[Link]
A.C. synthesized C-terminal α-synuclein peptides with differ-
I−V curve for the aerolysin K238A mutant in urea ent PTMs forms. C.C., L.F.K., P.M., S.F.M., H.L., and M.D.P.
condition, the ionic current signals of the different interpreted the data. C.C., H.L., and M.D.P. wrote the
PTMs on the peptide, the contour plot and confusion manuscript with input from all authors.
1512 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Author Contributions (13) Mnatsakanyan, R.; Shema, G.; Basik, M.; Batist, G.; Borchers,
# C. H.; Sickmann, A.; Zahedi, R. P. Detecting Post-Translational
P.M. and L.F.K. equally contributed.
Modification Signatures as Potential Biomarkers in Clinical Mass
Notes Spectrometry. Expert Rev. Proteomics 2018, 15, 515−535.
The authors declare the following competing financial (14) Dubois, B.; Feldman, H. H.; Jacova, C.; Hampel, H.;
interest(s): EPFL has filed a patent application Molinuevo, J. L.; Blennow, K.; DeKosky, S. T.; Gauthier, S.; Selkoe,
EP23182675.1. on technologies described herein. D.; Bateman, R.; Cappa, S.; Crutch, S.; Engelborghs, S.; Frisoni, G. B.;
A preprint version of this article is available on the bioRxiv Fox, N. C.; Galasko, D.; Habert, M.-O.; Jicha, G. A.; Nordberg, A.;
repository.78 Pasquier, F.; et al. Advancing Research Diagnostic Criteria for
Alzheimer’s Disease: The IWG-2 Criteria. Lancet Neurol. 2014, 13,
614−629.
ACKNOWLEDGMENTS (15) Janelidze, S.; Mattsson, N.; Palmqvist, S.; Smith, R.; Beach, T.
This research was supported by the Swiss National Science G.; Serrano, G. E.; Chai, X.; Proctor, N. K.; Eichenlaub, U.;
Foundation (to M.D.P. and PR00P3_193090 to C.C.), the Zetterberg, H.; Blennow, K.; Reiman, E. M.; Stomrud, E.; Dage, J. L.;
European Union’s Horizon 2020 Research and Innovation Hansson, O. Plasma P-Tau181 in Alzheimer’s Disease: Relationship
Program under the Marie Skłodowska-Curie Grant Agreement to Other Biomarkers, Differential Diagnosis, Neuropathology and
No. 665667, Peter and Traudl Engelhorn Foundation and Longitudinal Progression to Alzheimer’s Dementia. Nat. Med. 2020,
Synapsis Foundation (2019-CDA02 to C.C.). The authors also 26, 379−386.
(16) Kasianowicz, J. J.; Brandin, E.; Branton, D.; Deamer, D. W.
thank M. J. Marcaida for support with aerolysin production
Characterization of Individual Polynucleotide Molecules Using a
and R. Kolla for the discussion on the peptides’ character- Membrane Channel. Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 13770−
ization. 13773.
(17) Venkatesan, B. M.; Bashir, R. Nanopore Sensors for Nucleic
REFERENCES Acid Analysis. Nat. Nanotechnol. 2011, 6, 615−624.
(1) Marth, J. D. A Unified Vision of the Building Blocks of Life. Nat. (18) Ying, Y.-L.; Cao, C.; Long, Y.-T. Single Molecule Analysis by
Cell Biol. 2008, 10, 1015−1015. Biological Nanopore Sensors. Analyst 2014, 139, 3826−3835.
(2) Wang, Y.-C.; Peterson, S. E.; Loring, J. F. Protein Post- (19) Cao, C.; Long, Y.-T. Biological Nanopores: Confined Spaces
Translational Modifications and Regulation of Pluripotency in for Electrochemical Single-Molecule Analysis. Acc. Chem. Res. 2018,
Human Stem Cells. Cell Res. 2014, 24, 143−160. 51, 331−341.
(3) Walsh, G.; Jefferis, R. Post-Translational Modifications in the (20) Jain, M.; Koren, S.; Miga, K. H.; Quick, J.; Rand, A. C.; Sasani,
Context of Therapeutic Proteins. Nat. Biotechnol. 2006, 24, 1241− T. A.; Tyson, J. R.; Beggs, A. D.; Dilthey, A. T.; Fiddes, I. T.; Malla,
1252. S.; Marriott, H.; Nieto, T.; O’Grady, J.; Olsen, H. E.; Pedersen, B. S.;
(4) Pagel, O.; Loroch, S.; Sickmann, A.; Zahedi, R. P. Current Rhie, A.; Richardson, H.; Quinlan, A. R.; Snutch, T. P.; et al.
Strategies and Findings in Clinically Relevant Post-Translational Nanopore Sequencing and Assembly of a Human Genome with Ultra-
Modification-Specific Proteomics. Expert Rev. Proteomics 2015, 12, Long Reads. Nat. Biotechnol. 2018, 36, 338−345.
235−253. (21) Laszlo, A. H; Derrington, I. M; Ross, B. C; Brinkerhoff, H.;
(5) Restrepo-Pérez, L.; Joo, C.; Dekker, C. Paving the Way to Single- Adey, A.; Nova, I. C; Craig, J. M; Langford, K. W; Samson, J. M.;
Molecule Protein Sequencing. Nat. Nanotechnol. 2018, 13, 786−796. Daza, R.; Doering, K.; Shendure, J.; Gundlach, J. H Decoding Long
(6) Petrera, A.; von Toerne, C.; Behler, J.; Huth, C.; Thorand, B.; Nanopore Sequencing Reads of Natural DNA. Nat. Biotechnol. 2014,
Hilgendorff, A.; Hauck, S. M. Multiplatform Approach for Plasma 32, 829−833.
Proteomics: Complementarity of Olink Proximity Extension Assay (22) Houghtaling, J.; List, J.; Mayer, M. Nanopore-Based, Rapid
Technology to Mass Spectrometry-Based Protein Profiling. J. Characterization of Individual Amyloid Particles in Solution:
Proteome Res. 2021, 20, 751−762. Concepts, Challenges, and Prospects. Small 2018, 14, 1802412.
(7) Wu, D.; Milutinovic, M. D.; Walt, D. R. Single Molecule Array (23) Brinkerhoff, H.; Kang, A. S. W.; Liu, J.; Aksimentiev, A.;
(Simoa) Assay with Optimal Antibody Pairs for Cytokine Detection Dekker, C. Infinite Re-Reading of Single Proteins at Single-Amino-
in Human Serum Samples. Analyst 2015, 140, 6277−6282. Acid Resolution Using Nanopore Sequencing. bioRxiv, July 14, 2021.
(8) Banerjee, A.; Mastrangelo, C. H. Resurgence of Electron DOI: 10.1101/2021.07.13.452225.
Quantum Tunneling Sensors. Micro 2022, 2, 679−698. (24) Yan, S.; Zhang, J.; Wang, Y.; Guo, W.; Zhang, S.; Liu, Y.; Cao,
(9) Ohshiro, T.; Tsutsui, M.; Yokota, K.; Furuhashi, M.; Taniguchi, J.; Wang, Y.; Wang, L.; Ma, F.; Zhang, P.; Chen, H.-Y.; Huang, S.
M.; Kawai, T. Detection of Post-Translational Modifications in Single Single Molecule Ratcheting Motion of Peptides in a Mycobacterium
Peptides Using Electron Tunnelling Currents. Nat. Nanotechnol. Smegmatis Porin A (MspA) Nanopore. Nano Lett. 2021, 21, 6703−
2014, 9, 835−840. 6710.
(10) Lashuel, H. A.; Mahul-Mellier, A.-L.; Novello, S.; Hegde, R. N.; (25) Alfaro, J. A.; Bohländer, P.; Dai, M.; Filius, M.; Howard, C. J.;
Jasiqi, Y.; Altay, M. F.; Donzelli, S.; DeGuire, S. M.; Burai, R.; van Kooten, X. F.; Ohayon, S.; Pomorski, A.; Schmid, S.; Aksimentiev,
Magalhães, P.; Chiki, A.; Ricci, J.; Boussouf, M.; Sadek, A.; Stoops, E.; A.; Anslyn, E. V.; Bedran, G.; Cao, C.; Chinappi, M.; Coyaud, E.;
Iseli, C.; Guex, N. Revisiting the Specificity and Ability of Phospho- Dekker, C.; Dittmar, G.; Drachman, N.; Eelkema, R.; Goodlett, D.;
S129 Antibodies to Capture Alpha-Synuclein Biochemical and et al. The Emerging Landscape of Single-Molecule Protein
Pathological Diversity. NPJ Parkinsons Dis. 2022, 8, 136. Sequencing Technologies. Nat. Methods 2021, 18, 604−617.
(11) Chiki, A.; DeGuire, S. M.; Ruggeri, F. S.; Sanfelice, D.; (26) Rosen, C. B.; Rodriguez-Larrea, D.; Bayley, H. Single-Molecule
Ansaloni, A.; Wang, Z.-M.; Cendrowska, U.; Burai, R.; Vieweg, S.; Site-Specific Detection of Protein Phosphorylation with a Nanopore.
Pastore, A.; Dietler, G.; Lashuel, H. A. Mutant Exon1 Huntingtin Nat. Biotechnol. 2014, 32, 179−181.
Aggregation Is Regulated by T3 Phosphorylation-Induced Structural (27) Ying, Y.-L.; Yang, J.; Meng, F.-N.; Li, S.; Li, M.-Y.; Long, Y.-T.
Changes and Crosstalk between T3 Phosphorylation and Acetylation A Nanopore Phosphorylation Sensor for Single Oligonucleotides and
at K6. Angew. Chem., Int. Ed. 2017, 56, 5202−5207. Peptides. Research 2019, 2019, 1−8.
(12) Chiki, A.; Zhang, Z.; Rajasekhar, K.; Abriata, L. A.; Rostami, I.; (28) Li, S.; Wu, X.; Li, M.; Liu, S.; Ying, Y.; Long, Y. T232K/K238Q
Krapp, L. F.; Boudeffa, D.; Dal Peraro, M.; Lashuel, H. A. Aerolysin Nanopore for Mapping Adjacent Phosphorylation Sites of a
Investigating Crosstalk Among PTMs Provides Novel Insight Into Single Tau Peptide. Small Methods 2020, 4, 2000014.
the Structural Basis Underlying the Differential Effects of Nt17 PTMs (29) Ensslen, T.; Sarthak, K.; Aksimentiev, A.; Behrends, J. C.
on Mutant Httex1 Aggregation. Front. Mol. Biosci. 2021, 8, 686086. Resolving Isomeric Posttranslational Modifications Using a Biological

1513 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

Nanopore as a Sensor of Molecular Shape. J. Am. Chem. Soc. 2022, (45) Meuvis, J.; Gerard, M.; Desender, L.; Baekelandt, V.;
144, 16060−16068. Engelborghs, Y. The Conformation and the Aggregation Kinetics of
(30) Zhang, L.; Gardner, M. L.; Jayasinghe, L.; Jordan, M.; Aldana, α-Synuclein Depend on the Proline Residues in Its C-Terminal
J.; Burns, N.; Freitas, M. A.; Guo, P. Detection of Single Peptide with Region. Biochemistry 2010, 49, 9345−9352.
Only One Amino Acid Modification via Electronic Fingerprinting (46) Hejjaoui, M.; Butterfield, S.; Fauvet, B.; Vercruysse, F.; Cui, J.;
Using Reengineered Durable Channel of Phi29 DNA Packaging Dikiy, I.; Prudent, M.; Olschewski, D.; Zhang, Y.; Eliezer, D.; Lashuel,
Motor. Biomaterials 2021, 276, 121022. H. A. Elucidating the Role of C-Terminal Post-Translational
(31) Restrepo-Pérez, L.; Wong, C. H.; Maglia, G.; Dekker, C.; Joo, Modifications Using Protein Semisynthesis Strategies: α-Synuclein
C. Label-Free Detection of Post-Translational Modifications with a Phosphorylation at Tyrosine 125. J. Am. Chem. Soc. 2012, 134, 5196−
Nanopore. Nano Lett. 2019, 19, 7957−7964. 5210.
(32) Fahie, M. A.; Chen, M. Electrostatic Interactions between (47) Fauvet, B.; Butterfield, S. M.; Fuks, J.; Brik, A.; Lashuel, H. A.
OmpG Nanopore and Analyte Protein Surface Can Distinguish One-Pot Total Chemical Synthesis of Human [Small Alpha]-
between Glycosylated Isoforms. J. Phys. Chem. B 2015, 119, 10198− Synuclein. Chem. Commun. 2013, 49, 9254−9256.
10206. (48) Haj-Yahya, M.; Fauvet, B.; Herman-Bachinsky, Y.; Hejjaoui,
(33) Wloka, C.; Van Meervelt, V.; van Gelder, D.; Danda, N.; Jager, M.; Bavikar, S. N.; Karthikeyan, S. V.; Ciechanover, A.; Lashuel, H. A.;
N.; Williams, C. P.; Maglia, G. Label-Free and Real-Time Detection of Brik, A. Synthetic Polyubiquitinated α-Synuclein Reveals Important
Protein Ubiquitination with a Biological Nanopore. ACS Nano 2017, Insights into the Roles of the Ubiquitin Chain in Regulating Its
11, 4387−4394. Pathophysiology. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 17726−
(34) Degiacomi, M. T.; Iacovache, I.; Pernot, L.; Chami, M.; 17731.
Kudryashev, M.; Stahlberg, H.; van der Goot, F. G.; Dal Peraro, M. (49) Bertoncini, C. W.; Jung, Y.-S.; Fernandez, C. O.; Hoyer, W.;
Molecular Assembly of the Aerolysin Pore Reveals a Swirling Griesinger, C.; Jovin, T. M.; Zweckstetter, M. Release of Long-Range
Membrane-Insertion Mechanism. Nat. Chem. Biol. 2013, 9, 623−629. Tertiary Interactions Potentiates Aggregation of Natively Unstruc-
(35) Cirauqui, N.; Abriata, L. A.; van der Goot, F. G.; Dal Peraro, M. tured α-Synuclein. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 1430−
Structural, Physicochemical and Dynamic Features Conserved within 1435.
the Aerolysin Pore-Forming Toxin Family. Sci. Rep. 2017, 7, 13932. (50) Kawahara, K.; Tanford, C. Viscosity and Density of Aqueous
(36) Cao, C.; Ying, Y.-L.; Hu, Z.-L.; Liao, D.-F.; Tian, H.; Long, Y.- Solutions of Urea and Guanidine Hydrochloride. J. Biol. Chem. 1966,
T. Discrimination of Oligonucleotides of Different Lengths with a 241, 3228−3232.
Wild-Type Aerolysin Nanopore. Nat. Nanotechnol. 2016, 11, 713− (51) Fologea, D.; Uplinger, J.; Thomas, B.; McNabb, D. S.; Li, J.
718. Slowing DNA Translocation in a Solid-State Nanopore. Nano Lett.
(37) Ouldali, H.; Sarthak, K.; Ensslen, T.; Piguet, F.; Manivet, P.; 2005, 5, 1734−1737.
Pelta, J.; Behrends, J. C.; Aksimentiev, A.; Oukhaled, A. Electrical (52) Yu, L.; Kang, X.; Li, F.; Mehrafrooz, B.; Makhamreh, A.; Fallahi,
Recognition of the Twenty Proteinogenic Amino Acids Using an A.; Foster, J. C.; Aksimentiev, A.; Chen, M.; Wanunu, M.
Aerolysin Nanopore. Nat. Biotechnol. 2020, 38, 176−181. Unidirectional Single-File Transport of Full-Length Proteins through
(38) Oliveira, L. M. A.; Gasser, T.; Edwards, R.; Zweckstetter, M.; a Nanopore. Nat. Biotechnol. 2023, 41, 1130.
Melki, R.; Stefanis, L.; Lashuel, H. A.; Sulzer, D.; Vekrellis, K.; (53) Cao, C.; Krapp, L. F.; Al Ouahabi, A.; König, N. F.; Cirauqui,
Halliday, G. M.; Tomlinson, J. J.; Schlossmacher, M.; Jensen, P. H.; N.; Radenovic, A.; Lutz, J.-F.; Peraro, M. D. Aerolysin Nanopores
Schulze-Hentrich, J.; Riess, O.; Hirst, W. D.; El-Agnaf, O.; Decode Digital Information Stored in Tailored Macromolecular
Mollenhauer, B.; Lansbury, P.; Outeiro, T. F. Alpha-Synuclein Analytes. Sci. Adv. 2020, 6, No. eabc2661.
Research: Defining Strategic Moves in the Battle against Parkinson’s (54) Sze, J. Y. Y.; Ivanov, A. P.; Cass, A. E. G.; Edel, J. B. Single
Disease. NPJ Parkinsons Dis. 2021, 7, 65. Molecule Multiplexed Nanopore Protein Screening in Human Serum
(39) Eusebi, P.; Giannandrea, D.; Biscetti, L.; Abraha, I.; Chiasserini, Using Aptamer Modified DNA Carriers. Nat. Commun. 2017, 8, 1552.
D.; Orso, M.; Calabresi, P.; Parnetti, L. Diagnostic Utility of (55) Sauvage, V.; Boizeau, L.; Candotti, D.; Vandenbogaert, M.;
Cerebrospinal Fluid α-Synuclein in Parkinson’s Disease: A Systematic Servant-Delmas, A.; Caro, V.; Laperche, S. Early MinIONTM
Review and Meta-Analysis: CSF α-Synuclein for PD Diagnosis. Mov. Nanopore Single-Molecule Sequencing Technology Enables the
Disord. 2017, 32, 1389−1400. Characterization of Hepatitis B Virus Genetic Complexity in Clinical
(40) Majbour, N. K.; Vaikath, N. N.; van Dijk, K. D.; Ardah, M. T.; Samples. PLoS One 2018, 13, No. e0194366.
Varghese, S.; Vesterager, L. B.; Montezinho, L. P.; Poole, S.; Safieh- (56) Wang, Y.; Zheng, D.; Tan, Q.; Wang, M. X.; Gu, L.-Q.
Garabedian, B.; Tokuda, T.; Teunissen, C. E.; Berendse, H. W.; van Nanopore-Based Detection of Circulating microRNAs in Lung
de Berg, W. D. J.; El-Agnaf, O. M. A. Oligomeric and Phosphorylated Cancer Patients. Nat. Nanotechnol 2011, 6, 668−674.
Alpha-Synuclein as Potential CSF Biomarkers for Parkinson’s Disease. (57) Wang, S.; Haque, F.; Rychahou, P. G.; Evers, B. M.; Guo, P.
Mol. Neurodegener. 2016, 11, 7. Engineered Nanopore of Phi29 DNA-Packaging Motor for Real-Time
(41) Conibear, A. C. Deciphering Protein Post-Translational Detection of Single Colon Cancer Specific Antibody in Serum. ACS
Modifications Using Chemical Biology Tools. Nat. Rev. Chem. Nano 2013, 7, 9814−9822.
2020, 4, 674−695. (58) Kukwikila, M.; Howorka, S. Nanopore-Based Electrical and
(42) Nakayasu, E. S.; Gritsenko, M.; Piehowski, P. D.; Gao, Y.; Label-Free Sensing of Enzyme Activity in Blood Serum. Anal. Chem.
Orton, D. J.; Schepmoes, A. A.; Fillmore, T. L.; Frohnert, B. I.; 2015, 87, 9149−9154.
Rewers, M.; Krischer, J. P.; Ansong, C.; Suchy-Dicey, A. M.; Evans- (59) Yu, J.; Cao, C.; Long, Y.-T. Selective and Sensitive Detection of
Molina, C.; Qian, W.-J.; Webb-Robertson, B.-J. M.; Metz, T. O. Methylcytosine by Aerolysin Nanopore under Serum Condition. Anal.
Tutorial: Best Practices and Considerations for Mass-Spectrometry- Chem. 2017, 89, 11685−11689.
Based Protein Biomarker Discovery and Validation. Nat. Protoc. 2021, (60) Fahie, M. A.; Yang, B.; Mullis, M.; Holden, M. A.; Chen, M.
16, 3737−3760. Selective Detection of Protein Homologues in Serum Using an
(43) Cao, C.; Cirauqui, N.; Marcaida, M. J.; Buglakova, E.; OmpG Nanopore. Anal. Chem. 2015, 87, 11143−11149.
Duperrex, A.; Radenovic, A.; Dal Peraro, M. Single-Molecule Sensing (61) Chuah, K.; Wu, Y.; Vivekchand, S. R. C.; Gaus, K.; Reece, P. J.;
of Peptides and Nucleic Acids by Engineered Aerolysin Nanopores. Micolich, A. P.; Gooding, J. J. Nanopore Blockade Sensors for
Nat. Commun. 2019, 10, 4918. Ultrasensitive Detection of Proteins in Complex Biological Samples.
(44) Schmid, A. W.; Fauvet, B.; Moniatte, M.; Lashuel, H. A. Alpha- Nat. Commun. 2019, 10, 2109.
Synuclein Post-Translational Modifications as Potential Biomarkers (62) Galenkamp, N. S.; Soskine, M.; Hermans, J.; Wloka, C.; Maglia,
for Parkinson Disease and Other Synucleinopathies. Mol. Cell. G. Direct Electrical Quantification of Glucose and Asparagine from
Proteomics 2013, 12, 3543−3558. Bodily Fluids Using Nanopores. Nat. Commun. 2018, 9, 4085.

1514 [Link]
ACS Nano 2024, 18, 1504−1515
ACS Nano [Link] Article

(63) Cao, C.; Yu, J.; Wang, Y.-Q.; Ying, Y.-L.; Long, Y.-T. Driven
Translocation of Polynucleotides Through an Aerolysin Nanopore.
Anal. Chem. 2016, 88, 5046−5049.
(64) Barbour, R.; Kling, K.; Anderson, J. P.; Banducci, K.; Cole, T.;
Diep, L.; Fox, M.; Goldstein, J. M.; Soriano, F.; Seubert, P.; Chilcote,
T. J. Red Blood Cells Are the Major Source of Alpha-Synuclein in
Blood. Neurodegener. Dis. 2008, 5, 55−59.
(65) Klatt, S.; Roberts, A.; Lothian, A.; Masters, C. L.; Cappai, R.;
Fowler, C.; Roberts, B. R. Optimizing Red Blood Cell Protein
Extraction for Biomarker Quantitation with Mass Spectrometry. Anal.
Bioanal. Chem. 2020, 412, 1879−1892.
(66) Wanunu, M.; Morrison, W.; Rabin, Y.; Grosberg, A. Y.; Meller,
A. Electrostatic Focusing of Unlabelled DNA into Nanoscale Pores
Using a Salt Gradient. Nat. Nanotechnol. 2010, 5, 160−165.
(67) Sorrentino, Z. A.; Giasson, B. I. The Emerging Role of α-
Synuclein Truncation in Aggregation and Disease. J. Biol. Chem. 2020,
295, 10224−10244.
(68) McGlinchey, R. P.; Lee, J. C. Cysteine Cathepsins Are Essential
in Lysosomal Degradation of α-Synuclein. Proc. Natl. Acad. Sci. U. S.
A. 2015, 112, 9322−9327.
(69) Hu, Z.-L.; Li, M.-Y.; Liu, S.-C.; Ying, Y.-L.; Long, Y.-T. A
Lithium-Ion-Active Aerolysin Nanopore for Effectively Trapping
Long Single-Stranded DNA. Chem. Sci. 2019, 10, 354−358.
(70) Kumar, S. T.; Jagannath, S.; Francois, C.; Vanderstichele, H.;
Stoops, E.; Lashuel, H. A. How Specific Are the Conformation-
Specific α-Synuclein Antibodies? Characterization and Validation of
16 α-Synuclein Conformation-Specific Antibodies Using Well-
Characterized Preparations of α-Synuclein Monomers, Fibrils and
Oligomers with Distinct Structures and Morphology. Neurobiol. Dis.
2020, 146, 105086.
(71) McGlinchey, R. P.; Lacy, S. M.; Huffer, K. E.; Tayebi, N.;
Sidransky, E.; Lee, J. C. C-Terminal α-Synuclein Truncations Are
Linked to Cysteine Cathepsin Activity in Parkinson’s Disease. J. Biol.
Chem. 2019, 294, 9973−9984.
(72) Afshar Bakshloo, M.; Kasianowicz, J. J.; Pastoriza-Gallego, M.;
Mathé, J.; Daniel, R.; Piguet, F.; Oukhaled, A. Nanopore-Based
Protein Identification. J. Am. Chem. Soc. 2022, 144, 2716−2725.
(73) Lucas, F. L. R.; Versloot, R. C. A.; Yakovlieva, L.; Walvoort, M.
T. C.; Maglia, G. Protein Identification by Nanopore Peptide
Profiling. Nat. Commun. 2021, 12, 5795.
(74) Borrebaeck, C. A. K. Precision Diagnostics: Moving towards
Protein Biomarker Signatures of Clinical Utility in Cancer. Nat. Rev.
Cancer 2017, 17, 199−204.
(75) Vicente Miranda, H.; Cássio, R.; Correia-Guedes, L.; Gomes,
M. A.; Chegão, A.; Miranda, E.; Soares, T.; Coelho, M.; Rosa, M. M.;
Ferreira, J. J.; Outeiro, T. F. Posttranslational Modifications of Blood-
Derived Alpha-Synuclein as Biochemical Markers for Parkinson’s
Disease. Sci. Rep. 2017, 7, 13713.
(76) Montal, M.; Mueller, P. Formation of Bimolecular Membranes
from Lipid Monolayers and a Study of Their Electrical Properties.
Proc. Natl. Acad. Sci. U. S. A. 1972, 69, 3561−3566.
(77) Mayer, M.; Kriebel, J. K.; Tosteson, M. T.; Whitesides, G. M.
Microfabricated Teflon Membranes for Low-Noise Recordings of Ion
Channels in Planar Lipid Bilayers. Biophys. J. 2003, 85, 2684−2695.
(78) Cao, C.; Magalhães, P.; Krapp, L. F.; Bada Juarez, J. F.; Mayer,
S.; Rukes, V.; Chiki, A.; Lashuel, H. A.; Dal Peraro, M. Deep
Learning-Assisted Single-Molecule Detection of Protein Post-Trans-
lational Modifications with a Biological Nanopore. bioRxiv, September
8, 2023. DOI: 10.1101/2023.09.08.555904 (accessed on November
13, 2023).

1515 [Link]
ACS Nano 2024, 18, 1504−1515