plexDIA for scalable single cell analysis

– improving quantification and sensitivity by

TIMS separation
Single cell proteomics in a biological research context needs to deliver high sensitivity
for deep proteome coverage while maintaining quantitative accuracy at scale.

Abstract
Labeling approaches outperform label-free analyses, however, multiplexing on the fragment Keywords:
level typically results in a reduced proteome depth due to its dependence on data dependent plexDIA, dia-PASEF,
timsTOF Ultra, nanoElute2
acquisition. Precursor based non-isobaric labeling, as the here shown plexDIA approach, uses
non-isobaric labeling and is hence compatible with data independent acquisition, overcoming
stochastic precursor sampling and reduces missing not at random type data incompleteness.

Here, we applied plexDIA in a fully automated 3-label approach performed on CFPAC-1 pancreatic
cancer, WM-989 melanoma, and THP-1 leukemia cell lines (126 single cells in total) using the
timsTOF Ultra in dia-PASEF mode. This approach allowed to analyze almost 100 samples per
day with a proteome depth of more than 3000 proteins per cell and quantification of > 80,000
precursors per individual plexDIA run. Comparison to bulk material demonstrated preserved fold
changes between the three cell types.

Introduction
Over the last few years single cell proteomics by mass spectrometry has significantly increased
the number of peptides and proteins that can be detected and confidently identified from single
mammalian cells. Using these rapid advances for biomedical research also requires simultaneously
sustaining an increasing quantitative accuracy and throughput. For example, studies [1] have
demonstrated the feasibility of calculating and interpreting protein correlations across single cells
if sufficient quantitative accuracy, precision, and throughput can be achieved and maintained.

Achieving high quantitative accuracy, precision, and throughput is the focus of the workflow
outlined here. Specifically, it demonstrates a workflow that allows for flexible and automated
sample preparation that can be adapted to any desired set of mass tags for multiplexing
single cells. The multiplexing enables increased throughput in linear proportion to the number

Christoph Krisp 1, Andrew Leduc 2, Luke Khoury 2, Nikolai Slavov 2, Torsten Müller 1, Daniel Hornburg 3, Markus Lubeck 1,
Gary Kruppa 4; 1 Bruker Daltonics GmbH & Co. KG, Bremen, Germany; 2 Departments of Bioengineering, Biology,
Chemistry and Chemical Biology, Single Cell Proteomics Center, and Barnett Institute, Northeastern University, Boston,
MA, USA; 3 Bruker Scientific LLC, San Jose, CA, USA; 4 Bruker s.r.o., Brno, Czech Republik.
 of labels used while preserving the protein coverage and quantitative accuracy as previously
demonstrated with plexDIA [2]. This workflow and the quantitative accuracy also benefit
significantly from the efficient ion utilization enabled by trapped ion mobility spectrometry,
which enables the isolation, fragmentation, and analysis of a large fraction of the ions delivered
to the mass spectrometer. Furthermore, this technology operates with short duty cycles,
increasing the frequency of sampling of the eluting precursor ions and fragments and therefore
increasing the robustness and precision of quantification. We further increase this frequency by
using multiple MS1 frames per duty cycle. The additional MS1 scans included in the duty cycles
result in MS1 full range scans spaced by 300 ms, which results in about 10 points across a
3 second elution peak. This sampling frequency is significantly higher than the 3 points across
a 3 second peak expected for a standard dia-PASEF method with about 1 s cycle time. The
benefits of this frequent sampling have been demonstrated previously [3].

While multiplexing can substantially increase the throughput of mass spectrometry-based

proteomics, especially for single cell applications, flexible and easy to implement, as well as
scalable and sensitive sample preparation strategies are key for highest data quality. Yet, many
of the sample preparation approaches for single-cell proteomics do not implement multiplexing
or implement it assuming a fixed number of labels used.
To relax this limitation and increase throughput, this application note demonstrates an nPOP
workflow that can accommodate different numbers of labels and different reagents simply by
programming different configurations of droplets deposited on the surface of glass slides. In this
workflow, each single cell is lysed, its proteins are digested and peptides are labeled in droplets on
the surface of unpatterned glass slides. Since the droplets have volumes of 8 to 39 nL, the amounts
of reagents used per single cell and the contact to surfaces can be minimized accordingly,
crucial to increasing the sensitivity and sample purity. This workflow can simultaneously process
thousands of single cells. The labeled peptides are then automatically pooled and transferred into
multi-well plates that can be plugged into auto samplers (e.g., nanoElute 2), thus making the
workflow compatible with full automation. Therefore, this workflow provides an automated option
for scaling up single-cell proteomics to the automated analysis of thousands of single cells.

Methods
The CFPAC-1 pancreatic cancer, WM-989 melanoma, and THP-1 leukemia cell lines, for
simplicity in the following termed PDAC, Melanoma, and Monocytes, were purchased
from ATCC. Cells were thawed from liquid nitrogen and suspended directly in 1X PBS at a
concentration of 300 cells per µL for cell sorting.

A
8 nL Single 13 nL 20 nL Pool labeled sets
DMSO cell digest mix label to 384 well plate

10 min 0.5-2 hours Overnight 1 hour 1-5 hours

1 2 3 4
B Label C
0
4 30
8
20
Cell diameter [µm3]

Sample
Melanoma 20
dropYPos

Monocyte
PDAC
40

10

60
0
Melanoma Monocyte PDAC
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
dropXPos Cell type

Figure 1
Workflow and meta information
A Schematic depicting sample preparation workflow and LC-MS/MS analysis strategy. B Schematic layout of cell positioning on the fluorocarbon-coated glass
slides used in this experiment with label randomization information. C Cell diameter distributions within the cell type groups taken from the cellenONE report.
 3-plexDIA 120 label-free 3-plexDIA
A

Samples per day (SPD)

100

Throughput
Sample 3, ∆8 n-plex

ng
Intensity 80

xi
le
Sample 2, ∆4 n=3

tip
60

ul
M
Sample 1, ∆0
40

20 n=1
Retention time 0
0 5 10 15 20
30 min run time Instrument time [h]
Label-free
Sample 1 Sample 2 Sample 3
Intensity

Intensity

Intensity
Retention time Retention time Retention time

30 min run time 30 min run time 30 min run time

90 min run time

Figure 2
3-plexDIA analysis
A Schematic highlighting the advantages of plexDIA over label-free analysis in sample throughput. B Illustration of precursors seen in a m/z versus ion
mobility (1/k0) heatmap at a given retention time with associated MS spectrum of a precursor seen in all three cell types (left) and precursors only present
in 2 of 3 cell types (right).

The nPOP sample preparation procedure was used to prepare single cells for multiplexed
analysis by plexDIA. Briefly, the CellenONE cell sorter and liquid handler robotic system
deposited single cells in 300 pL of 1x PBS into 8 nL droplets of 100% DMSO on the surface of
a fluorocarbon coated glass slide for cell lysis. Then, 13 nL of master mix containing 100 ng/µL
Promega trypsin gold, 5 mM HEPES pH 8.5 and 0.025% weight DDM was added to each single
cell droplet on the slide. Droplet evaporation was prevented by setting the relative humidity to
75% and plate temperature to the dew point for overnight digest. The next day, cells were labeled
by dissolving mTRAQ multiplexed labeling reagents in DMSO at a concentration of 1/40th unit per
µL and dispensing 20 nL of either d0, d4 or d8 mass tags to each single-cell containing droplet.
To facilitate the labeling reaction, 20 nL of 100 mM TEAB pH 8.5 was added and labeling proceeded
for 1 hour. Single cell samples were then pooled by the CellenONE with 50% water 50%
Acetonitrile and deposited to a 384 well plate. Samples were finally dried down and stored at -20°C
until resuspension in 1 µL of water for LC/MS analysis. The samples were injected onto a 25 cm
Aurora Ultimate column (75 µm, 1.7 µm, IonOpticks) using a nanoElute® 2. Peptides were eluted
within a 25 min active gradient (30 min total acquisition time) and detected on a timsTOF Ultra
in dia-PASEF mode using a 25 Da fixed window methods with 3 mass range focus switches per
TIMS ramp with 8 TIMS ramps per cycle. Acquired data were searched against canonical human
protein sequences including splice variants (Uniprot reviewed canonical sequences + isoforms)
using DIA-NN version 1.8.1 [4] setting additional commands, as described by Derks et al. [2]:
{fixed-mod mTRAQ, 140.0949630177, nK}, {channels mTRAQ,0,nK,0:0; mTRAQ, 4, nK,
4.0070994:4.0070994; mTRAQ, 8, nK, 8.0141988132:8.0141988132}, {peak-translation}, {original-
mods}, {report-lib-info}, {ms1-isotope-quant}. This search used the spectral library that was
previously generated from higher cell number plexDIA runs of Melanoma, PDAC, and Monocytes.
Data were acquired and processed in the Single Cell Proteomics Center of Professor Nikolai
Slavov at Northeastern University in Boston, MA, USA. Data analysis was performed using
QuantQC for R (https://github.com/SlavovLab/QuantQC).
 A B
4000
90,000

Protein groups
3000

Precursors
60,000
2000

30,000
1000

0 0
0 10 20 30 40 Melanoma Monocyte PDAC
Run Cell type

C Type D R2 = 0.74
negative ctrl 95.0 Run
15
single cells order
40
# of samples

precursor area)
Log2(summed
92.5 30
10
20
90.0 10
5

87.5

0
6.5 7.0 7.5 8.0 8.5 10 11 12
Log2(summed precursor area) Log2(cell volume [µm3])
Figure 3
Results of data processing
A Distribution of precursor identification rates across the 42 3-plexDIA experiments acquired at 96 SPD, B Boxplot
of protein group identifications per cell type acquired at 96 SPD, C histogram of DIA-NN 1.8.1 log10 (summed precursor
area) of each analyzed single cell and the negative controls, D Correlation of the estimated cell volume (µm3) versus the
log2 (summed precursor area) of each analyzed single cell.

Results and discussion

Samples prepared for plexDIA analysis were analyzed on a timsTOF Ultra. Multiplexing
increases the sample throughput per day and scales with the number of labels. Including
overhead time and a gradient of 30 min, label free analysis result in a throughput of 32 runs
per day (SPD), while a 3-plex approach allows for the analysis of 96 SPD and reducing the
overhead time per sample by two-thirds (Figure 2A). Intrinsically, these samples are much more
complex multiplying the number of detectable precursors with the number of plexes. plexDIA
is therefore benefitting from TIMS separation, leading to reduced presence of interfering
molecules of similar mass in precursor and fragment level. This enables accurate quantification,
increased precursor detection and better association of non-isobaric mass-tagged peptides, as
they appear at the same collisional cross section (CCS) value with mass-tag specific m/z offset.
Figure 2B shows examples of ion mobility resolved 3-plex labeled precursors at different 1/k0
and m/z positions in a mobility vs. mass-to-charge based heatmap.

In total, 42 plexDIA runs including 126 samples in a 3-plex manor with 41 PDACs, 41
monocytes, 39 melanoma cells and 5 negative controls (no cell) were analyzed at 96 SPD using
the nanoElute 2 timsTOF Ultra setup. Data processing was performed with DIA-NN 1.8.1 [4]
with additional settings for the used labels and to enable MS1 translation for MS1 based
quantification (see methods or Derks et al. [2]). Processing of the 42 runs resulted on average
in the identification and quantification of 80,000 precursors (sum of precursors across channels;
Figure 3A) in 25 min of active gradients. This translates to on average 3000 protein groups for
Monocytes, 3100 protein groups for Melanoma cells, 3150 protein groups for PDAC cells, and
in total 4486 protein groups identified in this experiment (Figure 3B). Intercalated MS1 scans in
the dia-PASEF method increased the number of data points per peak from mean 4.3 to 14.6.

The distribution of log10 summed precursor intensity per cell (Figure 3C) demonstrated two
maxima, the one at a value of 8 is manly derived from the comparably small monocytes
(mean log10 (summed precursor area) of 8.05) and the other at a value of 8.5 is represented
 by the larger Melanoma (mean log10 (summed precursor area) of 8.35) and PDAC cells (mean
log10 (summed precursor area) of 8.48). The negative controls on the other hand show log10
summed precursor intensity at least 10-fold lower than the single cells, demonstrating a
clean background and an accurate data extraction of individual channel information out of the
3-plexDIA samples.

The estimated volume per cell (µm3) correlated well with the sum of precursor intensities
calculated for each cell (Figure 3D) with a correlation score of 0.73. This indicates that with an
increase in cell volume the protein content increases proportionately.

Principal component analysis (Figure 4A) after k-nearest-neighbor (KNN) based data imputation
clearly distinguished the three cell types. The Monocytes formed one cluster, whereas the
Melanoma cells formed two, one representing most Melanoma cells and one small subcluster
of 7 cells. The DPAC cells may also form a subcluster.

Protein abundance ratios calculated between the three cell types based on single cell data were
compared to the ratios of bulk experiments performed in label free mode on these cells and
showed good agreement in protein abundance ratios for Monocytes/PDAC, Melanoma/PDAC
and Monocytes/Melanoma (Figure 4B).
Protein abundance profiles across the three cell types were compared, demonstrating the
expected large differences in protein abundances in the three cell lines (Figure 5A-C). Among
the proteins showing the largest abundance increase in melanoma compared to PDAC cell
line and the Monocyte cell line were the cell surface proteins neural cell adhesion molecule
L1 (L1CAM) and the CD44 antigen. L1CAM is discussed as a potential target for malignant
melanoma therapy [5] and CD44 is known to be at high abundance in metastatic melanoma [6].
The monocyte cell line showed highest abundance difference to the melanoma and PDAC cell
lines for carbonic anhydrase 2 (CA2) and adenylyl cyclase-associated protein 1, both are involved
in immune response regulation in monocyte [7].

Interleukin 18 and aldehyde dehydrogenase 1A1 were elevated in abundance in the PDAC cell
line compared to the Melanoma and Monocyte cell lines. IL18 has been shown to be elevated
in pancreatic diseases like pancreatitis but also pancreatic cancer [8]. Aldehyde dehydrogenases
are commonly elevated in solid tumors including pancreatic cancer.

Melanoma
0.1 Monocyte
PDAC
PC2

0 Figure 4
Single cell based PCA and correlation to bulk samples
A Sample projection in first and second principal component after KNN data
imputation. B Correlation of Log2 protein abundance ratios calculated on single cell
data versus bulk data acquired in label-free of the three cell types.
-0.10 -0.05 0 0.05 0.10
PC1

B Log2(Monocyte/PDAC) Log2(Melanoma/PDAC) Log2(Monocyte/Melanoma)

4.0 4.0 4.0

Bulk label-free

Bulk label-free

Bulk label-free

0 0 0

-4.0 -4.0 -4.0

-2.5 0 2.5 5.0 -5.0 -2.5 0 2.5 5.0 -2.5 0 2.5 5.0
Single cell Single cell Single cell
 A L1CAM CD44
2.5 2.0

log2(protein abundance)

log2(protein abundance)
1.0
0
0

-2.5 -1.0

-2.0
-5.0
Melanoma Monocyte PDAC Melanoma Monocyte PDAC
Cell type Cell type

B 5.0
CA2
4.0
CAP1
log2(protein abundance)

log2(protein abundance)
2.5 2.0

0
0

-1.0
-2.5

-4.0
Melanoma Monocyte PDAC Melanoma Monocyte PDAC
Cell type Cell type

IL18 ALDH1A1
C
2.0
log2(protein abundance)

log2(protein abundance)

2.0

0
0

-2.0
-2.0

Melanoma Monocyte PDAC Melanoma Monocyte PDAC

Cell type Cell type

Figure 5
Cell type specific protein abundance profiles
Selection of proteins showing cell type specific protein abundance profiles. A Neural cell adhesion molecule
L1 (L1CAM) and the CD44 antigen for the Melanoma cell line, B Carbonic anhydrase 2 (CA2) and adenylyl
cyclase-associated protein 1 (CAP1) for the Monocyte cell line, and C interleukin 18 (IL18) and aldehyde
dehydrogenase 1 A1 (ALDH1A1) for the PDAC cell line.

Conclusion
plexDIA for scalable single cell analysis at 96 samples per day speed.
TIMS separation of MOMA events for interference reduced MS1 and MS2 spectra.
Identification of > 80,000 precursors from a plexDIA run with mean 14.5 data points
per peak enabled by 4x MS1 scan intercalated dia-PASEF method.
More than 3000 protein groups identified per cell type and about 4400 protein groups
in total identified.
Sample clustering according to cell type with tissue specific protein abundance increases.
References
[1] Leduc A, et al. Genome Biol, 2022, 23:261
[2] Derks J, et al. Nat Biotechnol 2023, 41:50–59
[3] Wallmann G, et al. J Proteome Res. 2023, 22(10):3149-3158
[4] Demichev V, et al. Nat Commun 2022, 13:3944
[5] Ernst AK, et al. PLOS ONE, 2018, 13(2):e0192525
[6] Dietrich A, et al. Eur J Cancer. 1997, 33(6):926-30
[7] Lee S, et al. Cell Metab. 2014, 4, 19(3):484-97
[8] Li Z, et al. Cytokine Growth Factor Rev. 2019, 50:1-12

Further reading

Deep Proteomic Insights from bulk Setting and Maintaining the Single
to single cells Cell Proteomics Benchmark with the
www.bruker.com/en/applications/ timsTOF Ultra in action
academia-life-science/proteomics/single- www.bruker.com/de/news-and-events/
cell-proteomics.html webinars/2023/setting-and-maintaining-
the-single-cell-proteomics-benchmark.
html

Speed meets Sensitivity with the Everything you need. Maximum

timsTOF Ultra Sensitivity, Speed, and Robustness
www.bruker.com/en/products-and- with PepSep Ultra and Ultra XL
solutions/mass-spectrometry/timstof/ www.bruker.com/en/products-and-
timstof-ultra.html solutions/mass-spectrometry/lc-ms/
pepsep-columns.html

Immediately dive into results The gold standard for DIA proteomics
of your experiment analysis: Spectronaut®.
www.bruker.com/en/products-and- biognosys.com/software/spectronaut/
solutions/mass-spectrometry/ms-
software/proteoscape.html

Expanding capabilities, see the PASEF: discover the advantage

timsTOF SCP in action of SPEED with Confidence
www.youtube.com/ www.bruker.com/de/products-and-
watch?v=3zKvc73HmQ4&t=57s solutions/mass-spectrometry/timstof/
pasef.html
Bruker 4D-Proteomics™ Ecosystem

Benefit Feature

Universal range NanoElute 2 to deliver robustness and

of applications performance without compromise.

Ultra separation PepSep Ultra and PepSep Ultra XL deliver

ultra sensitivity and reproducibility at PASEF ®
speeds, ensuring high-quality results.

Single Cell cellenONE® system to provide hands-free

Handling end-to-end cell sorting and sample preparation
for single cells

Ultimate CSI ion sources: CaptiveSpray Ultra moves

performance boundaries and delivers robust performance
and simplicity for outstanding sensitivity.

Higher TIMS spreads sample complexity across

throughput, the mobility dimension for fast turnaround,
deeper profiles without compromise.

Enhanced Mobility offset, mass aligned (MOMA)

confidence with for confident identification and quantification
TIMS of isobaric or nearly isobaric peptides.

to change specifications without notice. © Bruker Daltonics 05-2024, LCMS-222, 1911577

Powerful Bruker ProteoScape™: Real-time data
insights processing for large experiments.

Bruker Daltonics is continually improving its products and reserves the right
Limitless Spectronaut® simplifies complex DIA
processing workflows with unmatched sensitivity and
accelerated data analysis.

4D-Proteomics™ Ecosystem by Bruker, empowering unmatched

speed, selectivity, sensitivity, and robustness for the analysis of
large cohorts of single cell cohorts.

For Research Use Only. Not for use in clinical diagnostic procedures.

Bruker Switzerland AG Bruker Scientific LLC

Fällanden · Switzerland Billerica, MA · USA
Phone +41 44 825 91 11 Phone +1 (978) 663-3660

[email protected] – www.bruker.com Learn more at www.bruker.com/timstof-ultra