Single-cell Proteomics in Complex Tissues using Microprobe Capillary Electrophoresis Mass Spectrometry

Abstract

Direct measurement of proteins produced by single cells promises to expand our understanding of molecular cell-to-cell differences (heterogeneity) and their contribution to normal and impaired development. High-resolution mass spectrometry (HRMS) is the modern technology of choice for the label-free identification and quantification of proteins, albeit usually in large populations of cells. Recent advances in microscale sample collection and processing, separation, and ionization have extended this powerful technology to single cells. This chapter describes a protocol based on microprobe capillary electrophoresis (CE) HRMS to enable the direct proteomic profiling of single cells embedded in complex tissues without the requirement for dissociation or whole-cell dissection. We here demonstrate the technology for identified individual cells in early developing embryos of Xenopus laevis and zebrafish as well as electrophysiologically identified single neurons in physiologically active brain slices from the mouse substantia nigra. Instructions are provided step-by-step to identify single cells using physiological or morphological cues, collect the content of the cells using microfabricated capillaries, and perform bottom-up proteomics using a custom-built CE electrospray ionization mass spectrometer equipped with a quadrupole time-of-flight or orbitrap mass analyzer. Results obtained by this approach have revealed previously unknown differences between the proteomic state of embryonic cells and neurons. The data from single-cell proteomics by microprobe CE-ESI-HRMS complements those from single-cell transcriptomics, thereby opening exciting potentials to deepen our knowledge of molecular mechanisms governing cell and developmental processes.

1. Introduction

The goal of single-cell proteomics is to obtain an unbiased account of the molecular state of individual cells, equipping systems biology with a unique analytical tool to investigate cell-to-cell differences (cell heterogeneity) during normal and impaired development. Using whole-tissue dissociation and high-throughput microfluidics (Briggs et al., 2018; Wagner & Klein, 2017; Zilionis et al., 2017), molecular amplification with sequencing is now quasi-routinely used to study genomic and transcriptomic differences between cells in tumors (Tan et al., 2019), developing embryos (Briggs et al., 2018), and brain tissues (Rohrback, Siddoway, Liu, & Chun, 2018), among other applications. However, in dynamic biological systems, such as differentiating cells, developing tissues (Peshkin et al., 2015), and the developing nervous system (Raj et al., 2018), transcription has been found to correlate relatively poorly with gene translation. Therefore, utilization of data from single-cell transcriptomics as a surrogate for protein production requires careful validation of the results, ideally using orthogonal technologies, usually antibody staining one gene at a time. The next frontier in systems cell biology is dependent on the development and availability of single-cell proteomic technologies, especially those that operate without chemical probes (e.g., no antibodies), are sensitive, quantitative, and scalable in cell size and throughput.

Single-cell high-resolution mass spectrometry (HRMS) is a blooming field of technologies capable of unbiased characterization of biomolecules in single cells. The current state of single-cell HRMS has been the focus of several recent reviews, including but not restricted to the following publications: (Acunha, Simo, Ibanez, Gallardo, & Cifuentes, 2016; Chen et al., 2016; Comi, Do, Rubakhin, & Sweedler, 2017; Couvillion et al., 2019; Herr, Kitamori, Landegren, & Kamali-Moghaddam, 2019; Passarelli & Ewing, 2013; Stanislav S. Rubakhin, Lanni, & Sweedler, 2013; S. S. Rubakhin, Romanova, Nemes, & Sweedler, 2011; Yang et al., 2017; Yin, Zhang, Liu, Gao, & Gu, 2018; Zenobi, 2013; L. W. Zhang & Vertes, 2018). Capillary electrophoresis (CE) offers particular benefits for single-cell HRMS, because it is compatible with volume/amount-limited extracts available from single cells, offers exquisite separation efficiency to reduce chemical complexity and aid molecular identification, and can be integrated to ESI-HRMS. The detection of alpha and beta globulins in single erythrocytes in the mid-1990s (Hofstadler, Swanek, Gale, Ewing, & Smith, 1995; Valaskovic, Kelleher, & McLafferty, 1996) pioneered this technology toward single-cell analyses. The development of specialized CE electrospray ionization (ESI) interfaces enabling trace sensitivity [see reviews in references (Krenkova & Foret, 2012; W. Zhang, Hankemeier, & Ramautar, 2017; Zhong, Zhang, Jiang, & Li, 2014)] has been critical to the success of single-cell CE-HRMS that followed a decade later.

We and others have developed microanalytical instruments based on CE-HRMS to explore cell heterogeneity in complex tissues. Using whole-cell dissection, metabolic differences have been found between single neurons in the central nervous system of Aplysia californica (Lapainis, Rubakhin, & Sweedler, 2009; Nemes, Knolhoff, Rubakhin, & Sweedler, 2011; Nemes, Rubakhin, Aerts, & Sweedler, 2013). We have combined cell dissection with new-generation CE-ESI and HRMS to uncover metabolic and proteomic differences between embryonic cells (blastomeres) in the 4-, 8-, and 16-cell embryo that reproducibly give rise to specific types of tissues in Xenopus laevis, the South African clawed frog (Lombard-Banek, Moody, & Nemes, 2016b; Lombard-Banek, Reddy, Moody, & Nemes, 2016; Onjiko, Moody, & Nemes, 2015; Onjiko, Morris, Moody, & Nemes, 2016). These studies revealed previously unavailable information on proteomic and metabolic differences between cells that occupy the dorsalventral, animal-vegetal, and left-right axes in the early-stage embryo and even led to the discovery of small molecules capable of altering tissue specification (Onjiko et al., 2015). Many of the observed proteomic differences were not detectable at the level of mRNA, underscoring the importance of complementing data from single-cell transcriptomics with those from single-cell proteomics for a deeper characterization of molecular differences between cells.

Most recently, we have enabled, for the first time, the direct metabolomic and proteomic characterization of single cells directly bound in complex tissues. The technology integrates capillary microprobe sampling of single cells with micro- (Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016) and nano-flow (Choi, Zamarbide, Manzini, & Nemes, 2017; Lombard-Banek, Moody, Manzini, & Nemes, 2019) CE-ESI interfaces coupled to (ultra-)high-resolution mass spectrometers equipped with time-of-flight (Choi et al., 2017; Lombard-Banek, Reddy, et al., 2016) or orbitrap mass analyzers (Choi, Polter, & Nemes, 2019; Lombard-Banek, Moody, et al., 2019; Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016). Compared to cell culturing, cell dissection, or whole tissue dissociation, in situ analysis by microprobe CE-ESI-HRMS realizes several advantages for experimentation and data interpretation. Direct analysis is able to accurately capture information on cell identity and location in tissues (e.g., by using optical imaging). The approach also leaves neighboring cells and tissues intact, facilitating broader studies in biology. Additionally, microprobe sampling imposes significantly less oxidative stress on cells (Onjiko, Portero, Moody, & Nemes, 2017), thus allowing investigators to more closely probe the innate molecular state of the cell than feasible via classical cell isolation. We have utilized this technology to characterize, for the first time, the metabolic (Onjiko, Portero, et al., 2017) and proteomic (Lombard-Banek, Moody, et al., 2019) state of cells directly embedded in live 1-to-128-cell embryos of X. laevis and 2-cell embryos of zebrafish (Lombard-Banek, Moody, et al., 2019). Microprobe CE-ESI-HRMS is also adaptable to single neurons in identified mouse brain regions (Choi et al., 2019). Combined, these examples have demonstrated CE-ESI-HRMS as an emerging tool for single-cell metabolomics and proteomics.

This chapter presents a step-by-step protocol to build and utilize microprobe CE-ESI-HRMS to quantify the proteomic state of single cells in complex tissues, using X. laevis and zebrafish embryos and the mouse brain as biological models. Trouble-shooting advice is provided for the most common errors (see Table 1). This proteomics protocol complements recent work from our lab, in which microprobe CE-ESI-HRMS has been described for single-cell metabolomics (Onjiko, Portero, & Nemes, 2018). Our proteomic workflow commences with the identification of the cell of interest, usually based on morphological, physiological, cell-type specific molecular information (e.g., via the use of molecular probes). Figure 1 shows identification of the midline dorsal-animal cell (D11) in the 16-cell embryo of X. laevis following stereotypical pigmentation and reproducible cell fate maps for the frog (Moody, 1987a, 1987b), randomly chosen cells in the 2-cell zebrafish embryo (where cell identification is not possible due to a lack of visual cues), and single neurons in the mouse substantia nigra. The membrane of the identified cell is punctured with a pulled micropipette to aspirate a pre-calibrated volume from the cell using an inline-connected microinjector. The collected sample is released into a microcentrifuge tube, where metabolites (Onjiko, Portero, et al., 2017) and proteins are extracted and processed (Lombard-Banek, Moody, et al., 2019). To identify proteins via a bottom-up strategy, the proteins are digested to peptides, and resulting peptides are separated by CE, ionized by ESI (Fig. 2), before sequencing the generated peptide ions using tandem MS (Fig. 3). The MS–MS/MS data are processed using established bioinformatic platforms (e.g., Proteome Discoverer, ProteinScape, or MaxQuant) to identify proteotypic peptides against the proteome of the species and to perform absolute/relative quantification using the signal intensity recorded for each proteotypic peptide. As presented here (see “Representative Results” below), proteomic microprobe CE-ESI-HRMS is scalable in space and time to cells of different sizes (Fig. 4) and operates label-free with a capability for identifying and quantifying large numbers of proteins in single cells (e.g., hundreds-to-thousands of different protein groups). These analytical characteristics make microprobe single-cell CE-ESI-HRMS an attractive technology to propel the systems biology interrogation of proteomic processes forward in cells embedded in complex tissues and even in live organisms, such as live developing embryos.

Table 1.

Troubleshooting advice for representative errors during microprobe CE-ESI-HRMS.

Observation	Suspected Causes	Potential Solutions
No/few peptides detected	Microprobe sampling failed (e.g., clogged sampling needle) Enzymatic digestion failed (e.g., enzyme expired)	Repeat sampling workflow, then measure the new sample
Unstable electrospray signal (e.g., large fluctuations in TIC or BPC)	The spraying regime transitioned from cone-jet (preferred) to a pulsating/nonaxial mode or destabilized by air currents. Unstable liquid supply to the CE-ESI interface (connection errors, check syringe). Bubbles forming inside the electrospray capillary destabilize the Taylor-cone	Optimize experimental parameters for the CE-ESI system, including separation conditions (CE voltage/current and BGE composition) and the ion source (sheath flow rate, electrospray potential, emitter-to-MS orifice distance) Check connections, then flush all capillaries (electrospray sheath and also CE separation) Check for air currents and optionally enclose the CE-ESI source in an environmental chamber [see examples in (Portero & Nemes, 2019)]
Low CE current	Accidental injection of a bubble into the CE capillary BGE electrolysis occurred during the separation	Flush the capillary with BGE for ~10–15 min, then repeat the experiment; optimize separation conditions to avoid electrolysis.
Variation in migration time > 10%	The composition of the BGE has changed (e.g., due to electrolysis) CE capillary walls have become contaminated (e.g., due to adsorbed compounds) CE capillary is broken	Repeat experiment with freshly made BGE Replace the CE capillary in the CE-ESI setup
Low number of proteins identified	Erroneous sample injection Mass spectrometer calibration off	Repeat the injection and the CE-ESI-HRMS measurement Calibrate the mass spectrometer, then repeat the experiment

Figure 1.

Analytical workflow for single-cell proteomics in (A) representative biological models (B) using microprobe CE-ESI-HRMS. The cell of interest (1) is identified under a stereomicroscope (2), its content aspirated into a pulled micropipette (3) directed by a three-axis micromanipulator (4) controlled by a microinjector (5). The sample (6) is processed for analysis in a custom-built CE-ESI platform (7) hyphenated to a HRMS instrument. The primary data are analyzed to identify/quantify proteins. Scale bars: 150 μm (X. laevis), 200 μm (zebrafish), 50 μm (mouse). [Figures adapted with permission from references (Choi et al., 2019; Lombard-Banek, Moody, et al., 2019; Onjiko, Portero, et al., 2017).]

Figure 2.

Schematics of co-axial sheath-flow CE-ESI interfaces enabling single-cell proteomics in the micro- and nano-flow regime. Key: Scale bars = 200 μm (black) and 20 μm (white). [Reproduced with permission from references (Choi et al., 2017; Lombard-Banek, Moody, et al., 2019; Nemes et al., 2013).]

Figure 3.

Identification of peptide ions using CE-ESI-HRMS executing tandem MS. (A) Peptides are electrophoretically separated (left panel), ionized, and their accurate m/z values are recorded (right panel). (B) Peptide ions are fragmented by tandem MS (right panel, inset) and identified against the proteome of the species under study. The example shows identification of m/z 572.33 at ~50 min separation as LGLGLELEA, the proteotypic product of the voltage-dependent anion-selective channel protein 2 (Vdac2) upon tryptic digestion. [Figure adapted with permission from reference (Lombard-Banek, Moody, et al., 2016a).]

Figure 4.

Investigation of proteomic cell heterogeneity in identified cells in live developing X. laevis embryos (panels A and B) and physiologically active dopaminergic neurons in the mouse substantia nigra (panel C). (A) Label-free quantification with fuzzy c-mean cluster analysis uncovered proteins with different spatiotemporal expression profiles as the midline animal-dorsal cell divided to form a neural tissue fated cell clone. Each cluster represents a set of proteins following a similar temporal trend. Each line represents a unique protein belonging to the cluster. The scale bar indicates the confidence as probability that the protein belongs to the cluster. Scale bar = 200 μm. [Figure adapted with permission from reference (Lombard-Banek, Moody, et al., 2019).] (B) TMT-based relative quantification of proteomic differences between cells of different tissue fates in the 16-cell X. laevis embryo. [Figure adapted with permission from reference (Lombard-Banek, Moody, et al., 2016b).] (C) Quantification of 95 different proteins in single dopaminergic neurons spanning a ~4-decade concentration range. Proteins known to be produced in dopaminergic neurons are highlighted.

2. Materials and Equipment

a). Frog Embryo Culture

• Adult male and female X. laevis (frogs): Protocols related to the care and handling of animals must comply with institutional and federal guidelines. Results presented in this chapter were obtained from studies approved by the Institutional Animal Care and Use Committee of the University of Maryland (IACUC #R-DEC-17–57) or the George Washington University (IACUC #A311).

• Embryos may be obtained by hormone-induced (gonadotropin) natural mating of sexually mature adult frogs or in vitro fertilization following standard protocols (Sive, Grainger, & Harland, 2000).

• 60 and 90 mm Petri dishes

• Incubator set to 14 °C (e.g., Heratherm, ThermoFisher Scientific, Waltham, MA)

• Embryo dejellying solution: 2% cysteine hydrochloride at pH 8 prepared by dissolving 4 g of crystalline cysteine hydrochloride in 200 mL of deionized (DI) water, adjusted to pH 8 dropwise with 10 M NaOH.

• 100% Steinberg’s solution (SS): Prepared by mixing the following salts in 5 L of DI water: 17 g NaCl, 0.25 g KCl, 0.4 g of Ca(NO₃)₂×4H₂O, 1.02 g MgSO₄×7H₂O, 3.3 g of Trizma hydrochloride (acid), and 0.375 g of Trizma base. Adjust the pH to 7.4 by adding more Trizma base if necessary. Autoclave the solution and store it at 4–18 °C.

• 10% Steinberg’s solution: Dilute 100 mL of 100% SS with 900 mL of DI water.

b). Zebrafish Embryo Culture

• Adult male and female zebrafish: Protocols related to the care and handling of animals must comply with institutional and federal guidelines. Results presented in this chapter were obtained from studies approved by the Institutional Animal Care and Use Committee of the George Washington University (IACUC #A311).

• Embryos are obtained by natural mating of sexually mature adult zebrafish following an established protocol (Dahm & Nüsslein-Volhard, 2002).

• 90 mm Petri dish

• E2 medium: Prepared by mixing the following salts in 1 L of DI water: 0.8766 g NaCl, 0.0373 g KCl, 0.1204 g MgSO₄, 0.0204 g KH₂PO₄, 0.0071g Na₂HPO₄, 0.1110 g CaCl₂, and 0.0588 g NaHCO₃.

c). Animal and Brain Section Preparation

• Adult mice: Protocols related to the care and handling of animals must comply with institutional and federal guidelines. Procedures related to the care and handling of animals presented in this chapter were approved by the Institutional Animal Care and Use Committee of the George Washington University (IACUC #A378). C57Bl6/J mice were bred in-house and were maintained on a 12 h light/dark cycle and provided with food and water ad libitum.

• Ice-cold HEPES solution for perfusion: Prepared by mixing the following salts in 1 L of DI water: 0.19 g KCl, 0.14 g NaH₂PO₄, 2.9 g NaHCO₃, 4.8 g HEPES, 4.5 g glucose, 1.0 g sodium ascorbate, 0.15 g thiourea, 0.33 g sodium pyruvate, 0.12g MgSO₄, and 0.22 g CaCl₂.

• A vibratome to section brain slices (e.g., automated vibratome model VT 1200S, Leica, Wetzlar, Germany)

d). Protein Collection and Processing

• Capillary puller (e.g., P-1000, Sutter Instrument, Novato, CA)

• A microinjector equipped with control pedals (e.g., PSI-100A, Warner Instruments, Hamden, CT)

• A three-axis translation stage for fine positioning of the microprobe with manual (e.g. MM 33, Warner Instruments) or motorized (e.g., model TransferMan 4r, Eppendorf, Hauppauge, NY) operation.

• 0.2–0.5 mL Protein LoBind microcentrifuge tubes

• 50 mM ammonium bicarbonate: Dissolve 0.1976 g of crystalline ammonium bicarbonate in 50 mL of DI water.

• Trypsin (0.5 μg/μL) in 50 mM ammonium bicarbonate

• Two heat blocks set to 60 °C and 37 °C, respectively.

• A vacuum concentrator (e.g., model CentriVap, LabConco, Kansas City, MO)

• Tandem-mass-tags (TMTs) or isotope tags for relative and absolute quantification (iTRAQ) for sample barcoding (optional)

• 1 M triethylammonium bicarbonate (TEAB) (e.g., catalogue #90114, ThermoFisher Scientific)

• 100 mM TEAB: Prepare by adding 100 μL of 1 M TEAB to 900 μL of DI water.

• Anhydrous acetonitrile

• 50% hydroxylamine (e.g., catalogue #90115, ThermoFisher Scientific)

• 5% hydroxylamine: Prepare by mixing 1 μL of 50% Hydroxylamine with 9 μL of 100 mM TEAB.

Embryonic Cells

• A stereomicroscope (e.g., model SMZ18, Nikon, Melville, NY).

• Sampling dishes for frog embryos: Prepare 2% w/v agarose by adding 2 g of agarose in 100 mL of 100% SS and dissolve by autoclaving at 120 °C. When cool enough to handle but still in a liquid state, coat the bottom of a 60 mm in diameter Petri dish to an ~1 mm in thickness. After the agarose has completely solidified, imprint wells using a 6-inch Pasteur pipet with a tip melted into a ball.

• Sampling dishes for zebrafish embryos: Prepare 5% w/v agarose by adding 10 g of agarose to 200 mL of E2 medium while heating the mixture. Pour the solution into a Petri dish to a ~10 mm thickness and let cool slightly, before imprinting the agarose with grooves (grove size: 2.5 mm, groove depth: 0.5 mm) using a mold (e.g., Z-MOLDS, World Precision Instruments, Sarasota, FL).

• Disposable transfer pipet

• Hair loop: Form a 2–3 mm loop by placing an ~10 cm long strand of fine hair into a 6-inch Pasteur pipet and secure it with melted paraffin. Sterilize the hair loop before use by dipping it in 70% ethanol.

• Sharp forceps (e.g., model Dumont #5, Fine Science Tools, Foster City, CA)

• Microprobe: Borosilicate capillary 0.5/1 mm inner/outer diameter pulled to a taper and cut with fine sharp forceps to a ~20 μm opening.

Mouse Brain Slices

• Anesthetic solution (ketamine, 20 mg/mL; dexmedetomidine 0.1 mg/mL): Prepare by mixing 200 μL of ketamine stock solution (100 mg/mL) and 200 μL of dexmedetomidine stock solution (0.5 mg/mL) with 800 μL of 0.9% saline solution for injection (e.g., part # NDC0409-4888-06; Hopira, Inc, Lake Forest, IL)

• Perfusion solution for mouse brain slices: Artificial cerebrospinal fluid (aCSF) prepared by mixing the following salts into 1 L of DI water: 1.8 g NaHCO₃, 0.19 g KCl, 0.14 g NaH₂PO₄, 0.27 g CaCl₂, 0.12 g MgSO₄, 2 g glucose, and 7.4 g NaCl.

• An upright microscope (e.g., model Eclipse FN1, Nikon, Melville, NY) to visually inspect the morphology of the brain slices.

• An automated temperature controller set to 31–33 °C (e.g., model TC-324C, Harvard Apparatus, Cambridge, MA) to maintain sustainable temperature for the brain slices.

• Microprobe/Patch-clamp pipette: Borosilicate capillary (1.1/1.5 mm ID/OD with filament, Item# BF-150-11-10, Sutter Instrument) pulled to a taper with a 1–3 μm opening and a 2–4 MΩ resistance.

• A patch-clamp pipette filled with 50 mM ammonium bicarbonate (used as intracellular solution to support electrophysiology and optimal proteomics processing by MS)

• A micro-manipulator (e.g., MP-285, Sutter Instrument, Novato, CA) to precision-position the patch pipette to the membrane of the neuron.

• (Optional) An electrophysiology setup for electrophysiological recording of neuronal activity [e.g., consisting of an integrated patch amplifier (e.g., IPA, Sutter Instrument) and a constant current stimulus isolator (e.g., A365RC, World Precision Instruments, Sarasota, FL)]

e). Proteomic Analysis using CE-ESI-HRMS

• Separation capillary: 110/40 μm OD/ID bare fused silica capillary (e.g., part no. 1068150596, Polymicro Technologies, Phoenix, AZ)

• nanoESI emitter: Borosilicate capillaries (e.g., part no. B100-75-10, 0.75/1 mm ID/OD, Sutter Instruments, Novato, CA) pulled to a taper with a ~20 μm opening or stainless-steel emitter that has a tapered tip (e.g., part number# MT320-100-3.5-5, 100/320 μm ID/OD, New Objective, Woburn, MA).

• High-voltage power supply for CE: Stable regulated high-voltage power supply outputting up to 30 kV (e.g., model Bertan 230–30R, Spellman High Voltage Electronics Corp., Hauppauge, NY).

• High-voltage power supply for electrokinetic pump/electrospray: Regulated high voltage power supply outputting up to 5 kV (e.g., model P450, Stanford Research Systems, Sunnyvale, CA).

• A digital voltmeter connected in parallel to a 10 kΩ resistor installed in the CE electrical circuit to monitor the CE current (see details in reference (Nemes et al., 2013).

• Angiotensin II stock solution (1 g/L): Prepare by dissolving crystalline angiotensin standard (e.g., part #AAJ66457LB0, FisherScientific) in water.

• Angiotensin II standard at 1×10⁻⁴ g/L prepared by diluting a 1 g/L stock solution by 10,000 folds in the sample solvent.

• Sample solvent (60% acetonitrile in water with 0.05% acetic acid): Mix 600 μL of acetonitrile with 400 μL of water and 0.5 μL of acetic acid.

• Sheath solution (10% acetonitrile in water with 0.05% acetic acid): Mix 5 mL acetonitrile with 25 μL of acetic acid in 44.975 mL of LC-MS grade water.

• Background electrolyte (BGE, 25% acetonitrile in water with 1 M formic acid): Mix 12.5 mL of organic solvent (methanol or acetonitrile), 1.887 mL of formic acid, and 35.613 mL of water for a total of 50 mL.

• A high-resolution mass spectrometer (e.g., model Orbitrap Q-Exactive Plus or Fusion Lumos, ThermoFisher Scientific or Impact HD, Bruker Daltonics, Billerica, MA)

3. Procedures

a). Embryo Culture

The goal of this step is to culture embryos to a desired developmental stage. Early-stage Xenopus laevis embryos are dejellied and raised in 100% SS. Zebrafish embryos are collected using a strainer and raised in E2 medium. Developmental stages are monitored under a stereomicroscope.

• Prepare consumables:

  • 2% cysteine solution (for X. laevis embryos)
  • 100% SS (for X. laevis embryos)
  • 10% SS (for X. laevis embryos)
  • E2 medium (for zebrafish embryos)
• Remove the jelly coat surrounding X. laevis embryos following protocols established elsewhere (Sive et al., 2000):

  • Gently decant all media from the dishes containing the freshly fertilized embryos.
  • Add enough cysteine solution to cover the embryos. Let the solution sit undisturbed for 2 min and gently swirl the dishes for another 2 min.
  • Transfer the embryos immediately into a large beaker and let them settle at the bottom. Pour off the liquid and rinse the embryos with 10% SS four times.
  • Remove most of the solution and transfer ~300–500 embryos into 90 mm Petri dishes filled with 100% SS. Place the dishes in a 14 °C incubator until experimentation.

b). Preparation of Brain Slices

The goal of this step is to prepare brain tissues in vitro for single-cell proteomics following optional electrophysiology recording of neuronal activity.

• Instrumentation:

  • An upright microscope
  • A vibratome
  • A micro-manipulator
  • (Optional) An electrophysiology setup
• Prepare consumables:

  • Anesthetic solution
  • HEPES ringer solution
  • Artificial cerebrospinal fluid (aCSF)
  • Intracellular solution
• Prepare brain tissue sections following an established protocol (Polter et al., 2014):

  • Anesthetize mouse of 21–35 postnatal days with an intraperitoneal injection of the anesthetic solution while perfusing ice-cold HEPES ringer solution.
  • Rapidly dissect the brain and prepare 220 μm thick horizontal slices of the brain substantia nigra using a vibratome.
  • Let the substantia nigra slices recover for ~1 h at 34 °C in HEPES ringer solution, and then place them at room temperature until use.

c). Microprobe Sampling of Single Cells in Tissues

The goal of microsampling is to collect material from the cell of interest. Figure 1 presents vignettes for sampling ~1–10 nL from individual cells in embryos as well as neurons in important biological models. In X. laevis embryos, cell types are identifiable based on morphology and stereotypical cell fate maps (Moody, 1987a, 1987b). Cells in zebrafish embryos, however, lack visual cues for identification; these cells have been randomly selected for analysis in our studies. In the mouse substantia nigra, neuron types may be distinguished based on electrical activity. To collect a desired portion of the cell, the cell membrane is pierced with a micropipette (pulled capillary) and its content is aspirated by applying negative pressure on the pipette. The collected content is released in a microcentrifuge tube for processing prior to instrumental analysis.

• Instrumentation:

  • A capillary puller
  • An upright or inverted microscope
  • A three-axis micromanipulator
  • A microinjector
• Prepare Consumables:

  • Transfer pipet
  • Sampling dish
  • Hair loop
  • Forceps
  • Large beaker (400–600 mL)
  • A bucket filled with ice
  • Borosilicate capillaries
  • 0.2–0.5 mL protein LoBind centrifuge tubes
  • Prepare the micropipettes: Pull borosilicate capillaries to a fine taper using a capillary puller. For example, to pull micropipettes for X. laevis embryos sampling we used the following parameters on a Sutter Instrument P-1000 capillary puller: heat = 498, pull = 40; velocity = 60, time = 150. Under a stereomicroscope, cut the tip of the pulled capillary using forceps so that the opening is ~20 μm.
  • 50 mM ammonium bicarbonate

  Note: Store the cut capillaries secured in a closed box until usage to prevent contamination.

• Microsample single cells in live tissues:

  • Select the cell of interest under a microscope if possible. Figure 1 shows identification of the midline animal-dorsal cell (D11) in the 16-cell X. laevis embryo, random sampling of a cell in zebrafish, and electrophysiology-driven identification of a neuron in the mouse substantia nigra.
  • Insert the microprobe into the cell and apply negative pressure to withdraw a desired portion of the cell content. Although any apparatus capable of generating negative pressure can be used (e.g., a syringe), we have found a microinjector particularly useful in helping to reproducibly aspirate calibrated volumes from the cells. For example, we routinely use this approach to collect ~1–20 nL from the D11 cell (~90 nL cell volume).
  • Transfer the aspirated cell content into a protein LoBind microcentrifuge tube by applying a pulse of pressure to eject the aspirate into 5 μL of 50 mM ammonium bicarbonate.
  • While sampling more cells, store the collected samples on ice until further processing.

d). Protein Processing

In this step, the collected protein content is processed using one of three common approaches for proteomic detection. Recent reviews have discussed workflows for bottom-up proteomics (Aebersold & Mann, 2003; Y. Y. Zhang, Fonslow, Shan, Baek, & Yates, 2013) and top-down proteomics (Catherman, Skinner, & Kelleher, 2014; Fornelli et al., 2018; Moradian, Kalli, Sweredoski, & Hess, 2014). Alternatively, a middle-down strategy may be followed (Moradian et al., 2014; Pandeswari & Sabareesh, 2019; Wu et al., 2012). In the bottom-up workflow, we typically digest proteins to peptides, which are easier to analyze by MS. For single cells, our laboratory usually eliminates reduction and alkylation steps to reduce protein losses during sample handling (Fig. 1).

• Prepare consumables:

  • Trypsin solution at 0.5 μg/μL in 50 mM ammonium bicarbonate
  • Block heater at 60 °C
  • Block heater at 37 °C
• Protein Digestion:

  • Denature the proteins by incubating the samples at 60 °C for ~15 min.
  • Equilibrate the samples back to room temperature (~5 min).
  • Add 5–50 ng of trypsin to the protein extracts and incubate at 37 °C for ~5 h.
  • After digestion, spin the samples briefly in a table-top microcentrifuge to collect the resulting peptide solution at the bottom of the vial.
  • Vacuum-dry the samples in a vacuum concentrator.

  Note: We advise against overnight digestion, which is typical in bottom-up proteomics, because the finite amounts of sample may dry with prolonged exposure to heat, thus disrupting digestion. Additional trouble-shooting advice is summarized in Table 1.

e). (Optional) Sample Barcoding for Multiplexed Quantification

To enable multiplexing quantification, peptides resulting from trypsin digestion may be labeled with isotopically encoded reagents such as tandem mass tags (TMTs) (Thompson et al., 2003) or isobaric tags for relative and absolute quantification (iTRAQ) (Ross et al., 2004), among other approaches. The benefit of multiplexing quantification is increased throughput, enabling the measurement of up to 11 samples (conditions) at once (McAlister et al., 2012) at present. We have successfully used 10-plexing TMT to relatively quantify proteins between single embryonic cells (Lombard-Banek, Moody, et al., 2016b). TMT labeling can also be used for mammalian cells (Budnik, Levy, Harmange, & Slavov, 2018) with this strategy adaptable to microprobe CE-ESI-MS to quantify translation in smaller cells and other biological specimens.

• Prepare consumables:

  • 100 mM TEAB
  • Anhydrous acetonitrile
  • TMT labeling reagent (or alternative, such as iTRAQ, AB SciX)
  • 5% hydroxylamine
• Labeling reaction for multiplexing:

  • Reconstitute the TMT reagents (ThermoFisher) in anhydrous acetonitrile following vendor instructions.
  • For whole cells, reconstitute the peptides in 10 μL of 100 mM TEAB.
  • Add 10 μL of the reconstituted TMT labeling reagent and incubate at room temperature for 1 h.
  • Quench the reaction with 2 μL of 5% hydroxylamine and incubate at room temperature for 15 min.
  • Mix the samples in equal volumes and dry to completeness using a vacuum concentrator.

f). Analysis by CE-ESI-HRMS

The goal of this step is to identify proteins extracted from single cells via bottom-up proteomics. Our laboratory has developed various co-axial sheath-flow CE-ESI interfaces to electrophoretically separate and ionize peptides for detection by MS. Representative designs are presented in Figure 2. In the blunt-tip interface, peptides migrate into a Taylor-cone maintained in the micro-flow regime on the tip of the metal emitter (Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016; Nemes et al., 2013). The tapered-tip design enhances detection sensitivity (260-zmol demonstrated) by decreasing the rate of the sheath-flow to the nano-flow regime, stably anchoring the Taylor-cone, and efficiently entraining generated droplets/ions into the mass spectrometer for model peptides [see (Choi et al., 2017)]. The electrokinetically pumped low-flow interface [built following reference (Sun et al., 2013)] encloses the CE capillary outlet into a pulled borosilicate interface to efficiently ionize peptide ions using a nanoelectrospray emitter with a ~700 zmol sensitivity (Lombard-Banek, Moody, et al., 2019). The eluting peptide ions are analyzed in a high-resolution tandem mass spectrometer executing collision-induced, higher-energy collisional, or electron-transfer dissociation for fragmentation (Fig. 3). Our single-cell proteomic approach has been demonstrated on mass spectrometers equipped with quadrupole time-of-flight, quadrupole orbitrap, and quadrupole-ion trap-orbitrap tribrid mass analyzers. We anticipate our CE-ESI design to be compatible with virtually any mass spectrometer that employs an atmospheric pressure interface, such as a micro- or nano-electrospray source. The following protocol describes two nano-flow CE-ESI interfaces in hyphenation to orbitrap mass spectrometers (see Fig. 2, middle and bottom panels).

• Instrumentation:

  • A capillary electrophoresis system (custom-built or commercial)
  • A high-resolution mass spectrometer
• Prepare consumables:

  • Separation capillary
  • Borosilicate capillary for the nanospray emitter
  • Background electrolyte (25% v/v methanol in water with 1 M formic acid)
  • Sample loading solution (60% v/v acetonitrile in water with 0.05 % v/v acetic acid)
  • Angiotensin II standard at 1× 10⁻⁴ g/L
• Construct the CE-nanoESI interface following detailed instructions provided elsewhere for sheath-flow (Nemes et al., 2013; Sun et al., 2013) or sheathless interfaces (Moini, 2001). Sheath-flow interfaces are known for ease of construction, reproducibility, stability, and durability. Sheathless interfaces are known for sensitivity. In what follows, main steps are provided for two sheath-flow interfaces that use different experimental designs (Fig. 2).

  • Cut a ~1 m long piece of fused silica capillary (40/110 μm ID/OD)
  • Feed the capillary inside the PEEK sleeve, then burn the polyimide coating, and clean the capillary end using isopropanol.
  • Install the CE capillary into a cross or T-junction connected to a tubing delivering the sheath flow, allowing the tip of the CE capillary to protrude <50 μm (Fig. 2, top panel), <25 μm (Fig. 2, middle panel), or 500 μm beyond the emitter tip or into the emitter (Fig. 2, bottom panel). In our experience, these CE-ESI designs are facile to construct, require minimal maintenance (e.g., diminished likelihood of clogging of the emitter tip), and provide high sensitivity for trace-level analysis. For example, the lower limit of detection for model peptides (angiotensin II) is 260 zmol for the tapered-tip metal emitter construct (Fig. 2, top panel) and ~700 zmol for the pulled borosilicate design (bottom panel). With less likelihood for clogging at the emitter tip due to a larger tip opening, the former design improves robustness, albeit at the expense of enhanced dilution of the sample at the emitter tip due to a higher co-axial sheath flow.
  Note: When using the electrokinetically pumped sheath-flow interface, ensure sufficient space around the CE capillary inside the pulled borosilicate capillary to avoid blockage.

  • Position the tip of the CE-ESI interface ~5 mm from the orifice of the mass spectrometer for the tapered-tip (Fig. 2, middle panel) and ~2 mm for the electrokinetic sheath-flow interface (Fig. 2, bottom panel).
  • Clean the setup by flushing the capillary overnight with LC-MS grade water.

    • Initialize the CE-ESI system for ~15 min before starting experiments:

      • Fill the capillary with the background electrolyte (25% acetonitrile with 1 M formic acid in water) by flushing it for ~15 min.
      • Supply the appropriate electrospray solution (see Fig. 2) through the sheath tubing ensuring no bubbles in the supply line.
      • Turn on the high-voltage power supplies (without outputting voltage) to the CE and ESI devices to warm up the instruments.
      • Turn on the digital voltmeter to monitor CE current (see details in (Nemes et al., 2013))
    • Confirm stable operation and validate sensitivity daily using chemical standards (e.g., angiotensin II) following our protocols reported elsewhere (Choi, Lombard-Banek, Munoz, Manzini, & Nemes, 2018; Choi et al., 2017; Lombard-Banek, Moody, & Nemes, 2016a; Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016). The following steps are recommended to test for reproducible separation [desired variance in peptide migration times < 10% relative standard deviation (RSD)], quantification (desired variance in peak intensity < 25% RSD), and detection sensitivity (desired lower limit of detection ~ 210 zmol) using the tapered-tip emitter design (Fig. 2, middle panel) (Choi et al., 2017) as follows:

      • Measure the angiotensin II standard under the same conditions (identical injection volume, CE separation potential, and mass spectrometer settings) each day prior to start the measurements of biological samples.
      • Process the acquired data to plot the selected ion electropherogram for angiotensin II
      • Integrate the area under the curve for the angiotensin II peak
      • Determine reproducibility by calculating the relative standard deviation of the peak areas measured across all standard measurements. We recommend analyzing at least 3 technical replicates (same standard analyzed in 3 independent experiments).
    • Single cell measurements:

      • Reconstitute the dried protein digest into 1–2 μL of the sample loading solution, vortex mix the sample, and centrifuge it at 10,000 × g for 2 min
      • Deposit 250 nL to 1 μL of the digest onto the sample-loading microvial (see Fig. 1).
      • Inject ~1–20 nL of the protein digest hydrodynamically in the CE capillary as described elsewhere (Nemes et al., 2013).
      • Place the inlet of the capillary back into the background electrolyte vial.
      • Initialize the separation by ramping the CE separation voltage from ground to ~10 kV.
      • Establish the electrospray as follows:

        • For the tapered-tip co-axial sheath interface (Fig. 2, middle panel): Turn on the electrospray voltage to ~2,500 V. Ramp up the CE separation voltage to ~22.5 kV. A CE current of 2–4 μA should result under these conditions.
        • For the electrokinetic sheath-flow interface (Fig. 2, bottom panel): Start by applying ~1,000–1,500 V to the platinum wire connected to the sheath solution to initiate the electrospray. Ramp up the separation voltage to ~20 kV. A CE current of 5–8 μA should result under these conditions.
        • To obtain the highest ionization efficiency, maintain the electrospray in the cone-jet spraying regime by adjusting the emitter tip to MS orifice distance and the electrospray potential. Use a long working distance camera and/or spray current measurements to characterize the spraying mode [see details in reference (Nemes, Marginean, & Vertes, 2007)].
  Note: Setting the ESI voltage too high may cause electrowetting or corona discharge at the tip of the emitter, which destabilizes the Taylor cone, thus decreasing spray stability and detection sensitivity. Actively monitor the electrospray to avoid electrical sparks/arcs at the emitter tip, which pose electrical shock hazard to the mass spectrometer. Additional trouble-shooting advice is summarized in Table 1.

  • Load data-dependent acquisition method and start MS data collection. For example, the following settings were used on a Q-Exactive Plus (ThermoFisher Scientific) to measure single cell extracts from frog embryos. MS¹ parameters were: mass range, m/z 350–1,600; mass resolution, 35,000 FWHM (m/z 200); C-trap maximum injection time, 50 ms; automatic gain control (AGC) target, 1×10⁶ ions; MS/MS (MS²) events were triggered for ions with signal intensity greater than 1.5×10³ counts and a charge state between +2–7 with peptide-like isotopic distribution. MS² acquisition parameters were: ass resolution, 17,500 FHWM (m/z 200); C-trap maximum injection time, 60 ms; C-trap AGC target, 5×10⁴ counts; dynamic exclusion time and mass tolerance, 9 s and 5 ppm; peptide isolation window, 1 Da; collision, HCD in nitrogen at 28% normalized collision energy (NCE).
  • Once the acquisition is done, usually in ~60–90 min, gradually lower the separation and ESI voltages to 0 V (ground) and set the capillary to flush with BGE before analyzing the next sample.

g). Data Processing

The goal of data processing is to identify and quantify proteins based on proteotypic peptides from the raw MS-MS/MS data (Fig. 3). We have successfully used publicly available software platforms to process single-cell proteomics data, including Proteome Discoverer (ThermoScientific) and MaxQuant (Tyanova, Temu, & Cox, 2016).

• Protein identification based on proteotypic peptides: In data-dependent acquisition, precursor peptide ions are selected for fragmentation typically by collision-induced dissociation (CID), higher collision dissociation (HCD), or electron transfer dissociation (ETD) (Fig. 3A). Peptides are identified by comparing the MS/MS spectra to in-silico generated tandem mass spectra from the proteome of the specimen using a search engine (Fig. 3B). Reference proteomes are obtained as FASTA files from resources such as UniProt (Renaux & UniProt, 2018), SwissProt, or experimentally determined based on mRNA expression (Smits et al., 2014; Wuehr et al., 2014). Data resulting from our protocol can be searched using a variety of software packages, including but not limited to SEQUEST (MacCoss, Wu, & Yates, 2002), Mascot (Perkins, Pappin, Creasy, & Cottrell, 1999), and Andromeda (Cox et al., 2011). Further details in data processing have been reviewed elsewhere (Aebersold & Mann, 2003; Walther & Mann, 2010; Y. Y. Zhang et al., 2013). Figure 3 exemplifies the detection of a peptide ion with m/z 572.33, which was identified as LGLGLELEA from the voltage-dependent anion-selective channel protein 2 (Vdac2) from X. laevis using the following search and MaxQuant ver. 1.6.2.10:

  • Protein database: X. laevis proteome (SwissProt, downloaded from UniProt) concatenated with an mRNA-derived proteome (PHROG1.0).
  • Search parameters (typically used by us): Enzyme, trypsin; missed cleavages allowed, maximum 2; MS¹ mass tolerance, 4.5 ppm; MS² mass tolerance, 20 ppm; Static modification, none (we typically exclude alkylation and reduction for single-cell proteomics to minimize sample preparation); Variable modifications, methionine (M) oxidation and glutamine (Q) and asparagine (N) deamidation; minimum peptide length, 5 amino acids; match between runs, enabled; protein false discovery rate (FDR), < 1% (calculated against the reverse-sequence proteome).

• Protein quantification: Our approach is compatible with label-free quantification approaches such as MaxLFQ (Cox et al., 2014) as well as label-based (multiplexing) strategies, including tandem-mass tags (TMT) (Thompson et al., 2003), isotope tags for relative and absolute quantification (iTRAQ)(Ross et al., 2004), or dimethyl leucine (Xiang, Ye, Chen, Fu, & Li, 2010), among other approaches. The data resulting from these strategies implementing microprobe CE-ESI-HRMS can be processed by following data analysis workflows that have been established in the community (see reviews (Tyanova et al., 2016; Y. Y. Zhang et al., 2013)).

4. Representative Results

Microprobe CE-ESI-HRMS achieves sufficient figures of merit for single-cell proteomics. The instrument is capable of low amol to ~200 zmol lower limit of detection for model peptides (Choi et al., 2017; Lombard-Banek, Moody, et al., 2019; Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016; Lombard-Banek, Yu, Swiercz, Marvar, & Nemes, 2019). This level of sensitivity enabled the identification of hundreds-to-thousands of proteins in limited amount of brain tissue (Choi et al., 2018; Choi et al., 2017), single embryonic cells (Lombard-Banek, Moody, et al., 2019; Lombard-Banek, Moody, et al., 2016b; Lombard-Banek, Reddy, et al., 2016), and most recently single neurons (Choi et al., 2019). Label-free quantification has provided the instrument with a ~4-to-6 log-order linear dynamic range for quantification on mass spectrometers equipped with time-of-flight and orbitrap mass analyzers (Choi et al., 2018; Choi et al., 2017; Lombard-Banek, Reddy, et al., 2016). Relative quantification by designer mass tags (TMTs) has been used to quantify protein expression between cells (Lombard-Banek, Moody, et al., 2016b). Moreover, with the recent combination of microprobe sampling with CE-ESI-HRMS, we have enabled the identification and quantification of metabolomes (Onjiko, Plotnick, Moody, & Nemes, 2017; Onjiko, Portero, et al., 2017; Portero & Nemes, 2019) and proteomes in single cells in freely developing embryos (Lombard-Banek, Moody, et al., 2019) and active brain substantia nigra slices (Choi et al., 2019).

Microprobe CE-ESI-HRMS has allowed us to uncover previously unknown proteomic differences between limited populations of cells and single cells (Fig. 4). By measuring only ~0.05% of the total proteome content of the cell, we have identified ~800 different protein groups and quantified ~450 protein groups via LFQ in single identified cells in live X. laevis embryos. The data revealed, for the first time, proteomic differences as the midline dorsal-animal cell (see cell in the 16-cell embryo, Fig. 4A, left panel) divided through the 32-, 64-, and 128-cell stage towards forming a neural-tissue fated cell clone. Fuzzy c-mean cluster analysis of the quantitative protein abundances classified the proteins into four distinct trends in spatial and temporal expression (Fig. 4A, right panel). For example, proteins in group #1 presented with increasing concentration, while those in group #3 were degraded as the clone developed. Using multiplexing quantification (TMTs), we have demonstrated the identification of ~1,200 proteins and quantification of ~950 proteins between three cell types, which give rise to neural (D11), epidermal (V11), and hindgut (V21) tissues. Intriguingly, molecular differences had not been detectable at the level of transcription between these cell types in earlier studies, thus highlighting the importance of directly measuring proteins in cells to help understand cell heterogeneity.

Microprobe CE-ESI-HRMS is scalable to smaller cells and other types of biological models. We have recently applied the technology to identify ~350 different proteins between single cells analyzed in 2-cell zebrafish embryos (Lombard-Banek, Moody, et al., 2019). LFQ of these proteins between the cells revealed detectable variability between protein expression, raising a potential for mapping out proteomic cell heterogeneity in this model as well. Moreover, as shown in Figure 4B, we have successfully scaled microprobe CE-ESI-HRMS to dopaminergic neurons in the mouse substantia nigra, which are ~5-times smaller in size and contain ~1,000-times less protein than cells in the 16-cell X. laevis embryos. The neuron type was identified using patch-clamp electrophysiology and confirmed based on its action potential. Despite measuring only ~pg amounts of protein digest, the study identified ~160 proteins and quantified ~100 proteins in these neurons. LFQ found the proteins to span ~4-log order in concentration (Fig. 4C). Some of the identified proteins are preferentially enriched in dopaminergic neurons such as tho complex subunit 2 (Thoc), prolow-density lipoprotein receptor related protein (Lrp1), and zinc finger and scan domain containing protein 2 (Zscan2), thus partially validating our technology. These and other examples from our laboratory have established microprobe CE-ESI-HRMS as a highly sensitive technology that is capable of identifying and quantifying proteins in spatial and temporal scalability, delivering exciting measurement capabilities to help understand molecular mechanisms of cell heterogeneity.

5. Conclusions

We have developed microprobe CE-ESI-HRMS to enable, for the first time, the proteomic characterization of single cells that are embedded in complex tissues, even including live specimens. Representative applications from our laboratory have included the interrogation of proteomic cell heterogeneity in 1-to-128-cell live embryos of Xenopus laevis and zebrafish and single neurons in live sections from the mouse substantia nigra. Compared to manual cell dissection, microprobe CE-ESI-HRMS has the advantage of preserving tissue morphology and the integrity of neighboring cells, is commensurate with cell identification based on optical and electrophysiological investigation, and it offers higher sensitivity by limiting the collection of salts with known interference during ionization and detection. We have used our custom-built CE-ESI-HRMS platforms for the quasi-routine identification/quantification of ~800 different protein groups and ~1,400 protein groups among single cells (Lombard-Banek, Moody, et al., 2019; Lombard-Banek, Moody, et al., 2016b). With a compatibility with recent advances in commercial CE instrumentation (e.g., model CESI, AB Sciex) and ESI interfaces (e.g. EMASS-II, CMP Scientific), we expect our protocol to be broadly adaptable by other investigators as well. Combined, these single-cell proteomics studies have revealed previously unavailable information on the proteomic state of single cells, essentially serving as a quantitative snapshot of molecular activity complementing results from single-cell transcriptomics. Single-cell proteomics by microprobe CE-ESI-HRMS expands the analytical toolbox of technologies capable of single-cell analyses, opening the door to the holistic characterization of molecular processes in the cell encompassing gene transcription, translation of transcripts, and production of metabolites. The results from these studies can be used to derive hypotheses and enhance our understanding of differences between cells that underlie states of health and disease.