Analytica Chimica Acta

High-throughput Single Cell Metabolomics and Cellular Heterogeneity Exploration by

Inertial Microfluidics coupled with Pulsed Electric Field-Induced Electrospray Ionization-
High Resolution Mass Spectrometry
--Manuscript Draft--

Manuscript Number: ACA-22-1439R1

Article Type: Full Length Article

Section/Category: MASS SPECTROMETRY

Keywords: Cellular heterogeneity; Single cell metabolomics; High-throughput; mass

spectrometry; Microfluidic device

Corresponding Author: Guowang Xu, Dr.

Dalian Institute of Chemical Physics, Chinese Academy of Sciences
Dalian, CHINA

First Author: Disheng Feng

Order of Authors: Disheng Feng

Hang Li

Tianrun Xu

Fujian Zheng

Chunxiu Hu

Xianzhe Shi

Guowang Xu, Dr.

Manuscript Region of Origin: CHINA

Abstract: Single cell metabolomics can obtain the metabolic profiles of individual cells and reveal
cellular heterogeneity. However, high-throughput single-cell mass spectrometry (MS)
analysis under physiological conditions remains a great challenge due to the presence
of complex matrix and extremely small cell volumes. Herein, a serpentine channel
microfluidic device which was designed to achieve continuous cell separation and
inertial focusing, was coupled with a pulsed electric field-induced electrospray
ionization-high resolution MS ( PEF-ESI-HRMS ) to achieve high-throughput single
cell analysis. The pulsed square wave electric field was applied to realize on-line cell
disruption and induce electrospray ionization. Single cells were analyzed under near-
physiological conditions at a throughput of up to 80 cells min -1 . More than 900
features were detected and approximately 120 metabolites were tentatively identified
from a single cell. Further, by continually analyzing more than 3000 MCF7 and HepG2
cells, discrimination of different cancer cells based on their individual metabolic profiles
was achieved by using the principal component analysis. The PEF-ESI-HRMS method
was also applied for the analysis of single yeast cells, and more than 40 metabolites
were annotated. This method is versatile and has good robustness, which is promising
for high-throughput single cell metabolomics analysis.

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Cover Letter(including Novelty Statement)

Dear Editor

Thank you for giving us the chance to improve our manuscript (Ms. No.: ACA-22-1439). Now we

have finished the revision based on the comments and suggestion from reviewers and you. I hope

this version is suitable for the publication in Anal. Chim. Acta.

Best regards

Prof. Dr. Guowang Xu

Response to Reviews

Response to Reviewers’ comments

Reviewer #1: Detailed comments are as follows:

1. The keywords section should start with the "cellular heterogene".
Response: Thank you for this suggestion, but "cellular heterogeneity" is a more widely used
keyword, so we used this word and moved it to the first place.

2.The title is not catchy and does not reflect essential contents.
Response: We have modified the title to be “High Throughput Single Cell Metabolomics and
Cellular Heterogeneity Exploration by Inertial Microfluidics coupled with Pulsed Electric
Field-Induced Electrospray Ionization-High Resolution Mass Spectrometry”.

3.Where is the practical application of this manuscript? It must be added.

Response: We have added the practical application regarding the discovery of heterogeneity in
metabolite levels within the same kind of cells, to the best of our knowledge this was rarely
reported (Page 18, line 20-59 and Page 19, line 1-7). The title of Section 3.3 has been modified as
“3.3 Practical applications of developed method in high-throughput analysis of single mammalian
cells”.

4. Check the grammar throughout the article and correct it. Proofread the article as many language
errors were identified.
Response: We have checked the grammar throughout the article and corrected the errors.

5.The introduction needs to be further revised to highlight the purpose of the study, you need to
introduce what others have studied and what needs further research. Besides, the following all of
references are recommended to be cited:
Green synthesis of DyBa2Fe3O7. 988/DyFeO3 nanocomposites using almond extract with dual
eco-friendly applications: Photocatalytic and antibacterial activities// Synthesis, characterization
and application of Co/Co3O4 nanocomposites as an effective photocatalyst for discoloration of
organic dye contaminants in wastewater and antibacterial properties///Photo-degradation of
organic dyes: simple chemical synthesis of Ni(OH)2 nanoparticles, Ni/Ni(OH)2 and Ni/NiO
magnetic nanocomposites/// Dy2BaCuO5/Ba4DyCu3O9.09 S‐ scheme heterojunction
nanocomposite with enhanced photocatalytic and antibacterial activities///Green sonochemical
synthesis of BaDy2NiO5/Dy2O3 and BaDy2NiO5/NiO nanocomposites in the presence of core
almond as a capping agent and their application as photocatalysts for the removal of organic dyes
in water///Control sonochemical parameter to prepare pure Zn0.35Fe2.65O4 nanostructures and
study their photocatalytic activity///Hydrothermal synthesis of DyMn2O5/Ba3Mn2O8
nanocomposite as a potential hydrogen storage material///A new nanocomposite superionic system
(CdHgI4/HgI2): synthesis, characterization and experimental investigation///Hydrothermal
synthesis of nickel hydroxide nanostructures and flame retardant poly vinyl alcohol and cellulose
acetate nanocomposites.
Response: We have revised the introduction, and highlighted the purpose of our study as suggested.
To the best of our knowledge, the important studies that are relevant to our work were cited in our
manuscript. However, the recommended references are not related to our study, therefore, we will
not cite them.

6. The conclusion must be more than just a summary of the manuscript.

Response: We have added more discussion in the Conclusions section (Page 20, line 40-49 and
Page 21, line 1-8).

7. The figure captions should be written with more informative.

Response: We have added more information in the figure captions.

8. What is the main purpose of EIC of phosphorylcholine?

Response: Phosphorylcholine is an important metabolite present in the cytoplasm, and the
appearance of its mass spectral signal indicates that single cells are successfully disrupted online
by our device and cell metabolites are ionized. So we used this EIC to demonstrate the high
throughput for single cell analysis.

9. The quality of figure 5 is not acceptable.

Response: We have modified Figure 5.

10. The potential outliers identified by Random Forest should be discussed. Please more explain
the details.
Response: Thank you for this suggestion, we have further implemented the outlier detection and
data visualization by using the unsupervised learning algorithm Isolated Forest, which is a
well-studied and widely used outlier detection method in machine learning [ X. Chen, F.X. Wu, J.
Chen, M. Li, DoRC: Discovery of rare cells from ultra-large scRNA-seq data, 2019 IEEE
International Conference on Bioinformatics and Biomedicine (BIBM), 2019, 111-116]. The
detailed discussion is added in the manuscript (Page 18, line 20-59 and Page 19, line 1-7).

Reviewer #2: To achieve high-throughput single cell analysis, the authors successfully designed
to achieve continuous cell separation and inertial focusing coupled with a pulsed electric
field-induced electrospray ionization-high resolution MS. The pulsed square wave electric field
was applied to realize on-line cell disruption and induce electrospray ionization. Single cells were
analyzed under near physiological conditions at up to 80 cells min-1 of throughput. More than 900
features were detected and approximately 120 metabolites were tentatively identified from a single
cell. The manuscript shows favorable organization, language arrangement, and references. All the
data were analyzed and interpreted appropriately. The manuscript contains sufficient data to
support its conclusion. The analytical method shows great novelty in high-throughput single-cell
metabolomics research for future biological studies.
Minor comments:
1: The authors could use more accurate description in main text of the discovered compounds like
ornithine (m/z 133.0972, [M+H]+), creatine (m/z 132.0768, [M+H]+) and lysine (m/z 147.1128,
[M+H]+).
Response: Thank you for this suggestion, we have added more description of the important
metabolites in the manuscript (Page 20, line 1-10).
2: The authors could consider to transfer important data from supporting information to the main
text for better understanding.
Response: As suggested by the reviewer, we have transferred the simulation results of streamlines
of cells with the same diameter to main text. The simulation results of the cells with different
diameters have been moved to supporting information. The information of Figure S3 has been
merged into Figure 5c to demonstrate the stability of our method.

3: If the tip of the nanospray emitter tended to be blocked during on-line cell disruption?
Response: For the nanospray emitters that worked for more than 3 hours under optimized
conditions, we did not find any clogging of the nozzles by microscopic observation. Recently we
have also performed experiments with continuous analysis for 5 hours and no clogging was
observed. The reason for this phenomenon may be due to the self-cleaning ability of the
electric-field-induced electrospray [G. Huang, G. Li, R.G. Cooks, Induced Nanoelectrospray
Ionization for Matrix-Tolerant and High-Throughput Mass Spectrometry, Angewandte Chemie
International Edition, 50 (2011) 9907-9910].
Graphical Abstract
Highlights

Highlights:

 A versatile chip-PEF-ESI-HRMS method was developed for single cell metabolomics

 More than 900 MS features and 120 metabolites were detected and identified

 This method was validated to be suitable for large-scale single cell analysis
Declaration of Interest Statement Click here to view linked References

Declaration of interests

☒ The authors declare that they have no known competing financial interests or personal relationships
that could have appeared to influence the work reported in this paper.

☐The authors declare the following financial interests/personal relationships which may be considered
as potential competing interests:
Credit Author Statement Click here to view linked References

Credit authorship contribution statement

Disheng Feng: Conceptualization, Methodology, Formal analysis, Investigation, Writing -

original draft, Visualization. Hang Li: Instrument. Tianrun Xu: Analysis. Fujian Zheng:

Software. Chunxiu Hu: Investigation. Xianzhe Shi: Methodology, Writing - review &

editing. Guowang Xu: Conceptualization, Project administration, Funding acquisition,

Supervision, Writing - review & editing.

Revised manuscript with changes marked

High-throughput Single Cell Metabolomics and Cellular

1
2
3 Heterogeneity Exploration by Inertial Microfluidics coupled with
4
5
6 Pulsed Electric Field-Induced Electrospray Ionization-High
7
8
9 Resolution Mass Spectrometry
10
11 Disheng Feng1,2, Hang Li1, Tianrun Xu1,2, Fujian Zheng1,2,3, Chunxiu Hu1,3, Xianzhe
12
13
14 Shi1,3*, Guowang Xu1,2,3*
15
16 1
CAS Key Laboratory of Separation Science for Analytical Chemistry, Dalian
17
18
19 Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, 116023, China
20
2
21 University of Chinese Academy of Sciences, Beijing 100049, China.
22
23 3
24
Liaoning Province Key Laboratory of Metabolomics, Dalian, China
25
26
27
28 *
Correspondence to:
29
30
31 Dr. Xianzhe Shi, CAS Key Laboratory of Separation Science for Analytical
32
33 Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences,
34
35
36 Dalian 116023, China. E-mail: shixianzhe@dicp.ac.cn.
37
38 Prof. Dr. Guowang Xu, CAS Key Laboratory of Separation Sciences for Analytical
39
40
41 Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences,
42
43 Dalian 116023, China. Tel. / Fax: 0086-411-84379530. E-mail: xugw@dicp.ac.cn
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 1
61
62
63
64
65
 ABSTRACT:
1
2
3 Single cell metabolomics can obtain the metabolic profiles of individual cells and
4
5 reveal cellular heterogeneity. However, high-throughput single-cell mass
6
7
8 spectrometry (MS) analysis under physiological conditions remains a great challenge
9
10 due to the presence of complex matrix and extremely small cell volumes. Herein, a
11
12
13
serpentine channel microfluidic device which was designed to achieve continuous cell
14
15 separation and inertial focusing, was coupled with a pulsed electric field-induced
16
17 electrospray ionization-high resolution MS (PEF-ESI-HRMS) to achieve
18
19
20 high-throughput single cell analysis. The pulsed square wave electric field was
21
22 applied to realize on-line cell disruption and induce electrospray ionization. Single
23
24
25 cells were analyzed under near-physiological conditions at a throughput of up to 80
26
27 cells min-1. More than 900 features were detected and approximately 120 metabolites
28
29
30 were tentatively identified from a single cell. Further, by continually analyzing more
31
32 than 3000 MCF7 and HepG2 cells, discrimination of different cancer cells based on
33
34
35
their individual metabolic profiles was achieved by using the principal component
36
37 analysis. The PEF-ESI-HRMS method was also applied for the analysis of single
38
39 yeast cells, and more than 40 metabolites were annotated. This method is versatile and
40
41
42 has good robustness, which is promising for high-throughput single cell metabolomics
43
44 analysis.
45
46
47 Key words: Cellular heterogeneity; Single cell metabolomics; High-throughput; Mass
48
49 spectrometry; Microfluidic device
50
51
52
53
54
55
56
57
58
59
60 2
61
62
63
64
65
 1. Introduction
1
2
3 The complex life activities of organisms are orchestrated by coordinated interactions
4
5 of a diverse range of individual cells [1]. The metabolites are the end products of
6
7
8 biochemical regulation, which can accurately represent the biochemical phenotype of
9
10 cells in real time. Due to the heterogeneity of cells and fast turnover rates of some
11
12
13
metabolites, the traditional large-scale assays that investigate the state of large number
14
15 of cells cannot accurately reflect the behavior of the individual cells [2, 3]. Therefore,
16
17 it is increasingly necessary to study metabolites in a single cell under physiological
18
19
20 environmental conditions.
21
22 Single-cell metabolomics can directly obtain the metabolite information of individual
23
24
25 cells, which provides insight into the relationship between physiological behavior of
26
27 single cells and their chemical compositions. The latest development in mass
28
29
30 spectrometry has ushered in a new era of single-cell metabolomics analysis.
31
32 Nanoelectrospray ionization mass spectrometry (nano-ESI-MS) has been used in
33
34
35
single plant cell metabolomics analysis [4]. However, single mammalian cells and
36
37 yeast cells are difficult to be directly analyzed by traditional nano-ESI-MS due to its
38
39 extremely small sample volume, viscous cytoplasm and the possible presence of cell
40
41
42 walls. Some modified nano-ESI-MS methods have been developed in the recent years.
43
44 Induced nano-ESI-MS by using alternating current high-voltage to produce inductive
45
46
47 electrospray was used to investigate the metabolism of intracellular metabolites of
48
49 single neurons and revealed a novel glutamate biosynthetic pathway in brain [5-7].
50
51
52
Pulsed direct current electrospray ionization mass spectrometry (pulsed-dc-ESI-MS)
53
54 can decrease the flow rate of sample to picoliters per minute and extend the
55
56 electrospray time to acquire rich MS2 information. The pulsed-dc-ESI-MS strategy
57
58
59
60 3
61
62
63
64
65
 was also combined with microwell-based droplet microextraction to study the
1
2 glucose-phosphate and phospholipids in single cells [8, 9]. However, most of the
3
4
5 abovementioned MS methods employ manual capillary probe sampling, severely
6
7 limiting the throughput of single-cell analysis.
8
9
10 For molecular biology research and actual clinical applications, it is necessary to
11
12 analyze a large number of individual cells in a short period of time. Therefore,
13
14
15 high-throughput single-cell analysis methods are urgently required. To achieve
16
17 high-throughput analysis, cell separation and focusing should be performed to ensure
18
19
that single cells are introduced one by one into the mass spectrometer. In the previous
20
21
22 work, a drop-on-demand inkjet cell printing was combined with probe ESI-MS. The
23
24 droplets containing single cells were produced by the inkjet sampling of cell
25
26
27 suapension [10]. But it is difficult to ensure that the number of cells contained in each
28
29 printed droplet is just one. Since then, a Dean flow assisted cell sorting system based
30
31
32 on spiral capillaries coupled with ESI-MS was developed, which can reduce the
33
34 agglomeration and uneven distribution of cells in the cell suapension [11]. Recently, a
35
36
37
label-free mass cytometry named CyESI-MS was proposed by coupling flow
38
39 cytometry to ESI-MS. As the core part of CyESI-MS, three coaxial capillaties were
40
41 used to deliver cell suspension, sheath fluid, and sheath gas, respectively. Cells were
42
43
44 dispersed and the intracellular metabolites were extracted by the surrounding sheath
45
46 fluid. The highest throughput of CyESI-MS is approximately 38 cells min-1 [12, 13],
47
48
49 but the sheath fluid at a flow rate of 10 µL min-1 may lead to severe dilution effect.
50
51 The increase in dead volume and dilution of metabolites may also result from the
52
53
54 large opening of the sheath fluid capillary. Technical advances in microfluidics
55
56 conjunction with mass spectrometry have enabled their use in single-cell
57
58
metabolomics analysis [14]. The microfluidic devices are composed of a network of
59
60 4
61
62
63
64
65
 microchannels with micrometer-scale dimensions comparable in length scale to
1
2 various kinds of cells, which can realize precise cell manipulation based on the
3
4
5 interactions between cells, fluids, and microchannels. In addition, the small geometry
6
7 of microfluidic devices allows the integration with various detectors. In a recent study,
8
9
10 a simple microfluidic chip consisting of 5 curved loops was applied for cell dispersing
11
12 and ordering and then a chip-nanoESI mass cytometry was established with a
13
14
15 throughput of about 40 cells min-1 [15]. However, the cells were suspended in
16
17 aqueous methanol solution which could affect cellular metabolism and thus could not
18
19
reveal authentic metabolite levels of cells under their physiological conditions [15,
20
21
22 16]. Until now, the reported single-cell mass cytometrymethods are capable of
23
24 analyzing only 500 to 1000 cells in an experiment [17]. To our knowledge, there are
25
26
27 no methods that can be used to analyze several thousands of cells in single experiment,
28
29 while analyzing a large number of cells is of great importance in obtaining
30
31
32 biologically meaningful data.
33
34 Herein, we developed a novel asymmetric serpentine channel microfluidic chip
35
36
37
coupled to pulsed electric field-induced electrospray ionization-high resolution mass
38
39 spectrometry (chip-PEF-ESI-HRMS) method for high-throughput single cell analysis
40
41 under near-physiological conditions. The microfluidic device was employed for
42
43
44 continuous cell separation and inertial focusing, which is sheathless and operated
45
46 without any external force fields. The applied pulsed electric field has dual functions,
47
48
49 online cell disruption and inducing electrospray ionization. It is worth mentioning that
50
51 single cells were suspended in an aqueous solution under near-physiological condition
52
53
54 (i.e., isotonic salt concentration) instead of a cytotoxic methanol solution, which can
55
56 maintain the cell viability. The throughput of our method is up to 80 cells min-1, the
57
58
whole system can be operated continuously and stably for more than 3 h, more than
59
60 5
61
62
63
64
65
 3000 single cells can be analyzed in one experiment. Approximately 120 metabolites
1
2 in cancer cells were annotated and two different cancer cells (MCF7 and HepG2 cells)
3
4
5 were successfully distinguished based on individual cell metabolic profiling. We also
6
7 applied this PEF-ESI-HRMS method for single yeast cell analysis and over 40
8
9
10 metabolites were tentatively identified. This method is versatile and can be used to
11
12 obtain metabolomic information from various kinds of single cells under near
13
14
15 physiological conditions in a high-throughput way.
16
17
18
19
20 2. Experimental
21
22 2.1 Chemicals and Materials
23
24
25 Ammonium bicarbonate was obtained from Fluka (Steinheim, Germany). The
26
27 ultrapure water was produced by a Milli-Q water purification system (Millipore,
28
29
30 Bedford, MA, USA). The nanoelectrospray emitter was made of fused silica
31
32 capillaries with 100 μm i.d. and 365 μm o.d. The capillary was firstly disposed on fire
33
34
35
and then the coating of the capillary was peeled off, then it was pulled by the P-2000
36
37 puller (Sutter Instruments, Novato, CA, USA) to make the emitter with a tip of about
38
39 10 μm. The borosilicate glass capillaries (BF150-86-10) with 0.86 mm inner diameter
40
41
42 (i.d.) and 1.5 mm outer diameter (o.d.) were purchased from Sutter Instrument
43
44 (Novato, CA, USA). The micropipettes were fabricated with a P-1000 puller (Sutter
45
46
47 Instruments, Novato, CA, USA). The fabricated micropipettes tips with an inner
48
49 diameter (i.d.) of approximately 2 μm were used for the the single yeast cell analysis.
50
51
52
53
54 2.2 Cell culture and sample preparation
55
56
57
58
59
60 6
61
62
63
64
65
 The mammalian cancer cell lines HepG2 and MCF7 were maintained in Dulbecco's
1
2 Modified Eagle's Medium (DMEM; Gibco, Waltham, MA, USA), supplemented with
3
4
5 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin solution in a
6
7 humidified incubator at 37°C with 5% CO2. All the cells used in this work were
8
9
10 cultured in Petri dishes (10 cm in diameter). For the cell analysis by the
11
12 high-throughput method with the asymmetric serpentine channel microfluidic chip,
13
14
15 the cells were trypsinized by using 0.25% Trypsin-EDTA solution (Gibco, Waltham,
16
17 MA, USA) and washed 3 times by PBS. Then the cells were resuspended in 150 mM
18
19
NH4HCO3 aqueous solution. The obtained cell suspension was diluted to different
20
21
22 concentrations (2.5 × 103, 5 × 103, 1 × 104, 2 × 104, 4 × 104 and 8×104 cells mL-1) and
23
24 drawn up into a syringe and the syringe was loaded onto a syringe pump to inject the
25
26
27 cell suspension into the microfluidic chip.
28
29 The yeast cells used were Saccharomyces cerevisiae strain H1246 and were kindly
30
31
32 provided by Qingdao Institute of Bioenergy and Bioprocess Technology, Chinese
33
34 Academy of Sciences. The synthetic complete (SC) medium containing 0.67% yeast
35
36
37
nitrogen base (w/v), 0.074% DO supplement (-Ura) (w/v) and 2% glucose (w/v) was
38
39 used for cultivating yeast. The yeast cells were inoculated into the centrifuge tubes
40
41 and incubated at 30°C with shaking at 200 rpm overnight. Then the cells were
42
43
44 centrifuged and washed with ultrapure water for three times to remove the culture
45
46 medium. The obtained yeast cell suspension was diluted to ~5×103 cells mL-1. Since
47
48
49 yeast cells are much smaller than mammalian cancer cells, a back-end injection
50
51 method was used to make yeast cells move to the tip of the spray emitter. Then the
52
53
54 pulsed electric field was applied to achieve the cell membrane electroporation and
55
56 metabolites ionization process.
57
58
59
60 7
61
62
63
64
65
 2.3 Microfluidic chip fabrication
1
2 The asymmetric serpentine channel microfluidic chip (Figure 1a) was fabricated by
3
4
5 using the softlithography technique. Briefly, SU-8 3035 (Microchem, Newton, MA)
6
7 was spun at 2,000 rpm for 60 s on a spin coater (KW-4A, Chemat Technology Inc.,
8
9
10 Northridge, CA, USA) to create a 50-μm-thick layer on a silicon wafer. The master
11
12 molds for PDMS replica were photolithographically defined by using a mask. The
13
14
15 mold was pre-treated with chlorotrimethylsilane (Sigma Aldrich, St. Louis, MO, USA)
16
17 vapor overnight to facilitate the release of PDMS. The PDMS prepolymer and curing
18
19
agent (RTV615, Momentive Performance Materials, NY, USA) were mixed
20
21
22 completely (parts A and B in 10:1 ratio) and the air bubbles generated were removed
23
24 by using a vacuum desiccator for 1 h. Then the mixtures were poured onto the silicon
25
26
27 master and the PDMS was cured in the oven at 80°C for 2 h. After curing, the PDMS
28
29 layer was peeled off the mold and the holes for inlet and outlet ports were punched
30
31
32 with a sharpened flat-tip needle. The device was cleaned via sonication in ethanol and
33
34 2-propanol before bonding with a galss slide. After plasma treatment and placement
35
36
37
onto the glass substrate, the devices were placed in the oven at 80°C for 30 min to
38
39 strengthen the bonding.
40
41
42
43
44 2.4 Pulsed electric field-induced electrospray ionization-high resolution mass
45
46 spectrometry
47
48
49 In this study, an asymmetric serpentine channel microfluidic chip cell focusing system
50
51 is connected to the pulsed electric field-induced ESI-MS to achieve the
52
53
54 high-throughput single cell metabolomics analysis. The schematic diagram of the
55
56 device is shown in Figure 1a. An inverted microscope (CKX53, Olympus, Tokyo,
57
58
Japan) was used to observe the single cells and the cell focusing process. A
59
60 8
61
62
63
64
65
 commercial NanoSpray Flex ion source (Thermo Scientific, San Jose, CA, USA) was
1
2 modified to accommodate with the micropipettes and the home-made high voltage
3
4
5 pulsed power supply. The setup of the microfluidic chip coupled to pulsed electric
6
7 field-induced electrospray ionization (ESI) source was shown in Figure 1b. The
8
9
10 microfluidic chip was placed on a home-made 3-dimensional adjustment frame with
11
12 high positioning accuracy. The inlet was connected to a syringe pump. The out flow
13
14
15 from the microchannel chip was connected with the nanoelectrospray emitter. High
16
17 voltage was applied to a copper plate which is placed beneath the nanoelectrospray
18
19
emitter. An adjustable high voltage power supply was constructed to provide square
20
21
22 wave pulsed high voltage. The power supply can operate stably for long time without
23
24 significant heating problem, which is superior to the previously reported conventional
25
26
27 high-voltage power supplies based on magnetic amplifier[18] .
28
29 The circuit diagram of the adjustable square wave pulse high voltage power supply is
30
31
32 shown in Figure S1. The power supply is mainly composed of a DC positive
33
34 high-voltage module (DW-P502-2CQ1, Dongwen High Voltage Power Supply,
35
36
37
Tianjin, China), a DC negative high-voltage module (DW-N502-2CQ1, Dongwen
38
39 High Voltage Power Supply, Tianjin, China) and a fast high voltage transistor switch
40
41 (HTS 121-01-HB-C, Behlke Power Electronics, Kronberg, Germany). The input
42
43
44 signal is generated by the function waveform generator (DG1022U, RIGOL, Beijing,
45
46 China). The principle is to generate a square wave pulse voltage through the function
47
48
49 waveform generator aforementioned and input it to the fast high voltage transistor
50
51 switch. By controlling the DC high voltage modules, the power supply can output
52
53
54 square wave pulse high voltage with voltage amplitude range from -5 kV to 5 kV and
55
56 square wave frequency range from 1 to 1000 Hz. Typically, a square wave high
57
58
59
60 9
61
62
63
64
65
 voltage with an amplitude of 5 kV, a frequency of 200 Hz, and a duty cycle of 50% is
1
2 used unless otherwise stated.
3
4
5 The Q Exactive HF mass spectrometer (Thermo Scientific, San Jose, CA, USA) was
6
7 utilized for the MS experiment. The setup photo of the microfluidic chip and the
8
9
10 pulsed electric field-induced ESI source coupled to mass spectrometer is shown in
11
12 Figure 1. The pulse voltage is generated by home-made high-voltage power supplies
13
14
15 (Figure 1b). The MS was operated in full-scan positive ion mode with the following
16
17 settings: mass range 70-1,050 m/z; resolution for MS1 scan 60,000 at m/z 200;
18
19
microscans: 1; sheath gas flow rate: 0; aux gas flow rate: 0; sweep gas flow rate: 0;
20
21
22 capillary temperature: 275°C; S-Lens RF level: 50. For high-throughput single-cell
23
24 analysis, the resolution for MS1 scan was set to 60,000 with a scan speed of 7 Hz to
25
26
27 achieve a balance between mass resolution and scanning speed. For the analysis of
28
29 yeast cell samples, the resolution for MS1 scan was 120,000 at m/z 200. The AGC
30
31
32 target was set to 5e5 and the maximum injection time was set to 50 ms.
33
34
35
36
37
2.5 Data processing and analysis
38
39 The Xcalibur software (version 4.0.27.10, Thermo Scientific, San Jose, CA, USA)
40
41 was used for the control of the MS system and data analysis. The obtained .raw files
42
43
44 were firstly processed in the Qual Browser tool of the Thermo Xcalibur software to
45
46 obtain the mass spectrum list. First, the peaks in the extracted ion chromatogram (EIC)
47
48
49 at m/z 184.0733 were used as the signal of single cell events, because this ion was
50
51 assigned to phosphorylcholine which is related to the damage of cell membranes. Ion
52
53
54 peaks with the same appearance time as the single cell events and the detection rate
55
56 greater than 25% among analyzed single cells were chosen as single cell mass spectral
57
58
features. Before subsequent statistical analysis, normalization was performed based on
59
60 10
61
62
63
64
65
 the total ion intensity of all detected ions in each mass spectrum. A peak alignment
1
2 program written in R language was used for peak matching of different single cell
3
4
5 samples. The metabolites were identified based on the accurate mass (mass error <5
6
7 ppm), MS2 spectra and the identification results obtained by LC-MS analysis of
8
9
10 population cells. Detailed information about the LC-MS analysis and metabolite
11
12 identification is described in the Supporting Information. The Human Metabolome
13
14
15 Database (HMDB, http://www.hmdb.ca) and mzCloud (https://www.mzcloud.org/)
16
17 library were also used to broaden the scope of metabolite identification [19]. Whether
18
19
HMDB reports that the metabolite is present in cells was also considered to determine
20
21
22 the m/z corresponding metabolites during the identification process. The results
23
24 acquired were then used for further statistical analysis. The data analysis and
25
26
27 visualization were performed by using the in-house Python script and the web-based
28
29 metabolomics tool (MetaboAnalyst 5.0; https://www.metaboanalyst.ca/) [20].
30
31
32
33
34 3. Results and Discussion
35
36
37 3.1 Design of asymmetric serpentine microfluidic chip
38
39 To achieve the high-throughput analysis, a cell separation and focusing system for
40
41
42 sample introduction is essential before MS analysis. Inertial focusing in microfluidics
43
44 can achieve the precise lateral equilibrium position control of cells and the regulation
45
46
47 of cell spacing without the use of sheath fluid or other external fields [21, 22]. By
48
49 using this sheathless cell focusing method, the sampling rates can be increased
50
51
52
because cells are aligned without dramatically increasing their velocity or the overall
53
54 volumetric flow rate of carrier fluid. Consequently, the metabolites in the cells are not
55
56
57
58
59
60 11
61
62
63
64
65
 overly diluted, thus making it possible to expand the detection range of metabolites, it
1
2 is a promising sample introduction method for single cell mass cytometry.
3
4
5 Based on the abovementioned work on microfluidic inertial focusing, a microfluidic
6
7 chip with asymmetric serpentine channels was designed, and it consists of 15 repeated
8
9
10 S-shaped units and a connected straight channel (Figure 2a). The microchannel in the
11
12 inlet region was widened to reduce the normal and shear stresses over cells surfaces
13
14
15 [23]. The height of the microchannel is 50 μm. Fluid flow simulation within the
16
17 proposed asymmetric serpentine microchannel was performed by using COMSOL
18
19
Multiphysics 5.6 (COMSOL AB, Stockholm, Sweden). The simulation process
20
21
22 comprises of two major steps, in the first step the flow velocity distribution was
23
24 computed by using laminar flow module under steady-state conditions and in the
25
26
27 second step the motion trajectories of particles in the fluid was computed using
28
29 particle tracing module. The outlet boundary condition was set to “freeze” and the
30
31
32 microchannel wall conditions were set to “bounce”. Water is used as the simulation
33
34 fluid and the injection flow rate was set to 1 µL min-1 to simulate the flow rate under
35
36
37
actual experiment conditions of electrospray. The drag force, the up, down and side
38
39 microchannel walls induced forces acting on the particles were considered when
40
41 simulating particle trajectories. The Stokes law was used for drag forces and wall
42
43
44 induced lift law was used for lift forces. As shown in Figure 2b, the axial velocity
45
46 magnitude plot in the cross-section of the channel indicated that the maximum
47
48
49 velocity occurs at the inside of the turn, and the minimum velocity occurs near the
50
51 outside of the turn due to the no-slip condition on the walls. The streamlines of cells
52
53
54 with the same diameter (ten 12 µm particles released at random positions at the
55
56 entrance and the release time normally distributed at 0.5 s were used as homogeneous
57
58
cell suspension model proximity model) after different times were simulated (Figure
59
60 12
61
62
63
64
65
 2a and S2a). It can be observed that at the outlet, the particles outflow almost at the
1
2 same position. Five spherical particles of different diameters are released at the inlet
3
4
5 cross-section of the microchannel in 0.02 s as cell mixture proximity models, and the
6
7 trajectories and positions of the first three particles after 3.55 s are shown in Figure
8
9
10 S2b. These results demonstrated that the microfluidic chip has a dispersing and
11
12 focusing effect on mammalian cancer cells.
13
14
15
16
17 3.2 Configuration and perfermance of chip-PEF-ESI-HRMS
18
19
The chip-PEF-ESI-HRMS contains four parts: a syringe pump for delivering cell
20
21
22 suspension, an asymmetric serpentine channel microfluidic chip for dispersing and
23
24 focusing cells, PEF-ESI source, and Q Exactive HF mass spectrometer. In order to
25
26
27 maintain cell viability and integrity, the cells were suspended in an isotonic solution
28
29 of ammonium bicarbonate (150 mM NH4HCO3), an MS-friendly volatile salt. The cell
30
31
32 suspension was injected into the microfluidic chip by a syringe pump. Then the cells
33
34 were dispersed and focused through asymmetric microchannels with curvature. A
35
36
37
movie of single cells passing through the microchannel is provided in Supplementary
38
39 Video S1. Once the ordered single cells reach the tip of nanospary emitter, high
40
41 voltage square-wave pulses generated by our home-made power supply were applied
42
43
44 for online cell disruption without the addition of cytotoxic organic solvents. The
45
46 electric field strength between the emitter and the MS is approximately 10 kV cm-1,
47
48
49 and the pulses width can be freely adjustable, which enables the irreversible cell
50
51 electroporation and disruption [24]. Meanwhile, high voltage square-wave pulses
52
53
54 induce the electrospary ionization processes and this phenomenon was also observed
55
56 in previous studies [5, 25]. As a result, the released cellular contents are immediately
57
58
59
60 13
61
62
63
64
65
 ionized and analyzed by the mass spectrometer which helps to obtain the real-time
1
2 information of metabolites from single live cells.
3
4
5 The performance of chip-PEF-ESI-HRMS was evaluated in single-cell metabolomics
6
7 analysis. The cell suspensions of two cancer cell lines (MCF7 and HepG2) were
8
9
10 introduced into the system. As shown in the total ion chromatogram (TIC) (Figure 3a),
11
12 pulse-like signals corresponding to the single cells were obtained when the cell
13
14
15 suspension was flowing. The peak at m/z 184.0733 in the extracted ion chromatogram
16
17 (EIC) is assigned to phosphorylcholine, an important metabolite predominantly
18
19
present in the cytoplasm. The appearance of this signal represents the disruption of the
20
21
22 cell and is therefore used as a marker for single cell mass spectrometry events.
23
24 Previous studies have demonstrated elevated levels of phosphorylcholine and total
25
26
27 choline compounds in a variety of cancers [26]. Moreover, the EIC of
28
29 phosphorylcholine (Figure 3b) has a good agreement with the TIC, indicating that the
30
31
32 pulsed electric field-induced ESI source has the ability to realize the online cell
33
34 fragmentation and metabolites ionization. From the MS data, two distinct single-cell
35
36
37
metabolic profiles of MCF7 (Figure 3c) and HepG2 (Figure 3d) were obtained, it can
38
39 be observed that the overall relative abundance of metabolites in HepG2 liver cancer
40
41 cells is higher.
42
43
44 It is important to note that for high-concentration cell suspension, the interparticle
45
46 interactions can limit the degree of focusing because cells can randomly disturb the
47
48
49 ideal parabolic flow of the fluid and the lift, and Dean drag forces can be altered [23].
50
51 Therefore, the flow rate and concentration of cell suspensions are two important
52
53
54 factors affecting the single cell mass cytometry. Different flow rates were investigated
55
56 to optimize the analysis throughput and improve system stability. When flow rate was
57
58
greater than 1 μL min-1, liquid leakage was observed at the interface between the
59
60 14
61
62
63
64
65
 microfluidic chip and the inlet stainless steel tube, which may cause the stainless steel
1
2 tube to dislodge when subjected to excessive pressure. Additionally, large volumetric
3
4
5 flow rates may lead to insufficient time for separation forces to act on cells. HepG2
6
7 cell suspension with a concentration of 2 × 104 cells mL-1 was infused at flow rates of
8
9
10 0.5 μL min-1, 0.75 μL min-1 and 1 μL min-1. The relationship between the flow rate of
11
12 the cell suspension and the number of single cell detected by MS per minute is shown
13
14
15 in Figure 4a and Table S1. When the flow rate increased, the number of single cell
16
17 detected also increased. The highest number of single-cell correlation pulse peaks per
18
19
unit time was obtained at 1 μL min-1. After comprehensively considering the pressure
20
21
22 that the microfluidic chip interface can withstand and the maximum analytical
23
24 throughput, 1 μL min-1 was selected as the appropriate flow rate. In order to adapt to
25
26
27 this relatively high injection flow rate, a square wave voltage with an amplitude of 5
28
29 kV Vpp and a frequency of 200 Hz was applied in the nanoelectrospray emmiter. In
30
31
32 addition, HepG2 and MCF7 cell suspensions with different concentrations (2.5 × 103,
33
34 5 × 103, 1 × 104, 2 × 104, 4 × 104 and 8×104 cells mL-1) were injected at a flow rate of
35
36
37
1 µL min-1 to optimize the concentration of cell suspension, and the results are shown
38
39 in Figure 4b and Table S2. The number of single cell MS signals per unit time
40
41 increased with the increase of cell concentration and a good linear correlation between
42
43
44 them was found. This result also indicated that single cells were seriatim detected by
45
46 the mass spectrometer. As shown in Figure 4c, the maximum throughput could reach
47
48
49 about 80 cells min-1 by applying MCF7 cell suspension with a concentration of 8×104
50
51 cells mL-1 at a flow rate of 1 μL min-1. The magnified EIC at m/z 184.0733 shows that
52
53
54 about 20 cells were analyzed within 15 s and the intensity of peaks for the ions
55
56 dropped to zero between two pulse peaks, which indicated that almost no mutual
57
58
interference existed between each single-cell mass spectrometry event. The difference
59
60 15
61
62
63
64
65
 in the width of each pulse peak may be due to the cell size. Compared to the
1
2 previously reported method [12], the mass spectrum obtained by our method is more
3
4
5 stable.
6
7 The mass spectral information at the highest point of each pulse peak was extracted
8
9
10 for further data processing and analysis. More than 900 features can be detected from
11
12 a single cell by using the chip-PEF-ESI based mass cytometry. Among them, about
13
14
15 120 metabolites were tentatively identified in a single cell based on the accurate mass
16
17 and secondary mass spectrum obtained from the LC-MS analysis results of population
18
19
cells (Table S3). These identified metabolites were classified by their chemical
20
21
22 characteristics and include most kinds of important functional metabolites such as
23
24 amino acids, acylcarnitines, carboxylic acids and their derivatives or analogues
25
26
27 (Figure 4d). In order to maintain the cells in their native state as much as possible, an
28
29 isotonic salt solution was used as the extractive solvent rather than organic solvents.
30
31
32 Therefore, the detected metabolites were mainly polar metabolites with good water
33
34 solubility while high-abundance lipids with poor water solubility were rarely found.
35
36
37
More than 80% of the annotated metabolites were also detected by the traditional
38
39 LC-MS method. Those metabolites only detected by the single cell method are mainly
40
41 carboxylic acids and their derivatives as well as several kinds of oxygenated organic
42
43
44 compounds. The reason could be that these metabolites may be lost, unstable and
45
46 degraded in the pretreatment process of population cells. Therefore, the chip-PEF-ESI
47
48
49 based mass cytometry showed not only high throughput but also high coverage of
50
51 polar metabolites. Furthermore, in order to test the stability and robustness of the
52
53
54
instrument system, the chip-PEF-ESI based mass cytometry was operated
55
56 continuously for about three hours and stable mass spectrometry signals were still
57
58 obtained without clogging of the microchannel or the nanospray emitter (Figure 3a).
59
60 16
61
62
63
64
65
 The reason for this phenomenon may be due to the self-cleaning ability of the
1
2 electric-field-induced electrospray [5].
3
4
5
6
7 3.3 Practical applications of the developed method in high-throughput analysis of
8
9
10 single mammalian cells
11
12 Traditional metabolomics studies generally use millions of cells as samples and
13
14
15 require relatively long sample pretreatment time. Due to the intra-species variation of
16
17 cells and the fast turnover rate of metabolites, this kind of strategy is inappropriate for
18
19
studying the heterogeneity of cells. In contrast, the metabolic profile of individual
20
21
22 cells can be obtained by using single-cell mass spectrometry, and thus their intra- and
23
24 interspecies metabolic heterogeneity can be better represented. In order to achieve the
25
26
27 statistical characterization of single-cell metabolic information and obtain more
28
29 biologically meaningful results, the analysis of a large number of single cells must be
30
31
32 performed. More than 3000 HepG2 cells and MCF7 cells were continually analyzed
33
34 by the chip-PEF-ESI based mass cytometry, and the mass spectrometry data from
35
36
37
1000 single cells of each type were used for further analysis to demonstrate the
38
39 practicality of this high-throughput method in distinguishing the metabolic difference
40
41 of various cancer cells. The two types of cells can be well distinguished based on the
42
43
44 PCA score plot (Figure 5a). The PCA score plot also reflect the heterogeneity within
45
46 the same kind of cells and the stability of the method in the whole analysis process.
47
48
49 Machine learning based t-SNE (t-distributed stochastic neighbor embedding) and
50
51 UMAP (uniform manifold approximation and projection) algorithm were further
52
53
54 applied to mine and visualize the complex single cell MS data sets. The separation of
55
56 MCF7 and HepG2 cells is clearly observed as shown in Figure S3a and S3b, and
57
58
revealed significant metabolic heterogeneity of two kinds of cancer cells. Modified
59
60 17
61
62
63
64
65
 raincloud plots (consisting of split-half violin plots and box plots) of the relative
1
2 contents of representative metabolites including leucine and creatine are shown in
3
4
5 Figure 5b. These plots show the distribution and key summary statistics of metabolites
6
7 levels in cells, which illustrate the difference in metabolite levels in two kinds of
8
9
10 cancer cells. To identify potential biomarkers that differentiate these two kinds of
11
12 cancer cells and evaluating their performance, classical univariate receiver operating
13
14
15 characteristic (ROC) curve analyses were performed. The scatter box plots and ROC
16
17 curves of several metabolites are shown in Figure S4. The results suggest that these
18
19
metabolites can serve as biomarkers to differentiate MCF7 and HepG2 cells. To
20
21
22 explore the metabolic heterogeneity within the same kind of cells, we further
23
24 implemented outlier detection and data visualization by using an unsupervised
25
26
27 learning algorithm Isolated Forest, a well-studied and widely used outlier detection
28
29 method in machine learning [27]. Thirty-two potential outliers among 1000 HepG2
30
31
32 cells were found and marked by red x in 3-dimensional (3D) PCA plot (Figure 5c).
33
34 The data points obtained from non-outlier single HepG2 cells are well clustered in the
35
36
37
PCA scoring plot, further demonstrating the good stability of the device when
38
39 analyzing large scale single cell samples. Modified raincloud plots are used for
40
41 detailed data visualization of outliers versus non-outlier cells in the same kind of cell,
42
43
44 and the relative amounts of metabolites in HepG2 cells that have a relatively large
45
46 impact on the outlier detection model, including gamma-aminobutyric acid and
47
48
49 aminoacetone, are shown in Figure 5d. Overall, the metabolite contents in outlier cells
50
51 of HepG2 cells was higher than those in non-outlier cells.
52
53
54 Similaryly, 61 potential outliers in 1000 MCF7 cells were identified and the results
55
56 are shown in Figure 5e. Several key metabolites closely related to energy metabolism,
57
58
such as valine and phosphate had relatively higher contents in the outlier MCF7 cells
59
60 18
61
62
63
64
65
 (Figure 5f). Based on the above two cell heterogeneity exploration results, the
1
2 probability of heterogeneous cells is about 5%. These results demonstrate that
3
4
5 high-throughput chip-PEF-ESI based mass cytometry has the potential to explore the
6
7 cellular metabolic heterogeneity.
8
9
10
11
12 3.4 Practical applications of the developed method in metabolic profiling analysis of
13
14
15 single yeast cells
16
17 To further demonstrate the on-line cell disruption capability of the PEF-ESI source for
18
19
the cells with cell wall, the baker's yeast (Saccharomyces cerevisiae strain H1246)
20
21
22 was used as the model cell. The cell walls of this kind of yeast cells have an elastic
23
24 structure and are mainly composed of β-1,3-glucans, β-1,6-glucans, mannoproteins
25
26
27 and chitin [28, 29]. Traditionally, enzymatic hydrolysis, grinding and ultrasound were
28
29 often used to break the yeast cell wall, however, these methods are unsuitable for
30
31
32 single yeast analysis. Recently, the manipulation and analysis of single yeast cells by
33
34 using cell electro-migration and electroporation techniques has been reported [30], but
35
36
37
it required applying multiple steps of DC and pulse voltages which is complicated and
38
39 difficult to be operated in a high-throughput way. Previous studies had shown that the
40
41 yeast cell wall can be destructed by pulsed electric field [31, 32], therefore by
42
43
44 applying the square wave high-voltage pulse with an amplitude of 5 kV Vpp and
45
46 frequency of 20 Hz generated by our home-made power supply, single yeast cells
47
48
49 were rapidly electroporated, and the intracellular metabolites were successfully
50
51 released and ionized. The TIC and EICs of ornithine (m/z 133.0972), creatine (m/z
52
53
54 132.0768) and lysine (m/z 147.1128) obtained from single yeast cells are shown in
55
56 Figure 6a and 6b, the strong mass spectral signal peaks in the initial stage may
57
58
originate from the electroporation and rupture of cells. Ornithine is one of the
59
60 19
61
62
63
64
65
 products of the action of arginase on arginine, which plays an important role in the
1
2 urea cycle [33]. Creatine is a kind of nitrogenous organic acid which is associated
3
4
5 with the energy metabolism. Lysine is also an important amino acid that is taken up in
6
7 large quantities by yeast cells, which can lead to metabolic reorganization and make
8
9
10 the cells resistant to oxidants [34]. More than 40 metabolites including other
11
12 important amino acids were detected from single yeast cells in positive ion mode
13
14
15 which illustrates the potential application of this method for the metabolomics
16
17 analysis of cells with cell walls.
18
19
20
21
22 4. Conclusions
23
24
25 In summary, a versatile chip-PEF-ESI based mass cytometry method was established
26
27 for high-throughput single-cell metabolomics analysis under near physiological
28
29
30 conditions. The novel combination of PEF-ESI-HRMS with the inertial microfluidic
31
32 device has significant advantages in providing higher analytical throughput and stable
33
34
35
electrospray, while maintaining the analytical sensitivity. The analysis of ca. 80 single
36
37 cancer cells can be finished within 1 minute, more than 900 features were extracted
38
39 from single HepG2 or MCF7 cell, and 120 metabolites were tentatively identified, the
40
41
42 discrimination of these two types of cancer cell lines was achieved with this
43
44 high-throughput method. Moreover, the metabolic profiling outliers within the same
45
46
47 kind of cells were further identified by the isolation forest algorithm, which illustrated
48
49 the cells with metabolic heterogeneity can be defined. More importantly, this method
50
51
52
has a good analytical robustness which allows more than 3000 cells to be
53
54 continuously analyzed in a single experiment. The PEF-ESI ionization source also has
55
56 the capability to analyze the cells with cell wall such as yeast cell, and more than forty
57
58
59
60 20
61
62
63
64
65
 important metabolites such as amino acids can be detected in single yeast cells.
1
2 Therefore, this chip-PEF-ESI based method is expected to provide inspiration for the
3
4
5 new generation of high-throughput mass cytometry and play acceleration roles for
6
7 large-scale single-cell metabolomics analysis to aid in bioscience and clinical study.
8
9
10
11
12 Appendix A. Supplementary data
13
14
15 Supplementary data to this article can be found online.
16
17
18
19
20
ACKNOWLEDGMENTS
21
22 This work was funded by the foundation (No. 21934006) from the National Natural
23
24 Science Foundation of China, the innovation program (DICP&QIBEBT UN201806
25
26
27 and DICP I202141) of science and research from the DICP, CAS. and the CAS “Light
28
29 of West China” Program.
30
31
32
33
34 References
35
36
37
[1] E. Armingol, A. Officer, O. Harismendy, N.E. Lewis, Deciphering cell-cell
38 interactions and communication from gene expression, Nat Rev Genet, 22 (2021)
39 71-88.
40 [2] R. Zenobi, Single-cell metabolomics: analytical and biological perspectives,
41 Science, 342 (2013) 1243259.
42
43 [3] L. Zhang, A. Vertes, Single-Cell Mass Spectrometry Approaches to Explore
44 Cellular Heterogeneity, Angew. Chem. Int. Ed. Engl., 57 (2018) 4466-4477.
45 [4] T. Fujii, S. Matsuda, M.L. Tejedor, T. Esaki, I. Sakane, H. Mizuno, N. Tsuyama, T.
46 Masujima, Direct metabolomics for plant cells by live single-cell mass
47
48
spectrometry, Nat. Protoc., 10 (2015) 1445-1456.
49 [5] G. Huang, G. Li, R.G. Cooks, Induced Nanoelectrospray Ionization for
50 Matrix-Tolerant and High-Throughput Mass Spectrometry, Angewandte Chemie
51 International Edition, 50 (2011) 9907-9910.
52 [6] H. Zhu, G. Zou, N. Wang, M. Zhuang, W. Xiong, G. Huang, Single-neuron
53
54 identification of chemical constituents, physiological changes, and metabolism
55 using mass spectrometry, Proc. Natl. Acad. Sci. U. S. A., 114 (2017) 2586-2591.
56 [7] H. Zhu, N. Wang, L. Yao, Q. Chen, R. Zhang, J. Qian, Y. Hou, W. Guo, S. Fan, S.
57 Liu, Q. Zhao, F. Du, X. Zuo, Y. Guo, Y. Xu, J. Li, T. Xue, K. Zhong, X. Song, G.
58 Huang, W. Xiong, Moderate UV Exposure Enhances Learning and Memory by
59
60 21
61
62
63
64
65
 Promoting a Novel Glutamate Biosynthetic Pathway in the Brain, Cell, 173 (2018)
1 1716-1727 e1717.
2 [8] Z. Wei, X. Xiong, C. Guo, X. Si, Y. Zhao, M. He, C. Yang, W. Xu, F. Tang, X.
3
4 Fang, S. Zhang, X. Zhang, Pulsed Direct Current Electrospray: Enabling
5 Systematic Analysis of Small Volume Sample by Boosting Sample Economy,
6 Anal. Chem., 87 (2015) 11242-11248.
7 [9] J. Feng, X. Zhang, L. Huang, H. Yao, C. Yang, X. Ma, S. Zhang, X. Zhang,
8
9
Quantitation of Glucose-phosphate in Single Cells by Microwell-Based Nanoliter
10 Droplet Microextraction and Mass Spectrometry, Anal. Chem., 91 (2019)
11 5613-5620.
12 [10] F. Chen, L. Lin, J. Zhang, Z. He, K. Uchiyama, J.-M. Lin, Single-Cell Analysis
13 Using Drop-on-Demand Inkjet Printing and Probe Electrospray Ionization Mass
14
15 Spectrometry, Anal. Chem., 88 (2016) 4354-4360.
16 [11] Q. Huang, S. Mao, M. Khan, L. Zhou, J.M. Lin, Dean flow assisted cell ordering
17 system for lipid profiling in single-cells using mass spectrometry, Chem.
18 Commun. (Camb.), 54 (2018) 2595-2598.
19
[12] H. Yao, H. Zhao, X. Zhao, X. Pan, J. Feng, F. Xu, S. Zhang, X. Zhang, Label-free
20
21 Mass Cytometry for Unveiling Cellular Metabolic Heterogeneity, Anal. Chem.,
22 91 (2019) 9777-9783.
23 [13] H. Yao, H. Zhao, X. Pan, X. Zhao, J. Feng, C. Yang, S. Zhang, X. Zhang,
24 Discriminating Leukemia Cellular Heterogeneity and Screening Metabolite
25
26
Biomarker Candidates using Label-Free Mass Cytometry, Anal. Chem., 93 (2021)
27 10282-10291.
28 [14] D. Feng, T. Xu, H. Li, X. Shi, G. Xu, Single-cell Metabolomics Analysis by
29 Microfluidics and Mass Spectrometry: Recent New Advances, Journal of
30 Analysis and Testing, 4 (2020) 198-209.
31
32 [15] S. Xu, M. Liu, Y. Bai, H. Liu, Multi-Dimensional Organic Mass Cytometry:
33 Simultaneous Analysis of Proteins and Metabolites on Single Cells, Angew.
34 Chem. Int. Ed. Engl., 60 (2021) 1806-1812.
35 [16] S.T. Nguyen, H.T.-L. Nguyen, K.D. Truong, Comparative cytotoxic effects of
36
37
methanol, ethanol and DMSO on human cancer cell lines, Biomedical Research
38 and Therapy, 7 (2020) 3855-3859.
39 [17] I. Lanekoff, V.V. Sharma, C. Marques, Single-cell metabolomics: where are we
40 and where are we going?, Curr. Opin. Biotechnol., 75 (2022) 102693.
41 [18] G. Li, S. Yuan, Y. Pan, Y. Liu, G. Huang, Binding States of Protein–Metal
42
43 Complexes in Cells, Anal. Chem., 88 (2016) 10860-10866.
44 [19] D.S. Wishart, Y.D. Feunang, A. Marcu, A.C. Guo, K. Liang, R. Vazquez-Fresno,
45 T. Sajed, D. Johnson, C. Li, N. Karu, Z. Sayeeda, E. Lo, N. Assempour, M.
46 Berjanskii, S. Singhal, D. Arndt, Y. Liang, H. Badran, J. Grant, A. Serra-Cayuela,
47
Y. Liu, R. Mandal, V. Neveu, A. Pon, C. Knox, M. Wilson, C. Manach, A.
48
49 Scalbert, HMDB 4.0: the human metabolome database for 2018, Nucleic Acids
50 Res., 46 (2018) D608-D617.
51 [20] Z. Pang, J. Chong, G. Zhou, D.A. de Lima Morais, L. Chang, M. Barrette, C.
52 Gauthier, P.E. Jacques, S. Li, J. Xia, MetaboAnalyst 5.0: narrowing the gap
53
54
between raw spectra and functional insights, Nucleic Acids Res., 49 (2021)
55 W388-W396.
56 [21] D. Di Carlo, D. Irimia, R.G. Tompkins, M. Toner, Continuous inertial focusing,
57 ordering, and separation of particles in microchannels, Proc. Natl. Acad. Sci. U. S.
58 A., 104 (2007) 18892-18897.
59
60 22
61
62
63
64
65
 [22] J. Oakey, R.W. Applegate, Jr., E. Arellano, D. Di Carlo, S.W. Graves, M. Toner,
1 Particle focusing in staged inertial microfluidic devices for flow cytometry, Anal.
2 Chem., 82 (2010) 3862-3867.
3
4 [23] D. Di Carlo, J.F. Edd, D. Irimia, R.G. Tompkins, M. Toner, Equilibrium
5 separation and filtration of particles using differential inertial focusing, Anal.
6 Chem., 80 (2008) 2204-2211.
7 [24] B. Geboers, H.J. Scheffer, P.M. Graybill, A.H. Ruarus, S. Nieuwenhuizen, R.S.
8
9
Puijk, P.M.v.d. Tol, R.V. Davalos, B. Rubinsky, T.D.d. Gruijl, D. Miklavčič, M.R.
10 Meijerink, High-Voltage Electrical Pulses in Oncology: Irreversible
11 Electroporation, Electrochemotherapy, Gene Electrotransfer, Electrofusion, and
12 Electroimmunotherapy, Radiology, 295 (2020) 254-272.
13 [25] G. Li, S. Yuan, S. Zheng, Y. Liu, G. Huang, In Situ Living Cell Protein Analysis
14
15 by Single-Step Mass Spectrometry, Anal. Chem., 90 (2018) 3409-3415.
16 [26] K. Glunde, M.A. Jacobs, Z.M. Bhujwalla, Choline metabolism in cancer:
17 implications for diagnosis and therapy, Expert Rev. Mol. Diagn., 6 (2006)
18 821-829.
19
[27] X. Chen, F.X. Wu, J. Chen, M. Li, DoRC: Discovery of rare cells from ultra-large
20
21 scRNA-seq data, 2019 IEEE International Conference on Bioinformatics and
22 Biomedicine (BIBM), 2019, pp. 111-116.
23 [28] F.M. Klis, P. Mol, K. Hellingwerf, S. Brul, Dynamics of cell wall structure in
24 Saccharomyces cerevisiae, FEMS Microbiol. Rev., 26 (2002) 239-256.
25
26
[29] P. Orlean, Architecture and biosynthesis of the Saccharomyces cerevisiae cell
27 wall, Genetics, 192 (2012) 775-818.
28 [30] Z. Li, Z. Wang, J. Pan, X. Ma, W. Zhang, Z. Ouyang, Single-Cell Mass
29 Spectrometry Analysis of Metabolites Facilitated by Cell Electro-Migration and
30 Electroporation, Anal. Chem., 92 (2020) 10138-10144.
31
32 [31] V. Ganeva, B. Galutzov, J. Teissie, Evidence that pulsed electric field treatment
33 enhances the cell wall porosity of yeast cells, Appl. Biochem. Biotechnol., 172
34 (2014) 1540-1552.
35 [32] K.I. Kocurek, J. Havlikova, E. Buchan, A. Tanner, R.C. May, H.J. Cooper,
36
37
Electroporation and Mass Spectrometry: A New Paradigm for In Situ Analysis of
38 Intact Proteins Direct from Living Yeast Colonies, Anal. Chem., 92 (2020)
39 2605-2611.
40 [33] M. Ohashi, R. Nasuno, S. Isogai, H. Takagi, High-level production of ornithine
41 by expression of the feedback inhibition-insensitive N-acetyl glutamate kinase in
42
43 the sake yeast Saccharomyces cerevisiae, Metabolic Engineering, 62 (2020) 1-9.
44 [34] V. Olin-Sandoval, J.S.L. Yu, L. Miller-Fleming, M.T. Alam, S. Kamrad, C.
45 Correia-Melo, R. Haas, J. Segal, D.A. Peña Navarro, L. Herrera-Dominguez,
46 Lysine harvesting is an antioxidant strategy and triggers underground polyamine
47
metabolism, Nature, 572 (2019) 249-253.
48
49
50
51
52
53
54
55
56
57
58
59
60 23
61
62
63
64
65
 Figure Legends
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 Figure 1. a) Schematic diagram of the microfluidics chip assisted high-throughput
36
37
38 single cell mass spectrometry analysis device. b) Photo of the asymmetric serpentine
39
40 channel microfluidic chip and ESI source coupled to mass spectrometer.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 24
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29 Figure 2. a) Schematic illustration of geometry of the serpentine channel. The height
30
31 of the microchannel is approximately 50 μm. b) Secondary flow distribution and
32
33
34
vectors in cross-section of the channel at 8th large turn and 9th small turn. c) The
35
36 streamlines of cells with the same diameter (ten 12 µm particles released at random
37
38 positions at the entrance and release time normally distributed in 0.5 s) after 4 s.
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 25
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33 Figure 3. Different cell suspensions composed of MCF7 and HepG2 cancer cell lines
34
35
analyzed by PEF-ESI-HRMS. a) TIC and b) EIC of phosphorylcholine (m/z
36
37
38 184.0733) acquired in the positive ion mode. Two distinct single-cell metabolic
39
40 profiles of c) MCF and d) HepG2 were obtained from the MS data.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 26
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34 Figure 4. Effect of the flow rate and cell concentration of the HepG2 and MCF7 cell
35
36 suspension on single-cell MS analysis. a) Relationship between the flow rate of the
37
38
39 cell suspension and the number of single cell MS signal per minute. b) Number of
40
41 single cell MS signal has a linear relationship with the cell concentration. c) EICs of
42
43 phosphorylcholine (m/z 184.0733) obtained by analyzing MCF7 cell suspensions with
44
45
46 concentrations of 8 × 104 cells mL-1 at flow rates of 1 µL min-1and the magnified
47
48 spectrum is at 1.5 to 1.75 min. d) Classifications of detected metabolites in positive
49
50
51 ion mode.
52
53
54
55
56
57
58
59
60 27
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48 Figure 5. a) PCA plot based on the first two principal components of the single
49
50 HepG2 and MCF7 cells. b) Modified raincloud plots of the relative contents of
51
52
53
leucine and creatine in MCF7 and HepG2 single cells. Upper part is split-half violin
54
55 plot and lower part is box plot. c) Outliers in HepG2 cells (marked by red x in 3D
56
57 PCA plot) identified by Isolation Forest. d) Modified raincloud plots of the relative
58
59
60 28
61
62
63
64
65
 contents of gamma-aminobutyric acid and aminoacetone in outliers and non-outliers
1
2 of HepG2 cells. e) Outliers in MCF7 cells (marked by red x in 3D PCA plot)
3
4
5 identified by Isolation Forest. f) Modified raincloud plots of the relative contents of
6
7 valine and phosphate in outliers and non-outliers of MCF7 cells.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 29
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14 Figure 6. a) Schematic illustration of the setup for the analysis of single yeast cell. b)
15
16
TIC and EICs of ornithine (m/z 133.0972), creatine (m/z 132.0768) and lysine (m/z
17
18
19 147.1128) obtained from single yeast cells by using a square-wave voltage with an
20
21 amplitude of 5 kV Vpp and frequency of 20 Hz.
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 30
61
62
63
64
65
Electronic Supplementary Material (online publication only)

Click here to access/download

Electronic Supplementary Material (online publication
only)
Supplementary Video S1.avi
Electronic Supplementary Material (online publication only)

Click here to access/download

Electronic Supplementary Material (online publication
only)
ACA support info.docx
Revised manuscript  (Clean version) Click here to view linked References

High-throughput Single Cell Metabolomics and Cellular

1
2
3 Heterogeneity Exploration by Inertial Microfluidics coupled with
4
5
6 Pulsed Electric Field-Induced Electrospray Ionization-High
7
8
9 Resolution Mass Spectrometry
10
11 Disheng Feng1,2, Hang Li1, Tianrun Xu1,2, Fujian Zheng1,2,3, Chunxiu Hu1,3, Xianzhe
12
13
14 Shi1,3*, Guowang Xu1,2,3*
15
16 1
CAS Key Laboratory of Separation Science for Analytical Chemistry, Dalian
17
18
19 Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, 116023, China
20
2
21 University of Chinese Academy of Sciences, Beijing 100049, China.
22
23 3
24
Liaoning Province Key Laboratory of Metabolomics, Dalian, China
25
26
27
28 *
Correspondence to:
29
30
31 Dr. Xianzhe Shi, CAS Key Laboratory of Separation Science for Analytical
32
33 Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences,
34
35
36 Dalian 116023, China. E-mail: shixianzhe@dicp.ac.cn.
37
38 Prof. Dr. Guowang Xu, CAS Key Laboratory of Separation Sciences for Analytical
39
40
41 Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences,
42
43 Dalian 116023, China. Tel. / Fax: 0086-411-84379530. E-mail: xugw@dicp.ac.cn
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 1
61
62
63
64
65
 ABSTRACT:
1
2
3 Single cell metabolomics can obtain the metabolic profiles of individual cells and reveal
4
5 cellular heterogeneity. However, high-throughput single-cell mass spectrometry (MS)
6
7
8 analysis under physiological conditions remains a great challenge due to the presence
9
10 of complex matrix and extremely small cell volumes. Herein, a serpentine channel
11
12
13
microfluidic device which was designed to achieve continuous cell separation and
14
15 inertial focusing, was coupled with a pulsed electric field-induced electrospray
16
17 ionization-high resolution MS (PEF-ESI-HRMS) to achieve high-throughput single cell
18
19
20 analysis. The pulsed square wave electric field was applied to realize on-line cell
21
22 disruption and induce electrospray ionization. Single cells were analyzed under near-
23
24
25 physiological conditions at a throughput of up to 80 cells min-1. More than 900 features
26
27 were detected and approximately 120 metabolites were tentatively identified from a
28
29
30 single cell. Further, by continually analyzing more than 3000 MCF7 and HepG2 cells,
31
32 discrimination of different cancer cells based on their individual metabolic profiles was
33
34
35
achieved by using the principal component analysis. The PEF-ESI-HRMS method was
36
37 also applied for the analysis of single yeast cells, and more than 40 metabolites were
38
39 annotated. This method is versatile and has good robustness, which is promising for
40
41
42 high-throughput single cell metabolomics analysis.
43
44 Key words: Cellular heterogeneity; Single cell metabolomics; High-throughput; Mass
45
46
47 spectrometry; Microfluidic device
48
49
50
51
52
53
54
55
56
57
58
59
60 2
61
62
63
64
65
 1. Introduction
1
2
3 The complex life activities of organisms are orchestrated by coordinated interactions of
4
5 a diverse range of individual cells [1]. The metabolites are the end products of
6
7
8 biochemical regulation, which can accurately represent the biochemical phenotype of
9
10 cells in real time. Due to the heterogeneity of cells and fast turnover rates of some
11
12
13
metabolites, the traditional large-scale assays that investigate the state of large number
14
15 of cells cannot accurately reflect the behavior of the individual cells [2, 3]. Therefore,
16
17 it is increasingly necessary to study metabolites in a single cell under physiological
18
19
20 environmental conditions.
21
22 Single-cell metabolomics can directly obtain the metabolite information of individual
23
24
25 cells, which provides insight into the relationship between physiological behavior of
26
27 single cells and their chemical compositions. The latest development in mass
28
29
30 spectrometry has ushered in a new era of single-cell metabolomics analysis.
31
32 Nanoelectrospray ionization mass spectrometry (nano-ESI-MS) has been used in
33
34
35
single plant cell metabolomics analysis [4]. However, single mammalian cells and yeast
36
37 cells are difficult to be directly analyzed by traditional nano-ESI-MS due to its
38
39 extremely small sample volume, viscous cytoplasm and the possible presence of cell
40
41
42 walls. Some modified nano-ESI-MS methods have been developed in the recent years.
43
44 Induced nano-ESI-MS by using alternating current high-voltage to produce inductive
45
46
47 electrospray was used to investigate the metabolism of intracellular metabolites of
48
49 single neurons and revealed a novel glutamate biosynthetic pathway in brain [5-7].
50
51
52
Pulsed direct current electrospray ionization mass spectrometry (pulsed-dc-ESI-MS)
53
54 can decrease the flow rate of sample to picoliters per minute and extend the electrospray
55
56 time to acquire rich MS2 information. The pulsed-dc-ESI-MS strategy was also
57
58
59
60 3
61
62
63
64
65
 combined with microwell-based droplet microextraction to study the glucose-
1
2 phosphate and phospholipids in single cells [8, 9]. However, most of the
3
4
5 abovementioned MS methods employ manual capillary probe sampling, severely
6
7 limiting the throughput of single-cell analysis.
8
9
10 For molecular biology research and actual clinical applications, it is necessary to
11
12 analyze a large number of individual cells in a short period of time. Therefore, high-
13
14
15 throughput single-cell analysis methods are urgently required. To achieve high-
16
17 throughput analysis, cell separation and focusing should be performed to ensure that
18
19
single cells are introduced one by one into the mass spectrometer. In the previous work,
20
21
22 a drop-on-demand inkjet cell printing was combined with probe ESI-MS. The droplets
23
24 containing single cells were produced by the inkjet sampling of cell suapension [10].
25
26
27 But it is difficult to ensure that the number of cells contained in each printed droplet is
28
29 just one. Since then, a Dean flow assisted cell sorting system based on spiral capillaries
30
31
32 coupled with ESI-MS was developed, which can reduce the agglomeration and uneven
33
34 distribution of cells in the cell suapension [11]. Recently, a label-free mass cytometry
35
36
37
named CyESI-MS was proposed by coupling flow cytometry to ESI-MS. As the core
38
39 part of CyESI-MS, three coaxial capillaties were used to deliver cell suspension, sheath
40
41 fluid, and sheath gas, respectively. Cells were dispersed and the intracellular
42
43
44 metabolites were extracted by the surrounding sheath fluid. The highest throughput of
45
46 CyESI-MS is approximately 38 cells min-1 [12, 13], but the sheath fluid at a flow rate
47
48
49 of 10 µL min-1 may lead to severe dilution effect. The increase in dead volume and
50
51 dilution of metabolites may also result from the large opening of the sheath fluid
52
53
54 capillary. Technical advances in microfluidics conjunction with mass spectrometry have
55
56 enabled their use in single-cell metabolomics analysis [14]. The microfluidic devices
57
58
are composed of a network of microchannels with micrometer-scale dimensions
59
60 4
61
62
63
64
65
 comparable in length scale to various kinds of cells, which can realize precise cell
1
2 manipulation based on the interactions between cells, fluids, and microchannels. In
3
4
5 addition, the small geometry of microfluidic devices allows the integration with various
6
7 detectors. In a recent study, a simple microfluidic chip consisting of 5 curved loops was
8
9
10 applied for cell dispersing and ordering and then a chip-nanoESI mass cytometry was
11
12 established with a throughput of about 40 cells min-1 [15]. However, the cells were
13
14
15 suspended in aqueous methanol solution which could affect cellular metabolism and
16
17 thus could not reveal authentic metabolite levels of cells under their physiological
18
19
conditions [15, 16]. Until now, the reported single-cell mass cytometrymethods are
20
21
22 capable of analyzing only 500 to 1000 cells in an experiment [17]. To our knowledge,
23
24 there are no methods that can be used to analyze several thousands of cells in single
25
26
27 experiment, while analyzing a large number of cells is of great importance in obtaining
28
29 biologically meaningful data.
30
31
32 Herein, we developed a novel asymmetric serpentine channel microfluidic chip coupled
33
34 to pulsed electric field-induced electrospray ionization-high resolution mass
35
36
37
spectrometry (chip-PEF-ESI-HRMS) method for high-throughput single cell analysis
38
39 under near-physiological conditions. The microfluidic device was employed for
40
41 continuous cell separation and inertial focusing, which is sheathless and operated
42
43
44 without any external force fields. The applied pulsed electric field has dual functions,
45
46 online cell disruption and inducing electrospray ionization. It is worth mentioning that
47
48
49 single cells were suspended in an aqueous solution under near-physiological condition
50
51 (i.e., isotonic salt concentration) instead of a cytotoxic methanol solution, which can
52
53
54 maintain the cell viability. The throughput of our method is up to 80 cells min -1, the
55
56 whole system can be operated continuously and stably for more than 3 h, more than
57
58
3000 single cells can be analyzed in one experiment. Approximately 120 metabolites in
59
60 5
61
62
63
64
65
 cancer cells were annotated and two different cancer cells (MCF7 and HepG2 cells)
1
2 were successfully distinguished based on individual cell metabolic profiling. We also
3
4
5 applied this PEF-ESI-HRMS method for single yeast cell analysis and over 40
6
7 metabolites were tentatively identified. This method is versatile and can be used to
8
9
10 obtain metabolomic information from various kinds of single cells under near
11
12 physiological conditions in a high-throughput way.
13
14
15
16
17 2. Experimental
18
19
20 2.1 Chemicals and Materials
21
22 Ammonium bicarbonate was obtained from Fluka (Steinheim, Germany). The ultrapure
23
24
25 water was produced by a Milli-Q water purification system (Millipore, Bedford, MA,
26
27 USA). The nanoelectrospray emitter was made of fused silica capillaries with 100 μm
28
29
30 i.d. and 365 μm o.d. The capillary was firstly disposed on fire and then the coating of
31
32 the capillary was peeled off, then it was pulled by the P-2000 puller (Sutter Instruments,
33
34
35
Novato, CA, USA) to make the emitter with a tip of about 10 μm. The borosilicate glass
36
37 capillaries (BF150-86-10) with 0.86 mm inner diameter (i.d.) and 1.5 mm outer
38
39 diameter (o.d.) were purchased from Sutter Instrument (Novato, CA, USA). The
40
41
42 micropipettes were fabricated with a P-1000 puller (Sutter Instruments, Novato, CA,
43
44 USA). The fabricated micropipettes tips with an inner diameter (i.d.) of approximately
45
46
47 2 μm were used for the the single yeast cell analysis.
48
49
50
51
52
2.2 Cell culture and sample preparation
53
54 The mammalian cancer cell lines HepG2 and MCF7 were maintained in Dulbecco's
55
56 Modified Eagle's Medium (DMEM; Gibco, Waltham, MA, USA), supplemented with
57
58
59
60 6
61
62
63
64
65
 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin solution in a humidified
1
2 incubator at 37°C with 5% CO2. All the cells used in this work were cultured in Petri
3
4
5 dishes (10 cm in diameter). For the cell analysis by the high-throughput method with
6
7 the asymmetric serpentine channel microfluidic chip, the cells were trypsinized by
8
9
10 using 0.25% Trypsin-EDTA solution (Gibco, Waltham, MA, USA) and washed 3 times
11
12 by PBS. Then the cells were resuspended in 150 mM NH4HCO3 aqueous solution. The
13
14
15 obtained cell suspension was diluted to different concentrations (2.5 × 103, 5 × 103, 1 ×
16
17 104, 2 × 104, 4 × 104 and 8×104 cells mL-1) and drawn up into a syringe and the syringe
18
19
was loaded onto a syringe pump to inject the cell suspension into the microfluidic chip.
20
21
22 The yeast cells used were Saccharomyces cerevisiae strain H1246 and were kindly
23
24 provided by Qingdao Institute of Bioenergy and Bioprocess Technology, Chinese
25
26
27 Academy of Sciences. The synthetic complete (SC) medium containing 0.67% yeast
28
29 nitrogen base (w/v), 0.074% DO supplement (-Ura) (w/v) and 2% glucose (w/v) was
30
31
32 used for cultivating yeast. The yeast cells were inoculated into the centrifuge tubes and
33
34 incubated at 30°C with shaking at 200 rpm overnight. Then the cells were centrifuged
35
36
37
and washed with ultrapure water for three times to remove the culture medium. The
38
39 obtained yeast cell suspension was diluted to ~5×103 cells mL-1. Since yeast cells are
40
41 much smaller than mammalian cancer cells, a back-end injection method was used to
42
43
44 make yeast cells move to the tip of the spray emitter. Then the pulsed electric field was
45
46 applied to achieve the cell membrane electroporation and metabolites ionization
47
48
49 process.
50
51
52
53
54 2.3 Microfluidic chip fabrication
55
56 The asymmetric serpentine channel microfluidic chip (Figure 1a) was fabricated by
57
58
using the softlithography technique. Briefly, SU-8 3035 (Microchem, Newton, MA)
59
60 7
61
62
63
64
65
 was spun at 2,000 rpm for 60 s on a spin coater (KW-4A, Chemat Technology Inc.,
1
2 Northridge, CA, USA) to create a 50-μm-thick layer on a silicon wafer. The master
3
4
5 molds for PDMS replica were photolithographically defined by using a mask. The mold
6
7 was pre-treated with chlorotrimethylsilane (Sigma Aldrich, St. Louis, MO, USA) vapor
8
9
10 overnight to facilitate the release of PDMS. The PDMS prepolymer and curing agent
11
12 (RTV615, Momentive Performance Materials, NY, USA) were mixed completely (parts
13
14
15 A and B in 10:1 ratio) and the air bubbles generated were removed by using a vacuum
16
17 desiccator for 1 h. Then the mixtures were poured onto the silicon master and the PDMS
18
19
was cured in the oven at 80°C for 2 h. After curing, the PDMS layer was peeled off the
20
21
22 mold and the holes for inlet and outlet ports were punched with a sharpened flat-tip
23
24 needle. The device was cleaned via sonication in ethanol and 2-propanol before bonding
25
26
27 with a galss slide. After plasma treatment and placement onto the glass substrate, the
28
29 devices were placed in the oven at 80°C for 30 min to strengthen the bonding.
30
31
32
33
34 2.4 Pulsed electric field-induced electrospray ionization-high resolution mass
35
36
37
spectrometry
38
39 In this study, an asymmetric serpentine channel microfluidic chip cell focusing system
40
41 is connected to the pulsed electric field-induced ESI-MS to achieve the high-throughput
42
43
44 single cell metabolomics analysis. The schematic diagram of the device is shown in
45
46 Figure 1a. An inverted microscope (CKX53, Olympus, Tokyo, Japan) was used to
47
48
49 observe the single cells and the cell focusing process. A commercial NanoSpray Flex
50
51 ion source (Thermo Scientific, San Jose, CA, USA) was modified to accommodate with
52
53
54 the micropipettes and the home-made high voltage pulsed power supply. The setup of
55
56 the microfluidic chip coupled to pulsed electric field-induced electrospray ionization
57
58
(ESI) source was shown in Figure 1b. The microfluidic chip was placed on a home-
59
60 8
61
62
63
64
65
 made 3-dimensional adjustment frame with high positioning accuracy. The inlet was
1
2 connected to a syringe pump. The out flow from the microchannel chip was connected
3
4
5 with the nanoelectrospray emitter. High voltage was applied to a copper plate which is
6
7 placed beneath the nanoelectrospray emitter. An adjustable high voltage power supply
8
9
10 was constructed to provide square wave pulsed high voltage. The power supply can
11
12 operate stably for long time without significant heating problem, which is superior to
13
14
15 the previously reported conventional high-voltage power supplies based on magnetic
16
17 amplifier[18] .
18
19
The circuit diagram of the adjustable square wave pulse high voltage power supply is
20
21
22 shown in Figure S1. The power supply is mainly composed of a DC positive high-
23
24 voltage module (DW-P502-2CQ1, Dongwen High Voltage Power Supply, Tianjin,
25
26
27 China), a DC negative high-voltage module (DW-N502-2CQ1, Dongwen High Voltage
28
29 Power Supply, Tianjin, China) and a fast high voltage transistor switch (HTS 121-01-
30
31
32 HB-C, Behlke Power Electronics, Kronberg, Germany). The input signal is generated
33
34 by the function waveform generator (DG1022U, RIGOL, Beijing, China). The principle
35
36
37
is to generate a square wave pulse voltage through the function waveform generator
38
39 aforementioned and input it to the fast high voltage transistor switch. By controlling the
40
41 DC high voltage modules, the power supply can output square wave pulse high voltage
42
43
44 with voltage amplitude range from -5 kV to 5 kV and square wave frequency range
45
46 from 1 to 1000 Hz. Typically, a square wave high voltage with an amplitude of 5 kV, a
47
48
49 frequency of 200 Hz, and a duty cycle of 50% is used unless otherwise stated.
50
51 The Q Exactive HF mass spectrometer (Thermo Scientific, San Jose, CA, USA) was
52
53
54 utilized for the MS experiment. The setup photo of the microfluidic chip and the pulsed
55
56 electric field-induced ESI source coupled to mass spectrometer is shown in Figure 1.
57
58
The pulse voltage is generated by home-made high-voltage power supplies (Figure 1b).
59
60 9
61
62
63
64
65
 The MS was operated in full-scan positive ion mode with the following settings: mass
1
2 range 70-1,050 m/z; resolution for MS1 scan 60,000 at m/z 200; microscans: 1; sheath
3
4
5 gas flow rate: 0; aux gas flow rate: 0; sweep gas flow rate: 0; capillary temperature:
6
7 275°C; S-Lens RF level: 50. For high-throughput single-cell analysis, the resolution for
8
9
10 MS1 scan was set to 60,000 with a scan speed of 7 Hz to achieve a balance between
11
12 mass resolution and scanning speed. For the analysis of yeast cell samples, the
13
14
15 resolution for MS1 scan was 120,000 at m/z 200. The AGC target was set to 5e5 and
16
17 the maximum injection time was set to 50 ms.
18
19
20
21
22 2.5 Data processing and analysis
23
24 The Xcalibur software (version 4.0.27.10, Thermo Scientific, San Jose, CA, USA) was
25
26
27 used for the control of the MS system and data analysis. The obtained .raw files were
28
29 firstly processed in the Qual Browser tool of the Thermo Xcalibur software to obtain
30
31
32 the mass spectrum list. First, the peaks in the extracted ion chromatogram (EIC) at m/z
33
34 184.0733 were used as the signal of single cell events, because this ion was assigned
35
36
37
to phosphorylcholine which is related to the damage of cell membranes. Ion peaks with
38
39 the same appearance time as the single cell events and the detection rate greater than
40
41 25% among analyzed single cells were chosen as single cell mass spectral features.
42
43
44 Before subsequent statistical analysis, normalization was performed based on the total
45
46 ion intensity of all detected ions in each mass spectrum. A peak alignment program
47
48
49 written in R language was used for peak matching of different single cell samples. The
50
51 metabolites were identified based on the accurate mass (mass error <5 ppm), MS2
52
53
54 spectra and the identification results obtained by LC-MS analysis of population cells.
55
56 Detailed information about the LC-MS analysis and metabolite identification is
57
58
described in the Supporting Information. The Human Metabolome Database (HMDB,
59
60 10
61
62
63
64
65
 http://www.hmdb.ca) and mzCloud (https://www.mzcloud.org/) library were also used
1
2 to broaden the scope of metabolite identification [19]. Whether HMDB reports that the
3
4
5 metabolite is present in cells was also considered to determine the m/z corresponding
6
7 metabolites during the identification process. The results acquired were then used for
8
9
10 further statistical analysis. The data analysis and visualization were performed by using
11
12 the in-house Python script and the web-based metabolomics tool (MetaboAnalyst 5.0;
13
14
15 https://www.metaboanalyst.ca/) [20].
16
17
18
19
20 3. Results and Discussion
21
22 3.1 Design of asymmetric serpentine microfluidic chip
23
24
25 To achieve the high-throughput analysis, a cell separation and focusing system for
26
27 sample introduction is essential before MS analysis. Inertial focusing in microfluidics
28
29
30 can achieve the precise lateral equilibrium position control of cells and the regulation
31
32 of cell spacing without the use of sheath fluid or other external fields [21, 22]. By using
33
34
35
this sheathless cell focusing method, the sampling rates can be increased because cells
36
37 are aligned without dramatically increasing their velocity or the overall volumetric flow
38
39 rate of carrier fluid. Consequently, the metabolites in the cells are not overly diluted,
40
41
42 thus making it possible to expand the detection range of metabolites, it is a promising
43
44 sample introduction method for single cell mass cytometry.
45
46
47 Based on the abovementioned work on microfluidic inertial focusing, a microfluidic
48
49 chip with asymmetric serpentine channels was designed, and it consists of 15 repeated
50
51
52
S-shaped units and a connected straight channel (Figure 2a). The microchannel in the
53
54 inlet region was widened to reduce the normal and shear stresses over cells surfaces
55
56 [23]. The height of the microchannel is 50 μm. Fluid flow simulation within the
57
58
59
60 11
61
62
63
64
65
 proposed asymmetric serpentine microchannel was performed by using COMSOL
1
2 Multiphysics 5.6 (COMSOL AB, Stockholm, Sweden). The simulation process
3
4
5 comprises of two major steps, in the first step the flow velocity distribution was
6
7 computed by using laminar flow module under steady-state conditions and in the second
8
9
10 step the motion trajectories of particles in the fluid was computed using particle tracing
11
12 module. The outlet boundary condition was set to “freeze” and the microchannel wall
13
14
15 conditions were set to “bounce”. Water is used as the simulation fluid and the injection
16
17 flow rate was set to 1 µL min-1 to simulate the flow rate under actual experiment
18
19
conditions of electrospray. The drag force, the up, down and side microchannel walls
20
21
22 induced forces acting on the particles were considered when simulating particle
23
24 trajectories. The Stokes law was used for drag forces and wall induced lift law was used
25
26
27 for lift forces. As shown in Figure 2b, the axial velocity magnitude plot in the cross-
28
29 section of the channel indicated that the maximum velocity occurs at the inside of the
30
31
32 turn, and the minimum velocity occurs near the outside of the turn due to the no-slip
33
34 condition on the walls. The streamlines of cells with the same diameter (ten 12 µm
35
36
37
particles released at random positions at the entrance and the release time normally
38
39 distributed at 0.5 s were used as homogeneous cell suspension model proximity model)
40
41 after different times were simulated (Figure 2a and S2a). It can be observed that at the
42
43
44 outlet, the particles outflow almost at the same position. Five spherical particles of
45
46 different diameters are released at the inlet cross-section of the microchannel in 0.02 s
47
48
49 as cell mixture proximity models, and the trajectories and positions of the first three
50
51 particles after 3.55 s are shown in Figure S2b. These results demonstrated that the
52
53
54 microfluidic chip has a dispersing and focusing effect on mammalian cancer cells.
55
56
57
58
3.2 Configuration and perfermance of chip-PEF-ESI-HRMS
59
60 12
61
62
63
64
65
 The chip-PEF-ESI-HRMS contains four parts: a syringe pump for delivering cell
1
2 suspension, an asymmetric serpentine channel microfluidic chip for dispersing and
3
4
5 focusing cells, PEF-ESI source, and Q Exactive HF mass spectrometer. In order to
6
7 maintain cell viability and integrity, the cells were suspended in an isotonic solution of
8
9
10 ammonium bicarbonate (150 mM NH4HCO3), an MS-friendly volatile salt. The cell
11
12 suspension was injected into the microfluidic chip by a syringe pump. Then the cells
13
14
15 were dispersed and focused through asymmetric microchannels with curvature. A
16
17 movie of single cells passing through the microchannel is provided in Supplementary
18
19
Video S1. Once the ordered single cells reach the tip of nanospary emitter, high voltage
20
21
22 square-wave pulses generated by our home-made power supply were applied for online
23
24 cell disruption without the addition of cytotoxic organic solvents. The electric field
25
26
27 strength between the emitter and the MS is approximately 10 kV cm-1, and the pulses
28
29 width can be freely adjustable, which enables the irreversible cell electroporation and
30
31
32 disruption [24]. Meanwhile, high voltage square-wave pulses induce the electrospary
33
34 ionization processes and this phenomenon was also observed in previous studies [5, 25].
35
36
37
As a result, the released cellular contents are immediately ionized and analyzed by the
38
39 mass spectrometer which helps to obtain the real-time information of metabolites from
40
41 single live cells.
42
43
44 The performance of chip-PEF-ESI-HRMS was evaluated in single-cell metabolomics
45
46 analysis. The cell suspensions of two cancer cell lines (MCF7 and HepG2) were
47
48
49 introduced into the system. As shown in the total ion chromatogram (TIC) (Figure 3a),
50
51 pulse-like signals corresponding to the single cells were obtained when the cell
52
53
54 suspension was flowing. The peak at m/z 184.0733 in the extracted ion chromatogram
55
56 (EIC) is assigned to phosphorylcholine, an important metabolite predominantly present
57
58
in the cytoplasm. The appearance of this signal represents the disruption of the cell and
59
60 13
61
62
63
64
65
 is therefore used as a marker for single cell mass spectrometry events. Previous studies
1
2 have demonstrated elevated levels of phosphorylcholine and total choline compounds
3
4
5 in a variety of cancers [26]. Moreover, the EIC of phosphorylcholine (Figure 3b) has a
6
7 good agreement with the TIC, indicating that the pulsed electric field-induced ESI
8
9
10 source has the ability to realize the online cell fragmentation and metabolites ionization.
11
12 From the MS data, two distinct single-cell metabolic profiles of MCF7 (Figure 3c) and
13
14
15 HepG2 (Figure 3d) were obtained, it can be observed that the overall relative abundance
16
17 of metabolites in HepG2 liver cancer cells is higher.
18
19
It is important to note that for high-concentration cell suspension, the interparticle
20
21
22 interactions can limit the degree of focusing because cells can randomly disturb the
23
24 ideal parabolic flow of the fluid and the lift, and Dean drag forces can be altered [23].
25
26
27 Therefore, the flow rate and concentration of cell suspensions are two important factors
28
29 affecting the single cell mass cytometry. Different flow rates were investigated to
30
31
32 optimize the analysis throughput and improve system stability. When flow rate was
33
34 greater than 1 μL min-1, liquid leakage was observed at the interface between the
35
36
37
microfluidic chip and the inlet stainless steel tube, which may cause the stainless steel
38
39 tube to dislodge when subjected to excessive pressure. Additionally, large volumetric
40
41 flow rates may lead to insufficient time for separation forces to act on cells. HepG2 cell
42
43
44 suspension with a concentration of 2 × 104 cells mL-1 was infused at flow rates of 0.5
45
46 μL min-1, 0.75 μL min-1 and 1 μL min-1. The relationship between the flow rate of the
47
48
49 cell suspension and the number of single cell detected by MS per minute is shown in
50
51 Figure 4a and Table S1. When the flow rate increased, the number of single cell detected
52
53
54 also increased. The highest number of single-cell correlation pulse peaks per unit time
55
56 was obtained at 1 μL min-1. After comprehensively considering the pressure that the
57
58
microfluidic chip interface can withstand and the maximum analytical throughput, 1 μL
59
60 14
61
62
63
64
65
 min-1 was selected as the appropriate flow rate. In order to adapt to this relatively high
1
2 injection flow rate, a square wave voltage with an amplitude of 5 kV Vpp and a
3
4
5 frequency of 200 Hz was applied in the nanoelectrospray emmiter. In addition, HepG2
6
7 and MCF7 cell suspensions with different concentrations (2.5 × 103, 5 × 103, 1 × 104, 2
8
9
10 × 104, 4 × 104 and 8×104 cells mL-1) were injected at a flow rate of 1 µL min-1 to
11
12 optimize the concentration of cell suspension, and the results are shown in Figure 4b
13
14
15 and Table S2. The number of single cell MS signals per unit time increased with the
16
17 increase of cell concentration and a good linear correlation between them was found.
18
19
This result also indicated that single cells were seriatim detected by the mass
20
21
22 spectrometer. As shown in Figure 4c, the maximum throughput could reach about 80
23
24 cells min-1 by applying MCF7 cell suspension with a concentration of 8×104 cells mL-
25
26
27
1
at a flow rate of 1 μL min-1. The magnified EIC at m/z 184.0733 shows that about 20
28
29 cells were analyzed within 15 s and the intensity of peaks for the ions dropped to zero
30
31
32 between two pulse peaks, which indicated that almost no mutual interference existed
33
34 between each single-cell mass spectrometry event. The difference in the width of each
35
36
37
pulse peak may be due to the cell size. Compared to the previously reported method
38
39 [12], the mass spectrum obtained by our method is more stable.
40
41 The mass spectral information at the highest point of each pulse peak was extracted for
42
43
44 further data processing and analysis. More than 900 features can be detected from a
45
46 single cell by using the chip-PEF-ESI based mass cytometry. Among them, about 120
47
48
49 metabolites were tentatively identified in a single cell based on the accurate mass and
50
51 secondary mass spectrum obtained from the LC-MS analysis results of population cells
52
53
54 (Table S3). These identified metabolites were classified by their chemical
55
56 characteristics and include most kinds of important functional metabolites such as
57
58
amino acids, acylcarnitines, carboxylic acids and their derivatives or analogues (Figure
59
60 15
61
62
63
64
65
 4d). In order to maintain the cells in their native state as much as possible, an isotonic
1
2 salt solution was used as the extractive solvent rather than organic solvents. Therefore,
3
4
5 the detected metabolites were mainly polar metabolites with good water solubility while
6
7 high-abundance lipids with poor water solubility were rarely found. More than 80% of
8
9
10 the annotated metabolites were also detected by the traditional LC-MS method. Those
11
12 metabolites only detected by the single cell method are mainly carboxylic acids and
13
14
15 their derivatives as well as several kinds of oxygenated organic compounds. The reason
16
17 could be that these metabolites may be lost, unstable and degraded in the pretreatment
18
19
process of population cells. Therefore, the chip-PEF-ESI based mass cytometry showed
20
21
22 not only high throughput but also high coverage of polar metabolites. Furthermore, in
23
24 order to test the stability and robustness of the instrument system, the chip-PEF-ESI
25
26
27 based mass cytometry was operated continuously for about three hours and stable mass
28
29 spectrometry signals were still obtained without clogging of the microchannel or the
30
31
32 nanospray emitter (Figure 3a). The reason for this phenomenon may be due to the self-
33
34 cleaning ability of the electric-field-induced electrospray [5].
35
36
37
38
39 3.3 Practical applications of the developed method in high-throughput analysis of single
40
41 mammalian cells
42
43
44 Traditional metabolomics studies generally use millions of cells as samples and require
45
46 relatively long sample pretreatment time. Due to the intra-species variation of cells and
47
48
49 the fast turnover rate of metabolites, this kind of strategy is inappropriate for studying
50
51 the heterogeneity of cells. In contrast, the metabolic profile of individual cells can be
52
53
54
obtained by using single-cell mass spectrometry, and thus their intra- and interspecies
55
56 metabolic heterogeneity can be better represented. In order to achieve the statistical
57
58 characterization of single-cell metabolic information and obtain more biologically
59
60 16
61
62
63
64
65
 meaningful results, the analysis of a large number of single cells must be performed.
1
2 More than 3000 HepG2 cells and MCF7 cells were continually analyzed by the chip-
3
4
5 PEF-ESI based mass cytometry, and the mass spectrometry data from 1000 single cells
6
7 of each type were used for further analysis to demonstrate the practicality of this high-
8
9
10 throughput method in distinguishing the metabolic difference of various cancer cells.
11
12 The two types of cells can be well distinguished based on the PCA score plot (Figure
13
14
15 5a). The PCA score plot also reflect the heterogeneity within the same kind of cells and
16
17 the stability of the method in the whole analysis process.
18
19
Machine learning based t-SNE (t-distributed stochastic neighbor embedding) and
20
21
22 UMAP (uniform manifold approximation and projection) algorithm were further
23
24 applied to mine and visualize the complex single cell MS data sets. The separation of
25
26
27 MCF7 and HepG2 cells is clearly observed as shown in Figure S3a and S3b, and
28
29 revealed significant metabolic heterogeneity of two kinds of cancer cells. Modified
30
31
32 raincloud plots (consisting of split-half violin plots and box plots) of the relative
33
34 contents of representative metabolites including leucine and creatine are shown in
35
36
37
Figure 5b. These plots show the distribution and key summary statistics of metabolites
38
39 levels in cells, which illustrate the difference in metabolite levels in two kinds of cancer
40
41 cells. To identify potential biomarkers that differentiate these two kinds of cancer cells
42
43
44 and evaluating their performance, classical univariate receiver operating characteristic
45
46 (ROC) curve analyses were performed. The scatter box plots and ROC curves of several
47
48
49 metabolites are shown in Figure S4. The results suggest that these metabolites can serve
50
51 as biomarkers to differentiate MCF7 and HepG2 cells. To explore the metabolic
52
53
54 heterogeneity within the same kind of cells, we further implemented outlier detection
55
56 and data visualization by using an unsupervised learning algorithm Isolated Forest, a
57
58
well-studied and widely used outlier detection method in machine learning [27]. Thirty-
59
60 17
61
62
63
64
65
 two potential outliers among 1000 HepG2 cells were found and marked by red x in 3-
1
2 dimensional (3D) PCA plot (Figure 5c). The data points obtained from non-outlier
3
4
5 single HepG2 cells are well clustered in the PCA scoring plot, further demonstrating
6
7 the good stability of the device when analyzing large scale single cell samples. Modified
8
9
10 raincloud plots are used for detailed data visualization of outliers versus non-outlier
11
12 cells in the same kind of cell, and the relative amounts of metabolites in HepG2 cells
13
14
15 that have a relatively large impact on the outlier detection model, including gamma-
16
17 aminobutyric acid and aminoacetone, are shown in Figure 5d. Overall, the metabolite
18
19
contents in outlier cells of HepG2 cells was higher than those in non-outlier cells.
20
21
22 Similaryly, 61 potential outliers in 1000 MCF7 cells were identified and the results are
23
24 shown in Figure 5e. Several key metabolites closely related to energy metabolism, such
25
26
27 as valine and phosphate had relatively higher contents in the outlier MCF7 cells (Figure
28
29 5f). Based on the above two cell heterogeneity exploration results, the probability of
30
31
32 heterogeneous cells is about 5%. These results demonstrate that high-throughput chip-
33
34 PEF-ESI based mass cytometry has the potential to explore the cellular metabolic
35
36
37
heterogeneity.
38
39
40
41 3.4 Practical applications of the developed method in metabolic profiling analysis of
42
43
44 single yeast cells
45
46 To further demonstrate the on-line cell disruption capability of the PEF-ESI source for
47
48
49 the cells with cell wall, the baker's yeast (Saccharomyces cerevisiae strain H1246) was
50
51 used as the model cell. The cell walls of this kind of yeast cells have an elastic structure
52
53
54 and are mainly composed of β-1,3-glucans, β-1,6-glucans, mannoproteins and chitin
55
56 [28, 29]. Traditionally, enzymatic hydrolysis, grinding and ultrasound were often used
57
58
to break the yeast cell wall, however, these methods are unsuitable for single yeast
59
60 18
61
62
63
64
65
 analysis. Recently, the manipulation and analysis of single yeast cells by using cell
1
2 electro-migration and electroporation techniques has been reported [30], but it required
3
4
5 applying multiple steps of DC and pulse voltages which is complicated and difficult to
6
7 be operated in a high-throughput way. Previous studies had shown that the yeast cell
8
9
10 wall can be destructed by pulsed electric field [31, 32], therefore by applying the square
11
12 wave high-voltage pulse with an amplitude of 5 kV Vpp and frequency of 20 Hz
13
14
15 generated by our home-made power supply, single yeast cells were rapidly
16
17 electroporated, and the intracellular metabolites were successfully released and ionized.
18
19
The TIC and EICs of ornithine (m/z 133.0972), creatine (m/z 132.0768) and lysine (m/z
20
21
22 147.1128) obtained from single yeast cells are shown in Figure 6a and 6b, the strong
23
24 mass spectral signal peaks in the initial stage may originate from the electroporation
25
26
27 and rupture of cells. Ornithine is one of the products of the action of arginase on arginine,
28
29 which plays an important role in the urea cycle [33]. Creatine is a kind of nitrogenous
30
31
32 organic acid which is associated with the energy metabolism. Lysine is also an
33
34 important amino acid that is taken up in large quantities by yeast cells, which can lead
35
36
37
to metabolic reorganization and make the cells resistant to oxidants [34]. More than 40
38
39 metabolites including other important amino acids were detected from single yeast cells
40
41 in positive ion mode which illustrates the potential application of this method for the
42
43
44 metabolomics analysis of cells with cell walls.
45
46
47
48
49 4. Conclusions
50
51
52
In summary, a versatile chip-PEF-ESI based mass cytometry method was established
53
54 for high-throughput single-cell metabolomics analysis under near physiological
55
56 conditions. The novel combination of PEF-ESI-HRMS with the inertial microfluidic
57
58
59
60 19
61
62
63
64
65
 device has significant advantages in providing higher analytical throughput and stable
1
2 electrospray, while maintaining the analytical sensitivity. The analysis of ca. 80 single
3
4
5 cancer cells can be finished within 1 minute, more than 900 features were extracted
6
7 from single HepG2 or MCF7 cell, and 120 metabolites were tentatively identified, the
8
9
10 discrimination of these two types of cancer cell lines was achieved with this high-
11
12 throughput method. Moreover, the metabolic profiling outliers within the same kind of
13
14
15 cells were further identified by the isolation forest algorithm, which illustrated the cells
16
17 with metabolic heterogeneity can be defined. More importantly, this method has a good
18
19
analytical robustness which allows more than 3000 cells to be continuously analyzed in
20
21
22 a single experiment. The PEF-ESI ionization source also has the capability to analyze
23
24 the cells with cell wall such as yeast cell, and more than forty important metabolites
25
26
27 such as amino acids can be detected in single yeast cells. Therefore, this chip-PEF-ESI
28
29 based method is expected to provide inspiration for the new generation of high-
30
31
32 throughput mass cytometry and play acceleration roles for large-scale single-cell
33
34 metabolomics analysis to aid in bioscience and clinical study.
35
36
37
38
39 Appendix A. Supplementary data
40
41 Supplementary data to this article can be found online.
42
43
44
45
46 ACKNOWLEDGMENTS
47
48
49 This work was funded by the foundation (No. 21934006) from the National Natural
50
51 Science Foundation of China, the innovation program (DICP&QIBEBT UN201806 and
52
53
54 DICP I202141) of science and research from the DICP, CAS. and the CAS “Light of
55
56 West China” Program.
57
58
59
60 20
61
62
63
64
65
 References
1
2 [1] E. Armingol, A. Officer, O. Harismendy, N.E. Lewis, Deciphering cell-cell
3
4 interactions and communication from gene expression, Nat Rev Genet, 22 (2021)
5 71-88.
6 [2] R. Zenobi, Single-cell metabolomics: analytical and biological perspectives,
7 Science, 342 (2013) 1243259.
8
9
[3] L. Zhang, A. Vertes, Single-Cell Mass Spectrometry Approaches to Explore
10 Cellular Heterogeneity, Angew. Chem. Int. Ed. Engl., 57 (2018) 4466-4477.
11 [4] T. Fujii, S. Matsuda, M.L. Tejedor, T. Esaki, I. Sakane, H. Mizuno, N. Tsuyama, T.
12 Masujima, Direct metabolomics for plant cells by live single-cell mass
13 spectrometry, Nat. Protoc., 10 (2015) 1445-1456.
14
15 [5] G. Huang, G. Li, R.G. Cooks, Induced Nanoelectrospray Ionization for Matrix-
16 Tolerant and High-Throughput Mass Spectrometry, Angewandte Chemie
17 International Edition, 50 (2011) 9907-9910.
18 [6] H. Zhu, G. Zou, N. Wang, M. Zhuang, W. Xiong, G. Huang, Single-neuron
19
identification of chemical constituents, physiological changes, and metabolism
20
21 using mass spectrometry, Proc. Natl. Acad. Sci. U. S. A., 114 (2017) 2586-2591.
22 [7] H. Zhu, N. Wang, L. Yao, Q. Chen, R. Zhang, J. Qian, Y. Hou, W. Guo, S. Fan, S.
23 Liu, Q. Zhao, F. Du, X. Zuo, Y. Guo, Y. Xu, J. Li, T. Xue, K. Zhong, X. Song, G.
24 Huang, W. Xiong, Moderate UV Exposure Enhances Learning and Memory by
25
26
Promoting a Novel Glutamate Biosynthetic Pathway in the Brain, Cell, 173
27 (2018) 1716-1727 e1717.
28 [8] Z. Wei, X. Xiong, C. Guo, X. Si, Y. Zhao, M. He, C. Yang, W. Xu, F. Tang, X.
29 Fang, S. Zhang, X. Zhang, Pulsed Direct Current Electrospray: Enabling
30 Systematic Analysis of Small Volume Sample by Boosting Sample Economy,
31
32 Anal. Chem., 87 (2015) 11242-11248.
33 [9] J. Feng, X. Zhang, L. Huang, H. Yao, C. Yang, X. Ma, S. Zhang, X. Zhang,
34 Quantitation of Glucose-phosphate in Single Cells by Microwell-Based Nanoliter
35 Droplet Microextraction and Mass Spectrometry, Anal. Chem., 91 (2019) 5613-
36
37
5620.
38 [10] F. Chen, L. Lin, J. Zhang, Z. He, K. Uchiyama, J.-M. Lin, Single-Cell Analysis
39 Using Drop-on-Demand Inkjet Printing and Probe Electrospray Ionization Mass
40 Spectrometry, Anal. Chem., 88 (2016) 4354-4360.
41 [11] Q. Huang, S. Mao, M. Khan, L. Zhou, J.M. Lin, Dean flow assisted cell ordering
42
43 system for lipid profiling in single-cells using mass spectrometry, Chem.
44 Commun. (Camb.), 54 (2018) 2595-2598.
45 [12] H. Yao, H. Zhao, X. Zhao, X. Pan, J. Feng, F. Xu, S. Zhang, X. Zhang, Label-free
46 Mass Cytometry for Unveiling Cellular Metabolic Heterogeneity, Anal. Chem.,
47
91 (2019) 9777-9783.
48
49 [13] H. Yao, H. Zhao, X. Pan, X. Zhao, J. Feng, C. Yang, S. Zhang, X. Zhang,
50 Discriminating Leukemia Cellular Heterogeneity and Screening Metabolite
51 Biomarker Candidates using Label-Free Mass Cytometry, Anal. Chem., 93 (2021)
52 10282-10291.
53
54
[14] D. Feng, T. Xu, H. Li, X. Shi, G. Xu, Single-cell Metabolomics Analysis by
55 Microfluidics and Mass Spectrometry: Recent New Advances, Journal of
56 Analysis and Testing, 4 (2020) 198-209.
57
58
59
60 21
61
62
63
64
65
 [15] S. Xu, M. Liu, Y. Bai, H. Liu, Multi-Dimensional Organic Mass Cytometry:
1 Simultaneous Analysis of Proteins and Metabolites on Single Cells, Angew.
2 Chem. Int. Ed. Engl., 60 (2021) 1806-1812.
3
4 [16] S.T. Nguyen, H.T.-L. Nguyen, K.D. Truong, Comparative cytotoxic effects of
5 methanol, ethanol and DMSO on human cancer cell lines, Biomedical Research
6 and Therapy, 7 (2020) 3855-3859.
7 [17] I. Lanekoff, V.V. Sharma, C. Marques, Single-cell metabolomics: where are we
8
9
and where are we going?, Curr. Opin. Biotechnol., 75 (2022) 102693.
10 [18] G. Li, S. Yuan, Y. Pan, Y. Liu, G. Huang, Binding States of Protein–Metal
11 Complexes in Cells, Anal. Chem., 88 (2016) 10860-10866.
12 [19] D.S. Wishart, Y.D. Feunang, A. Marcu, A.C. Guo, K. Liang, R. Vazquez-Fresno,
13 T. Sajed, D. Johnson, C. Li, N. Karu, Z. Sayeeda, E. Lo, N. Assempour, M.
14
15 Berjanskii, S. Singhal, D. Arndt, Y. Liang, H. Badran, J. Grant, A. Serra-Cayuela,
16 Y. Liu, R. Mandal, V. Neveu, A. Pon, C. Knox, M. Wilson, C. Manach, A.
17 Scalbert, HMDB 4.0: the human metabolome database for 2018, Nucleic Acids
18 Res., 46 (2018) D608-D617.
19
[20] Z. Pang, J. Chong, G. Zhou, D.A. de Lima Morais, L. Chang, M. Barrette, C.
20
21 Gauthier, P.E. Jacques, S. Li, J. Xia, MetaboAnalyst 5.0: narrowing the gap
22 between raw spectra and functional insights, Nucleic Acids Res., 49 (2021)
23 W388-W396.
24 [21] D. Di Carlo, D. Irimia, R.G. Tompkins, M. Toner, Continuous inertial focusing,
25
26
ordering, and separation of particles in microchannels, Proc. Natl. Acad. Sci. U.
27 S. A., 104 (2007) 18892-18897.
28 [22] J. Oakey, R.W. Applegate, Jr., E. Arellano, D. Di Carlo, S.W. Graves, M. Toner,
29 Particle focusing in staged inertial microfluidic devices for flow cytometry, Anal.
30 Chem., 82 (2010) 3862-3867.
31
32 [23] D. Di Carlo, J.F. Edd, D. Irimia, R.G. Tompkins, M. Toner, Equilibrium
33 separation and filtration of particles using differential inertial focusing, Anal.
34 Chem., 80 (2008) 2204-2211.
35 [24] B. Geboers, H.J. Scheffer, P.M. Graybill, A.H. Ruarus, S. Nieuwenhuizen, R.S.
36
37
Puijk, P.M.v.d. Tol, R.V. Davalos, B. Rubinsky, T.D.d. Gruijl, D. Miklavčič, M.R.
38 Meijerink, High-Voltage Electrical Pulses in Oncology: Irreversible
39 Electroporation, Electrochemotherapy, Gene Electrotransfer, Electrofusion, and
40 Electroimmunotherapy, Radiology, 295 (2020) 254-272.
41 [25] G. Li, S. Yuan, S. Zheng, Y. Liu, G. Huang, In Situ Living Cell Protein Analysis
42
43 by Single-Step Mass Spectrometry, Anal. Chem., 90 (2018) 3409-3415.
44 [26] K. Glunde, M.A. Jacobs, Z.M. Bhujwalla, Choline metabolism in cancer:
45 implications for diagnosis and therapy, Expert Rev. Mol. Diagn., 6 (2006) 821-
46 829.
47
[27] X. Chen, F.X. Wu, J. Chen, M. Li, DoRC: Discovery of rare cells from ultra-large
48
49 scRNA-seq data, 2019 IEEE International Conference on Bioinformatics and
50 Biomedicine (BIBM), 2019, pp. 111-116.
51 [28] F.M. Klis, P. Mol, K. Hellingwerf, S. Brul, Dynamics of cell wall structure in
52 Saccharomyces cerevisiae, FEMS Microbiol. Rev., 26 (2002) 239-256.
53
54
[29] P. Orlean, Architecture and biosynthesis of the Saccharomyces cerevisiae cell
55 wall, Genetics, 192 (2012) 775-818.
56 [30] Z. Li, Z. Wang, J. Pan, X. Ma, W. Zhang, Z. Ouyang, Single-Cell Mass
57 Spectrometry Analysis of Metabolites Facilitated by Cell Electro-Migration and
58 Electroporation, Anal. Chem., 92 (2020) 10138-10144.
59
60 22
61
62
63
64
65
 [31] V. Ganeva, B. Galutzov, J. Teissie, Evidence that pulsed electric field treatment
1 enhances the cell wall porosity of yeast cells, Appl. Biochem. Biotechnol., 172
2 (2014) 1540-1552.
3
4 [32] K.I. Kocurek, J. Havlikova, E. Buchan, A. Tanner, R.C. May, H.J. Cooper,
5 Electroporation and Mass Spectrometry: A New Paradigm for In Situ Analysis of
6 Intact Proteins Direct from Living Yeast Colonies, Anal. Chem., 92 (2020) 2605-
7 2611.
8
9
[33] M. Ohashi, R. Nasuno, S. Isogai, H. Takagi, High-level production of ornithine
10 by expression of the feedback inhibition-insensitive N-acetyl glutamate kinase in
11 the sake yeast Saccharomyces cerevisiae, Metabolic Engineering, 62 (2020) 1-9.
12 [34] V. Olin-Sandoval, J.S.L. Yu, L. Miller-Fleming, M.T. Alam, S. Kamrad, C.
13 Correia-Melo, R. Haas, J. Segal, D.A. Peña Navarro, L. Herrera-Dominguez,
14
15 Lysine harvesting is an antioxidant strategy and triggers underground polyamine
16 metabolism, Nature, 572 (2019) 249-253.
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 23
61
62
63
64
65
 Figure Legends
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 Figure 1. a) Schematic diagram of the microfluidics chip assisted high-throughput
36
37
38 single cell mass spectrometry analysis device. b) Photo of the asymmetric serpentine
39
40 channel microfluidic chip and ESI source coupled to mass spectrometer.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 24
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29 Figure 2. a) Schematic illustration of geometry of the serpentine channel. The height
30
31 of the microchannel is approximately 50 μm. b) Secondary flow distribution and vectors
32
33
34
in cross-section of the channel at 8th large turn and 9th small turn. c) The streamlines
35
36 of cells with the same diameter (ten 12 µm particles released at random positions at the
37
38 entrance and release time normally distributed in 0.5 s) after 4 s.
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 25
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33 Figure 3. Different cell suspensions composed of MCF7 and HepG2 cancer cell lines
34
35
analyzed by PEF-ESI-HRMS. a) TIC and b) EIC of phosphorylcholine (m/z 184.0733)
36
37
38 acquired in the positive ion mode. Two distinct single-cell metabolic profiles of c) MCF
39
40 and d) HepG2 were obtained from the MS data.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 26
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34 Figure 4. Effect of the flow rate and cell concentration of the HepG2 and MCF7 cell
35
36 suspension on single-cell MS analysis. a) Relationship between the flow rate of the cell
37
38
39 suspension and the number of single cell MS signal per minute. b) Number of single
40
41 cell MS signal has a linear relationship with the cell concentration. c) EICs of
42
43 phosphorylcholine (m/z 184.0733) obtained by analyzing MCF7 cell suspensions with
44
45
46 concentrations of 8 × 104 cells mL-1 at flow rates of 1 µL min-1and the magnified
47
48 spectrum is at 1.5 to 1.75 min. d) Classifications of detected metabolites in positive ion
49
50
51 mode.
52
53
54
55
56
57
58
59
60 27
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48 Figure 5. a) PCA plot based on the first two principal components of the single HepG2
49
50 and MCF7 cells. b) Modified raincloud plots of the relative contents of leucine and
51
52
53
creatine in MCF7 and HepG2 single cells. Upper part is split-half violin plot and lower
54
55 part is box plot. c) Outliers in HepG2 cells (marked by red x in 3D PCA plot) identified
56
57 by Isolation Forest. d) Modified raincloud plots of the relative contents of gamma-
58
59
60 28
61
62
63
64
65
 aminobutyric acid and aminoacetone in outliers and non-outliers of HepG2 cells. e)
1
2 Outliers in MCF7 cells (marked by red x in 3D PCA plot) identified by Isolation Forest.
3
4
5 f) Modified raincloud plots of the relative contents of valine and phosphate in outliers
6
7 and non-outliers of MCF7 cells.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 29
61
62
63
64
65
 1
2
3
4
5
6
7
8
9
10
11
12
13
14 Figure 6. a) Schematic illustration of the setup for the analysis of single yeast cell. b)
15
16
17 TIC and EICs of ornithine (m/z 133.0972), creatine (m/z 132.0768) and lysine (m/z
18
19 147.1128) obtained from single yeast cells by using a square-wave voltage with an
20
21
22
amplitude of 5 kV Vpp and frequency of 20 Hz.
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 30
61
62
63
64
65