TECHNOLOGICAL INNOVATION AND RESOURCESSystematic Optimization of Automated
Phosphopeptide Enrichment for High-Sensitivity
Phosphoproteomics
Authors
Patricia Bortel, Ilaria Piga, Claire Koenig, Christopher Gerner, Ana Martinez-Val, and Jesper V. OlsenCorrespondence Graphical Abstract2024, Mol Cell Proteomics 23(5), 100
© 2024 THE AUTHORS. Published b
Molecular Biology. This is an open a
licenses/by/4.0/).
https://doi.org/10.1016/j.mcpro.2024ana.martinezdelval@cnic.es;
jesper.olsen@cpr.ku.dk
In Brief
Optimization of experimental
parameters for automated on-
bead phosphoproteomics
sample preparation for low-input
samples was conducted. Key
factors assessed included
glycolic acid concentration,
ammonium hydroxide
percentage, peptide-to-bead
ratio, binding time, and buffer
volumes. Using 30 μg of
peptides, 16,000
phosphopeptides were
confidently identified within
30 min on an Orbitrap Exploris
480, or 32,000 phosphopeptides
when using Orbitrap Astral and
0.5 million HeLa cells. An
alternative enrichment strategy
based on sequential enrichment
with stepwise addition of beads
boosted coverage by 20%.Highlights
• Systematic optimization of parameters for automatized phosphopeptide enrichment.
• Sequential enrichment and pooling enhance phosphoproteome sensitivity.
• Sequential enrichment by stepwise addition of beads boosts phosphoproteome coverage by 20%.
• Optimized workflow on Orbitrap-Astral yields 32,000 phosphopeptides from 0.5 M cells.754
y Elsevier Inc on behalf of American Society for Biochemistry and
ccess article under the CC BY license (http://creativecommons.org/
.100754
TECHNOLOGICAL INNOVATION AND RESOURCESSystematic Optimization of Automated
Phosphopeptide Enrichment for High-Sensitivity
Phosphoproteomics
Patricia Bortel1,2 , Ilaria Piga3 , Claire Koenig3, Christopher Gerner1,4 ,
Ana Martinez-Val3,* , and Jesper V. Olsen3,*Improving coverage, robustness, and sensitivity is
crucial for routine phosphoproteomics analysis by
single-shot liquid chromatography-tandem mass spec-
trometry (LC-MS/MS) from minimal peptide inputs. Here,
we systematically optimized key experimental parame-
ters for automated on-bead phosphoproteomics sample
preparation with a focus on low-input samples.
Assessing the number of identified phosphopeptides,
enrichment efficiency, site localization scores, and
relative enrichment of multiply-phosphorylated peptides
pinpointed critical variables influencing the resulting
phosphoproteome. Optimizing glycolic acid concentra-
tion in the loading buffer, percentage of ammonium
hydroxide in the elution buffer, peptide-to-beads ratio,
binding time, sample, and loading buffer volumes
allowed us to confidently identify >16,000 phosphopep-
tides in half-an-hour LC-MS/MS on an Orbitrap Exploris
480 using 30 μg of peptides as starting material.
Furthermore, we evaluated how sequential enrichment
can boost phosphoproteome coverage and showed that
pooling fractions into a single LC-MS/MS analysis
increased the depth. We also present an alternative
phosphopeptide enrichment strategy based on stepwise
addition of beads thereby boosting phosphoproteome
coverage by 20%. Finally, we applied our optimized
strategy to evaluate phosphoproteome depth with the
Orbitrap Astral MS using a cell dilution series and were
able to identify >32,000 phosphopeptides from 0.5 million
HeLa cells in half-an-hour LC-MS/MS using narrow-
window data-independent acquisition (nDIA).
Protein phosphorylation is a highly dynamic post-trans-
lational modification (PTM) that plays a critical role in
regulating cellular signal transduction pathways. Protein
phosphorylation has been the objective of extensive studiesFrom the 1Faculty of Chemistry, Department of Analytical Chemistry, a
Vienna, Vienna, Austria; 3Proteomics Program, Faculty of Health a
Research, University of Copenhagen, Copenhagen, Denmark; 4Joint
Vienna, Vienna, Austria
*For correspondence: Ana Martinez-Val, ana.martinezdelval@cnic.es; Je
Present address for Ana Martinez-Val: Centro Nacional de Investigacion
© 2024 THE AUTHORS. Published by Elsevier Inc on behalf of American Society for Bio
This is an open access article under the CC BY license (http://creativecommons.org/liceby the mass spectrometry (MS)-based proteomics commu-
nity, with the analysis of thousands of phosphorylation sites
across different cellular contexts (1–5). However, there is
still ongoing research to try to identify the functionality of
most of them (6) since it has been suggested that 75% of
the proteome might be phosphorylated (7). Quantitative
mass spectrometry has proved to be the best platform for
retrieving large-scale information about the identification,
quantification, and localization of phosphorylation sites in
complex systems (5). The very deep proteomes described
so far (8) or the recent development of highly sensitive mass
spectrometers (9) indicate the potential to explore the
phosphoproteome without the need for specific enrichment
of phosphopeptides prior to MS analysis. However, the
capacity to explore the functional phosphoproteome is still
impacted by the relatively low abundance of phosphorylated
peptides and their sub-stoichiometric nature (10). Therefore,
to this day, deep phosphoproteomics relies on the enrich-
ment of phosphopeptides prior to LC-MS/MS analysis. In
this regard, the phosphoproteomics technology has taken a
significant leap in recent years with significant improve-
ments in sensitivity, making it possible to perform phos-
phoproteomics analysis of minute samples (11), even as low
as single spheroids (12). Moreover, the incorporation of
magnetic beads into workflows on automated sample
preparation platforms nowadays allows for robust high-
throughput sample preparation, making phosphoproteo-
mics applicable to large-scale studies (13–15) and clinical
sample cohorts (16). Alternatively, novel high-throughput
methods combining suspension trapping on micro-columns
with affinity material tips have also allowed for stream-
lining the sample preparation workflows for phosphopro-
teomics (17).nd 2Vienna Doctoral School in Chemistry (DoSChem), University of
nd Medical Sciences, Novo Nordisk Foundation Center for Protein
Metabolome Facility, University of Vienna and Medical University of
sper V. Olsen, jesper.olsen@cpr.ku.dk.
es Cardiovasculares Carlos III (CNIC), Madrid 28029, Spain.
Mol Cell Proteomics (2024) 23(5) 100754 1
chemistry and Molecular Biology.
nses/by/4.0/). https://doi.org/10.1016/j.mcpro.2024.100754
TABLE 1
Overview of supplemental files
Experiment Supplemental file
Bead volume S1
Bead binding time S2
GA in loading buffer S3
NH4OH in elution buffer S4
Peptide input amount S5
Sample volume S6
Loading buffer volume S7
Sequential enrichment, new beads S8
Sequential enrichment, 6 rounds S9
Sequential enrichment, pooled fractions S10
Sequential enrichment, increasing beads S11
HeLa dilution, Astral S12
HeLa phosphosite annotations S13
Systematic Optimization for PhosphoproteomicsCurrently, the most popular phosphopeptide enrichment
strategies rely on affinity-based methods either by immobi-
lized metal affinity chromatography (IMAC) (18), or metal oxide
affinity chromatography (MOAC) (19, 20). Both IMAC and
MOAC strategies rely on transition metals (Ti, Zr, Fe) that,
either chelated on a substrate (Ti-IMAC or Zr-IMAC), or as
metal oxides (TiO2), enable the selective binding of phos-
phopeptides. The effectiveness of these strategies relies on
multiple factors, including the ratio of peptide-to-beads
(binding capacity) (21), the loading buffer composition (bind-
ing conditions) (22), the washing buffer composition, and
sample-bead binding time, among others (23, 24). For
instance, the inclusion of competitive non-phosphopeptide
binders, such as glycolic acid (GA) or 2,5-dihydoxybenzoic
acid (DHB) in the binding buffer, as well as a high concen-
tration of trifluoroacetic acid (TFA), can minimize the binding of
highly-acidic or sialic-acid containing peptides (25–27),
significantly increasing the phosphopeptide enrichment effi-
ciency of IMAC and MOAC strategies. Moreover, different
strategies based on the binding of phosphopeptides to the
metal matrix have been presented to increase the depth and
diversity of the purified phosphopeptide population. For
instance, Thingholm et al. (28) showed that multiply-
phosphorylated peptides can be separately purified by
sequential elution, based on the higher affinity between the
metal matrix and peptides with several phosphate groups. On
the other hand, the combination of different bead types has
been suggested as a method to enrich complementary sets of
phosphopeptides based on their multiplicity and acidity (29).
Moreover, with the advent of data-independent acquisition
(DIA) strategies and the implementation of software tools
capable of analyzing phosphoproteomics datawithout the need
for spectral libraries (30, 31), the depth obtained from single-
shot phosphoproteomics analyses has increased significantly.
The combination of short LC gradients with single-shot DIA
nowadays provides analysis of deep (phospho)proteomes in a
high-throughput manner. In this regard, the improvements
achieved by technical sample preparation optimizations have
been overshadowed by the increased sensitivity and coverage
of detected phosphopeptides by DIA approaches.
In this work, we systematically evaluated how key experi-
mental parameters affect phosphopeptide enrichment with a
special focus on low-input samples, which were subsequently
analyzed using DIA. We specifically tested the impact of using
different (i) bead-to-peptide ratios, (ii) sample-bead-binding
times, (iii) concentrations of glycolic acid in the loading
buffer, (iv) percentages of ammonium hydroxide (NH4OH) in
the elution buffer, (v) peptide input amounts, (vi) sample vol-
umes and (vii) loading buffer volumes. To identify the optimal
phosphopeptide enrichment conditions, the effects of the
different parameters were evaluated in terms of phospho-
peptide enrichment efficiency, relative purification of multiply-
phosphorylated peptides, and coverage of well-localized
phosphosites (class I phosphosites, localization probability2 Mol Cell Proteomics (2024) 23(5) 100754>0.75) as a proxy for the quality of the DIA-MS/MS spectra.
Then, based on the optimized experimental parameters, we
devised different strategies to increase the phosphoproteome
depth of the analysis by sequential enrichment strategies,
either by repetitive enrichment using the non-bound fraction or
by modifying the peptide-to-bead ratio. Finally, we applied our
optimized strategy for the phosphoproteomics analysis of a
cell dilution series on a state-of-the-art Orbitrap-Astral MS to
determine the limits of deep phosphoproteomics analysis with
strongly downscaled cell input amounts.EXPERIMENTAL PROCEDURES
Experimental Design and Statistical Rationale
All experiments in this manuscript were performed as experimental
replicates. For method optimization on the Orbitrap Exploris, all experi-
ments (fromphosphopeptide enrichment toMSanalysis)wereperformed
in triplicates (n = 3). For the cell dilution experiment on theOrbitrap Astral,
all experiments (from cell lysis to MS analysis) were performed in qua-
druplicates (n=4). For theEGF treatmentexperiment, theexperimentwas
performed using 12 experimental replicates. An overview of all
Supplementary Files with experimental data can be found in Table 1.
Cell Culture and Cell Lysis
A549 and HeLa cells were cultured in P15 dishes in DMEM (Gibco,
Invitrogen) supplemented with 10% fetal bovine serum (FBS, Gibco)
and 100 μg/ml streptomycin (Invitrogen) until 90% confluence was
reached.
A549 cells were washed twice with phosphate-buffered saline
(PBS) (Gibco, Life Technologies) and lysed using 600 μl 95 ◦C hot lysis
buffer (5% sodium dodecyl sulfate (SDS), 5 mM Tris(2-carboxyethyl)
phosphine (TCEP), 10 mM chloroacetamide (CAA), 100 mM Tris pH
8.5). Cells were scraped and collected in a falcon tube and the lysate
was incubated at 95 ◦C, 500 rpm for 10 min.
HeLa cells were first detached with 0.05% trypsin-EDTA (Gibco,
Invitrogen) and counted via a trypan blue viability assay (10 μl of cell
suspension was diluted in 1:1 (v/v) ratio with trypan blue stain 0.4% (v/
v)) by using an automated cell counter (Corning). For cell count esti-
mation, the average count of five images acquired with the CytoS-
MART software was calculated. Cell dilutions with respectively; 1 ×
106, 0.5 × 106, 0.2 × 106, 0.1 × 106, 0.05 × 106, and 0.01 × 106 cells
Systematic Optimization for Phosphoproteomicswere collected in Eppendorf tubes with four replicates for each con-
dition. Cell pellets were washed with PBS, lysed with 50 μl of 95 ◦C hot
lysis buffer, and incubated at 95 ◦C, 500 rpm for 10 min.
The lysates were cooled to room temperature and sonicated by
microtip probe sonication (Vibra-Cell VCX130, Sonics). Sonication
parameters were set to a total runtime of 2 min with pulses of 1 s on
and 1 s off at an amplitude of 80%.
For the EGF stimulation experiment, HeLa cells were grown in 24 p6
dishes until 80% confluence. Cells were serum-starved for 6 h upon
treatment. Treated HeLa cells were stimulated with 100 ng/ml of EGF
for 8 min. Cells were lysed in the plates with 80 μl of boiling lysis buffer
(5% SDS, 100 mM Tris pH 8.5, 5 mM TCEP, and 10 mM CAA) and
scraped to distribute the lysis buffer among the plates. The lysates
were incubated at 95 ◦C for 10 min with mixing (500 rpm).
Determination of Protein Concentration via BCA-Assay
Protein concentration was determined utilizing the Pierce BCA
Protein Assay Kit (Thermo Scientific) according to the manufacturer’s
protocol for 96-well plates setup.
For the cell dilution experiment in the Orbitrap Astral data, mean
peptide input amounts were estimated based on 25% recovery after
PAC digestion from BCA measurement of protein concentration.
Automatized Protein Aggregation Capture (PAC)-Based Protein
Digestion
Protein digestion was performed according to an adapted version
of the Protein Aggregation Capture (PAC) based digestion (32) on a
KingFisher Flex System (Thermo Scientific) (33) with MagReSyn Hy-
droxyl beads (ReSyn Biosciences). KingFisher deep-well plates were
prepared for washing steps, containing 1 ml of 95% Acetonitrile (ACN)
or 70% Ethanol (EtOH). For each sample, 300 μl of digestion solution
(50 mM ammonium bicarbonate (ABC)-buffer) containing Lys-C and
Trypsin at an enzyme-to-substrate ratio of 1:500 and 1:250, respec-
tively, were prepared and transferred to KingFisher plates. Samples
were mixed with 100 mM Tris-buffer to obtain a total volume of 300 μl,
transferred to KingFisher plates and ACN was added to a final volume
percentage of 70%. The storage solution from the Hydroxyl beads
was replaced with 70% ACN. Finally, beads were added to the
samples at a protein beads ratio of 1:2. Protein aggregation was
carried out in two steps of 1 min mixing at medium speed, followed by
a 10 min pause each. Sequential washes were performed in 2.5 min at
a slow speed without releasing the beads from the magnet. Digestion
was set to 100 cycles of agitation for 45 s and pausing of 6 min
overnight at 37 ◦C. Protease activity was quenched by acidification
with TFA to a final volume percentage of 1%.TABLE
Experimental design for the optimization of
Parameter optimization
Peptide
input
[μg]
Sample V
[μl] LB V [μ
Standard 30 3.29 200
Peptide μg 5–30 Standard Standa
Sample V Standard 7.5–120 Standa
LB V Standard 15 100–40
GA in LB Standard Standard Standa
Bead V Standard Standard Standa
Bead binding Standard Standard Standa
NH4OH in EB Standard Standard Standa
Evaluated parameters included peptide input (Peptide μg), sample volu
acid in the loading buffer (GA in LB), Zr-IMAC HP bead volume (Bead
hydroxide in the elution buffer (NH4OH in EB).Processing of HeLa samples was performed similarly but with some
adaptations. The ratio of MagReSyn hydroxyl beads to protein was
16:1, the time of digestion was 6 h in 200 μl of 50 mM triethy-
lammonium bicarbonate, and the Lys-C and Trypsin to substrate ra-
tios were 1:100 and 1:50, respectively. After digestion, samples were
acidified with 50 μl of 10% formic acid, peptides were concentrated in
SpeedVac at 45 ◦C until 20 μl and directly processed for phospho-
peptide enrichment without Sep-Pak desalting.
Sep-Pak Desalting
Desalting was performed on Sep-Pak C18 cartridges (C18 Classic
Cartridge, Waters). The cartridges were conditioned with 900 μl 100%
ACN and 3× 900 μl 0.1% TFA followed by sample loading, and
washing 3× with 900 μl 0.1% TFA. Peptides were eluted with 150 μl
40% ACN followed by 150 μl 60% ACN. The acetonitrile was evap-
orated in a SpeedVac at 45 ◦C and peptide concentration was
determined by measuring absorbance at 280 nm on a NanoDrop
2000C spectrophotometer (Thermo Fisher Scientific).
Phosphopeptide Enrichment
Phosphopeptide enrichment was performed on a KingFisher Flex
System (Thermo Scientific) using MagReSyn zirconium-based immo-
bilized metal affinity chromatography (Zr-IMAC HP) beads (ReSyn
Biosciences).
Standard Phosphopeptide Enrichment Workflow–Samples were
mixed with 200 μl loading buffer (80% ACN, 5% TFA, 0.1 M glycolic
acid) and transferred to a KingFisher 96 deep-well plate. Additional
KingFisher plates were prepared containing 500 μl of loading buffer,
500 μl of washing buffer 2 (80% ACN, 1% TFA), or 500 μl of washing
buffer 3 (10% ACN, 0.2% TFA) each. For each sample, 5 μl of beads
(20 mg/ml) were suspended in 500 μl 100% ACN previously added to
the KingFisher plates. For elution, 200 μl of elution buffer (1% NH4OH)
was prepared and transferred to KingFisher plates. Beads were
washed in loading buffer for 5 min at 1000 rpm, incubated with the
samples for 20 min with mixing at medium speed, and subsequently
washed in loading buffer, washing buffer 2, and washing buffer 3 for
2 min with mixing at fast speed. Phosphopeptides were eluted from
the beads by mixing with elution buffer for 10 min at a fast speed.
When evaluating the effect of different experimental parameters, the
standard phosphopeptide enrichment workflow was altered as indi-
cated in the experimental design table (Table 2). For evaluating the
influence of varying sample volume while keeping the peptide input
the same, samples were diluted with 1% TFA to the final desired
concentrations.2
phosphopeptide enrichment parameters
l]
GA in LB
[mol/l]
Beads V
[μl]
Bead
binding
[min]
NH4OH in
EB [%(v/v)]
0.1 5 20 1
rd Standard Standard Standard Standard
rd Standard Standard Standard Standard
0 Standard Standard Standard Standard
rd 0–2 Standard Standard Standard
rd Standard 1–20 Standard Standard
rd Standard Standard 5–30 Standard
rd Standard Standard Standard 0.1–2
me (Sample V), loading buffer volume (LB V), concentration of glycolic
V), beads binding time (Bead binding) and percentage of ammonium
Mol Cell Proteomics (2024) 23(5) 100754 3
Systematic Optimization for PhosphoproteomicsStandard Sequential/Looped Enrichment Workflow–Standard
sequential enrichment was performed for 2 to 3 enrichment rounds by
re-starting the phosphopeptide enrichment workflow on the King-
Fisher Flex System while keeping samples, beads, and washing
buffers the same. For obtaining “pooled” samples from the sequential
enrichment, all rounds were carried out re-using the same elution
buffer. To retrieve the phosphopeptides captured in each round as
separate fractions, the elution buffer was removed after each round
and replaced by new elution buffer.
Sequential Phosphopeptide Enrichment Workflow Adaptations–
Parameters of the sequential phosphopeptide enrichment workflow
were altered as indicated in the experimental design table (Table 3).
For sequential enrichment with increasing molarity of glycolic acid in
the loading buffer, 64 μl loading buffer containing 8 M GA was added
to the sample in 200 μl standard (0.1 M GA) loading buffer after the first
round of enrichment to adjust to 2 M glycolic acid. Sequential
enrichment with additional Zr-IMAC HP beads per round was per-
formed by the addition of 1 μl or 2 μl Zr-IMAC HP beads after the first
and second rounds, respectively, to the beads plate containing 1 μl
starting amount of beads. Sample-bead incubation times of 1 min and
5 min were applied for testing sequential enrichment in combination
with short incubation times. For testing the effect of fresh beads in a
sequential approach, Zr-IMAC HP beads were removed after the first
round of enrichment, and replaced by the same amount of fresh beads
for a second enrichment step.
Sample Preparation for LC-MS/MS Analysis
Eluates containing phosphopeptides were acidified with 40 μl 10%
TFA to achieve a pH <2. Acidified eluates were transferred to Multi-
ScreenHTS-HV 96-well filtration plates (0.45 μm, clear, non-sterile,
Millipore), stacked on 96-well plates, and centrifuged for 1 min at
500g to remove in-suspension particles.
Evotip Pure (Evosep) were washed by adding 20 μl 100% ACN and
centrifuging for 1 min at 800g. Tips were pre-conditioned by the
addition of 20 μl 0.1% formic acid (FA) while soaking the tips in 100%
isopropanol and centrifuged for 1 min at 800g. Filtered samples were
added to the tips and loaded by centrifugation for 2 min at 500g.
Evotip preparation was completed by adding 20 μl of 0.1% FA,
centrifuging for 1 min at 800g, adding 200 μl of 0.1% FA, and
centrifuging for 10 s at 800g.
LC-MS/MS Analysis
Samples were analyzed using an IonOpticks Aurora column (15 cm–
75 μm-C18 1.6 μm) interfaced with the Orbitrap Exploris 480 Mass
Spectrometer (Thermo Scientific) or the Orbitrap Astral MassTABLE
Experimental design for the optimization o
Sequential Rounds
Peptide
input [μg]
Sample V
[μl]
GA in LB
[mol/l]
Standard 2 30 3.29 0.1
6 rounds 6 2.5–20 2.5–20 Standard
Bead
addition
Standard 5–30 Standard Standard
New beads Standard Standard Standard Standard
Fraction
Pooling
Standard 2.5–5 Standard Standard
The sequential enrichment workflow design was evaluated in terms of
beads), total number of enrichment rounds (6 rounds), addition of beads
approaches for LC-MS/MS analysis (Fractions).
4 Mol Cell Proteomics (2024) 23(5) 100754Spectrometer (Thermo Scientific) using a Nanospray Flex Ion Source
with an integrated column oven (PRSO-V2, Sonation) to maintain the
temperature at 50 ◦C. In all samples, spray voltage was set to 1.8 kV,
funnel RF level at 40, and heated capillary temperature at 275 ◦C.
Samples were separated on an Evosep One LC system using the pre-
programmed gradient for 40 samples per day (SPD).
For phosphoproteome analysis of A549 samples using DIA on the
Orbitrap Exploris 480 Mass Spectrometer, full MS resolutions were set
to 120,000 at m/z 200 and the full MS AGC target was 300% with a
maximum injection time (IT) of 45 ms. The AGC target value for frag-
ment spectra was set to 1000%. 49 windows of 13.7 m/z scanning
from 472 to 1143 m/z were employed with an overlap of 1 Da. MS2
resolution was set to 15,000, IT to 22 ms, and normalized collision
energy (NCE) to 27%.
For phosphoproteome analysis of HeLa samples using DIA on the
Orbitrap Astral MS, the MS was operated at a full MS resolution of
180,000 with a full scan range of 480 to 1080 m/z. The full scan AGC
target was set to 500%. Fragment ion scans were recorded at a fixed
resolution of 80,000 and with a maximum IT of 6 ms. 150 windows of
4 m/z scanning from 480 to 1080 m/z were used. The isolated ions
were fragmented using HCD with 27% NCE.
Data Analysis
LC-MS/MS runs were searched using Spectronaut (version 17.1 for
Orbitrap Exploris 480 data and version 18 for Orbitrap Astral data)
employing a direct DIA search strategy against the Homo sapiens
proteome UniProt Database (2022 version, 20,958 entries) supple-
mented with a database of common contaminants (246 entries). MS2
de-multiplexing was set to automatic. Carbamidomethylation of
cysteine was set as a fixed modification, whereas oxidation of
methionine, N-terminal protein acetylation, and phosphorylation of
serine, threonine, and tyrosine were set as variable modifications. The
enzyme/cleavage rule was set to Trypsin/P, the digest type to specific,
and a maximum number of two missed cleavages per peptide were
allowed. The maximum number of variable modifications per peptide
was set to 5 and method evaluation was turned on. PTM localization
was turned on, but the localization probability threshold was set to 0.
The MS1 and MS2 mass tolerance strategy was set to system default.
PSM, peptide, and protein group FDR were set to 0.01 (1%) and the
employed directDIA workflow was directDIA+ (Deep). FDR calculation
in Specronaut is based on mProphet (34). The precursor q-value and
PEP cutoffs were 0.01 and 0.2, respectively. The protein q-value
experiment and run wide cutoffs were 0.01 and 0.05, respectively. For
quantification, precursor filtering was set to “Identified (Q value)” and
performed at the MS2 level using the area. No imputation was3
f sequential phosphopeptide enrichment
Bead V
[μl]
Bead
binding
[min]
Bead
exchange
Collected fractions
5 20 No Separate Rounds
Standard Standard Standard Standard
1 + 1 + 2 Standard Standard Pooled & Rounds
Standard Standard Yes Standard
Standard Standard Standard Pooled & Rounds
reusing or exchanging the beads after the first enrichment round (New
with increasing enrichment round (Bead addition), and sample pooling
Systematic Optimization for Phosphoproteomicsperformed, and cross-run normalization was turned off. The results
were filtered for the best N fragments per peptide between 3 and 6.
Decoy generation was performed using the mutated strategy. If not
explicitly mentioned, all remaining parameters were set to the default
Spectronaut settings.
Searches of data from sequential enrichment sets were performed
by searching the enrichment rounds and/or fractions separately while
keeping replicates of the same peptide input amount in the same
analysis.
Precursor-level pivot tables were exported from Spectronaut for
phosphopeptide analysis. Tables were filtered to contain unique
modified sequences (i.e., phosphopeptide isomers—same stripped
sequence but different phosphorylation site—are kept as separated
entities) and unique modified sequences containing phosphorylation
sites were further filtered to preserve only those with a localization
score >0.75 in at least one replicate. The unfiltered precursor pivot
tables are available as Supplemental Files, which have been deposited
to the ProteomeXchange Consortium via the PRIDE partner repository
with the dataset identifier PXD045601. An overview of all
Supplementary Files can be found in Table 1.
Data was exported in long-format and imported into Perseus
(v1.6.5.0) where it was collapsed into phospho-sites or phospho-
peptides using the “peptide-collapse” plugin (v1.4.2) described in
Bekker-Jensen et al. (35). For collapse into phosphosites, the option
“Target PTM site-level” was used. By default, the localization cutoff
was kept at 0.75. When evaluating the localization probability distri-
bution, the localization cutoff was set to 0. Importantly, phopshosites
reported by “peptide-collapse” plugin must have been identified and/
or localized in at least two experimental replicates. Collapse into
phosphopeptides using the Perseus plugin was employed for quan-
tification purposes. For phosphopeptide collapse, the option “Mod-
Spec peptide-level” was used and localization cutoff was kept at 0.75.
Phosphopeptide collapse in Perseus groups together different phos-
pho-isomers.
All remaining processing steps were performed either in Perseus
(v1.6.5.0) or R (v3.6.2 or higher) implementing the packages
ComplexHeatmap (36), sitools (https://CRAN.R-project.org/pack
age=sitools), eulerr (https://CRAN.R-project.org/package=eulerr),
stringr (https://CRAN.R-project.org/package=stringr), ggplot2 (37), dplyr
(https://CRAN.R-project.org/package=dplyr) and tidyverse (38). Calcu-
lation of isoelectric point (pI) valueswasperformedusing the packagepIR
(http://github.com/ypriverol/pIR2015), considering N-terminal acetyla-
tion and phosphorylation of the peptides.
Network analysis was performed in Cytoscape (39) (v3.10.1) using
SIGNOR App (40) (v1.2) to import the EGFR signaling pathway and
Omics Visualizer (41) (v1.3.1) to add the phosphosite information to
each node of the network. Regulatory sites of transcription factors in
the dataset were determined based on the “PG ProteinDescriptions”
column of the Spectronaut report and the regulatory site annotation
obtained from Perseus using PhosphoSitePlus database.RESULTS AND DISCUSSION
Phosphopeptide Binding Conditions Affect the Population
of Purified phosphopeptides
To establish an optimized automated phosphopeptide
enrichment procedure for high-sensitivity samples (13), we
started by evaluating the main experimental parameters that
can affect phosphopeptide enrichment. This was done by
introducing modifications to our default automated protocol
for sensitive phosphoproteomics, which relies on the use of
magnetic Zr-IMAC HP beads in the KingFisher System. Theresulting phosphopeptide mixtures were subsequently
analyzed by DIA-MS on an Orbitrap Exploris 480 MS coupled
to an Evosep One LC system taking advantage of the higher
sensitivity of the Whisper gradients. All enrichments were
performed from a starting amount of 30 μg of purified peptides
from a whole cell tryptic digest of the A549 lung cancer cell
line, as it represents the optimal peptide input amount for
phosphopeptide enrichment and subsequent analysis with our
LC-MS/MS setup using Whisper gradients. By using this
amount, we ensured to have an adequate reference to assess
the effect of the changes in the experimental workflow. The
phosphopeptide enrichment protocol consists of three main
steps: (i) the binding of phosphopeptides to the beads, (ii) the
washing of the beads to remove non-specific interactions, and
(iii) the elution of phosphopeptides from the beads (Fig. 1A).
We focused on the first part of the protocol, the binding of
phosphopeptides to the beads, in which we evaluated the
following parameters: (i) the beads to peptide ratio, (ii) the
proportion of glycolic acid in the loading buffer, (iii) the binding
time, and (iv) the sample to loading buffer volume ratio. In the
last step of the protocol, we evaluated the effect of modifying
the concentration of ammonium hydroxide in the elution buffer
(Fig. 1B and Table 2).
The proteomics community has extensively evaluated the
beads-to-peptide ratio as this plays an important role in the
phosphopeptide enrichment efficiency. A high beads-to-
peptide ratio can lead to increased binding of non-
phosphorylated peptides due to nonspecific interactions
with the bead surface, thus reducing the selectivity of the
enrichment. In contrast, it has also been described that a too
low beads-to-peptide ratio can result in a higher fraction of
multiply-phosphorylated peptides identified (21). In this work,
we assessed the effect of using different bead amounts. This
is particularly relevant when performing enrichment with low
peptide amounts since it is difficult to proportionally scale
down the volume of beads, due to lack of reproducibility when
pipetting low volumes of beads. Therefore, starting from 30 μg
of peptides, we evaluated what the best compromise between
bead volume and good phosphopeptide recovery would be by
testing 1, 2, 5, 10, and 20 μl of beads corresponding to a
beads-to-peptide ratio of 0.7, 1.3, 3.3, 6.7 and 13.3 (Fig. 2A,
supplemental Table S1, and supplemental File S1). We
observed that the best outcome was obtained using a volume
of 5 μl (beads-to-peptide ratio of 3.3), which resulted in 16,193
phosphopeptides. Importantly, throughout this work, we only
report a phosphopeptide as valid for those where the phos-
phorylation was localized to an amino acid with a score >0.75.
When investigating enrichment efficiency, it should be noted
that there are different ways of calculation: either based on
counts (i.e., number of phosphopeptides versus total number
of peptides), or based on abundance (i.e., MS signal from
phosphopeptides versus total signal). In most figures in this
work (unless explicitly stated), enrichment efficiencies are re-
ported based on counts. However, this strategy can result in aMol Cell Proteomics (2024) 23(5) 100754 5
FIG. 1. Experimental design. A, schematic overview of the phosphopeptide enrichment workflow and the evaluated experimental param-
eters. Evaluated parameters included the peptide input, the sample volume, the loading buffer volume, the proportion of glycolic acid in the
loading buffer, the percentage of ammonium hydroxide in the elution buffer, the Zr-IMAC HP bead volume, and the sample-beads binding time.
B, schematic overview of the sequential phosphopeptide enrichment workflow and the evaluated experimental parameters. Evaluated pa-
rameters included the peptide input amount of the sample, the number of sequential enrichment rounds and the way of retrieving the eluate by
either reusing the elution buffer and obtaining a “Pooled” eluate or exchanging the elution buffer after each round and obtaining each enrichment
round as single fraction for LC-MS/MS analysis. Modified sequential enrichment approaches included exchanging, reusing or increasing the Zr-
IMAC HP bead volume.
Systematic Optimization for Phosphoproteomicsbiased interpretation of the quality of the phosphopeptide
enrichment performed. For instance, when comparing
enrichment efficiency based on counts with enrichment effi-
ciency based on abundance in the samples of our bead vol-
ume evaluation, we found that the enrichment efficiency
based on abundance was consistently higher than that based
on counts for all tested bead volumes (supplemental Fig. S1),
and in all cases >90%, reflecting that the enrichment strategy
prioritized phosphopeptides (supplemental Fig. S2) as ex-
pected and on par with previous works (13, 17, 22). We sug-
gest that this divergence could be a consequence of using a
DIA search strategy, which can favor the identification of non-
phosphopeptides, even at low abundance levels. However,
we used count-based enrichment since it gives more insights
into the nuances between the parameters tested. For
instance, increasing the bead volume to 20 μl slightly
decreased the overall number of phosphopeptides to 15,026
as well as the relative enrichment efficiency based on counts
(from 79% with 5 μl to 62% with 20 μl), whilst at abundance
level, the enrichment efficiency changed from 98 to 93%.6 Mol Cell Proteomics (2024) 23(5) 100754Moreover, reducing the bead volume to 1 μl resulted in higher
enrichment efficiency (84%) (Fig. 2E) and a trend towards
more acidic phosphopeptides (Fig. 2H), although lower overall
phosphoproteome coverage (12,962 phosphopeptides)
(Fig. 2, A and F). Our titration experiment also confirmed that
the selectivity of the enrichment in regard to the purification of
mono- or multiply-phosphorylated peptides is affected by the
available binding surface, as the absolute number of multiply-
phosphorylated peptides increased with decreasing bead
amount (Fig. 2, G and I).
We next evaluated the effect of changing the beads-to-
peptide binding time to explore whether shortening the bind-
ing time would lead to lower phosphoproteome depth and/or
bias the recovery towards multiply-phosphorylated peptides.
Shortening the binding time from 20 to 5 min had a slight
impact on the phosphoproteome depth achieved, with 15,700
phosphopeptides quantified after 5 min binding compared to
16,440 after 20 min binding (Fig. 2, B and F, supplemental
Table S2, and supplemental File S2). However, it seems that
there was a slight improvement in enrichment efficiency (from
FIG. 2. Evaluation of experimental parameters during phosphopeptide enrichment. A–D, barplots show the mean numbers of peptides
(light color) and phosphopeptides with loc. prob. >0.75 (dark color) identified across three experimental replicates using different (A) molarities of
Systematic Optimization for Phosphoproteomics
Mol Cell Proteomics (2024) 23(5) 100754 7
Systematic Optimization for Phosphoproteomics75% in 20 min to 77% in 5 min), likely due to more unspecific
binding to the beads with longer incubation time (Fig. 2E).
Moreover, only a marginal increase in the percentage of
multiply-phosphorylated peptides was observed by short-
ening the binding time (Fig. 2, G and J).
The use of non-phosphopeptide excluders during binding to
prevent binding of non-phosphopeptides with high affinity
towards IMAC-metal conjugates was evaluated next. In our
standard protocol, we originally used 0.1 M of GA as a
competitive binder in the loading buffer, as recommended by
the beads’ manufacturer. With this GA concentration, we
obtained an enrichment efficiency of 78% based on peptide
counts or 94% based on MS signal abundance. In line with
others (22), we observed that increasing the GA concentration
up to 2 M improved the overall enrichment efficiency (86%
based on peptide counts) and slightly lowered the median
phosphopeptide pI (Fig. 2H), but at the cost of reduced
phosphoproteome depth with 10,811 phosphopeptides
quantified against 15,929 in our standard protocol (Fig. 2, C
and F, supplemental Table S3, and supplemental File S3). In
contrast, removing the GA greatly decreased the enrichment
efficiency (to 74% based on peptide counts), while preserving
the phosphoproteome coverage (16,342 phosphopeptides
with 0 M GA) obtained with 0.1 M of GA (Fig. 2, C–F). Inter-
estingly, the high concentrations of GA (1 M and 2 M) biased
the enrichment towards multiply-phosphorylated peptide
species, providing up to 16% more multiply-phosphorylated
peptides than 0.1 M of GA (Fig. 2, G and K). This demon-
strated that in high concentrations, GA does not only compete
with acidic amino acids in its function as a competitive binder,
but also with phosphopeptides. Hence, the most competitive
phosphopeptides (multiply phosphorylated species) would
bind preferentially. Overall, we could confirm that multiply-
phosphorylated peptides have a higher affinity towards Zr-
IMAC HP beads, explaining why they are preferably recov-
ered with more competitive binding conditions such as 2 M
GA or shorter binding time.
Next, we questioned whether the lower multiply-
phosphorylated peptide recovery in standard enrichment
conditions (0.1 M GA, 20 min binding time) could be due to the
stronger binding of multiply-phosphorylated peptides to the
beads, preventing them from proper elution when present in a
large pool of singly-phosphorylated peptides. We evaluatedglycolic in the loading buffer, (B) percentage of ammonium hydroxide in
binding times. Each dot represents one experimental replicate. E–H, H
the loading buffer, percentage of ammonium hydroxide in the elution bu
selectivity of the enrichment in terms of identified phosphorylated and no
with loc. prob. >0.75 (G) percentage of multiply-phosphorylated peptides
peptides with loc. prob. >0.75 (H) median pI of phosphopeptides with l
parameter column for (E–G) and to the lowest value for (H). If not othe
replicates. I–L, barplots show the mean numbers of singly (light color), d
with loc. prob. >0.75 identified across three experimental replicates using
ammonium hydroxide in the elution buffer, (K) Zr-IMAC HP bead volum
mental replicate.
8 Mol Cell Proteomics (2024) 23(5) 100754different concentrations of ammonium hydroxide (NH4OH) for
phosphopeptide elution from the beads, ranging from 0.1 to
2% (v/v) (supplemental Table S4 and supplemental File S4).
Although subtle, we observed that the lowest NH4OH con-
centration tested (0.1%) resulted in lower multiply-
phosphorylated peptide recovery with 8% multiply-
phosphorylated phosphorylated peptides compared to 10%
multiply-phosphorylated peptides with 2% NH4OH (Fig. 2, D
and L). Enrichment efficiency was highest with 0.5% NH4OH
(81%) and slightly decreased with higher NH4OH concentra-
tions (79% with 1% NH4OH and 78% with 2% NH4OH) (Fig. 2,
D–E). Altogether, we conclude that the percentage of NH4OH
in the elution buffer does not significantly impact the elution of
bound phosphopeptides from the beads.
To test the influence of the sample input amount in the low
peptide range, we performed phosphopeptide enrichment
using 30, 10, and 5 μg of peptides. As expected, phospho-
peptide recovery is highly dependent on the sample input with
6888 phosphopeptides quantified from 5 μg compared to
12,480 phosphopeptides quantified using 30 μg of peptide
amount (Fig. 3A, supplemental Table S5, and supplemental
File S5). Also, the site localization scores scaled with sample
amount (Fig. 3B), reflecting that the capacity of the search
engine to localize phosphorylation sites is dependent on the
signal and therefore the quality of the MS2 spectra. The per-
centage of multiply-phosphorylated peptides barely increased
with increasing peptide amount (5% for 5 μg, 6% for 15 μg
and 7% for 30 μg) (Fig. 3C). Finally, phosphopeptides enriched
from different peptide input amounts highly overlapped, with
more abundant phosphopeptides being preferentially enriched
from all input amounts, and the phosphoproteome depth
increased as expected with higher input amounts, where lower
abundant phosphopeptides were detectable (Fig. 3, D and E).
Finally, we evaluated the sample volume-to-loading buffer
(LB) volume ratio. First, we diluted the sample while keeping
the LB volume constant (Table 2, Fig. 3F, supplemental
Table S6, and supplemental File S6), and second, we kept
the sample highly concentrated in a constant volume of 15 μl
while using increasing LB volumes (Table 2, Fig. 3G,
supplemental Table S7, and supplemental File S7). Increasing
the ratio sample volume-to-LB volume had a negative impact
on the phosphoproteome depth (16,132 phosphopeptides
with 7.5 μl sample volume versus 13,770 phosphopeptidesthe elution buffer, (C) Zr-IMAC HP bead volumes or (D) sample-bead
eatmaps show the influence of increasing molarity of glycolic acid in
ffer, Zr-IMAC HP bead volume or sample-bead incubation time on (E)
n-phosphorylated peptides (F) number of identified phosphopeptides
with loc. prob. >0.75 in the context of total identified phosphorylated
oc. prob. >0.75. Stars refer to the highest value within the respective
rwise indicated, all values represent the mean of three experimental
oubly (medium color), and triply (dark color) phosphorylated peptides
different (I) molarities of glycolic in the loading buffer, (J) percentage of
es, or (L) sample-bead binding times. Each dot indicates one experi-
FIG. 3. Evaluation of peptide input and sample/loading buffer volume effects. A, barplots show the mean number of peptides (light color)
and phosphopeptides with loc. prob. >0.75 (dark color) identified across three experimental replicates using different peptide input amounts.
Each dot represents one experimental replicate. B, violin plots show the range and distribution of the localization probability of phosphosites
identified using different peptide input amounts. C, barplots show the mean numbers of singly (light color), doubly (medium color), and triply (dark
color) phosphorylated peptides with loc. prob. >0.75 was identified across three experimental replicates using different peptide input amounts.
Each dot represents one experimental replicate. D, the Venn diagram shows uniquely and commonly identified phosphosites with loc. prob.
>0.75 among different peptide input amounts. E, Violin plots show the log2 mean intensities of uniquely and commonly identified phosphosites
with loc. prob. >0.75 among different peptide input amounts. F, barplots show the mean numbers of peptides (light color) and phosphopeptides
with loc. prob. >0.75 (dark color) identified across three experimental replicates using the same peptide input amount (30 μg) diluted in different
sample volumes, mixed with the same volume of loading buffer (200 μl). Each dot represents one experimental replicate. G, barplots show the
mean numbers of peptides (light color) and phosphopeptides with loc. prob. >0.75 (dark color) identified across three experimental replicates
using the same peptide input amount (30 μg) diluted in the same sample volume (15 μl), mixed with different volumes of loading buffer. Each dot
represents one experimental replicate.
Systematic Optimization for Phosphoproteomics
Mol Cell Proteomics (2024) 23(5) 100754 9
Systematic Optimization for Phosphoproteomicswith 120 μl sample volume), potentially due to the dilution of
the LB by addition of higher sample volumes during binding.
However, the dilution of LB in increasing volumes of the
sample had an impact on the enrichment by reducing the
binding of non-phosphorylated species (2691 non-
phosphorylated peptides with 7.5 μl sample volume versus
2258 non-phosphorylated peptides with 120 μl sample vol-
ume) (Fig. 3F). On the other hand, we observed that the vol-
ume of the loading buffer did not have an impact on the
phosphoproteome depth or enrichment efficiency, as long as
the sample was kept to a minimal volume (<30 μl) (Fig. 3G).
Altogether, in Zr-IMAC HP-based phosphopeptide enrich-
ment, we observed that the beads-to-peptide ratio, the con-
centration of GA in the loading buffer, the peptide input itself,
as well as the sample concentration can have a significant
impact on the resulting phosphoproteomes. On the contrary,
the binding time and percentage of NH4OH in the elution
buffer did not seem to have such a significant influence on the
phosphopeptide enrichment.
Our evaluation showed that for highly sensitive phospho-
proteomics, 5 μl of Zr-IMAC HP beads, 20 min binding time,
0.1 M GA in the loading buffer and 0.5% of NH4OH in the
elution buffer should be employed to obtain the best phos-
phopeptide enrichment. However, when multiply phosphory-
lated peptides are of interest, highly competitive binding
conditions such as using 1 μl of Zr-IMAC HP beads, 5 min
binding time, 2 M GA in the loading buffer and a high per-
centage of NH4OH in the elution buffer (2%) could be the best
choice.
Sequential Enrichment of the Phosphoproteome as a
Strategy to Increase the Depth of the Analysis
Our data so far showed that changing experimental pa-
rameters during phosphopeptide enrichment can have a sig-
nificant impact on the population of enriched
phosphopeptides. Exploring these differences has been sug-
gested before as a potential way to enhance the performance
of phosphopeptide enrichment strategies by performing
sequential enrichment. The most straightforward way to
perform sequential enrichment is to iterate the enrichment by
using the flow-through from the previous enrichment. This
strategy has been utilized before to increase the depth of the
phosphoproteome (14, 24, 42, 43). Therefore, we wanted to
evaluate how many sequential rounds of enrichment are
needed to efficiently deplete a sample for phosphopeptides,
and whether sequential enrichment is as efficient with high
peptide input amounts as it is with low peptide input amounts.
First, we evaluated whether the beads employed in one
round of phosphopeptide enrichment could be reused for a
second enrichment. Our data reflected the potential of reusing
the beads for sequential enrichment. Interestingly, reusing the
beads from the first enrichment in a second one resulted in a
higher phosphopeptide recovery in the second enrichment
round (10,928 phosphopeptides with new beads compared to10 Mol Cell Proteomics (2024) 23(5) 10075412,812 phosphopeptides with reused beads) and a better
enrichment efficiency, when compared to using new beads for
the second enrichment (52% with new beads compared to
60% with reused beads) (supplemental Fig. S3, supplemental
Table S8, and supplemental File S8).
Next, we tested more extensive sequential enrichment (up
to six rounds) for samples spanning 20 to 2.5 μg of peptide
input. Whilst the enrichment efficiency (based on phospho-
peptide intensities, measured as the percentage of the overall
identified MS signal intensity from phosphopeptides alone)
was above 90% in the first enrichment round for all amounts
tested, it abruptly decreased with lower peptide input amounts
in subsequent enrichment rounds down to 86% for 20 μg and
58% for 2.5 μg in round six (Fig. 4A, supplemental Fig. S4, A
and B, and supplemental File S9). Similarly, the phosphopro-
teome depth (phosphopeptide count relative to round 1)
decreased with each sequential enrichment, which was more
evident for lower input amounts (Fig. 4, A, C, and D, and
supplemental Table S9). The number of additional unique
phosphopeptides did not significantly increase after the third
enrichment round (Fig. 4F and supplemental Fig. S4A), even
though up to 2105 phosphopeptides were still enriched in the
sixth enrichment of 20 μg peptide input amount (Fig. 4, A–C).
The population of phosphopeptides enriched in each subse-
quent enrichment, especially from third and forward, was
mainly driven by abundance since the most abundant pep-
tides in the first enrichment round continued being enriched
subsequently (supplemental Fig. S4A). Unlike the phospho-
peptides, the non-phosphorylated peptides eluted differently
across the sequential enrichments (supplemental Fig. S4A). At
least three different trends were observed in the non-
phosphorylated peptides elution: (group 1) peptides eluting
mainly in the first fraction (enrichment round), (group 2) pep-
tides eluting mainly in the second fraction, and (group 3)
peptides with a consistent elution between fractions
(supplemental Fig. S4A). We decided to explore the nature of
these peptides further to understand the mechanisms behind
the non-phosphopeptide binding with Zr-IMAC HP beads.
When compared to a comprehensive proteome of the same
cell line (A549), it was evident that the non-phosphorylated
peptides that bound to the beads belonged to the most
abundant pool of peptides in the proteome (supplemental
Fig. S4C). In particular, the non-phosphorylated peptides
that showed a consistent elution across the first three frac-
tions (group 3) were more abundant in the proteome than the
rest, showing that their binding is most likely mainly driven by
abundance (supplemental Fig. S4C). We also calculated the
isoelectric point of these peptides and found them to be
generally acidic (pI ~5), although less than the phosphopep-
tides (pI ~3) (supplemental Fig. S4D). Altogether, these data
show that although the selectivity will strongly favor the
binding of phosphopeptides, the high abundance of non-
phosphorylated peptides can lead to unspecific binding dur-
ing sequential enrichment.
FIG. 4. Evaluation of an extensive 6-round sequential enrichment approach. A, barplots show the mean numbers of singly phosphorylated
(light color) and multiply-phosphorylated (dark color) peptides with loc. prob. >0.75 (dark color) identified across three experimental replicates
Systematic Optimization for Phosphoproteomics
Mol Cell Proteomics (2024) 23(5) 100754 11
Systematic Optimization for PhosphoproteomicsThroughout the course of this project, we observed that the
way DIA proteomics data, and in our case phosphoproteomics
data in particular, is analyzed by the search engine (i.e.,
Spectronaut) had an impact on the identifications. In spectral
library-free mode (direct-DIA) searches in Spectronaut, when
several files are searched together, the information from all of
them is used during the search, allowing data for spectral li-
brary inference to be obtained from one file and used during
peptide identification in the other files. This effect is of special
relevance when searching together high-load samples and
low-load samples, and it is a strategy often employed in the
field of single-cell proteomics to boost identifications. In our
experiments, searching all six fractions together using direct-
DIA led to an increase in the number of identified (phospho)
peptides in all fractions (supplemental Fig. S4E). In contrast,
when the search of each enriched fraction was done sepa-
rately using the evaluation mode, most of the identified pep-
tides were found uniquely in the first enrichment
(supplemental Fig. S4F).
We previously observed that the site localization score
decreased with lower peptide input and hence phosphopep-
tide intensity (Fig. 3B). Since we observed a constant
decrease in median phosphopeptide intensity with each
subsequent enrichment (Fig. 4E), we evaluated whether the
site localization scores also worsened in each subsequent
enrichment (supplemental Fig. S5, A–D). Interestingly, such a
trend was not observed for lower input amounts
(supplemental Fig. S5, C and D). Potentially, this could be due
to the lower number of phosphopeptides identified in the last
enrichment rounds when starting with 2.5 or 5 μg, which likely
represents the more abundant phosphopeptides which are
therefore easier to localize (supplemental Fig. S6A).
Conversely, for higher input amounts, the population of
phosphopeptides in the last enrichment might include less
abundant phosphopeptides that result in worse localization
scores.
Finally, we hypothesized that sequential enrichment might
eventually deplete the most abundant phosphopeptides,
allowing other phosphopeptide species to be enriched.
Interestingly, we observed that while the overall intensity in the
population of singly phosphorylated peptides decreased over
time, the multiply-phosphorylated counterpart increasedusing different peptide input amounts for a sequential six-round enrichm
enrichment round and analyzed separately via LC-MS/MS. Each dot r
enrichment efficiency across three experimental replicates in each roun
percentage. Each dot represents one experimental replicate. B–D, Heatm
percentage of multiply-phosphorylated peptides with loc. prob. >0.75 in
with loc. prob. >0.75 (C) number of identified phosphopeptides with loc. p
enrichment depth in percentage in terms of number of identified phosp
peptide input amount. E, line plots show the medians of log2 mean inten
with loc. prob. >0.75 identified in each enrichment round upon differen
cumulative phosphosites with loc. prob. >0.75 per peptide input amoun
lative” refers to the addition of phosphosites that were not identified in th
replicate.
12 Mol Cell Proteomics (2024) 23(5) 100754towards the last fractions, especially for low input amounts
(Fig. 4E and supplemental Fig. S6). Although the number of
multiply-phosphorylated peptides decreased with each
enrichment (Fig. 4B), this increment in the intensity of multiply-
phosphorylated peptides could be due to the higher affinity of
those peptides when the overall population of phosphopep-
tides is depleted.
Overall, the highest increase in phosphoproteome depth
when doing sequential enrichment originated mostly from the
second enrichment (Fig. 4, D and F). However, doing
sequential enrichment involves not only more sample prepa-
ration time, but also an increase in subsequent MS mea-
surement time. Moreover, there is no standardized approach
on how to handle multiple enrichments from a quantitative
phosphoproteomics perspective, considering the high
redundancy of phosphopeptides identified across the
sequentially enriched fractions and that their intensity is rela-
tive to their environment. Therefore, we hypothesized that a
potential solution to benefit from the increase in depth of
sequential enrichment could be achieved by pooling the
fractions prior to MS analysis. We explored this possibility for
highly sensitive analysis, using 2.5 and 5 μg of peptide as
starting amounts for enrichment. Interestingly, we observed a
gain of 7% (from 9750 phosphopeptides in one single
enrichment to 10,511 phosphopeptides in a pooled sample,
supplemental Table S10, supplemental File S10) when
combining first and second enrichment from 5 μg of input
peptides (Fig. 5A). The gain in IDs in the pooled sample could
be potentially due to a boost in the intensity (Fig. 5B). How-
ever, we did not observe such a significant gain when using
2.5 μg of peptide (only 2% increase in phosphopeptides)
(Fig. 5, A and B). Furthermore, we explored the impact of
pooling from a quantitative perspective by calculating the CVs
of the pooled fractions and comparing it to separate enrich-
ments or the cumulative strategy (i.e., First and second
enrichment analyzed separately by LC-MS/MS, and the
resulting data merged in-silico afterward). Reassuringly, CVs
were not significantly affected by pooling the samples and
remained with a median of around 20% (Fig. 5C).
Next, we tried to further exploit the benefit of pooling
sequential enrichment fractions by modifying the conditions of
each enrichment to favor complementary populations ofent. Each fraction (round) was obtained as eluate after the respective
epresents one experimental replicate. Line plots represent the mean
d per peptide input amount based on phosphopeptide intensities in
aps show, for each peptide input amount and enrichment round, the (B)
the context of the total number of identified phosphorylated peptides
rob. >0.75 in the context of total identified phosphorylated peptides (D)
hopeptides with loc. prob. >0.75 relative to round 1 of the respective
sities across replicates of singly and multiply-phosphorylated peptides
t peptide input amounts. F, line plots represent the mean number of
t and enrichment round across three experimental replicates. “Cumu-
e previous enrichment round(s). Each dot represents one experimental
FIG. 5. Strategies for pooling sequential enrichment samples. A, barplots show the numbers of phosphopeptides with loc. prob. >0.75
(dark color), 3/3 valid intensity values among replicates (medium color) or with a CV <0.2 among replicate intensities (light color) identified in a
two-round sequential enrichment approach using 2.5 μg (pink) or 5 μg (purple) peptide input amount. Each dot represents one replicate. Each
fraction (round) was either obtained as eluate after the respective enrichment round and analyzed separately via LC-MS/MS (“Round 1” and
“Round 2”) or obtained as a pooled eluate by reusing the elution buffer from the first enrichment round (“Pooled”). “Cumulative” refers to the
cumulation of unique phosphopeptide IDs identified in the separate fractions (“Round 1” & “Round 2”) during data analysis. B, boxplots show the
log2 mean intensities of identified phosphopeptides with loc. prob. >0.75 per fraction and peptide input amount. C, density plots show the
distribution of CVs across replicates of phosphopeptide (loc. prob. >0.75) intensities per fraction (columns) and peptide input (rows) after
normalization. The labels within the density plots show the median CVs.
Systematic Optimization for Phosphoproteomicsphosphopeptides. In particular, we applied our observation
that the amount of beads used inversely correlates with the
number of multiply-phosphorylated peptides identified in a
sample. We hypothesized that, when the amount of beads is
limited, the competitive binding conditions lead to the favored
binding of multiply-phosphorylated peptides. Consequently,
to take advantage of this in a sequential enrichment strategy,
we designed the following experiment: first enrichment with
1 μl of beads, second enrichment adding 1 μl of new beads,
and third enrichment adding 2 μl of new beads. We either
analyzed each enrichment round fraction separately and
cumulated the phosphopeptides during data analysis(cumulative approach) or pooled them into one fraction by
reusing the same elution buffer aliquot (pooled approach).
Additionally, we performed the standard enrichment strategy
with 5 μl of beads as a comparison (Fig. 6). The results
revealed that there was a significant gain when using this
sequential strategy. The pooled sequential approach
improved the phosphoproteome depth compared to a single
enrichment in standard conditions when starting with input
amounts of at least 15 μg (from 9247 to 11,356 phospho-
peptides for 15 μg, and from 12,568 to 14,085 phosphopep-
tides for 30 μg, supplemental Table S11 and supplemental File
S11). Moreover, the pooled approach yielded moreMol Cell Proteomics (2024) 23(5) 100754 13
FIG. 6. Refined sequential enrichment pooling approach with increasing Zr-IMAC HP bead volume. Barplots show the numbers of
phosphopeptides with loc. prob. >0.75 was identified using a three-round sequential enrichment approach with increasing Zr-IMAC HP bead
volume for different peptide input amounts. “Normal” represents a standard single-round enrichment with 5 μl beads. “Pooled” represents a
sequential enrichment for three rounds with increasing bead volume (Round 1: 1 μl beads, Round 2: +1 μl beads, Round 3: +2 μl beads) and
rounds pooled into the same elution buffer. “Round 1”, “Round 2” and “Round 3” represent the identifications in the respective separately
collected and analyzed fractions. “Cumulative” refers to the cumulation of unique phosphopeptides identified in the separate fractions (“Round
1”, “Round 2”, “Round 3”) during data analysis.
Systematic Optimization for Phosphoproteomicsphosphopeptide IDs than the cumulative approach for all input
amounts (6385 versus 5343 phosphopeptides for 5 μg, 11,356
versus 10,442 phosphopeptides for 15 μg and 14,085 versus
13,004 phosphopeptides for 30 μg). Interestingly, we
observed that there was no such gain when using lower input
amounts (i.e., 5 μg). This might indicate that the beads-to-
peptide ratio was not optimized for such low amounts and
that more optimization might be required to make this strategy
beneficial. Overall, we were able to confirm that sequential
enrichment approaches can significantly increase phospho-
peptide identifications compared to a standard one-round
enrichment and observed that pooling fractions from multi-
ple sequential enrichment rounds can outperform their sepa-
rate analysis in terms of phosphopeptide identifications.Combination of Sequential Enrichment With LC-MS/MS
analysis on an Orbitrap Astral Mass Spectrometer
Finally, the evolution of mass spectrometers is one of the
most significant aspects of phosphoproteomics leading to
improvements in sensitivity, speed, and depth of analysis.
Therefore, we decided to complete our systematic evaluation
by benchmarking the optimized phosphoproteomics workflow
using the latest generation high-end proteomics-grade MS
instrumentation, the Orbitrap Astral mass spectrometer.14 Mol Cell Proteomics (2024) 23(5) 100754To evaluate the performance of the Orbitrap Astral for
phosphoproteomics using the optimized phosphopeptide
enrichment workflow in settings best representing typical
biological experiments, we performed the phosphopeptide
enrichment starting from different numbers of HeLa cells.
Consequently, the resulting phosphoproteome coverage re-
flects how sensitive the protocol is to the number of input
cells. Our experiment used four replicates for six different
numbers of cells, ranging from 1 million cells to 10,000 cells.
The different cell samples were lysed in SDS-buffer and
digested using PAC-based trypsin digestion, followed by
phosphopeptide enrichment without any desalting step to
minimize losses. The phosphopeptide enrichment was per-
formed using the optimized parameters described in this
project. Two rounds of sequential enrichment were performed,
eluates were pooled together for analysis and evotipped
samples were analyzed by a 40-SPD Whisper flow method
with half-an-hour LC gradient time on the Orbitrap Astral using
narrow-window data-independent acquisition (nDIA) with
narrow DIA isolation windows (9).
This resulting data reflects the higher sensitivity of the
Orbitrap Astral MS with more than 35,000 phosphopeptides
from 1 M cells or >32,000 phosphopeptides mapping to
26,000 class I phosphosites quantified in at least two samples
when starting with 0.5 million cells (Fig. 7, A and B, and
FIG. 7. Two-round sequential enrichment for analysis of a HeLa dilution series for LC-MS/MS analysis on an Orbitrap Astral Mass
Spectrometer. A, barplots show the mean numbers of peptides (light color), phosphopeptides (medium color) and phosphopeptides with loc.
Systematic Optimization for Phosphoproteomics
Mol Cell Proteomics (2024) 23(5) 100754 15
Systematic Optimization for Phosphoproteomicssupplemental Tables S12 and S13; supplemental File S12).
This is approximately equivalent to 34 μg of purified tryptic
peptides, which is an improvement of approximately two-fold
when compared to the previous coverage achieved from 30 μg
of purified peptides in the Orbitrap Exploris 480 with up to
16,500 phosphopeptides (Fig. 2). The Orbitrap Astral allows a
comprehensive coverage of the phosphoproteome for
amounts as low as 50,000 cells (7967 phosphopeptides). With
fewer cells, the output dropped below the sensitivity limits of
our workflow (Fig. 7, A and B and supplemental Fig. S7, A and
B). We also evaluated the site localization scores obtained
from the phosphopeptides and observed a similar trend as the
one observed for Orbitrap Exploris 480 data (Fig. 3B) with a
clear drop in localization scores for lower cell inputs
(supplemental Fig. S7C). The enrichment efficiency calculated
as a function of overall intensities was >90% as expected
(supplemental Fig. S7D).
To further evaluate whether the depth achieved in the
Orbitrap Astral single-shot phosphoproteome improved the
downstream functional analysis, we mapped the number of
known regulatory sites for transcriptional factors contained in
our data (Fig. 7C and supplemental File S13). We could
identify relevant regulatory sites in transcription factors such
as Jun or Atf2 which were distributed across the full abun-
dance range of the obtained phosphoproteome (Fig. 7D). The
depth of the Orbitrap Astral single-shot phosphoproteome
also provided information for 2398 known kinase substrates
from 277 kinases (supplemental File S13). Finally, we evalu-
ated whether the coverage of known signaling pathways was
improved in this analysis using the EGFR signaling pathway as
an example. Not only could we obtain information for almost
the entire network, but we could also map the most relevant
regulatory sites involved in EGFR signaling (Fig. 7E).
To evaluate the quantification quality of the phosphopro-
teomics data derived from the Orbitrap Astral, we compared
the phosphoproteome profiles of 12 replicates treated with
100 ng/ml of EGF for 8 min to activate the EGFR signaling
pathway to 12 of which were used as control samples
(Fig. 8A). With this experimental design, we first evaluated the
effect of the sample size on the identification depth as well as
on the completeness of the data. In this regard, we searched
the data from 3, 6, 9, and 12 replicates per condition sepa-
rately in Spectronaut (Fig. 8A). As expected, the global num-
ber of phosphosites identified by Spectronaut increased with
a higher number of replicates (Fig. 8B), but this increase wasprob. >0.75 (dark color) identified across four experimental replicates
enrichment. Each dot represents one experimental replicate. B, barplot
phosphosites (loc. prob. >0.75) (dark color) identified across four experim
sequential enrichment. Each dot represents one experimental replicate. C
million HeLa cells phosphopeptide enrichment experiment. Known reg
sentative sites plotted in panel (D) are labeled. D, Extracted Ion Chromato
E, EGFR signaling pathway network obtained from SIGNORApp (40). The
1 million HeLa cells phosphopeptide enrichment experiment, which are la
to a phosphorylation site. Pink sections indicate which phosphosite is k
16 Mol Cell Proteomics (2024) 23(5) 100754not translated to the completeness of the dataset, since the
depth of the analysis decreased when filtering for phospho-
sites identified in all replicates (Fig. 8B). When 75% of valid
values were allowed, the depth of the phosphoproteome was
similar regardless of the number of replicates used in the
search (Fig. 8C). Median coefficients of variation (CVs) be-
tween the replicates were consistent independently of the
number of replicates that were used in the calculation
(Fig. 8D). This observation indicates that the full workflow,
from cell culture to MS acquisition and phosphopeptide
analysis, was consistent across all replicates, since intro-
ducing more replicates did not increase the variability of the
measured CVs. This repeatability was also observed in a
sample correlation analysis where the Pearson correlation
among samples from the same condition was >0.95
(supplemental Fig. S8A).
Finally, we also evaluated the impact of the number of
replicates in DIA-based phosphoproteomics perturbation ex-
periments on the statistical sensitivity. Phosphosites with less
than 2-thirds of valid values in at least one condition were
filtered out. The data was nornalized and statistically signifi-
cant phosphosites were identified using a two-samples t test
with Benjamini-Hochberg adjusted correction for FDR control.
As expected, the statistical sensitivity increased with a higher
number of replicates, resulting in a significantly higher number
of differentially expressed phosphosites (Fig. 8E). Regardless,
relevant EGFR substrates and other relevant sites involved in
EGFR downstream signaling (Fig. 8, F and G), were consis-
tently regulated across the majority of the replicates analyzed,
although, as expected, the statistical significance of the fold-
change was more robust when using at least 6 replicates.CONCLUSION
Our extensive optimization of phosphopeptide enrichment
conditions elucidated that the key parameters, including
beads-to-peptide ratio, binding time, competitive-binder
concentration, and sample volume, markedly influence phos-
phopeptide enrichment efficiency, phosphopeptide recovery,
and phosphosite localization scores. Particularly, our findings
underscore that multiply-phosphorylated peptides exhibit
enhanced affinity towards Zr-IMAC HP beads, leading to their
preferential enrichment under competitive binding conditions
such as high glycolic acid concentration or low bead volumes.
In line with others (44), we observed that the highestusing different cell input amounts in a 2-round pooled sequential
s show the mean numbers of phosphosites (light color), and class I
ental replicates using different cell input amounts in a 2-round pooled
, rank plot based on log2 intensities of phosphosites identified in the 1
ulatory sites from transcription factors are highlighted in red. Repre-
grams at MS2 fragment level of selected sites highlighted in panel (C).
size of the nodes indicates the number of phosphosites identified in the
beled in the outer ring. Each section of the outer ring (41) corresponds
nown to be regulatory (51).
FIG. 8. Response to EGF treatment as a quantification benchmark of phosphoproteome analysis on an Orbitap Astral. A, experimental
design. 12 replicates of HeLa cells treated with EGF for 8 min were compared to 12 replicates of non-treated HeLa cells. Samples were analyzed
in a nested manner using either 3, 6, 9 or 12 replicates of each condition. B, average number of class I phosphosites identified in Spectronaut
Systematic Optimization for Phosphoproteomics
Mol Cell Proteomics (2024) 23(5) 100754 17
Systematic Optimization for Phosphoproteomicsphosphoproteome coverage does not necessarily correspond
to the highest enrichment efficiency and therefore recommend
adapting enrichment conditions accordingly to the specific
needs when aiming for either the highest phosphoproteome
depth, enrichment efficiency, or proportion of multiply-
phosphorylated peptides.
Throughout our investigation of enrichment efficiency, we
found that enrichment efficiency calculated from abundance
exhibited consistently higher values than that calculated from
peptide counts. A potential explanation for the comparably
low enrichment efficiency based on counts in our DIA-based
study in contrast to other data-dependent acquisition (DDA)
based studies reporting values >90% (13, 17, 22) could be
that DIA provides higher peptide identification rates than DDA
(45–47), increasing the detection of low-abundant non-phos-
phorylated peptides and possibly biasing count-based
enrichment efficiency towards lower values than expected.
Enrichment efficiency based on abundance could therefore
serve as a potentially more nuanced representation of peptide
versus phosphopeptide ratio in phosphoproteomics samples,
especially in DIA-based studies.
We propose sequential phosphopeptide enrichment as a
powerful strategy to further amplify the depth of phospho-
peptide analysis. Our study indicates that while initial enrich-
ment rounds demonstrate higher enrichment efficiency,
subsequent rounds do not, providing only minimal improve-
ments in phosphoproteome depth after the third round.
Therefore, a sequential approach with two enrichment rounds
seems to be favorable for most applications in a range from
20 μg to 2.5 μg of peptide input, although more rounds might
offer further improvement for high peptide input amounts.
Importantly, the post-acquisition analysis in the search en-
gine (i.e., Spectronaut) of separate fractions or experimental
conditions has so far been the standard method for method
optimization. However, in the case of sequential enrichment,
the data analysis of one phosphopeptide entity with multiple
LFQ intensities derived from independent enrichments is not
trivial. It can result in higher variability and/or increased CVs,
hindering subsequent data interpretation, especially when
performing label-free quantification. Our data shows that
pooling fractions into a single LC-MS/MS analysis is a good
alternative to circumvent these issues while decreasing the
LC-MS/MS analysis time. In this regard, we present anusing 3, 6, 9 or 12 replicates for the search. Striped columns reflect the nu
class I phosphosites identified in Spectronaut using 3, 6, 9 or 12 replicat
quantified in at least 75% of the replicates used in the analysis. D, histo
replicates measured in the different analyses. At the top, allowing 3 o
replicates. At the bottom, allowing three or more valid values. Dashed ve
number of significantly regulated phosphosites plotted against the q-valu
F, heatmap of relevant sites in the EGFR signaling pathway. First heatm
replicates). Second heatmap shows the log2 fold changes (EGF versus C
versus 9, 6 versus 6 or 3 versus 3 replicates. The third heatmap shows th
using 12 versus 12, 9 versus 9, 6 versus 6 or 3 versus 3 replicates. G, Bar
pathway in control (light color) and EGF treated samples (dark color).
18 Mol Cell Proteomics (2024) 23(5) 100754improved strategy based on the incremental addition of beads
and subsequent fraction pooling, which offers up to 20%
boosted phosphoproteome coverage compared to standard
enrichment, while maintaining high sample-throughput and
straightforward data analysis.
Finally, we report that our optimized phosphoproteomics
pipeline can be translated to the newest generation of mass
spectrometers, such as the Orbitrap Astral, that can increase
the phosphopeptide coverage by 2-fold, allowing for deep
phosphoproteomics analysis without the need to scale up the
starting cell amounts. Evaluation of the repeatability of our
optimized workflow in combination with data acquisition on
the Orbitrap Astral indicated that the full workflow, from cell
culture to MS acquisition and phosphopeptide analysis, yiel-
ded repeatable deep phosphoproteomes across all replicates.
Moreover, we observed that the regulation of many phos-
phosites involved in relevant signaling pathways can be
already measured using only three biological replicates,
although at least six replicates are recommended for statisti-
cal analysis. Future research might explore the refinement of
enrichment conditions that simultaneously maximize both
enrichment efficiency and phosphoproteome depth. Further-
more, as bead type/composition (13, 22, 48), bead metal pu-
rity (43) and affinity material format (49) have also been shown
to influence phosphopeptide yield, enrichment efficiency, and
multiplicity of enriched phosphopeptides, further comparison
and optimization of the beads themselves in terms of capacity
and availability of binding sites could potentially further
improve phosphopeptide enrichment. Combining optimal
enrichment conditions with strategic application-tailored
pooling of sequential enrichments could pave the way for
more comprehensive phosphoproteomics investigations,
allowing this workflow to be translated to relevant applica-
tions, such as single-spheroid analysis for high-throughput
drug screening (12) or in vivo signaling using tissues (50).
Although more specific applications might require further
optimization, this work can serve as a roadmap for sensitive
phosphoproteomics analysis.DATA AVAILABILITY
The mass spectrometry proteomics data and supplemental
files have been deposited to the ProteomeXchangember of phosphosites quantified in all replicates. C, average number of
es for the search. Striped columns reflect the number of phosphosites
grams of the coefficient of variation at the phosphosite level between
r more valid values. At the top, allowing only values quantified in all
rtical lines indicate the median value for the coefficient of variation. E,
e for the analysis using 6 (light blue), 9 (brown) or 12 replicates (yellow).
ap shows the relative intensities (plotted as z-score per site across
ontrol) measured in the different experimental designs: 12 versus 12, 9
e -log10 of the q-value obtained from the two-sample t test performed
plots of the mean log2 intensity of relevant sites in the EGFR signaling
Systematic Optimization for PhosphoproteomicsConsortium via the PRIDE partner repository with the dataset
identifier PXD045601.
Supplemental Data—This article contains supplemental
data.
Acknowledgments—The authors would like to thank Justin
Jordaan, Isak Gerber and Stoyan Stoychev (ReSyn Bio-
sciences) for their valuable input on experimental design and
results.
Funding and additional information—Work at The Novo
Nordisk Foundation Center for Protein Research (CPR) is
funded in part by a donation from the Novo Nordisk Foun-
dation (NNF14CC0001). This work has also been funded as
part of EPIC-XS project under the grant agreement no. 823839
funded by the Horizon 2020 programme of the European
Union. P.B. was supported by the “International Exchange
Program” (Vienna Doctoral School in Chemistry, DoSChem),
the funding program “Internationale Kommunikation” (Öster-
reichische Forschungsgemeinschaft, ÖFG) and the “Eras-
mus+ Traineeship Mobility” program. C.K. is supported by the
Marie Skłodowska Curie European Training Network
“PUSHH” (grant number No. 861389).
Author contributions—A. M.-V, J. V. O., C. K., and P. B.
conceptualization; P. B. methodology; P. B., C. K., and I. P.
investigation; P. B. and A. M.-V. formal analysis; P. B., A. M.
V., and J. V. O. writing–original draft; P. B., I. P., C. K., C. G., A.
M. V., and J. V. O. writing–review and editing.
Conflict of interest—The authors declare that they have no
conflicts of interest with the contents of this article.
Abbreviations—The abbreviations used are: DHB, 2,5-
dihydoxybenzoic acid; DIA, data-independent acquisition;
FA, formic acid; GA, glycolic acid; IMAC, immobilized metal
affinity chromatography; MS, mass spectrometry; MOAC,
metal oxide affinity chromatography; nDIA, narrow-window
data-independent acquisition; PBS, phosphate-buffered sa-
line; PTM, post-translational modification; TCEP, Tris(2-
carboxyethyl)phosphine; TFA, trifluoroacetic acid.
Received November 30, 2023, and in revised form, March 8, 2024
Published, MCPRO Papers in Press, March 27, 2024, https://doi.org/
10.1016/j.mcpro.2024.100754
REFERENCES
1. Humphrey, S. J., Karayel, O., James, D. E., and Mann, M. (2018) High-
throughput and high-sensitivity phosphoproteomics with the EasyPhos
platform. Nat. Protoc. 13, 1897–1916
2. Olsen, J. V., and Mann, M. (2013) Status of large-scale analysis of post-
translational modifications by mass spectrometry. Mol. Cell. Proteomics
12, 3444–3452
3. Lundby, A., Andersen, M. N., Steffensen, A. B., Horn, H., Kelstrup, C. D.,
Francavilla, C., et al. (2013) In vivo phosphoproteomics analysis reveals
the cardiac targets of β-adrenergic receptor signaling. Sci. Signal. 6, rs114. Ville´n, J., Beausoleil, S. A., Gerber, S. A., and Gygi, S. P. (2007) Large-scale
phosphorylation analysis of mouse liver. PNAS 104, 1488–1493
5. Olsen, J. V., Blagoev, B., Gnad, F., Macek, B., Kumar, C., Mortensen, P.,
et al. (2006) Global, in vivo, and site-specific phosphorylation dynamics in
signaling networks. Cell 127, 635–648
6. Ochoa, D., Jarnuczak, A. F., Vie´itez, C., Gehre, M., Soucheray, M., Mateus,
A., et al. (2020) The functional landscape of the human phosphopro-
teome. Nat. Biotechnol. 38, 365–373
7. Sharma, K., D'Souza, R. C. J., Tyanova, S., Schaab, C., Wis´niewski, J. R.,
Cox, J., et al. (2014) Ultradeep human phosphoproteome reveals a
distinct regulatory nature of Tyr and Ser/Thr-based signaling. Cell Rep. 8,
1583–1594
8. Bekker-Jensen, D. B., Kelstrup, C. D., Batth, T. S., Larsen, S. C., Haldrup,
C., Bramsen, J. B., et al. (2017) An optimized shotgun strategy for the
rapid generation of comprehensive human proteomes. Cell Syst. 4, 587–
599.e4
9. Guzman, U. H., Martinez-Val, A., Ye, Z., Damoc, E., Arrey, T. N., Pashkova,
A., et al. (2024) Ultra-fast label-free quantification and comprehensive
proteome coverage with narrow-window data-independent acquisition.
Nat. Biotechnol. https://doi.org/10.1038/s41587-023-02099-7
10. Wu, R., Haas, W., Dephoure, N., Huttlin, E. L., Zhai, B., Sowa, M. E., et al.
(2011) A large-scale method to measure absolute protein phosphorylation
stoichiometries. Nat. Methods 8, 677–683
11. Yang, S., Han, Y., Li, Y., Zhang, L., Yan, G., Yuan, J., et al. (2023) Rapid and
high-sensitive phosphoproteomics elucidated the spatial dynamics of the
mouse brain. Anal. Chem. 95, 10703–10712
12. Martínez-Val, A., Fort, K., Koenig, C., van der Hoeven, L., Franciosa, G.,
Moehring, T., et al. (2023) Hybrid-DIA: intelligent data acquisition in-
tegrates targeted and discovery proteomics to analyze phospho-
signaling in single spheroids. Nat. Commun. 14, 3599
13. Leutert, M., Rodríguez-Mias, R. A., Fukuda, N. K., and Ville´n, J. (2019) R2-
P2 rapid-robotic phosphoproteomics enables multidimensional cell
signaling studies. Mol. Syst. Biol. 15, e9021
14. Tape, C. J., Worboys, J. D., Sinclair, J., Gourlay, R., Vogt, J., McMahon, K.
M., et al. (2014) Reproducible automated phosphopeptide enrichment
using magnetic TiO2 and Ti-IMAC. Anal. Chem. 86, 10296–10302
15. Koenig, C., Martinez-Val, A., Naicker, P., Stoychev, S., Jordaan, J., and
Olsen, J. V. (2023) Protocol for high-throughput semi-automated label-
free- or TMT-based phosphoproteome profiling. STAR Protoc. 4, 102536
16. Emdal, K. B., Palacio-Escat, N., Wigerup, C., Eguchi, A., Nilsson, H., Bek-
ker-Jensen, D. B., et al. (2022) Phosphoproteomics of primary AML pa-
tient samples reveals rationale for AKT combination therapy and p53
context to overcome selinexor resistance. Cell Rep. 40, 111177
17. Chen, C.-W., Tsai, C.-F., Lin, M.-H., Lin, S.-Y., and Hsu, C.-C. (2023)
Suspension trapping-based sample preparation workflow for in-depth
plant phosphoproteomics. Anal. Chem. 95, 12232–12239
18. Zhou, H., Ye, M., Dong, J., Corradini, E., Cristobal, A., Heck, A. J. R., et al.
(2013) Robust phosphoproteome enrichment using monodisperse
microsphere-based immobilized titanium (IV) ion affinity chromatography.
Nat. Protoc. 8, 461–480
19. Gates, M. B., Tomer, K. B., and Deterding, L. J. (2010) Comparison of metal
and metal oxide media for phosphopeptide enrichment prior to mass
spectrometric analyses. J. Am. Soc. Mass Spectrom. 21, 1649–1659
20. Leitner, A. (2010) Phosphopeptide enrichment using metal oxide affinity
chromatography. Trends Anal. Chem. 29, 177–185
21. Li, Q., Ning, Z., Tang, J., Nie, S., and Zeng, R. (2009) Effect of peptide-to-
TiO2 beads ratio on phosphopeptide enrichment selectivity. J. Prote-
ome Res. 8, 5375–5381
22. Arribas Diez, I., Govender, I., Naicker, P., Stoychev, S., Jordaan, J., and
Jensen, O. N. (2021) Zirconium(IV)-IMAC revisited: improved performance
and phosphoproteome coverage by magnetic microparticles for phos-
phopeptide affinity enrichment. J. Proteome Res. 20, 453–462
23. Jensen, S. S., and Larsen, M. R. (2007) Evaluation of the impact of some
experimental procedures on different phosphopeptide enrichment tech-
niques. Rapid Commun. Mass Spectrom. 21, 3635–3645
24. Li, J., Wang, J., Yan, Y., Li, N., Qing, X., Tuerxun, A., et al. (2022) Compre-
hensive evaluation of different TiO2-based phosphopeptide enrichment
and fractionation methods for phosphoproteomics. Cells 11, 2047
25. Palmisano, G., Lendal, S. E., and Larsen, M. R. (2011) Titanium dioxide
enrichment of sialic acid-containing glycopeptides. Methods Mol. Biol.
753, 309–322Mol Cell Proteomics (2024) 23(5) 100754 19
Systematic Optimization for Phosphoproteomics26. Palmisano, G., Lendal, S. E., Engholm-Keller, K., Leth-Larsen, R., Parker, B.
L., and Larsen, M. R. (2010) Selective enrichment of sialic acid-containing
glycopeptides using titanium dioxide chromatography with analysis by
HILIC and mass spectrometry. Nat. Protoc. 5, 1974–1982
27. Larsen, M. R., Thingholm, T. E., Jensen, O. N., Roepstorff, P., and
Jørgensen, T. J. D. (2005) Highly selective enrichment of phosphorylated
peptides from peptide mixtures using titanium dioxide microcolumns.
Mol. Cell. Proteomics 4, 873–886
28. Thingholm, T. E., Jensen, O. N., Robinson, P. J., and Larsen, M. R. (2008)
SIMAC (sequential elution from IMAC), a phosphoproteomics strategy for
the rapid separation of monophosphorylated from multiply phosphory-
lated peptides. Mol. Cell. Proteomics 7, 661–671
29. Tsai, C.-F., Hsu, C.-C., Hung, J.-N., Wang, Y.-T., Choong, W.-K., Zeng, M.-
Y., et al. (2014) Sequential phosphoproteomic enrichment through com-
plementary metal-directed immobilized metal ion affinity chromatog-
raphy. Anal. Chem. 86, 685–693
30. Bekker-Jensen, D. B., Bernhardt, O. M., Hogrebe, A., Martinez-Val, A.,
Verbeke, L., Gandhi, T., et al. (2020) Rapid and site-specific deep phos-
phoproteome profiling by data-independent acquisition without the need
for spectral libraries. Nat. Commun. 11, 787
31. Lou, R., Liu, W., Li, R., Li, S., He, X., and Shui, W. (2021) DeepPhospho
accelerates DIA phosphoproteome profiling through in silico library
generation. Nat. Commun. 12, 6685
32. Batth, T. S., Tollenaere, M. X., Rüther, P., Gonzalez-Franquesa, A., Prab-
hakar, B. S., Bekker-Jensen, S., et al. (2019) Protein aggregation capture
on microparticles enables multipurpose proteomics sample preparation.
Mol. Cell. Proteomics 18, 1027–1035
33. Bekker-Jensen, D. B., Martínez-Val, A., Steigerwald, S., Rüther, P., Fort, K.
L., Arrey, T. N., et al. (2020) A compact quadrupole-orbitrap mass
spectrometer with FAIMS interface improves proteome coverage in short
LC gradients. Mol. Cell. Proteomics 19, 716–729
34. Reiter, L., Rinner, O., Picotti, P., Hüttenhain, R., Beck, M., Brusniak, M.-Y.,
et al. (2011) mProphet: automated data processing and statistical vali-
dation for large-scale SRM experiments. Nat. Methods 8, 430–435
35. Martinez-Val, A., Bekker-Jensen, D. B., Hogrebe, A., and Olsen, J. V. (2021)
Data processing and analysis for DIA-based phosphoproteomics using
Spectronaut. Methods Mol. Biol. 2361, 95–107
36. Gu, Z., Eils, R., and Schlesner, M. (2016) Complex heatmaps reveal patterns
and correlations in multidimensional genomic data. Bioinformatics 32,
2847–2849
37. Wickham, H. (2016) ggplot2. Springer International Publishing, Cham,
Switzerland
38. Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L., François, R.,
et al. (2019) Welcome to the tidyverse. J. Open Source Softw. 4, 168620 Mol Cell Proteomics (2024) 23(5) 10075439. Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D.,
et al. (2003) A software environment for integrated models of biomole-
cular interaction networks. Genome Res. 13, 2498–2504
40. Marinis, I. D., Lo Surdo, P., Cesareni, G., and Perfetto, L. (2022) SIGNO-
RApp: a Cytoscape 3 application to access SIGNOR data. Bioinformatics
38, 1764–1766
41. Legeay, M., Doncheva, N. T., Morris, J. H., and Jensen, L. J. (2020) Visualize
omics data on networks with Omics visualizer, a Cytoscape app.
F1000Res. 9, 157
42. Palmisano, G., Parker, B. L., Engholm-Keller, K., Lendal, S. E., Kulej, K.,
Schulz, M., et al. (2012) A novel method for the simultaneous enrichment,
identification, and quantification of phosphopeptides and sialylated gly-
copeptides applied to a temporal profile of mouse brain development.
Mol. Cell. Proteomics 11, 1191–1202
43. Ye, J., Zhang, X., Young, C., Zhao, X., Hao, Q., Cheng, L., et al. (2010)
Optimized IMAC-IMAC protocol for phosphopeptide recovery from
complex biological samples. J. Proteome Res. 9, 3561–3573
44. [preprint] Oliinyk, D., Will, A., Schneidmadel, F. R., Humphrey, S. J., and
Meier, F. (2023) μPhos: a scalable and sensitive platform for functional
phosphoproteomics. bioRxiv. https://doi.org/10.1101/2023.04.04.535617
45. Bruderer, R., Bernhardt, O. M., Gandhi, T., Xuan, Y., Sondermann, J.,
Schmidt, M., et al. (2017) Optimization of experimental parameters in
data-independent mass spectrometry significantly increases depth and
reproducibility of results. Mol. Cell. Proteomics 16, 2296–2309
46. Michalski, A., Cox, J., and Mann, M. (2011) More than 100,000 detectable
peptide species elute in single shotgun proteomics runs but the majority is
inaccessible todata-dependentLC-MS/MS.J.ProteomeRes.10, 1785–1793
47. Steger, M., Demichev, V., Backman, M., Ohmayer, U., Ihmor, P., Müller, S.,
et al. (2021) Time-resolved in vivo ubiquitinome profiling by DIA-MS re-
veals USP7 targets on a proteome-wide scale. Nat. Commun. 12, 5399
48. Post, H., Penning, R., Fitzpatrick, M. A., Garrigues, L. B., Wu, W., MacGil-
lavry, H. D., et al. (2017) Robust, sensitive, and automated phospho-
peptide enrichment optimized for low sample amounts applied to primary
hippocampal neurons. J. Proteome Res. 16, 728–737
49. Ruprecht, B., Koch, H., Medard, G., Mundt, M., Kuster, B., and Lemeer, S.
(2015) Comprehensive and reproducible phosphopeptide enrichment
using iron immobilized metal ion affinity chromatography (Fe-IMAC) col-
umns. Mol. Cell. Proteomics 14, 205–215
50. Snieckute, G., Ryder, L., Vind, A. C., Wu, Z., Arendrup, F. S., Stoneley, M.,
et al. (2023) ROS-induced ribosome impairment underlies ZAKα-medi-
ated metabolic decline in obesity and aging. Science 382, eadf3208
51. Hornbeck, P. V., Zhang, B., Murray, B., Kornhauser, J. M., Latham, V., and
Skrzypek, E. (2015) PhosphoSitePlus, 2014: mutations, PTMs and reca-
librations. Nucleic Acids Res. 43, D512–D520