In the nearly
15 years since
electron transfer dissociation (ETD) was introduced,
1
bioanalytical mass spectrometry (MS) and MS-driven proteomics
have continued to develop as the premier tools for characterizing
proteome composition, structure, and function.
2
The new era of proteomics has focused on throughput, sampling depth,
and reproducibility, often relying on traditional collision-based
dissociation because of its speed and ease of implementation.
3−5
That said, the development of alternative dissociation methods remains
an active field in proteome research, offering complementary analyses
to collisional activation and valuable insights for analytes that
can be intractable with traditional methods.
6
Because of its applicability to a wide range of biomolecules and
its compatibility with a variety of instrument platforms, ETD is among
the most prominent of these alternative approaches. Concentrating
on developments in the last 5 years, this review examines the role
ETD plays in modern proteomic experiments, including its use in characterizing
intact proteins, post-translational modifications (PTMs), and structural
aspects of the proteome (Figure 1
). We focus on ETD instrumentation and methodological
developments and how they are being employed to answer biologically
driven problems, and we also include brief discussions about other
ion–ion reaction techniques. Finally, we offer our perspective
on the future of ETD development and comment on how ETD will continue
to be a major contributor to proteome research in years to come.
Figure 1
Diverse
roles of electron transfer dissociation. ETD is a valuable
tandem MS method for characterizing many aspects of the proteome and
is used in bottom-up, middle-down, and top-down proteomic methods.
Structural Proteomics image is reprinted in part by permission from
Macmillan Publishers Ltd.: NATURE, Liu, F.; Rijkers, D. T. S.; Post,
H.; Heck, A. J. R. Nat. Methods
2015, 12, 1179–1184 (ref (255)). Copyright 2015. Deep
Proteome Sequencing image is reproduced in part from Proteomics Beyond
Trypsin, Tsiatsiani, L; Heck, A.J.R. FEBS J., Vol. 282, Issue 14, pp 2612–2626 (ref
(189)). Copyright 2015 Wiley.
Principles of ETD
ETD is an ion–ion
reaction involving multiply charged precursor
peptide (or protein) cations and singly charged radical reagent anions.
An electron from the reagent anion is transferred to the cation, resulting
in an odd-electron cation that undergoes free-radical-driven cleavage.
Similar to its ion-electron predecessor, electron capture dissociation
(ECD),
7
this cleavage dissociates N–Cα bonds along the peptide backbone to generate even electron
c-type and odd electron z•-type product ions that
provide sequence information on the peptide precursor.
8
Here we briefly cover basic principles of ETD to provide
context for recent developments. For in-depth discussions of radical-mediated
gas-phase chemistry and reaction pathways involved in ETD (and related
techniques), we direct readers to several excellent recent reviews.
9−11
Other reviews have given a more historical perspective on how the
introduction of ETD provided new dimensions in proteomic analyses
as well.
12,13
The success of ETD reactions, i.e.,
the ability to generate sequence
informative fragment ions, is largely dependent on (1) the reagent
anion used as an electron donor and (2) the precursor cation charge
density. Several small-molecule reagent anions have been explored
within the context of ETD reactions, leading to either peptide deprotonation
(proton transfer) or deposition of an electron from the reagent anion
onto the peptide (electron transfer), the latter of which promotes
peptide backbone cleavage.
14
Anthracene,
9,10-diphenyl-anthracene, azobenzene, and azulene have demonstrated
suitability for ETD reactions,
1,14−16
but fluoranthene is generally regarded as one of the more favorable
ETD reagents for generating c- and z•-type ions.
17−19
In addition to the chemical properties of the reagent anion,
precursor
cation charge density governs successful generation of product ions
in ETD reactions, with higher charge density being much more favorable.
20−22
The gas-phase structure of peptide precursor ions is directly related
to its charge density: precursors with higher charge density are more
linear and those with lower charge density are more compact. When
ETD reactions are conducted with lower charge density precursors,
peptide backbone cleavage can occur but c- and z•-type product ions are held together
by noncovalent interactions
present in the more compact structures. Through this process, called
nondissociative electron transfer dissociation (ETnoD),
23
the radical product ion complex appears in the
mass spectrum as an intact species (i.e., [M + nH](n−1)•), yielding no sequence
information. One way to combat this challenge involves chemical modification
of peptides to increase their charge density and, thus, improve fragmentation.
24,25
ETD has also been paired with collision-based fragmentation methods
within a given experiment, where ETD is employed for higher charge
(z ≥ 3), low m/z precursors and collisional dissociation is used for doubly protonated
and higher m/z ions.
26,27
These so-called decision-tree methods, where the fragmentation method
most suited for a given precursor is chosen in real-time, have proven
valuable for increasing identifications and improving confidence of
peptide spectral matches in a wide variety of analyses. Indeed, this
theme of complementary use of ETD and collision-based fragmentation
is present throughout this review and has been discussed previously.
8,28
Supplemental
Activation for ETD Reactions
Supplemental
activation is another approach to increase efficiency of ETD by introducing
extra energy into the reaction, and Figure 2
summarizes the most common supplemental
activation strategies. Drawing on several years of research in ECD
methodology,
29−34
the goal of these methods is to disrupt noncovalent interactions
that contribute to ETnoD (usually via collisions or photons) to promote
product ion generation, especially for low charge density precursors.
Supplemental activation schemes that utilize collisions are popular
and easily implemented, but they are generally limited to post-ETD
activation. Pre-ETD activation is generally ineffective because the
bath gas pressures in ion traps where ETD is usually performed are
high enough (∼3 mTorr) to promote collisional cooling of activated
precursors, which causes them to relax back into more compact structures.
Collisional activation during ETD reactions is a nonviable supplemental
activation approach because it prevents sufficient overlap of the
cation and anion clouds, which is required for the ion–ion
reaction to occur.
35
Several iterations
of post-ETD activation have been explored,
36−41
with the most widely adopted being gentle resonant excitation of
ETnoD products, a process termed ETcaD.
42
Most recently, the Heck group introduced EThcD, which involves broadband
activation of all ETD products with high-energy collisional dissociation
(HCD).
43
Opposed to ETcaD, which produces
mainly c/z-type product ions, EThcD generates both c/z- and b/y-type
product ions by activating both the ETnoD products and remaining unreacted
precursor ions. EThcD produces series of complementary fragments that
increase peptide backbone coverage (i.e., the number of inter-residue
bonds explained by observed fragments) for improved confidence in
peptide identifications, and it boosts the number of peptides that
can be identified in LC–MS/MS analyses in complex mixtures
compared to ETD and ETcaD.
Figure 2
Supplemental activation strategies for ETD.
ETD spectra from low-charge
density precursors can have modest product ion generation (top left),
but supplemental activation strategies can increase precursor-to-product
ion conversion. In ETcaD (green), ETnoD products are resonantly excited
to disrupt noncovalent interactions and release sequence-informative
c/z-type product ions. EThcD (blue) uses broadband beam-type collisional
activation of all ETD reaction products to generate b/y- and c/z-type
fragments. Both ETcaD and EThcD perform supplemental activation steps
after the ETD reaction is finished. AI-ETD (red) leverages infrared
photoactivation simultaneous to the ETD reaction to promote formation
of mostly c/z•-type fragments. Each strategy significantly
improves the product ion yield of ETD reactions. Several hybrid MS
systems have ETD capabilities and various supplemental activation
options (bottom left, supplemental activation options indicated with
colored circles). Note, implementations of some methods on the FTICR
and TOF systems may differ slightly from the originally described
technique, e.g., AI-ETD on the quadrupole-FTICR is only available
for postreaction photoactivation.
Two challenges of post-ETD supplemental activation include
increased
levels of hydrogen migration between ETD product ions and additional
time needed to perform the activation, which adds to the overhead
time required per MS/MS scan. Hydrogen migrations occur in the radical
ETnoD complex, where even electron c-type ions and odd electron z•-type ions are held
together for several milliseconds
or more by noncovalent interactions. While in this complex, radical
z•-type ions can abstract a hydrogen atom from c-type
products, resulting in a population of odd-electron c+•-type and even electron z+-type
ions.
42,44−46
Mixtures of c/z•-type and c+•/z+-type product ions, which are present
in ETcaD and EThcD, can complicate spectra and challenge both manual
and automated spectral interpretation. The additional scan times required
for ETcaD and EThcD differs depending on instrument configuration,
but ETcaD often affects scan times more dramatically (tens of milliseconds),
while EThcD can add ∼5–15 ms to each MS/MS scan.
The combination of these challenges associated with post-ETD collision-based
supplemental activation led to investigations into photoactivation
to improve ETD efficiency. Photoactivation provides greater flexibility
for this purpose because it cannot only be used for pre- or postreaction
activation, but more importantly, it can also be used to activate
ions concurrent to the ion–ion reaction. Pre-ETD photoactivation
is rarely used because collisional cooling in the ion trap still occurs,
but ideal ion cloud overlap for ETD reactions can be readily achieved
while simultaneously irradiating the trapping volume of the reaction
cell, meaning photoactivation can happen concomitant with ETD. Activated
ion ETD (AI-ETD) uses infrared photoactivation (10.6 μm) concurrent
with ETD reactions to disrupt noncovalent interactions and minimize
ETnoD, significantly increasing sequence-informative product ion generation.
47,48
The concurrent nature of activation in AI-ETD requires no additional
time to complete and also minimizes the hydrogen migration seen in
ETcaD and EThcD.
47
These benefits translate
to increased peptide identifications in LC–MS/MS experiments
compared to ETD, ETcaD, and EThcD.
47−49
In general, AI-ETD produces
mainly ETD-specific c/z•-type product ions, although
increases in y-type ions (which are generated in standard ETD reactions
as well
50
) are also observed. Furthermore,
AI-ETD has been combined with a short (∼4 ms) postactivation
using infrared multiphoton activation (IRMPD) to generate spectra
rich in both c/z•- and b/y-type fragments, much
like EThcD.
48
This strategy, termed AI-ETD+,
utilizes short settling times built into existing scan sequences to
perform its postreaction IRMPD activation following standard AI-ETD,
meaning it also adds no additional time to the scan sequence but can
further improve peptide backbone coverage and identifications.
Other wavelengths have been explored for ETD supplemental photoactivation
to increase product ion yield, too, most notably 193 nm photons used
in ultraviolet photodissociation (UVPD). This hybrid reaction scheme
is called ETUVPD and can be used for both broadband activation of
all ETD products and specific activation of charge reduced species,
although activation typically occurs postreaction.
51
The 193 nm photons used in ETUVPD are significantly more
energetic than the infrared photons used in AI-ETD (6.4 vs 0.12 eV),
resulting in UVPD-driven product ion generation in addition to ETD
products. Observed fragment ion types included a/x-, b/y-, and c/z-type
products, with product ion distributions showing intermediate percentages
of fragment types between those seen in ETD and UVPD. Overall, ETUVPD
can provide a boost in both product ion generation and sequence coverage
over ETD alone but has not been widely implemented.
ETD-Equipped Instrumentation
One
of the main drivers for the prevalence of ETD has been its
amenability to numerous instrumentation platforms. The development
of ETD as an ion–ion reaction that could be performed in rf
trapping devices was the technological advance that brought the radically
driven fragmentation of ECD to a wider audience. Since its introduction
on a quadrupole linear ion trap (QLT), ETD has been incorporated into
numerous instrument layouts, and we estimate that well over 1000 ETD-enabled
commercial instruments have been installed in various laboratories
around the world.
Implementation of ETD on hybrid MS systems
that have multiple mass
analyzers has greatly improved its utility by offering flexibility
in reaction conditions, timing of ion manipulations, and high-resolution
mass analysis of precursor and product ions. The first commercially
available ETD-enabled hybrid mass spectrometer was a quadrupole linear
ion trap-Orbitrap (QLT-OT) system that used a negative chemical ionization
(CI) source at the back of the instrument to generate radical anions,
similar to the original description of ETD on a standalone QLT system.
19
This QLT-OT design served as a major development
platform for ETD technologies for years, including decision tree and
supplemental activation methods (described above), and was also the
first Orbitrap platform to employ the compact, high-field Orbitrap
analyzer that decreased the required transient time to achieve a desired
resolution,
52
a significant benefit for
ETD MS/MS spectra of highly charged precursor ions.
Although
robust for ETD development, the QLT-OT hybrid system was
not necessarily the ideal implementation for ETD. In 2013, our group
described a modified system in which the rf quadrupole collision cell
used exclusively for HCD was replaced with a multipurpose dissociation
cell (MDC) that was specifically designed to improve ETD reactions
on the QLT-OT system.
53
The MDC had the
same dimensions as the HCD collision cell, but it had four independently
controllable segments (for simultaneous separate storage of precursor
cations and reagent anions) and could perform charge-sign independent
trapping for ETD reactions via a secondary rf voltage applied the
end lenses of the cell. The net result was an ion–ion reaction
cell capable of storing nearly 10 times as many precursor ions than
a standard QLT and performing ion–ion reactions at least twice
as fast as the QLT due to higher ion cloud densities. Larger precursor
ion populations significantly improved product ion signal-to-noise
(S/N) in ETD MS/MS spectra, and the faster reaction times increased
throughput for LC–MS/MS analyses. Importantly, this provided
a direct route for introducing a photons into the ETD reaction cell
for AI-ETD
54
as well, which is not easily
accomplished in the QLT due to instrument geometry. The Heck group
also investigated performing ETD reactions in the HCD collision cell
itself.
55
They created a Z-shaped potential
using a static dc gradient, providing two potential minima for different
charge signs to allow charge sign independent trapping of cations
and anions. In this strategy, increasing the accumulation time (i.e.,
population size) of reagent anions caused overlap of the cation and
anion clouds in the middle of the HCD cell due to space charge, which
resulted in ETD. This implementation ultimately proved far less efficient
than ETD in the QLT because it required long anion injection times,
but it provided a way to increase the size of precursor cation populations
by using a bigger reaction cell (similar to the MDC). It also enabled
access to photoactivation schemes for ETD and was used for the ETUVPD
experiments discussed above.
Another approach to increasing
product ion S/N in ETD MS/MS spectra
is to perform a single mass analysis on product ions generated from
several ETD reactions. This idea of mass analyzing “multiple
fills” is prohibited by the position of the reagent anion CI
source at the back of the QLT-OT systems because there is no storage
region available in the instrument that is not also utilized to transfer
reagent anions prior to each ETD reaction. To circumvent this limitation,
Hunt and co-workers pioneered the design of a front-end ETD source
that permits reagent anion generation near the inlet at the front
of the instrument.
56
Using the front-end
ETD source, both precursor cations and reagent ions can be introduced
into the QLT using only the front ion optics of the instrument, allowing
products from several ETD reactions to be stored in the C-trap behind
the QLT prior to mass analysis for improved spectral quality. Importantly,
the front-end ETD source also represents a significant improvement
in reagent anion generation over the previously used negative chemical
ionization source. The front-end source uses a glow discharge operated
in a relatively high pressure region, providing a stable and robust
source of electrons for reagent anion generation and eliminating most
of the failures seen with filament-based CI sources.
Many of
ETD-specific improvements investigated on QLT-OT systems
have recently become commercially available on one of the newest generations
of Orbitrap hybrid instruments, a quadrupole-Orbitrap-quadrupole linear
ion trap (q-OT-QLT) Tribrid MS system.
57
Perhaps most notably, the use of the front-end reagent anion source
has made ETD much more robust and user-friendly, expanding its use
to more nonexpert laboratories. ETD reactions on larger precursor
ion populations is now possible as well, even though the QLT is still
used as the ETD reaction cell. In an approach called high-capacity
ETD, the accumulation of precursor and reagent ions has been modified
to allow for reactions on larger precursor ion populations, which
has proved especially beneficial for analysis of intact proteins.
58
This provides many of the same benefits for
product ion S/N seen in the MDC work without requiring specialized
hardware modifications. Furthermore, standardization of ETD reactions
through calibrated reaction times, also first developed on the QLT-OT
platform, has simplified ETD method development and use on the newer
system.
59
For supplemental activation,
the q-OT-QLT is the first system with commercially available EThcD
methods (and also has ETcaD). Moreover, the instrument geometry of
the q-OT-QLT platform removed many of the constraints that prevented
photoactivation in the QLT on the earlier systems, and a simple and
robust implementation of AI-ETD was recently described that eliminated
many of the challenges of previous AI-ETD setups.
48
Overall, the q-OT-QLT system represents one of the most
mature ETD-equipped MS platforms with many state-of-the-art ETD technologies.
Hybrid Orbitrap systems are far from the only mass spectrometers
to provide ETD capabilities, however, and other platforms offer many
benefits not achievable on the q-OT-QLT. ETD was first coupled with
Fourier transform ion cyclotron resonance (FTICR) hybrid MS systems
in 2008 using a hexapole ion trap to perform ETD reactions,
60
which is now available in commercial systems.
This implementation uses auxiliary rf potentials on the end lenses
of the hexapole ion trap to enable mutual storage of cations and anions,
with the duration of mutual storage governing the ion–ion reaction
time (similar to work described by McLuckey and co-workers
61
). Recently, Hendrickson et al. at the National
High Magnetic Field Laboratory described the first 21 Telsa (T) FTICR
hybrid MS system, which incorporates a QLT at the front of the instrument
to enable, among other benefits, ETD reactions.
62
This instrument utilizes the front-end ETD reagent anion
source described above, has supplemental activation capabilities,
and has an external quadrupole trap that allows for multiple fills
of MS/MS product ion populations prior to delivery to the ICR cell
for mass analysis. The combination of ultrahigh resolution and multiple-fill
ETD have shown significant benefits for intact protein analysis, where
an impressive ∼90% sequence coverage was achieved for carbonic
anhydrase (29 kDa) with ETD alone.
63
Furthermore,
Weisbrod et al. showed that the number of fills used for ETD mass
spectra can be controlled on-the-fly based on precursor ion molecular
weight using instrument control software, with the number of fills
scaled to appropriately fit smaller or larger protein species.
63
ETD has also been implemented on commercially
available quadrupole-ion
mobility-time-of-flight (q-IM-TOF) MS platforms. Ion mobility spectrometry
enables rapid gas-phase separation of ions based on their mobility
through a carrier gas and a number of approaches have been described
to accomplish mobility separations through various means.
64
Coupling ETD with ion mobility spectrometry
provides a direct avenue to study detailed higher-order structures
of peptides, proteins, and protein complexes. The ETD-capable q-IM-TOF
system includes four traveling wave (T-wave) ion guides, with a quadrupole
between T-wave cells 1 and 2 and a time-of-flight mass analyzer at
the back. Precursor cations and reagent anions are sequentially generated,
mass selected using the quadrupole, and stored in T-wave cell 2. Following
precursor and reagent ion accumulation, ETD reactions are conducted
by propelling precursor cations through the cloud of reagent anions
using a traveling wave voltage, the amplitude and velocity of which
control the ion–ion reaction time (typically tens of milliseconds).
65,66
ETD product ions can then be subjected to traveling wave ion mobility
separations and supplemental activation via beam-type collisions prior
to TOF mass analysis. Alternatively, broadband ETD activation of all
eluting precursor cations (i.e., no mass selection of individual precursors)
in a data-independent acquisition fashion, called MSETD, has been demonstrated on
this system with relatively simple mixtures,
as well.
67
The q-IM-TOF platform has been
largely utilized for a number of structural proteomics studies discussed
further later in this review.
Beyond commercially available
instrument platforms, more specialized
instruments equipped with ETD have also been developed. Field asymmetric
waveform ion mobility spectrometry (FAIMS) and ETD have been combined
on a research-grade QLT platform to improve analysis of post-translational
modifications (via a device that is available for some commercial
systems),
68
and Valentine and co-workers
recently introduced an ETD-enabled ion trap system modified with a
low-pressure drift tube.
69
Outside of ion
mobility interfaces, two groups have recently coupled matrix-assisted
ionization, an ionization technique that can produce multiply charged
cations without laser ablation or high voltage, with multiple MS systems
to improve characterization of modified peptides and intact proteins
generated via ambient ionization using ETD and collisional activation.
70,71
Other instrumentation developments have enabled studies designed
to provide insight into the fundamentals of ETD reactions and other
gas phase chemistries. He et al. described a dual-polarity ion trap
to simultaneously manipulate and analyze cations and anions. This
system provided a more refined study of ETD reactions by enabling
detection of positive and negative mode products from the same reaction
to investigate not only peptide dissociation products but also how
the reagent anions are modified.
72
Oomens
and colleagues coupled an ion trap mass spectrometer to a variety
of infrared laser sources at the Free Electron Lasers for Infrared
eXperiments (FELIX) laboratory, allowing for flexible experimental
designs in using infrared ion spectroscopy and IRMPD to characterize
biomolecules, including ETD-generated products.
73
This instrument has enabled examination of z•-type ETD products with infrared ion
spectroscopy, allowing accurate
determination of the products and mechanisms associated with ETD-like
fragmentation.
74
They have also commented
on the nature of hydrogen migration in ETD products
74
and have identified structures of both the ETD-inducing
fluoranthene radical anion in addition to a closed-shell proton-transfer-prone
fluoranthene anion, which has provided insights into the reaction
mechanism involved in proton transfer.
75
Moreover, this instrument setup allows for unique experiments, such
as characterization of the degree of activation of small molecules
in the active sites of homogeneous catalysts to investigate how ligand
environments affect reactivity.
76
Clearly the development of ETD instrumentation continues to push
the field forward for both fundamental and applied investigations.
Improved hardware and modified instrumentation have been a driving
force in ETD development since its inception, and we expect future
work will focus on both increasing sensitivity in ETD spectra (i.e.,
product ion S/N) and reducing ion–ion reaction times. Availability
of ETD technologies, including both new instrument platforms and strategies
like EThcD and AI-ETD, has been key in advancing proteome characterization,
and below we discuss how current ETD technologies have enabled analyses
of intact proteins, PTMs, cross-linked peptides, and other biomolecules
(Figure 1
).
ETD and
Top-Down Proteomics
ETD is ideally suited for large, highly
charged precursor cations,
making it a valuable method for top-down proteomics. The top-down
approach analyzes intact proteins (rather than proteolytically derived
peptides) using tandem MS, aiming to characterize sequence truncations,
splice variants, single nucleotide polymorphisms, and combinatorial
patterns of PTMs that contribute to proteome complexity. For unambiguous
sequence elucidation of proteoforms one must achieve extensive fragmentation
of the protein backbone,
77
making ETD an
attractive method for improving sequence coverage and confidence in
proteoform assignments. Advances in top-down proteomics span a diverse
swath of technologies, including protein purification, both online
and offline protein separations, and data analysis platforms, but
ETD remains a critical component in the recent progress of top-down
proteomic approaches (along with other alternative dissociation methods).
78
ETD provides complementary dissociation
of intact proteins compared
to collision-based fragmentation,
79−81
so the two methods are
often used in tandem in top-down experiments. The combination of ETD
and collisional dissociation has enabled large-scale, discovery-based
profiling of proteoforms, including the study by Tran et al. that
identified more than 1000 unique gene products and 3000 proteoforms.
82
More recently, Kelleher and colleagues expanded
this approach to generate the largest top-down study to date, identifying
more than 1220 proteins and 5000 proteoforms from the human lung cancer
cell line H1299.
83
Despite the clear benefits,
using multiple dissociation methods for each precursor ion in these
large-scale top-down experiments limits throughput, usually requiring
a significant investment in acquisition time. Durbin et al. presented
a modified approach to top-down proteomic data acquisition to address
this issue by guiding fragmentation of intact protein precursors to
limit redundant sampling (Figure 3
).
84
A valuable extension
of decision-tree methods, this automated acquisition scheme, called
Autopilot, uses m/z and charge state
to judiciously select ETD or collision-based dissociation. It also
calculates precursor intact mass and performs online database searches
for real-time identifications to prevent sampling of multiple precursor
ions from the same protein species, especially those that have already
been confidently identified. The Autopilot approach was applied to
a large-scale top-down analysis of human fibroblast proteomes, allowing
for quantitative comparison of nearly 1600 proteoforms in approximately
one-fifth of the total acquisition time of previous experiments, representing
a significant advance for top-down analyses of complex systems.
85
Figure 3
Autopilot data acquisition workflow on an example protein.
A full
precursor scan is taken, followed by HCD fragmentation of the 9+ charge
state on the detected mass 7246.26 Da. After an online search, the
software determines more analysis should be performed as the P-score (1.8 × 10–47)
is not below
the cutoff. An ETD scan of the highest charge state is taken and searched.
The fragment ions are combined and the final P-score
of 5.0 × 10–102 is below the cutoff. All charge
states of the 7246.26 Da species are permanently excluded from further
fragmentation and the system goes in search of the next target mass.
Reproduced from Durbin, K. R.; Fellers, R. T.; Ntai, I.; Kelleher,
N. L.; Compton, P. D. Anal. Chem.
2014, 86, 1485–1492 (ref (84)). Copyright 2017 American
Chemical Society.
Even though intact proteins
tend to carry many charges that, in
theory, make them good candidates for electron-driven dissociation,
ETD efficiency can still suffer if overall precursor charge density
is low. In recent years both EThcD and AI-ETD have been explored for
top-down proteomic applications with the goal of increasing the number
and S/N of sequence-informative product ions and protein sequence
coverage (analogous to peptide backbone coverage).
86,87
EThcD often increases the number of sequencing ions over ETD alone,
especially for higher m/z precursors,
but benefits are typically reduced for high charge density intact
protein precursors (low m/z).
88,89
That said, EThcD is still a valuable technique to improve intact
protein characterization, as shown by Brunner et al. for more confident
assignment of phosphosites in a 17.5 kDa phosphoprotein.
86
AI-ETD shows promise as a suitable supplemental
activation method for top-down proteomics, as it improves ETD-driven
activation for precursor ions of all m/z and charge states.
87−90
Similar to its performance in peptide dissociation, AI-ETD of intact
proteins generates mainly c/z•-type product ions
with some y-type fragments (and few b-type fragments), maintaining
a similar distribution of product ion types to ETD while increasing
the number and S/N of product ions observed.
87−89
The similarity
in product ion distribution between ETD and AI-ETD shows the additional
energy imparted by the concurrent IR photoactivation in AI-ETD serves
to disrupt weaker, noncovalent interactions but rarely drives vibrational
dissociation of backbone bonds, which differs from EThcD (where the
postreaction collisional activation generally increases the number
of b/y-type ions observed). Comprehensive sequence coverage (75–100%)
of proteins up to ∼20 kDa has been achieved with AI-ETD,
88
making it comparable to the extensive top-down
fragmentation reported with UVPD.
91
Online
separations with capillary zone electrophoresis have been coupled
with AI-ETD to improve characterization of the Mycobacterium
marinum secretome,
92
and more
thorough high-throughput LC–MS/MS top-down proteomic experiments
with AI-ETD are forthcoming.
Even as top-down proteomics continues
to progress, current technology
has difficulty detecting proteins with molecular weights greater than
∼30 kDa. Raising this size barrier is a current topic in the
field. In 2013, Tsybin and co-workers showed that ETD on a TOF MS
system could provide extensive sequence information for moderately
sized proteins (∼30 kDa) and structural motifs embedded in
large proteins (up to ∼80 kDa).
93
Just this year, Anderson et al. demonstrated the benefits of the
ETD-equipped 21T FTICR hybrid MS system to characterize complex mixtures
of human proteins, using a combination of collisional dissociation
and multiple-fill ETD.
94
ETD spectra typically
provided superior confidence scores and proteoform characterization,
and they identified 684 unique gene products and 3238 unique proteoforms,
a considerable portion of which were >30 kDa. As instrumentation
and
other offline approaches continue to improve, AI-ETD may also play
a role in expanding the molecular weight range that can be characterized
in top-down proteomics experiments, as it was recently shown to outperform
HCD, ETD, and EThcD for ∼30–70 kDa proteins.
89
Other groups have focused on more clinical-based
applications of
ETD and top-down proteomics. Coelho Graça et al. showed that
a targeted method to look for ETD fragments from intact hemoglobin
ions from blood samples could provide a simple and flexible methodology
that required only hours from sample collection to results, making
it suitable for application in clinical laboratories.
95
They followed up the proof-of-concept study by comparing
rare hemoglobin β chain variants that enabled fast and reliable
determination of uncommon hemoglobin proteoforms useful for hemoglobin
disorder diagnoses.
96
Other groups have
worked to bring ETD to spatial analyses of tissues for potential clinical
use. Using liquid extraction surface analysis (LESA) MS, Cooper and
co-workers used ETD to reliably identify intact proteins in healthy
and diseased human liver tissue, helping to distinguish potential
protein biomarkers of nonalcoholic liver disease.
97,98
Schey et al. combined spatially directed tissue microextraction
to analyze intact proteins from specific locations in ocular lens,
brain, and kidney tissues using LC–MS/MS with ETD, identifying
several proteoforms that were difficult to assign with bottom-up methods.
99
From our perspective, ETD offers a number of
benefits for intact protein analysis, especially when used in concert
with collisional dissociation. We anticipate that top-down proteomics
will utilize ETD for years to come and will continue to incorporate
new ETD technologies like EThcD and AI-ETD as they become available.
Characterizing
Post-Translational Modifications with ETD
Protein post-translational
modification is a dynamic and important
process that regulates a diverse range of cellular functions.
100,101
MS-based approaches are the premier tools for global PTM analysis,
boasting high sensitivity, considerable throughput, amenability to
diverse classes of PTMs, and the capacity to localize modifications
to a single residue. Realization of these benefits, however, requires
high-quality MS/MS spectra that have extensive fragmentation along
the peptide/protein backbone, in addition to retention of the modifications
themselves on the subsequent fragments. ETD has been central in analyses
of many PTMs because it drives the radical dissociation of peptide
backbone bonds while retaining labile modifications intact at modified
residues. The breadth of work surrounding technology development for
PTM analysis is far too great to cover here, especially because considerable
efforts in these fields concentrate on offline methods, e.g., enrichment
strategies, chromatographic separations, and data analysis. In the
following section we focus on how ETD methodology specifically has
been applied to characterize three specific classes of PTMs in recent
years. We direct readers to recent reviews that cover broader technologies
for analysis of phosphorylation,
102−104
glycosylation,
105−107
and PTMs in general
100,101
for further insights. We also
note that ETD has been valuable for PTM analysis beyond the scope
discussed here, including ubiquitylation,
108−110
ADP-ribosylation,
111−113
and arginine methylation.
114−118
Phosphorylation
Protein phosphorylation is a rapid
and reversible means to modulate protein activity and transduce signals.
Regulation of phosphorylation is a central mechanism in cell health
and disease. Phosphorylation most often occurs on serine and threonine
residues, with tyrosine phosphorylation also occurring with moderate
frequency. The labile nature of phosphoryl groups can limit utility
of collision-based dissociation for analysis of phosphorylated peptides
and proteins because neutral losses, rather than peptide backbone
fragmentation, are energetically favored pathways. Indeed, this shortcoming
of collisional activation was a major impetus in the original development
of ETD. Modern phosphoproteomic experiments often employ decision
tree methods combining collision-based fragmentation and ETD to increase
the number and confidence of phosphopeptide identifications,
119−123
and more advanced decision tree algorithms have been designed to
trigger ETD MS/MS scans based on phosphoric acid neutral losses or
evaluation of phosphosite assignments in collisional activation spectra.
124−126
These acquisition strategies are often paired with new informatic
approaches to integrate ETD spectra into phosphosite assignment algorithms,
and analysis of a large synthetic phosphopeptide library with multiple
fragmentation methods has aided such developments.
127
Neutral loss-triggered decision trees have also been employed
for analysis of pyro-phosphorylation, a more rare version of serine
and threonine phosphorylation, where EThcD was used to confidently
localize pyro-phosphorylation sites and rule out possibilities of
multiple standard (e.g., “non-pyro”) phosphosites.
128
As expected, EThcD and AI-ETD supplemental
activation methods have significantly benefitted ETD-based phosphoproteomic
experiments. Increases in the number of sequencing ions and peptide
backbone coverage provided by both methods greatly improves the ability
to identify and unambiguously localize phosphosites in phosphopeptides,
especially for low charge density precursors (which are more prevalent
in phosphoproteomic experiments).
90,129
Development
of EThcD and AI-ETD for phosphoproteomics also led to modifications
of the phosphoRS localization algorithm
130
that uses observed product ions to statistically evaluate probabilities
of phosphosite locations.
90,129
Both EThcD and AI-ETD
have been used for analyses of intact phosphoproteins,
86,90
and AI-ETD has shown superior performance for localizing phosphosites
in the multiply phosphorylated protein α-casein (∼23.5
kDa, eight sites). Lössl et al. used combinations of ETD and
EThcD methods for integrated bottom-up and top-down phosphoproteomic
experiments (among several other MS approaches) to decipher how the
number and order of individual phosphorylation events impact protein
behavior at a mechanistic level, especially in multiple-protein systems.
131
A recent study from Tamara et al. showed that
gas-phase phosphate transfer can readily occur between proteins in
a complex, enabling high-precision elucidation of binding sites between
phosphoproteins and their binding partners, an interesting finding
that used EThcD to localize these phosphate-transfer sites.
132
ETD can characterize N-phosphorylation
(e.g., lysine and arginine)
as well.
133−135
Although less common than S/T/Y O-phosphorylation,
these alternative phosphorylation events regulate signaling mechanisms
in bacterial systems, and more studies are emerging to suggest they
may also play a role in eukaryotic signaling. The lability of the
phosphoramidate bond leads to neutral losses that considerably increase
false localizations in collision-based fragmentation spectra of N-phosphorylated
peptides. Conversely, phosphoarginine and phospholysine are sufficiently
stable under most ETD conditions, and ETD often enables confident
identification and localization of N-phosphorylation. That said, some
parameters (e.g., the number of nonmobile protons in the precursor
cation) must be considered to account for possible gas-phase rearrangements
that can affect reliability of phosphosite localization.
136
Clausen and co-workers recently used ETD to
characterize the role arginine phosphorylation in protein turnover
in Bacillus subtilis, providing new insights into
protein degradation pathways in Gram-positive bacteria.
137
Glycosylation
Glycoproteomics is
one of the arenas
of modern proteomics where ETD has the greatest impact. Protein glycosylation
is a prevalent, chemically complex, and biologically diverse PTM involved
in a wide array of intra- and intercellular functions. It is highly
heterogeneous modification that accounts for the greatest proteome
diversity over any other PTM. Analysis of intact glycopeptides is
imperative to glycoproteome characterization because multiple glycans
can modify a given glycosite (i.e., glycan microheterogeneity), which
makes glycan identity at a given site crucial to the biological context
of the modification. Collision-based fragmentation of intact glycopeptides
usually only reveals information about the glycan component of the
precursor and offers little sequence information about the peptide
backbone. ETD, on the other hand, results in nearly exclusive dissociation
of the peptide backbone while leaving the glycan moiety intact, allowing
for peptide sequence elucidation and site-specific analysis of the
glycan modification.
Consistent with themes discussed throughout
this review, ETD is usually paired with collision-based dissociation
in glycoproteomic experiments, although we note that this coupling
is especially prevalent in glycoproteomics to capitalize on the complementary
nature of fragmentation of the two methods for peptide and glycan
characterization.
138−146
A product ion-dependent decision tree method has become a powerful
approach for glycopeptide analyses, triggering ETD scans for precursors
when glycan oxonium ions are observed in HCD spectra.
147−149
The benefits of this approach are 2-fold: (1) the faster scan rates
of HCD MS/MS enable increased sensitivity/greater sampling depth,
acquiring slower ETD scans only for precursor ions that have a high
likelihood of being glycopeptides (where ETD adds value), and (2)
paired HCD and ETD MS/MS spectra for potential glycopeptide ions are
automatically acquired, aiding in the interpretation of glycopeptide
identifications.
Analysis of N-linked glycosylation, where the
glycan is attached
at an asparagine residue, is generally more prevalent and somewhat
more straightforward. Here, glycans can be readily cleaved off enzymatically
for separate analysis of glycans and peptides that can inform intact
glycopeptide characterization with ETD. Additionally, modifications
happen within the confines of a sequence motif, N-X-S/T (where X is
a residue other than proline). O-Glycosylation at serine and threonine
residues is more challenging to characterize because cleavage of glycans
from the peptide backbone is less straightforward (limiting ability
to leverage deglycoproteomic data to inform intact glycopeptide experiments),
and the lack of consensus motif makes unambiguous glycosite localization
more difficult (i.e., multiple potential modifications sites in a
given peptide are more likely). For example, Parker et al. combined
intact glycopeptide analysis (using ETD and collisional dissociation)
with glycomics of released N-glycans and deglycoproteomic experiments
to characterize 863 unique N-glycopeptides, corresponding to 276 N-glycosites
on 161 proteins in rat brain lysates.
142
Xu et al. identified ∼1150 unique N-linked glycopeptides
representing 348 N-glycosites on 270 protein in Arabidopsis influorescence tissue
using various ETD-based methods.
145
Neither study targeted O-linked glycopeptides.
In one of the largest intact glycopeptide characterizations to-date,
Trinidad et al. used ETcaD to identify 2100 N-glycopeptides representing
∼700 N-glycosites on 375 proteins.
143
This is compared to identification of only 463 O-glycopeptides on
122 proteins, with ∼45% of O-glycosites left ambiguously defined
in the same study, highlighting the discrepancy in N- and O-linked
glycopeptide analyses.
Several groups have explored strategies
to use ETD in concert with
other approaches for O-glycopeptide characterization.
998
Several years ago, Thaysen-Anderson and co-workers
demonstrated the ability of ETD to characterize densely O-glycosylated
mucin-type peptides and provided reasonable performance in unambiguous
O-glycosite localization.
999
Darula et
al. used a combination of enrichments (jacalin, a lectin specific
for N-acteylgalactosamine seen in O-glycans, ion exchange chromatography,
and hydrophilic interaction chromatography) in addition to partial
deglycosylation of O-glycans via exoglycosidases to improve ETD fragmentation
and glycosite localization of mucin core-1, core-2, and core-3 oligosaccharides.
1000
In recent years, the Medzihradszky and Wuhrer
groups have followed up with these methods to characterize mucin core-1
type O-glycoproteins in human serum and blood samples with ETD.
150,151
Futhermore, removing N-glycans via peptide N-glycosidase F can provide
significant benefits for the characterization of O-glycosylation sites
with a combination of collisional dissociation and ETD, as underscored
by methods investigated by Houcel et al.
152
A SimpleCell strategy, which uses genetic engineering to simplify
O-glycans to a single truncated O-GalNAc residue, has also been coupled
with ETD methods to profile the O-glycoproteome of mammalian cell
lines, and similar strategy was used to investigate O-mannosylation
in yeast.
1001−1004
Windwarer and Altmann used ETD in combination with extensive offline
fractionation, generation of large proteolytic glycopeptides, and
direct infusion analysis of individual fractions to characterize occupancy
of O-glycosylation sites in bovine fetuin, a common glycoprotein standard.
153
Beyond general O-glycosylation, ETD has
also significantly impacted
analysis of O-GlcNAcylation, a common and specific form of O-glycosylation
in which a single β-linked N-acetylglucosamine is attached to
serine and threonine residues.
154
Burlingame
and co-workers recently highlighted the essential contribution ETD
offers for O-GlcNAc analysis by retaining the labile monosaccharide
modification on product ions for unambiguous localization of modification
sites.
155
Studies of O-GlcNAcylation often
examine the modification within the context of cross-talk of between
it and phosphorylation, as they both modify serine and threonine residues,
156
and ETD plays a central role in the ability
to elucidate the interplay between both modifications. Trinidad et
al. performed one of the first large-scale characterization of site-specific
O-GlcNAc and phosphorylation cross-talk, identifying 1750 and 16 500
sites of O-GlcNAcylation and phosphorylation, respectively, from murine
brain tissue.
1005
Since then, several studies
have expanded those investigations to interrogate O-GlcNAc cross-talk
in Alzheimer’s disease,
157,158
regulation of circadian
clock mechanisms in mice and drosophila,
159
and nutrient-sensitive intracellular processes that have significant
implications for downstream metabolic regulation.
160
As with most applications of ETD, the value of supplemental
application
cannot be overstated for glycoproteome characterization with ETD.
ETcaD has been employed in several large-scale glycopeptide characterization
studies,
143−145
and the benefits of EThcD and AI-ETD for
intact glycopeptide analysis are just beginning to be explored. Yu
et al. demonstrated the ability of EThcD to generate both glycan and
peptide fragmentation in a single MS/MS spectrum based on the hybrid
combination of electron-driven and collision-based dissociation (Figure 4
).
161
They used EThcD to characterize several glycoproteomic
samples, ultimately showing it to be well-suited for large-scale glycopeptide
studies by identifying ∼1000–1200 unique N-glycopeptides
from rat carotid arteries. A second study by Yu et al. used a combination
of approaches, including EThcD, to report the first description of
O-linked glycosylation on mouse insulin chains among other proteins
and signaling peptides, providing new methods to investigate the role
of glycosylation in diabetes.
162
Glover
et al. used EThcD and HCD-triggered EThcD methods to combine phosphoproteome
and glycoproteome analyses, focusing on sialylated and phosphorylated
glycopeptides contained in standard phosphopeptide enrichments.
163
In all, they characterized ∼4000 phosphopeptides
and ∼1000 sialylated glycopeptides from a single enrichment
of rat smooth muscle cells, and they also identified glycopeptides
with mannose-6-phosphate glycans. Parker et al. noted that EThcD improved
glycopeptide spectral quality but commented that further optimizations
involving spectral scoring and interpretation were probably needed
to maximize its utility.
164
Pitteri and
co-workers used ETcaD and EThcD-based methods to assess the quality
of intact glycopeptide enrichment strategies, ultimately identifying
829 unique glycoforms across 208 N-glycosites in 95 proteins from
human plasma.
165
Also we recently described
the first application of AI-ETD to glycoproteomics,
166
showing that AI-ETD enabled the largest intact glycopeptide
analysis to date by identifying >7500 localized unique N-glycopeptides. Various
pre- and post-ETD
supplemental activation schemes characterized the effect of glycosylation
on fragmentation of intact proteins, as well, and distinguished high
mannose glycans in RNase B (∼14.7 kDa).
167
Figure 4
EThcD MS/MS of 3+ charge state precursor ion at m/z 1577.9 of bovine fetuin triantennary
N-glycopeptide
KLCPDCPLLAPLNDSR (AA 126–141). Starred peaks (*) in the spectra
were deconvoluted and annotated in the inset. Reproduced in part from J. Am. Soc.
Mass Spectrom., Electron-Transfer/Higher-Energy
Collision Dissociation (EThcD)-Enabled Intact Glycopeptide/Glycoproteome
Characterization, Vol. 28, 2017, 1751–1764,
Yu, Q.; Wang, B.; Chen, Z.; Urabe, G.; Glover, M. S.; Shi, X.; Guo,
L. W.; Kent, K. C.; Li, L. (ref (161)) with permission of Springer.
In an effort to enable more quantitative comparisons
in large-scale
glycoproteomic experiments, Bertozzi and co-workers recently described
a new strategy to called isotope-targeted glycoproteomics (IsoTaG),
where stable isotopes are incorporated through labeled azido sugars.
168
Intact azidosugar-labeled glycopeptides can
then be enriched and analyzed via ETD and EThcD based methods. IsoTaG
using Ac4GalNAz or Ac4ManNAz sugars enabled identification of 1375
N- and 2159 O-glycopeptides across 15 cell lines,
169
and the approach was extended through use of alkynyl sugars
as metabolic labels in analysis of the sialylated glycoproteome, identifying
699 intact glycopeptides from 192 glycoproteins in PC-3 cells.
170
ETD has enabled characterization of several
noncanonical glycosylation
modifications as well.
143
The Heck group
combined high-resolution native mass spectrometry and EThcD-based
glycopeptide analyses to study human ethyropoeitn and plasma properdin
as models for therapeutic proteins and plasma protein markers.
171
They qualitatively and quantitatively monitored
coappearing proteoforms in these proteins in addition to revealing
PTM localizations, relative abundances, and glycan structures in a
site-specific manner. Furthermore, this synergistic approach lead
to the discovery of three new C-mannosylation sites, a noncanonical
type of glycosylation involving linkage of the glycan through a carbon–carbon
bond. In a second study they integrated native mass spectrometry and
EThcD-based glycoproteomics workflows to detail the structural microheterogeneity
of human complement C9 protein (∼65 kDa), revealing ∼50
distinct signals via native MS and 15 co-occurring proteoforms that
included known N- and C-glycosylation and new evidence of O-glycosylation.
172
Pronker et al. used ETcaD in combination with
X-ray diffraction, structure guided mutations, and several biophysical
assays to characterize the structural basis of myelin-associated glycoprotein
adhesion and signaling, which included identification of N-glycosylation
and a tryptophan C-mannosylation.
173
ETD
and EThcD revealed that cysteine S-linked N-acetylglucosamine (S-GlcNAcylation)
occurs in mammalian systems, as well, including murine and human systems.
174
Finally, ETD has been valuable for characterizing
sites of glycation (nonenzymatic addition of sugars to amino acids)
175−177
in addition to peptidoglycans, which are mostly glycans cross-linked
with nonstandard amino acids.
178
In terms of scale and scope, glycoproteomic methods have lagged
behind analyses of other PTMs, but ETD is quickly making large-scale
glycopeptide analysis a realistic endeavor. Methods like EThcD and
AI-ETD, which offer complementary dissociation of glycan and peptide
components of glycopeptides in a single MS/MS scan, will be essential
in enabling global glycoproteomic profiling of complex systems as
the field continues to advance.
Disulfide Bonds
Disulfide bonds, naturally occurring
intramolecular cross-links between cysteine residues in proteins,
play an important role in stabilizing tertiary and quaternary structure,
often dictating proper biological function. Disulfide bonds are especially
enriched in secreted and membrane proteins and are an important feature
in the complexity of the proteome. They represent a challenge for
MS/MS approaches, both peptide and intact protein, because both the
disulfide and the peptide backbone must be fragmented to access sequence
information within the disulfide-enclosed regions. ETD reactions can
preferentially cleave disulfide bonds,
179,180
lending ETD
methods high utility for characterizing this PTM. Often ETD is used
to open disulfide bonds
181
and subsequent
collisional activation provides sequence information on the peptide(s)
present;
182
however, in many cases ETD
alone can generate sufficient c/z•-type product
ions through radical cascades initiating at N–Cα bond cleavage that propagates to
cleave multiple disulfide bonds.
183
Liu et al. developed an EThcD workflow
specifically designed to provide a precise, yet generic approach for
disulfide bond mapping.
184
In EThcD of
disulfide-bonded peptides, the ETD reaction preferentially leads to
the cleavage of the S–S bonds (in addition to some peptide
backbone fragmentation). Supplemental HCD activation then dissociates
unreacted and charge-reduced precursor ions of the disulfide-cleaved
peptides to provide further peptide backbone fragmentation. They also
described a software platform called SlinkS to process the relatively
complex spectra generated by this process. Note, pepsin is often the
protease of choice in peptide-based disulfide-bond mapping because
it is active at low pH, which prevents unwanted disulfide reshuffling
during sample preparation.
Other approaches involve more nuanced
tactics to study disulfide
bonds, including targeted methods to dissociate intrachain disulfide
bond containing peptides with ETD based on prior analysis of reduced/alkylated
samples and assigning disulfide bond connectivity via extracted ion
chromatograms of disulfide bonded peptide pairs that are generated
together following ETD reactions.
185,186
In all, ETD-based
analyses of disulfide bonds have been used in recent studies to investigate
neuropeptides,
185
secreted proteins,
182
HIV envelope proteins,
186
and protein storage stability,
187
among other things.
Alternative Proteases, Middle-Down Proteomics,
and Peptidomics
The vast majority of proteomics experiments
utilize trypsin to
proteolytically cleave proteins C-terminal to lysine and arginine
residues, generating peptides for further analysis with LC–MS/MS.
This bottom-up, or shotgun, approach with trypsin is particularly
well-suited for ubiquitously available, high-throughput collision-based
dissociation methods; reliance on characterization of exclusively
tryptic peptides, however, limits the depth and coverage of the proteome
that can be sampled. Use of alternative proteases (e.g., LysC, GluC,
ArgC, AspN, and chymotrypsin) in a complementary fashion with trypsin
can significantly increase sequence coverage across the proteome,
but many proteases generate peptides that are not conducive to collision-based
dissociation. Thus, ETD has directly benefited deep sequencing efforts
and has contributed to much greater proteome coverage via multiple-protease
approaches. Two recent reviews provide a broader perspective of the
field beyond ETD.
188,189
In the first large-scale
multiple protease study to leverage ETD,
Swaney et al. showed that ETD methods could offer improved characterization
of ArgC, AspN, GluC, and LysC peptides when compared with collisional
dissociation, and the combination of identifications from these complementary
proteases with tryptic peptide identifications enabled a 3-fold increase
in protein sequence coverage across the yeast proteome.
190
In 2015, Nardiello et al. followed-up on this
idea by using collisional activation and ETD to characterize peptides
derived from digestions of standard proteins using trypsin and chymotrypsin.
They showed that the combination of both dissociation methods and
proteases could increase sequence coverage and aid in identification
of PTMs, species-specific residues, and single-point amino acid modifications
in natural protein variants.
191
Trypsin
and chymotrypsin were also used in tandem by Somasundaram et al. to
identify protein C-termini, where they derivatized carboxylic acids
and used collisional dissociation and ETD to improve C-termini characterization.
192
Multiple proteases (trypsin, chymotrypsin,
subtilisin, and AspN) and combinations of collisional activation and
ETD were used to improve characterization of the density and complexity
of glycosylation in polymeric mucin MUC2, which has a highly O-glycosylated
mucin domain.
193
One of the more robust
applications of multiple protease characterization with ETD in recent
years came from Guthals et al.
194
Using
overlapping peptides from multiple digests and corroborating b/y/c/z•-type fragments
from collisional activation and ETD,
they de novo assembled peptide sequences averaging
nearly 70 amino acids in length at 99% sequencing accuracy.
Tsiatsaiani et al. investigated the chromatographic separation
properties, ETD and HCD fragmentation behavior, and (phospho)proteome
sequence coverage of LysargiNase, a protease that mirrors the proteolytic
activity of trypsin by cutting N-terminal to lysine and arginine residues
with high specificity.
195
Interestingly,
LysargiNase peptides fragment as near mirror images to tryptic peptides,
with LysargiNase peptides generating predominantly N-terminal (c-
and b-type) product ions using ETD and HCD, respectively (Figure 5
). ETD of LysargiNase
peptides provided especially informative sequence ladders (although
HCD performance was worse than for trypsin); and overall, analysis
of LysargiNase generated peptides provided complementary characterization,
adding coverage in both proteome and phosphoproteome analysis. In
a similar vein, Aebersold and co-workers introduced arginyl-tRNA protein
transferase (ATE)-mediated LysC/AspN proteolysis that generates arginylated
peptides with basic amino acids on both termini.
196
Dissociation of these peptides generates near complete
sequence ladders from both N- and C-terminal ends, and ETD generates
complete c- and y-type fragments in addition to other product ion
types (e.g., z-type).
Figure 5
Fragmentation characteristics
of proteolytic K/R(X)n and (X)nK/R
peptides. N-terminal or C-terminal protons drive the formation of
opposite but complementary ion patterns for the alike peptides during
ETD, with basic residues depicted in red. (a) ETD MS/MS spectra of
LysargiNase (top) and tryptic peptides (bottom) with a single basic
residue or with multiple basic residues. b) Fragmentation heat maps
based on ion intensities for K/R(X)n and (X)nK/R peptides during HCD
and ETD. Normalized relative intensity values were calculated for
peptides 6–20 amino acids long. Series numbers are matched
to the sequence orientation. Reproduced in part from Tsiatsiani, L.;
Giansanti, P.; Scheltema, R. A.; van den Toorn, H.; Overall, C. M.;
Altelaar, A. F. M.; Heck, A. J. R. J. Proteome Res.
2017, 16, 852–861 (ref (195)) Copyright 2017 American
Chemical Society.
Middle-down proteomics
is an extension of the alternative protease
approach, with the goal of generating large peptides that cover the
middle mass range (∼3–10 kDa) between tryptic peptides
and intact proteins. In theory, the middle-down approach should allow
for the detection of some PTM combinations and provide more extensive
proteome coverage not achievable with shotgun approaches. Either nontraditional
proteases or limited trypsin digestion can be used to accomplish this
goal.
197
Whichever approach is leveraged,
ETD is well-suited to sequence these large peptides. In 2012, Kelleher
and colleagues described a method for restricted enzymatic proteolysis
using the outer membrane protease T (OmpT), a member of the novel
omptin protease family derived from the Escherichia coli K12 outer membrane.
198
OmpT produced
peptides with >6.3 kDa mass on average, and ETD showed highly complementary
dissociation behavior to collisional dissociation for these peptides.
In all, the OmpT middle-down workflow enabled identification of 3697
unique peptides from 1038 proteins, including those with PTMs, and
closely related protein isoforms could be readily differentiated.
Cristobal et al. recently focused on optimizing several aspects of
middle-down workflows, including proteolysis with GluC and AspN (and
a nonenzymatic formic acid induced cleavage), chromatographic separations
with larger pore-sized particles and peptide dissociation with HCD,
ETD, and EThcD.
199
They found the generation
of larger peptides suitable for true middle-down analysis was not
readily achievable with GluC or AspN, which highlights the value of
new middle-down proteases like OmpT and other limited proteolysis
approaches. Regardless, EThcD performed best for the analysis of midsized
peptides, providing high MS/MS success rates and higher peptide sequence
coverage (up to 95% peptide backbone coverage) compared to HCD and
ETD.
Peptidomics likewise involves analysis of nontryptic peptides,
as it focuses on endogeonously produced protein fragments. Opposed
to the other methods described in the section, proteases are rarely
used to prepare peptidomic samples, so a wide range of endogenous,
nontryptic protein fragments are present that require a combination
of dissociation methods to properly characterize.
200
Collision-based dissociation and ETD are used together
in a wide range of peptidomic experiments, including large-scale identification
of secretory peptides in human endocrine cells,
201
microproteins and endogenous peptides in murine brain tissues,
202,203
and native peptides present in urine of healthy women during pregnancy.
204
Peptidomics also enables screening of natural
peptide products that have medicinal or therapeutic uses. Gucinski
and Boyne used ETD to identify multiple forms of protamine sulfate,
a complex peptide drug product with multiple basic peptides.
205
Juba, Bishop, and co-workers used ETD-based
methods to screen Komodo dragon (Varanus komodoensis) and American alligator (Alligator
mississippiensis) plasma for cation antimicrobial peptides (CAMPs), potential targets
for development of future antibacterial theurapeutics.
206,207
ETD has been used in characterizations of venoms, as well, which
are a rich source for discovering new peptide and protein products
with biotechnological applications.
208
Cases
Where ETD is Particularly Helpful
Major Histocompatibility
Complex/Human Leukocyte Antigen Peptides
The major histocompatibility
complex (MHC), referred to in humans
in the human leukocyte antigen complex (HLA), is a group of proteins
that are central to immune response. They function by presenting peptides
at the cell surface for recognition by different T-cells for recognition.
The MHC peptides presented are generated via proteolysis and can be
derived from cytosolic proteins (MHC Class I proteins) or from antigens
that are internalized into antigen presenting cells via phagocytosis
(MHC Class II proteins). Class I peptides are typically 8–11
residues in length, while Class II peptides range from 10 to 25 amino
acids; much like the challenges seen in peptidomics, these endogeonously
derived MHC peptides can be difficult to characterize, especially
because they can have PTMs that play important roles in their recognition.
MS-based approaches to study MHC peptides were pioneered by Hunt and
co-workers more than 2 decades ago,
209
but
the introduction of ETD and related techniques has greatly aided discovery
of new properties of MHC peptides in recent years.
The majority
of work has focused on the Class I immunopeptidome. Combinations of
ETD and collisional dissociation are leveraged to study phosphorylation
of HLA-I peptides,
210,211
and Mommen et al. showed in
2014 that EThcD could significantly improve both sequence coverage
and PTM characterization, including phosphorylation, for global analyses
of HLA-I peptides. A follow up study of HLA-II peptides also demonstrated
the value of ETD-based approaches in combination with other dissociation
methods to characterize the complex nature of the HLA class II system.
212
Further studies with EThcD have indicated new
PTMs on HLA-I peptides, including O-GlcNAc,
213
extended O-GlcNAc structures,
214
and
dimethylated arginine,
215
and Liepe et
al. shed new insight into the frequency and abundance of proteasome-catalyzed
peptide splicing events in Class I peptides.
216
ETD is also leveraged for the study of drug binding interactions
with Class I and II peptides,
217
and Malaker
et al. recently showed that Class II peptides have 17 different glycoforms,
indicating a role for diverse and complex glycosylation patterns in
MHC recognition.
218
Histones
Histones
are fundamental protein components
of chromatin, which is the structural framework for chromosomal DNA
in the cell nucleus. The modification states and sequence variants
of histones are directly involved in gene expression, making histones
supremely important in epigenetic regulation. Bottom-up analyses of
histones can be challenging since N-terminal regions are lysine and
arginine rich (resulting in small, difficult to analyze tryptic peptides)
and combinatorial patterns of PTMs on histone tails (i.e., the “histone
code”) are important to understanding regulation states, meaning
that middle-down and top-down methods are particularly valuable for
histone characterizations.
219−221
ETD thus provides obvious benefits
for histone analysis, and indeed, it has been highly utilized for
histone studies.
Bottom-up and middle-down proteomic analysis
of histones with ETD are used to study histones in a variety of capacities,
including cellular reprogramming of histone H4 in induced pluripotent
stem cells,
222
the role of H2A modifications
in vertebrate embryo development,
223
histone
ADP-ribosylation in DNA damage response,
224
and histone modification states in transformed cell line vs primary
cell monocyte-derived macrophages.
225
Schriemer
and co-workers recently described a new alternative protease called
neprosin that is well-suited for middle-down histone analysis.
226
Neprosin cleaves C-terminal to proline residues,
making relatively large peptides that depend more heavily on ETD for
characterization. Importantly, it also creates relatively large peptides
from histone tails (1–38 for H3 and 1–32 for H4) for
analysis of co-occurring PTMs. Various separation methods have been
tested for middle-down workflows for histone characterization with
ETD fragmentation, with weak cation exchange chromatography,
227,228
capillary electrophoresis,
229
and ion
mobility
230
all being demonstrated as compatible
online separation techniques. Data-independent ETD methods have been
investigated with some success for middle-down histone characterization
as well.
231
Quantitative comparisons of
histones and their modifications can be accomplished through several
strategies enabled by bottom-up and middle-down approaches.
232−234
The top down approach offers some advantages for histone analysis
because intact characterization best captures the biological information
present in histone PTM patterns and provides insight into the complexity
of the histone proteome; that said, technology development is still
needed to make top-down methods more accessible.
235
Advances in ETD-based methods are making significant inroads
toward this goal, and three recent studies, all using front end ETD
sources in various capacities, have noticeably improved the ability
to analyze intact histones on chromatographic time scales.
236−238
Furthermore, Molden et al. showed how combinations of bottom-up
and top-down approaches can comprehensively profile histone changes.
239
Antibody Characterization
Monoclonal
antibodies (mAbs)
and related biological molecules are an important and growing class
of human therapeutics; more than 30 immunoglobulins (Igs) have been
approved for the treatment of cancer, immunological diseases, and
infectious diseases.
240
The specificity
and affinity of mAbs are among their key advantages, but engineering
these molecules to have the desired traits requires thorough characterization.
Their large size (two heavy chains at ∼50 kDa and two light
chains at ∼25 kDa for a total mass of ∼150 kDa), number
of disulfide bonds (which are critical for structure and activity),
sequence differences in key variable regions, and presence of PTMs
(e.g., glycosylation) make comprehensive characterization challenging,
but as discussed above, ETD-based methods offer several benefits for
many of these features.
Bottom-up, middle-down, and top-down
methods have all been explored for mAb characrterization, and ETD
plays a role in each approach. Extending their de novo sequencing approach to mAb
analysis, Guthals et al. recently showed
that a combination of multiple proteases and multiple fragmentation
methods could sequence mAb from human serum with no sequence database
required, enabling discovery of new mAbs.
241
Hunt and others have used online proteolysis with pepsin to analyze
mAbs in bottom-up and middle-down approaches,
242−244
using ETD to characterize large peptides derived from limited proteolysis,
including those with intact disulfide bonds. Several laboratories
have explored using alternative proteases (including immunoglobulin
G-degrading enzyme of Streptococcus pyogenes [IdeS]
and secreted aspartic protease 9 [Sap9]) or simple reduction of disulfide
bonds to characterize subunits and intact chains of IgG with a variety
of fragmentation methods, often employing ETD for large subunits that
require extensive fragmentation.
245−249
Others have investigated top-down methods
to characterize fully intact mAbs (i.e., no proteolysis) with some
success (Figure 6
),
and they rely on ETD to derive sequence information from the ∼150
kDa precursor ions, although sequence coverage has been limited to
∼30%.
250−253
Figure 6
Intact
mAb analysis with ETD. Product ion abundance analysis of
ETD top-down MS of adalimumab with 10 ms ETD duration and trastuzumab
with 10 and 25 ms duration. In the case of trastuzumab, the color-coded
histogram demonstrates the improvement in sequence coverage obtained
through the combination of 10 ms (magenta) and 25 ms (cyan) ETD MS/MS
data. Whereas medium and large product ions (e.g., z117 and z63) are
generally produced using 10 ms duration, smaller ions such as z16,
z17, and z18 are detected only in the case of a longer ETD duration,
presumably as the result of secondary fragmentation (i.e., refragmentation
of larger product ions). Note the ability of ETD to generate fragments
within a disulfide bridge region. Reprinted from J. Proteomics, Vol. 159, Fornelli,
L.; Ayoub, D.; Aizikov, K.;
Liu, X.; Damoc, E.; Pevzner, P. A.; Makarov, A.; Beck, A.; Tsybin,
Y. O. Top-down analysis of immunoglobulin G isotypes 1 and 2 with
electron transfer dissociation on a high-field Orbitrap mass spectrometer,
pp 67–76 (ref (253)). Copyright 2017, with permission from Elsevier.
Structural Characterization Using ETD
Structural characterization of the proteome is a natural extension
from recent advances in instrumentation, top-down proteomics, and
PTM analyses. Beyond providing information about the presence and
abundance of proteins and their various proteoforms, MS contributions
to molecular and structural biology are gaining momentum and offering
new ways to examine protein–protein interactions, small molecule–protein
interactions, and other studies of large biomolecular assemblies (recently
reviewed comprehensively by Lössl et al.
254
). ETD offers several benefits to this effort, including
analysis of cross-linked peptides, peptides and proteins labeled via
hydrogen–deuterium exchange, and proteins ionized via native
electrospray ionization.
Cross-Linked Peptides
Chemical cross-linking
proteomics
provides an avenue to investigate protein–protein interactions
(i.e., the “interactome”) and protein conformations.
Cross-linking reagents have two reactive groups directed toward chemical
reactivity with specific amino acid residues, and the reactive groups
are connected by a spacer arm that can vary in length. Cross-linking
reagents are introduced prior to proteolysis to introduce intra- and
interprotein cross-links that help discern structural information.
Following proteolysis, cross-linked peptides are analyzed with tandem
MS (sometimes via multiple stages) to elucidate which sequences exist
within a given proximity in the proteome. Note, the resolution of
proximity measurements is defined by the length of the spacer group
in the cross-linking reagent. In 2015, Heck and co-workers introduced
a proteome-scale workflow for studying protein assemblies via cross-linked
peptides, and they used a back-to-back combination of collisional
activation and ETD MS/MS scans on the same precursor to identify 1622
intraprotein and 200 interprotein cross-links (134 protein–protein
interactions) at a 1% false discovery rate (Figure 7
).
255
They recently
followed up on this strategy by optimizing fragmentation conditions
for cross-linked peptides, showing that MS2-MS3 collisional dissociation in concert
with ETD MS2 scans
provided the best characterization of cross-linked peptides.
256
Rappsilber and colleagues extended studies
of collision- and ETD-based methods to cross-linked peptides for development
of decision tree methods including EThcD,
257,258
and other groups have worked to develop cross-linking reagents that
improve ETD performance.
259,260
Figure 7
Strategy to identify
cross-linked peptides. (a) Schematic structure
of a DSSO interpeptide cross-link (left) and its specific fragmentation
pattern under CID (right). The four signature MS/MS fragment ions
are derived via the equation presented below the structure (i.e.,
the Δm principle). (b) The XlinkX workflow
to identify interpeptide cross-links. The MS precursor ion is subjected
to sequential CID-ETD fragmentation. Only CID spectra are used to
obtain the precursor masses of both linked peptides by the Δm principle. The four
signature fragment ions resulting
from cross-linker cleavage are represented by purple peaks in the
MS/MS fragmentation spectra. Subsequently, CID spectra are used to
match b- and y-ions, and ETD spectra are used to search for c- and
z-ion series. Reprinted by permission from Macmillan Publishers Ltd.:
NATURE, Liu, F.; Rijkers, D. T. S.; Post, H.; Heck, A. J. R. Nat. Methods
2015, 12, 1179–1184
(ref (255)). Copyright
2015.
Ion Mobility and Native
Proteomics
As noted above,
ion mobility separates ions based on their movement through a carrier
gas, and complementing ion mobility with ETD provides access to detailed
higher-order structures of peptides, proteins, and protein complexes.
Lermyte and Sobott used a q-IM-TOF instrument to investigate native
proteins and protein complexes with ETD.
261
They demonstrated that post-ion-mobility collisional activation
of ETD products could release noncovalently bound product ions to
improve sequence coverage of exposed regions of proteins, while pre-ETD
activation of tetramer complexes caused unfolding without monomer
ejection that showed efficient fragmentation in some regions which
are not sequenced under more gentle MS conditions (Figure 8
). They also used incremental
increases in pre-ETD collision-activation to trace initial steps of
gas-phase protein unfolding. Sobott and co-workers also used ion mobility,
ETD, and supplemental activation to study the topology of the native
form of tetrameric alcohol dehydrogenase (∼150 kDa) by showing
that regions of ETD fragmentation mapped to exposed regions on the
complex.
262
Furthermore, they used ion
mobility separations to study the stability of ETnoD complexes, showing
that more linear complexes are more likely to release c/z-type products
(which has implications for understanding supplemental activation
in ETD reactions at the intact protein level).
263
Figure 8
Using ETD and ion mobility to discern structure. (A) Charge state
distribution observed in native ESI of the ADH tetramer, with the
26+ charge state isolated in the quadrupole for subsequent top-down
dissociation. (B) ETD products at a sampling cone voltage of 40 V
and a supplemental activation of (top) 10 V and (bottom) 70 V applied
in the transfer cell and (C) at sampling cone voltage 120 V and supplemental
activation 10 V. (D) Crystal structure of ADH tetramer (only one subunit
shown for clarity). The fragmentation sites observed with a sampling
cone of 80 V, without supplemental activation, are shown in red. The
additional cleavage sites observed at a higher sampling cone voltage
(causing partial unfolding) are shielded in the native structure and
shown in blue. Reproduced from Electron transfer dissociation provides
higher-order structural information on native and partially unfolded
protein complexes Lermyte, F.; Sobott, F. Proteomics, Vol. 15, pp 2813–2822 (ref (261)).
Copyright 2015 Wiley.
Combinations of ETD and collisional
activation have been valuable
in studying electrostatic interactions (i.e., salt bridges) in native
protein structures too. In this approach, differences between pre-
or postreaction collisionally activated ETD spectra and nonactivated
ETD spectra indicate presence of electrostatic interactions in standard
proteins where known salt bridge contacts in solution occur.
264,265
ETD of native proteins complexes can also identify differences in
fragmentation between subunits that arise from asymmetric charge portioning,
in addition to characterizing domains of secondary-structure present
in dimers, ejected monomers, and monomers obtained directly from electrospray
ionization.
266
Other approaches can
be used in tandem with ETD to investigate
protein structures. Kaltashov and co-workers coupled online ion exchange
chromatographies with native electrospray ionization to characterize
complex and heterogeneous therapeutic proteins and protein conjugates
with intact conformational integrity, using ETD to provide online
top-down structural analysis for identification PTMs that could not
be identified via intact mass analysis alone.
267
Cassou et al. showed that electrothermal supercharging
can increase the precursor ion charge state distributions for proteins
in buffered aqueous solutions used for native electrospray, and the
higher charge state species facilitated collection of high-quality
ETD spectra for structural analysis of intact proteins.
268
Xie and Sharp used online size exclusion chromatography
to ensure that peptide isomers eluted together and then quantified
relative amounts of each isomer based on the presence of fragment
ions in a single ETD MS/MS spectrum.
269
Hydrogen–Deuterium Exchange
Hydrogen–deuterium
exchange (HDX) is a noncovalent labeling approach for mapping protein
structure based on the principle that hydrogens on amino acid residues
at solvent-exposed regions of the protein backbone will exchange with
deuterium in D2O, while amino acids that are buried in
the folded protein core (or by an interacting protein) will not. The
mass difference in exchanged deuterium atoms can then be measured
by MS, with site-specific assignment enabled by tandem MS. Collision-based
dissociation of HDX-labeled peptides and proteins can cause hydrogen
scrambling that obscures assignment of heavy labeled residues, while
ETD minimized hydrogen scrambling to enable measurement of deuterium
incorporation with single-residue resolution.
270−273
HDX strategies that leverage ETD are used at both the peptide
and protein level. Masson et al. used HDX-ETD in a bottom-up strategy
to screen inhibitors of the oncogene phosphoinositide 3-kinase (PI3K)
catalytic p110α subunit, highlighting its potential use in pharmaceutical
development for screening therapeutics.
274
Seger et al. engineered new disulfide bonds in human growth hormone
and used HDX-ETD to investigate conformational and functional consequences
of new bond positions, showing how a different disulfide bond could
stabilize the protein.
275
HDX-ETD has also
played a valuable role in determination of site-specific changes in
enzyme activities (like coagulation factors) upon cofactor binding,
276
structural changes upon binding of epidermal
growth factor receptor inhibitors in cancer treatments,
277
and oligomerization of apolipoprotein E, which
can be a risk factor in Alzheimer’s disease.
278
In an interesting study, Borchers and co-workers showed
how specific phosphorylation events can affect protein structure using
top-down HDX-ETD experiments, providing an avenue to study how PTMs
affect protein activity and binding via structural changes.
279
HDX can be combined with ion mobility
(and subsequent beam type
supplemental activation) with minimal hydrogen scrambling to expand
the structural information obtained in an experiment.
280,281
Interestingly, Rand et al. showed that this combination allows gas-phase
HDX inside a mass spectrometer.
282
The
reaction occurs in milliseconds, offers complementary information
to solution phase HDX and can be combined with ETD to provide orthogonal
modes of structure characterization on the chromatographic time scale.
Note, a combination of HDX and ETD also enabled analysis of the structure
of antibodies in middle-down and top-down approaches discussed above.
244,248,252
Beyond HDX, ETD has been used
with other surface labeling techniques, as well,
283−285
although HDX is the more common approach to use ETD because of the
hydrogen scrambling concerns with collisional activation.
Other
Uses of ETD and Related Ion–Ion Reactions
Ion–ion
reactions like ETD can be used for proteome characterization
in unique ways beyond those discussed above. Many of these approaches
involve the ability to generate radical peptide cations and radical
product ions that can be further probed to elucidate peptide structural
information via radical ion chemistry. The body of work involving
various manipulation of peptide radical cations for analytical means
is extensive and technical, making it difficult to sufficiently review
it in the space available here. Instead, we discuss specific applications
peptide radical chemistry that we find particularly interesting and
useful, and we point readers to recent reviews for further information.
9,286,287
We also describe other uses
of ETD and related ion–ion reactions for proteomics and briefly
cover how ETD is coupled with quantitative strategies.
One valuable
nontraditional application of ETD is for differentiating
peptide isomers and isomeric residues.
288
Lebedev et al. and Xiao et al. both recently described methods to
perform HCD on z•-type generated from ETD to create
diagnostic w-ions for distinguishing leucine and isoleucine residues
in peptides.
289,290
Lebedev and co-workers further
extended this study to EThcD methods for more globally applicable
leucine/isoleucine discrimination,
291
although
they also described limitations and considerations when using these
approaches.
292
Lyon et al. recently showed
that ETD reactions can identify peptide isomers through a strategy
that leverages hemolytic cleavage of carbon–iodine bonds that
drives radical directed dissociation upon supplemental activation.
293
Additionally, Turacek and colleagues have used
radical product ions, created via ETD, to extensively study structure
and ion chemistry of peptides with a variety of approaches for theoretical
and analytical purposes.
294−300
Proton
Transfer and Other Ion–Ion Reactions
As mentioned
in the Principles of ETD section,
the choice of reagent anion used in ion–ion reactions governs
the amount of electron transfer versus proton transfer between the
anion and precursor cation populations. Proton transfer reactions
(PTR), pioneered by McLuckey, Stephenson, and co-workers, have several
analytical uses in proteomics, including signal concentration into
a small distribution of charge states and simplification of spectra
of highly charged precursors.
301−304
Building on the work of Stephenson and McLuckey,
Hunt and co-workers showed in 2005 that ETD and PTR are a powerful
combination for sequencing intact proteins, using ETD to drive sequence
informative product ion generation and then performing PTR on highly
charged ETD products to decongest spectra for straightforward product
ion assignments.
17
The Hunt group has recently
expanded on this sequential ion–ion reaction approach to Orbitrap
systems with front end ETD sources, using ion–ion proton transfer
reactions (PTR) to simplify ETD spectra and to disperse fragment ions
over the entire mass range in a controlled manner (Figure 9
).
305
Additionally, multiple fills of ETD-PTR product ions can be collected
prior to mass analysis in this approach, considerably enhancing observed
ion current and product ion S/N without the need for time-consuming
averaging of data from multiple mass measurements. The ETD/PTR technique
proved extremely valuable in middle down analyses of antibodies,
243
and they further extended its utility in a
recent top-down characterization of histones by incorporating parallel
ion parking
35,306
during the PTR reactions to
specifically control reaction of precursor ions to remain in a targeted
product m/z range.
237
The value of these strategies is clear for analysis of
large peptides and proteins, and we expect this technology will be
widely implemented in top-down experiments in coming years. Unsurprisingly,
similar approaches have recently been extended to combinations of
UVPD and PTR reactions for top-down analyses of denatured and native
proteins.
307,308
We note PTR has proven useful
for gas-phase purification in bottom-up quantitative proteomic experiments
as well.
309,310
Figure 9
ETD and PTR reactions (labeled here as
ion–ion proton transfer,
IIPT) combined for top-down intact protein analysis. (A) ETD spectrum
recorded on (M + 26H)26+ ions from apomyoglobin using a
reaction time of 5 ms. (B–E) Spectra obtained by performing
PTR reactions on the ETD fragment ions in (A) for 20, 40, 80, and
160 ms, respectively. Reprinted from Int. J. Mass Spectrom., Vol. 377, Anderson, L.
C.; English, A. M.; Wang,
W.-H.; Bai, D. L.; Shabanowitz, J.; Hunt, D. F. A Protein derivatization
and sequential ion/ion reactions to enhance sequence coverage produced
by electron transfer dissociation mass spectrometry, pp 617–624
(ref (305)). Copyright
2015, with permission from Elsevier.
PTR can also aid in structural proteomic experiments. Bush
and
co-workers have used PTR reactions to reduce the charge states of m/z-selected, native-like
ions of proteins
and protein complexes, which helps interpretation of complicated mass
spectra that often represent contributions from multiple, coexisting
species.
311
Since its introduction for
this purpose in 2015, PTR has enabled several detailed studies of
gas-phase protein structure, folding, and dynamics.
312−314
Jhingree et al. performed similar experiments but used ETnoD products
instead of PTR products to study the effects of charge reduction on
protein structure.
315
Moreover, the Sobott
group investigated how ETD and PTR can be balanced to generate sequencing
ions and spectral decongestion.
316
McLuckey and co-workers have recently used a variety of gas-phase
ion–ion chemistry approaches for less traditional proteomic
applications.
317
Such methods include converting
cations to anions in the gas-phase,
318
performing
1,3-dipolar cycloadditions between azides and alkynes (click chemistry)
through ion–ion reactions,
319
creating
dehydroalanine residues that provide specific backbone cleavages in
peptide and protein cations
320,321
and mapping of cysteine
modification states and disulfide bonds through ion–ion oxidation
reactions.
322,323
Of particular note, they demonstrated
the ability to form peptide bonds in the gas phase, providing a unique
means for generating peptide linkages that is fast (<1 s), efficient
(tens of percent), and flexible.
324,325
Ion–ion
reactions were also harnessed by Brodbelt and co-workers to derivative
peptides with 4-formyl-1,3-benezenedisulfonic acid (FBDSA) anions
to improve collisional dissociation of phosphopeptides and UVPD fragmentation
efficiencies.
326,327
Negative Electron Transfer
Dissociation
Negative electron
transfer dissociation (NETD) is the negative mode analogue of ETD,
where precursor anions are sequenced through reactions with radical
reagent cations. In NETD, radical reagent cations oxidize peptide
precursor anions by abstracting an electron, which drives electron
rearrangement steps that promote cleavage of the C–Cα backbone bond and produce odd-electron
a•
- and even-electron x-type product ions.
328
Analysis of peptide anions offers several benefits, including
the ability to sequence acidic peptides that do not readily ionize
upon positive mode electrospray ionization
329−332
and access to acid-labile PTMs that are difficult to characterize
via positive mode approaches.
NETD is a notably valuable technique
for peptide anion analysis because collisional activation often fails
to reproducibly provide sequencing information in MS/MS spectra, making
electron-driven fragmentation one of the main approaches to characterizing
peptides in the negative mode. In fact, NETD enabled the first large-scale
negative mode analysis of the proteome and demonstrated that systematic
analysis of peptide anions at a proteome scale was a viable approach.
333
NETD spectra often contain more side-chain
neutral losses than their positive mode ETD counterparts, but the
losses can have diagnostic value.
334
Specific
cleavages in NETD can also be used to create diagnostic fragments
for peptide anion identification, such as enhanced c/z-type ion formation
N-terminal to tyrosine residues.
335
NETD suffers from nondissociative negative electron transfer (NETnoD)
to a similar or even more severe degree than ETD for low charge-density
precursors.
336
We and others have demonstrated
that supplemental activation for NETD using concurrent photoactivation
(AI-NETD) is superior to collision-based supplemental activation (NETcaD)
(Figure 10
),
337,338
and AI-NETD allowed characterization of nearly the entire yeast
proteome exclusively in the negative mode
338
and the largest study of the human proteome via peptide anion analysis.
339
The radical cation of fluoranthene is a suitable
NETD reagent for proteomic and phosphoproteomic analysis,
340
but we recently demonstrated that the radical
SF5
•+ cation can be a preferred NETD
reagent instead of fluoranthene as it decreases NETnoD, improves spectral
quality through the generation of more a•- and x-type
product ions, and increases peptide identifications in LC–MS/MS
analyses.
341
Although AI-NETD is generally
a better strategy to improve NETD analyses, the use of SF5
•+ reagent is a valuable approach when instrument
modifications for AI-NETD are not available. Note, these studies also
required modification of spectral interpretation tools, like the Open
Mass Spectrometry Search Algorithm (OMSSA) and Byonic, to assign peptide
spectral matches to NETD spectra.
333,339
Figure 10
Fragment
map of peptides identified with both NETD and AI-NETD.
Here, each row is a unique peptide and each subcolumn corresponds
a peptide backbone bond. The numbers in parentheses to the left show
peptide length in number of residues, and all peptides shown here
are z = −2. With NETD, a•- and x-type fragments decrease in number and intensity as
precursor
charge density decreases (i.e., as peptide length increases). AI-NETD
maintains superior fragment ion generation even with decreasing precursor
charge density, greatly increasing peptide dissociation and sequence
coverage compared with NETD. This research was originally published
in Molecular & Cellular Proteomics. Riley, N. M.; Rush, M. J.
P.; Rose, C. M.; Richards, A. L.; Kwiecien, N. W.; Bailey, D. J.;
Hebert, A. S.; Westphall, M. S.; Coon, J. J. The Negative Mode Proteome
with Activated Ion Negative Electron Transfer Dissociation (AI-NETD). Mol. Cell. Proteomics
2015, 14, 2644–266 (ref (330)). Copyright the American Society for Biochemistry and
Molecular
Biology.
NETD has currently been implemented
on a handful of MS systems
but has yet to be commercialized. Although it has admittedly niche
applications compared to positive mode ETD analyses, the availability
of NETD on commercially available platforms in the coming years, in
addition to the NETD-compatible informatics tools now available, will
boost its application to more biological problems (e.g., labile PTMs
like tyrosine sulfation
342
and histidine
phosphorylation
343
) and, thus, its impact
on proteomics. It is also worth noting that NETD has proven valuable
beyond the realm of proteomics, as well, benefiting MS-based oligonucleotide
and glycosaminoglycan analyses.
344−351
Quantitative Proteomic Strategies
With the various
implementations and applications of ETD, it is important to be able
to leverage common quantitative proteomic approaches while using ETD
for characterization. Perhaps the most straightforward quantitative
approach is label-free quantitation, which has been a component of
several ETD experiments discussed above.
85,120,203,234,236
Stable isotope labeling is another
widely used approach to enable multiplexed quantitation, and several
strategies can be employed to incorporate stable isotopes into samples.
352
ETD is readily coupled with MS1-based
labeling approaches, including dimethyl labeling, stable isotope labeling
in amino acid cell culture (SILAC), and neutron-encoded (NeuCode)
SlLAC.
233,353−357
In NeuCode SILAC, quantitative channels
are spaced very closely together (several to tens of mDa) to allow
controlled masking or revealing of quantitative peaks based on resolving
power. This is typically employed at the MS1 level, but
quantitative information can also be revealed in high-resolution ETD
MS/MS spectra of both peptides
356
and proteins.
357
Isobaric labels, e.g., tandem mass tags
(TMT) and isobaric tag for relative and absolute quantification (iTRAQ),
are another approach to chemical labeling that enable mutliplexed
quantitative comparisons at the MS/MS level. iTRAQ labels were first
shown to be compatible with ETD, although the amount plexing was limited
and required a second collisional activation step to generate the
fully array of reporter ions for quantitative comparisons.
358,359
TMT labels were modified to accommodate the fragmentation pathways
of ETD, a process that required switching heavy carbon and nitrogen
atoms to account for relocation of the proximal heavy carbon from
the reporter ion to the balance region upon ETD fragmentation.
360
This modification serendipitously led to the
discovery that neutron-encoded mass differences could be exploited
to increase the plexing of TMT reagents, and this phenomenon also
led to the rise of the NeuCode SILAC approach.
361
Just this year Li and co-workers investigated EThcD for
TMT multiplexed quantification, looking for sequence information and
reporter ion intensities for quantitative comparisons in the same
MS/MS spectra in global proteomic and phosphoproteomic experiments.
362
They showed through careful optimization and
balancing of parameters that EThcD could generate extensive sequence-informative
product ions while also preserving the presence of reporter ion signal
and that this improved analyses over a combination of separate ETD
and HCD scans.
New Informatic Tools for ETD and Related
Technologies
In addition to improvements in ETD instrumentation
and methodology,
development of informatics tools to process ETD data is crucial to
its utility in proteomics. Statistical analyses of ETD spectra have
provided insight into how to design search algorithms to properly
assess ETD spectra,
363,364
and a number of widely available
search engines exist to process ETD spectra.
365
Incorporation of ETD search capabilities into flexible pipelines
that offer many dimensions of data analysis (e.g., MaxQuant
366
and OpenMS
367
) is
important and enables straightforward integration of ETD methods with
a variety of other tools. Moreover, resources like the Web-based tutorial
recently presented by Hunt and colleagues for training in ETD spectral
interpretation are valuable in making ETD methods more broadly accessible
to a growing proteomics community.
368
ProteomeTools,
a project that analyzed >330 000 synthetic tryptic peptides
representing essentially all canonical human gene products with ETD,
ETcaD, and EThcD (in addition to collisional dissociation approaches),
represents another fantastic resource for development of software
tools ranging from intelligent decision-tree acquisition routines
to search engines that will benefit ETD-based analyses and proteomic
efforts in general.
369
Even with
the improvements made in ETD spectral interpretation,
an important and often overlooked strategy to improve scoring of ETD
spectra is the removal of interfering ions prior to searching.
370−372
Charge-reduced precursor ions and neutral loss peaks resulting from
intact peptide radical products in ETD are present
373
but do not contribute sequence information, and they can
be systematically cleaned from spectra while retaining sequence informative
c/z•-type product ions. This can make a considerable
difference in peptide identifications because presence of unexplained
signal can penalize scoring in many search algorithms that only look
for sequencing ions derived from expected peptide backbone cleavages.
Development of new algorithms to integrate ETD spectra into de novo sequencing strategies
and PTM analyses are also
emerging. The majority of de novo sequencing approaches,
such as pNovo+, rely on pairing complementary dissociation from HCD
and ETD spectra.
374−376
NovoExD, recently described by Yan et al.,
enables de novo sequencing from ETD (or ECD) spectra
alone,
377
and the increasing prevalence
of supplemental activation methods like EThcD and AI-ETD may drive
development of similar approaches that do not require HCD complements.
Glycoproteomic applications are another active arena for development
of ETD-enabled informatic tools. In 2013, Desaire and co-workers introduced
GlycoPep Detector, a Web-based tool designed to identify intact N-glycopeptides
from ETD spectra. Their glycoproteomic search engine applies filtering
functions followed by correlation of glycopeptide compositions with
the ETD spectra and intensity-weighted scoring based on independent
assessment of multiple ion series (c-, z-, and y-type ions). They
followed GlycoPep Detector with another tool called GlycoPep Evaluator
to improve false discovery rate calculations in ETD-based intact glycopeptide
analysis.
378
Around the same time, Tang
and co-workers described a search strategy for N-glycopeptides that
combined scoring of collision-based dissociation spectra and ETD spectra
to identify intact glycopeptides from complex samples (e.g., human
serum).
379
Most recently, Lee et al. described
the benefits of the Byonic search algorithm for automated N-glycopeptide
profiling,
380
and although they did not
use ETD in that particular study, Byonic has been widely used in recent
glycoproteomic experiments, including several of those discussed above.
Investigations into ETD spectra of O-glycopeptides specifically
have improved data interpretation. Darula et al. showed that altering
scoring of O-glycopeptide ETD spectra with Protein Prospector can
sizeably improve identifications by weighting product ion scores based
on several spectral features.
381
Zhu et
al. conducted a similar study on ETD spectra from O-glycopeptides,
finding that flexible scoring of c- or z-type ion series based on
the precursor ion and inclusion of multiply charged c/z-type product
ions can significantly benefit glycopeptide identifications.
382
Beyond spectral scoring, the databases used
for querying spectra can influence data quality and Chalkley and Baker
recently employed a reference glycosite database based on known glycosites
to improve identification of glycopeptide ETD spectra for both N-
and O-glycopeptides.
383
Informatics
tools for analysis of ETD-based top-down proteomic
data continue to improve as well. Kelleher et al. launched the freely
available ProSight Lite in 2015, which is a simple and intuitive platform
to characterize proteoforms based on MS/MS spectra.
384
ProSight Lite is compatible with a range of dissociation
methods, including ETD, EThcD, and AI-ETD, and it has provided a straightforward,
flexible alternative to the commercially available ProSight software.
MASH Suite Pro, created by Ge and co-workers, is a freely available
comprehensive software package for top-down proteomics that is capable
of processing high-resolution MS and MS/MS data (including ETD) using
two deconvolution algorithms, enables PTM and sequence variant characterization,
and provides relative quantitation of multiple proteoforms in different
experimental conditions.
385
Several other
recently developed top-down proteomics software platforms, including
Informed-Proteomics and Protein Goggle, can process ETD spectra as
well.
386−388
In addition to development of these
data analysis pipelines, new
strategies for spectral interpretation of MS/MS spectra of intact
proteins are being developed. The C-score, for example,
assesses proteoform identification and characterization to improve
how both collisional dissociation and ETD spectra are interpreted
in high-throughput top-down proteomic experiments.
389
Sobott and co-workers recently introduced a method to examine
the prevalence of ETD and PTR products ion MS/MS spectra of intact
proteins, providing both an avenue to compare reaction conditions
on different instrument platforms and a strategy for understanding
how different reaction conditions and supplemental activation schemes
effect product ions.
390
They have also
used relative quantities of fragmentation products in top-down ETD
spectra to estimate ETD reaction rates on highly charged intact protein
cations.
391
Finally, Pevzner and co-workers
described de novo sequencing approaches for intact
proteins based on generating high-quality sequence tags, both with
combinations of bottom-up and top-down spectra
392
and with only intact protein MS/MS spectra.
393
Looking Forward
The application
of ETD to a diverse array of biological questions
highlights its impact on proteome characterization. Its ease of implementation
on a variety of instrument platforms and a concerted effort among
many researchers to advance ETD methodology for widespread use laid
a foundation for this success. As more complex questions about the
proteome come to the forefront, ETD is poised to play an important
role in research ranging from epigenetic and post-translational regulation
of health and disease to the roles of protein–protein interactions
and higher order protein structure in molecular biology.
ETD
technology is fundamentally dependent on continued progress
in MS instrumentation, and its utility will expand as mass spectrometers
generally improve in sensitivity and resolution. ETD will specifically
benefit from advances in robust reagent ion sources, data acquisitions
schemes to improve product ion S/N (e.g., the multiple fill approach
and spectral decongestion via PTR), and supplemental activation implementations,
all of which have been active areas of development in recent years.
ETD is particularly valuable when coupled with collisional dissociation,
especially in real-time decisions tree strategies for characterizing
modified peptides and intact proteins in complex samples, and the
implementation of ETD on the newest generations of hybrid instruments
will only extend this further. From our perspective, supplemental
activation methods, especially EThcD and AI-ETD, are the future of
ETD as well, as they offer substantial benefits with few drawbacks.
We expect that the majority of ETD research in the coming years will
use supplemental activation, especially as the various methods become
default options on commercially available instruments.
One intriguing
area of growth for ETD is in reaction cell design
for both speed and S/N considerations. Collision-based fragmentation
has more markedly benefitted from the speed and parallelized acquisition
schemes on new instruments thus far, but little work has focused on
how to optimize ETD reaction cell design to capitalize on these faster
instruments. Improved reaction cells will not only minimize ion–ion
reaction times, but they will also allow efficient storage of large
precursor and reagent ion populations and provide easy access to supplemental
activation methods, ideally infrared photoactivation for concurrent
rather than postreaction activation. Most of the reaction cells in
use today were described in the early years after the introduction
of ETD, and it is now high time for a concentrated effort in improving
the speed at which ETD spectra can be acquired.
In all, ETD
is arguably the most valuable and widely used alternative
dissociation method in peptide and protein characterization, and its
utility is expanding to other biomolecules as well. The ETD community
is active and continues to grow, especially as more ETD-enabled instruments
make it into laboratories around the world. Success will depend on
a sustained push to improve hardware and data acquisition strategies,
and these technologies must reach beyond the expert laboratories and
developers of ETD methodologies. We foresee ETD contributing to new
insights in a wide swath of proteomic experiments in the coming years
as it further expands to users across the realm of proteome research.