Enhanced single-cell metabolomics by capillary electrophoresis electrospray ionization-mass spectrometry with field amplified sample injection

Abstract

Single-cell metabolomics provides information on the biochemical state of an individual cell and its relationship with the surrounding environment. Characterization of metabolic cellular heterogeneity is challenging, in part due to the small amounts of analytes and their wide dynamic concentration ranges within individual cells. CE-ESI-MS is well suited to single-cell assays because of its low sample-volume requirements and low detection limits. While the volume of a cell is in the picoliter range, after isolation, the typical volume of the lysed cell sample is on the order of a microliter; however, only nanoliters are injected into the CE system, with the volume mismatch limiting analytical performance. Here we developed an approach for the detection of intracellular metabolites from a single neuron using field amplified sample injection (FASI) CE-ESI-MS. Through the application of FASI, we achieved 100- to 300-fold detection limit enhancement compared to hydrodynamic injections. We further enhanced the analyte identification and quantification accuracy via introduction of two internal standards. As a result, the relative standard deviations of migration times were reduced to <5%, aiding identification. Finally, we successfully applied FASI CE-ESI-MS to the untargeted profiling of metabolites of Aplysia californica pleural sensory neurons with <50 μm diameter cell somata. As a result, twenty one neurotransmitters and metabolites have been quantified in these neurons.

1. Introduction

Single-cell ‘omics, a relatively new approach in analytical and molecular biology investigations, has advanced the discovery of cellular heterogeneity by focusing on the analysis of endogenous compounds, such as gene transcripts, metabolites, lipids, peptides, and proteins [1–3]. Genetic, structural, functional, or environmental differences lead to cellular chemical heterogeneity and variability. Analysis of single cells expands our understanding of the mechanisms and manifestations of biological heterogeneity and variability of different biological systems, including ones formed by morphologically similar cells [4]. This knowledge may help explain an organism’s differential susceptibility to diseases and response to treatments. Single-cell ‘omics has facilitated studies in areas ranging from neuroscience to cancer and stem cell research [5, 6].

Along with other ‘omics techniques, metabolomics provides relevant information on cellular phenotype, representing the interplay of internal biological processes and external environmental influences in healthy and diseased states [7]. Comparative metabolomics aids in the discovery of biomarkers for different pathologies, such as diabetes and organ failure [8–10], cardiovascular disease [11], and various cancers [12–18].

Small analyte amounts are among the major challenges in single-cell analysis, restricting the use of many metabolomics approaches and limiting the depth of metabolome coverage. Generally, mammalian cells have diameters of ~10 μm with ~500 fL volumes containing small amount of metabolites [19], thus requiring atto- to zeptomole limits of detection and minimization of analyte loss during sample handling and separation [20]. GC, LC, and CE analyte separation methods (followed by MS detection) are typically used in metabolomics analysis [21–27], with CE being the most effective for small-volume samples, especially for polar analytes [22, 24].

How does one optimize CE-MS protocols for single-cell analysis? The minute volumes of individual cells present challenges in maximizing the chemical information obtained in a measurement. A common and useful approach is to improve the detection limits of the measurement itself. For example, sheathless porous tip interfaces for CE-MS have demonstrated up to two orders of magnitude sensitivity improvement and thus, become suitable for metabolic profiling [28–31]. An additional approach is to minimize the dilution of analytes during sample preparation where, in most cases, individual cells are lysed in a volume of at least a microliter, with only a small fraction injected into the CE capillary. There are several on-line analyte concentration methods that address the volume mismatch, including pH junctions [32], sweeping [33], isotachophoresis [34], and the combination of different on-line concentration methods [35]. Stacking is an approach that utilizes the difference of an analyte’s electrophoretic velocities in different environments, leading to its accumulation at the boundary of the sample matrix and a separation buffer [36]. With the optimization of the conductivities of sample matrix and buffer, analytes can be stacked, resulting in improved detection sensitivity [35]. Large volume sample stacking (LVSS) and field amplified sample injection (FASI) are two typical techniques used in CE [37]. Sensitivity enhancements of 100-fold have been reported with LVSS [35], and it has successfully been applied to both untargeted and targeted single-cell metabolomics [22, 38, 39]. FASI has also demonstrated similar enhancements up to 1000-fold [35].

Here we report efficient and sensitive FASI CE-MS, sample desalting, and low-volume manipulations that combine to optimize the measurements of single-cell metabolites. The described approach successfully addresses several classic FASI-associated issues, including salt interferences, leading to migration time-shifts and analyte competition during sample loading. To demonstrate the capabilities of the FASI CE-MS approach, metabolites in small pleural sensory neurons of the sea slug Aplysia californica have been detected, characterized, and quantified.

2. Experimental

2.1. Chemicals

Stock solutions (25–100 mM) of metabolite standards were prepared using LC-MS grade water and methanol and stored at −20 °C until use. Most chemicals were purchased from MilliporeSigma (St. Louis, MO) unless otherwise specified. Methanol, isopropanol, formic acid (FA), and acetonitrile (ACN) were obtained from Thermo Fisher Scientific (Waltham, MA). Sample extracts were kept in polymerase chain reaction tubes from MidSci (St. Louis, MO).

2.2. Animals and single-neuron isolation

Adult Aplysia californica (180–250 g) were purchased from the National Resource for Aplysia (Rosenstiel School of Marine & Atmospheric Science, University of Miami, FL), and kept at 14 °C in continuously circulated and aerated aquarium filled with sea water prepared in house using Instant Ocean Sea Salt (Instant Ocean Spectrum Brands, Blacksburg, VA). Before dissection, animals were anesthetized by injecting 390 mM MgCl₂ solution into the vascular cavity. The injection volume of the MgCl₂ solution was equal to approximately one-third of each animal’s body weight. The central nervous system was surgically isolated and placed into artificial seawater (ASW) comprising: 460 mM NaCl, 10 mM KCl, 10 mM CaCl₂, 22 mM MgCl₂, 26 mM MgSO₄, and 10 mM 4-(2-hydroxyethyl)-1-piperazineethane sulfonic acid, and supplemented with 100 U/mL penicillin G, 100 μg/mL streptomycin and 100 μg/mL gentamicin, pH 7.8. Ganglia and adjacent nerves were dissected and treated in 1% protease type IX in ASW-antibiotic solution for 60–100 min (depending on animal size) at 34 °C. After treatment and wash in fresh ASW-antibiotic solution, the sensory neurons were manually isolated from the pleural ganglia using sharp tungsten needles (World Precision Instruments, Inc., Sarasota, FL). The isolated single neurons were quickly (2–5 s) washed with deionized water to minimize transfer of extracellular inorganic salts from the ASW into final samples. Washed single neurons were placed into 10 μL of methanol to facilitate analyte extraction and quench enzymatic processes occurring ex vivo. The samples were stored in the solution at −20 °C until analysis. Before analysis, the samples were dried in a SpeedVac vacuum concentrator (Genevac, Ipswich, Suffolk, UK) and resuspended in 50 μL of a mixture containing isopropanol and ACN (0.8:0.2, v:v) providing analyte solubilization as well as further extraction of metabolites. The samples were centrifuged for 5 min at 16000 × g, and the supernatant was further dried in the SpeedVac vacuum concentrator. Finally, the samples were reconstituted in 10 μL of a solution containing 95% methanol, 5% water, and 0.01% FA.

2.3. FASI-CE-MS platform and single-neuron analysis

Standard- and single cell-containing samples were assayed using an in-house assembled CE platform hyphenated to a high-resolution mass spectrometer (qTOF maXis 4G, Bruker Corp., Billerica, MA) through a custom-built coaxial sheath-flow ESI source [40–42]. The ESI source consisted of a micro-tee assembly (part P-875, IDEX Health & Science, Oak Harbor, WA) with a platinum alloy emitter (10% iridium, 0.0055” ID × 0.003” wall × 1” length, part 29910E, Johnson Matthey Inc., Wayne, PA). The sheath liquid (50% methanol in water solution with 0.1% FA) was supplied through the emitter at a 750 nL/min flow rate using a syringe pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA) and a fused silica delivery capillary for the sheath flow (75 μm ID, 363 μm OD, Polymicro Technologies, Phoenix, AZ). The CE separation was performed on 75–80 cm, 40 μm ID, and 105 μm OD fused silica capillary (part TSP040105, Polymicro Technologies,) with a separation voltage of 20 kV. The background solution was 1.0% FA in water. FASI was performed by electrokinetic injection of the sample solution at 20 kV for 30 s. For the hydrodynamic injection, the sample was injected by elevating the separation capillary inlet by 13 cm for 1 min. The emitter was grounded, and the capillary voltage of the mass spectrometer was set at 2000 V to establish the cone-jet spray. The dry gas was set at 180 ⁰C with a flow rate of 3 L min⁻¹. The mass spectrometer was calibrated regularly in the mass range of m/z 50–500 by infusing sodium formate (15 mM) via the ESI source. Recalibration of acquired data sets was performed offline with endogenous sodium formate clusters. Tandem MS experiments were accomplished with 20 to 40 eV collision energies to facilitate the molecular identifications. To determine the molecular features, the obtained data were first processed by the XCMS [43] package written in R [44] using the following settings: 30 ppm mass error, signal-to-noise ratio equal to 3, and filtered with an arbitrary threshold of 1000 counts. Signals related to the analyte’s isotopic distribution and background interference were manually eliminated. Finally, the detected accurate molecular masses were matched to data available at online metabolite databases (METLIN [45], the Human Metabolome Database [46], and MycompoundID) [47]) with an allowed error of 5 ppm.

To evaluate the migration time variation caused by different salt content, the metabolites were measured in a sample matrix comprised of 0.01% FA in 95% methanol with 0, 0.1, 0.5, 1, and 2 mM NaCl (n = 5 samples, with each sample run twice). The repeatability of peak area was also evaluated using fresh 10 nM standard mixtures examined on four consecutive days. The calibration curves of the 21 metabolites were constructed using 5, 10, 50, 100, 500 and 1000 nM standard mixtures (n = 3). The limit of quantification (LOQ) for each metabolite was evaluated by a signal-to-noise ratio (s/n) of 5 [48].

3. Results and discussion

3.1. FASI

Qualitative and quantitative measurements of low amounts of metabolites in a single cell are challenging. Cell diameters span the range of a few micrometers to hundreds of micrometers and corresponding cell volumes vary from fL to nL, which is a million-fold volume range. The wide range of analyte concentrations in different cells adds to the technical demands of the analysis. Therefore, single-cell measurements require a sensitive analytical approach that offers a wide dynamic range.

Here, we optimized FASI to improve the detection limits in single-cell ESI-MS analysis. Because analyte stacking efficiency is determined by the conductivity difference between the background solution and sample matrix, the percent composition of organic solvents and FA in the solutions differ [49–51]. Generally, increasing the organic solvent percentage and decreasing the FA content improves detection sensitivity because this reduces conductivity. However, this also leads to increased Joule heating, and an increased likelihood of air bubble formation. After method optimization, we found that a 0.01% FA, 4.99% water, and 95% methanol solution was the most effective. This solution was used in experiments comparing the FASI approach to the classic hydrodynamic sample injection method. In these measurements, ~6 nL of a standard mixture containing lysine, histidine, and arginine (all at 100 nM) was loaded onto a CE column by a 60 s hydrodynamic injection and analyzed. The standard mixture, diluted to 10 nM for all analytes, was loaded during 60 s by FASI. We observed that 307-, 191-, and 215-fold detection sensitivity enhancement was achieved for lysine, histidine, and arginine, respectively, by FASI CE-ESI analysis of the more diluted samples (Figure 1 and Figure S1). Table S1 shows the enhancement factors of 15 metabolites with FASI.

Figure 1.

Comparison of hydrodynamic injection and FASI for CE-MS analysis of a lysine, histidine, and arginine mixture. (A) System schematic. (B) The extracted ion electropherograms of lysine, histidine, and arginine standards with (i) 6 nL of 100 nM lysine, histidine, and arginine solution injected via the hydrodynamic approach and (ii) 10 nM lysine, histidine, and arginine standard solution with FASI.

To make a direct comparison between hydrodynamic injection and FASI, we used our standard sample volume of 500 nL [41]; we expect that performing FASI from a larger sample volume would further increase signal enhancement, and likely prevent electrolytic degradation of molecular species in proximity to the walls of the stainless steel vials. However, the smaller vial is a more realistic comparison given our standard use of the 500 nL vials for individual cell measurements.

3.2. Inorganic salt interference in single-cell CE-MS analysis

Inorganic salts present inside single cells and the associated extracellular physiology saline impact FASI performance, especially when working with cells collected from a marine organism with cation concentrations reaching ~500 mM. Sodium and potassium can be at concentrations greater than 100 mM, are found both inside and outside the cell, and impact sample stacking, ionization efficiency, and metabolite detection [52]. A quick wash of the single cell with deionized water is helpful in reducing sample salt content, and accordingly, ion competition during electrokinetic injection [40]. Because the volume of the extracellular liquid depends on the operator’s skills during cell isolation, washing is an important step that helps increase the repeatability of measurements between cells. Cell membrane stability should be considered during this quick hypoosmotic washing step to prevent cell lysis. However, because the FASI approach requires high sample purity, without salts, which cannot be achieved by the cell rinse alone, we introduced a salt precipitation step (Figure 2A). The solubility of sodium chloride differs with the solvents used [53]. For example, isopropanol (IPA) and ACN are poor solvents for NaCl (Figure 2B) [53]; therefore, IPA and ACN were tested as analyte extraction solvents that can also precipitate and remove inorganic salts. Generally, inorganic salt solubility in ACN is lower than in IPA, resulting in improved detection sensitivity for metabolites with lower mobility. However, a decrease in detection sensitivity for some polar metabolites with higher mobility, such as lysine, arginine, and histidine, was observed (Figure S2), likely due to their poor solubility in ACN. Considering these results and the solubility of the amino acids and inorganic salts, as well as salt-related analyte injection competition, a 4:1 mixture of IPA to ACN was chosen. Sodium is the most abundant inorganic salt in our sample, and will form sodium formate adducts in the presence of the FA in our CE separation buffer system. The most abundant sodium formate adduct is [Na(NaCOOH)₃]⁺ with an m/z of 226.9534 in our CE-MS condition. We can evaluate the sodium content in the sample before and after extraction using different organic solvents by noting the peak area of this specific sodium formate adduct (Figure S3). To evaluate the effect of salt precipitation on the detection of endogenous lysine, histidine, and arginine, we used samples containing single pleural sensory neurons. As a result of incorporating this step, the signal intensity of these metabolites increased 4.5- to 6-fold (Figure 2C).

Figure 2.

Optimization of endogenous amino acid measurements from individual neurons using a salt precipitation. (A) Experimental schematic depicting microphotograph of a pleural sensory neuron (~50 μm in soma diameter). (B) The respective solubilities of relevant solvents. (C) The extracted mass electropherograms of lysine, histidine, and arginine detected from the same pleural sensory neuron: (i) without salt precipitation, and (ii) with salt precipitation.

3.3. Metabolite identification in untargeted FASI CE-MS analysis of single cells

The enhanced detection sensitivity of the FASI-based platform increased metabolite coverage in single-cell CE-MS analysis. On average, more than one thousand molecular features were detected from a single cell when analyzed by the XCMS software; however, as with many metabolomics studies, only a small fraction are assignable to known metabolites (Table S2). Possible reasons for unassignable peaks include chimeric MS2 spectra, salt adducts, or overlap of isotopic envelopes. In-source degradation of some metabolites is observed for some single-cell CE-MS analyses (Figure S4). Our laboratory-built CE system improves analyte separations [54, 55], alleviating reported issues related to peak overlap. Additionally, to improve confidence in our identifications, the migration times of endogenous analytes and related standards were compared. Migration times of metabolites in CE measurements are affected by many factors, including the conductivity of the sample and changes in electroosmotic flow, which can be impacted by proteins adsorbing onto the capillary walls. Here, we addressed migration-time issues using a two internal standard (IS) correction approach with the dipeptides alanine-alanine and proline-leucine (see Supporting Information for details on migration time correction calculations).

To evaluate the effectiveness of the two IS method, we tested several amino acid standard solutions spiked with different NaCl concentrations. The relative standard deviation (RSD) of the migration time variation caused by different salt content significantly decreased, from 5–15% to less than 2.5% (n = 5 samples, and each sample was tested twice) (Figure 3).

Figure 3.

The RSDs of migration times of different metabolites measured in the sample matrix consisting of 0.01% FA in 95% methanol with 0, 0.1, 0.5, 1, and 2 mM NaCl.

3.4. FASI CE‐MS metabolite detection and quantification in single-cell samples

Using the FASI CE-MS system, we can quantify a range of metabolites from individual cells. As with our standards, we used alanine-alanine and proline-leucine to correct for migration time variability and to address issues related to sample preparation. The linear working ranges were from 5 nM to at least ~1000 nM for the evaluated metabolites (Table 1). For the calibration curves of the 21 metabolites, each metabolite was measured at 5, 10, 50, 100, 500 and 1000 nM of standard mixtures (n = 3) (Figure S5). The repeatability of peak area and intensity was also evaluated using fresh 10 nM standard mixtures examined on four consecutive days. These experiments revealed that the analytical repeatability of the peak area was 15±2%, and the intensity was 13±3%. The positioning of the capillary inlet may affect the repeatability of the peak area and intensity [56]. Injecting from a larger volume would mitigate the risk of capillary-to-vial contact but would result in higher analyte dilution from the cell.

Table 1.

Details of the calibration curves for the 21 metabolites.

Metabolite	m/z ([M+H]+)	Fitting of external calibration standard curve	Correlation coefficient (R2)	Linear range (nM)	LOQ (s/n=5)
Glycine	76.0388	y = 4.05E−07x + 2.52E−05	0.990	5–500	3.2
Alanine	90.0544	y = 2.45E−05x + 2.13E−03	0.982	5–500	1.
Serine	106.0496	y = 2.25E−04x + 1.43E−02	0.990	5–500	0.2
Proline	116.0705	y = 5.95E−04x + 1.15E−02	0.999	5–500	0.9
Valine	118.0861	y = 2.80E−04x + 1.15E−02	0.999	5–1000	0.5
Threonine	120.0654	y = 2.19E−04x + 9.39E−03	0.990	5–1000	0.5
Leucine/Isoleucine	132.1017	y = 2.03E−03x + 8.80E−02	0.991	5–1000	0.2
Asparagine	133.0605	y = 1.90E−04x + 5.56E−03	0.987	5–1000	3.
Tyramine	138.0910	y = 5.15E−04x + 1.46E−02	0.998	5–1000	0.3
Glutamine	147.0760	y = 1.02E−04x + 2.19E−03	0.996	5–1000	5.
Lysine	147.1124	y = 6.37E−04x + 1.72E−02	0.999	5–1000	0.3
Glutamic acid	148.0611	y = 2.54E−04x + 9.94E−03	0.991	5–1000	1.2
Methionine	150.0578	y = 4.86E−04x + 2.11E−03	0.999	5–1000	3.3
Dopamine	154.0857	y = 4.50E−04x + 3.52E−04	0.999	5–1000	3.6
Histidine	156.0761	y = 7.55E−04x + 3.20E−02	0.990	5–1000	0.2
Tryptamine	161.1067	y = 9.61E−04x + 2.83E−02	0.997	5–1000	2.3
Phenylalanine	166.0855	y = 1.48E−03x + 5.22E−02	0.988	5–1000	0.2
Arginine	175.1181	y = 1.07E−03x + 2.72E−02	0.997	5–1000	0.3
Citrulline	176.1020	y = 4.60E−04x + 1.12E−02	0.997	5–1000	0.6
Serotonin	177.1013	y = 6.91E−05x + 4.02E−03	0.962	5–1000	2.
Tryptophan	205.0960	y = 7.77E−04x + 3.37E−02	0.987	5–1000	1.8

To demonstrate the FASI CE-MS system’s capabilities for the analysis of endogenous metabolites, isolated sensory neurons from the Aplysia nervous system were characterized. Neurons with a cell body diameter of ~50 μm were randomly isolated from the pleural sensory cluster. These neurons have 10- to 1000-fold less cell body volume compared to A. californica left pleural 1 (LPl1) and R2 neurons exhibiting diameters of somata ranging between 150 to 500 μm, previously analyzed and characterized by CE-MS [41]. We detected and identified 37 metabolites in the sensory pleural neurons (Figure 4).

Figure 4.

FASI CE-MS analysis of individual pleural sensory neurons. Extracted ion electropherograms of several detected metabolites. Inset: pleural sensory neuron with a ~50 μm diameter cell body.

We used FASI CE-MS to quantify selected metabolites inside individual neurons (Figure 5A). Among these metabolites, tyramine is a neuromodulator in A. californica [57], and it showed a surprisingly high concentration of 67 ± 25 mM, higher than others we measured. The results of the product ion scan for tyramine are shown in Figure S6. Dopamine was also detected in several neurons at 3 ± 1 mM. Both tyramine and dopamine are important neurotransmitters / modulators in A. californica [58, 59]. Neither automated or manual examination of the electropherograms revealed other transmitters / neuromodulators, such as γ-aminobutyric acid (GABA), and serotonin (Figure 5B). Arginine (1.2 ± 1.3 mM) and citrulline (<LOQ) were also detected in some of the pleural sensory neurons (Figure 5A). Similar results were obtained using other approaches for an entire pleural sensory cluster previously characterized [60–62]. These findings corroborate with data demonstrating the presence of the nitric oxide synthase pathway in the pleural sensory neurons [60–62].

Figure 5.

Quantitative FASI CE-MS analysis of metabolites in individual pleural sensory neurons. (A) Box plot demonstrating absolute concentrations of different metabolites in 6 pleural sensory neurons. (B) Representative extracted mass electropherograms of common neurotransmitters and metabolites detected in studied neurons. Red arrows indicate localization of metabolite signal matching by migration time and accurate m/z to signal of corresponding standard. The EICs showed no serotonin or GABA signals.

4. Conclusions

We created a robust and reproducible FASI CE-MS approach to characterize and quantify the cationic metabolites present in neurons. The use of FASI in CE-ESI-MS analysis improves metabolite detection, providing enhanced sensitivity (100- to 300-fold) and repeatability. Additionally, with the introduction of a sample pretreatment using a 0.8:0.2 IPA:ACN solvent mixture the negative impact of inorganic salts on analyte detection and characterization was reduced. These experimental advances were further enhanced by two IS normalization of analyte migration times occurring due to run-to-run changes in analyte electrophoretic mobilities. FASI CE-MS was applied to the detection, characterization, and quantification of metabolites in single neurons from A. californica. The observation of several functionally important metabolites, such as the tyramine, arginine, and citrulline, in the pleural sensory neurons demonstrate the capabilities of this microanalysis approach.

Because the number of detected signals in a single cell could not be identified with reasonable confidence, further improvement in the sensitivity of the FASI CE-MS system is needed for analyte fragmentation and identification in smaller and more chemically complex neuron samples. We expect the signal enhancement to further improve when paired with higher performance CE-ESI-MS systems, such as those that use a sheathless ESI interface. Although FASI provides enhanced signals for charged metabolites, this technique may not be as suitable for the analysis of all metabolites of interest, especially for metabolites that can be easily oxidized or are unstable under high voltage and high temperature conditions [63, 64]. The optimized FASI CE-MS approach can be applied to analysis of other cell types, including mammalian and insect neurons. Future advances in cell sampling, e.g., microjunction analyte extraction as well as automated sample injection, are expected to further enhance the throughput and accuracy of single-cell FASI CE-MS analysis.

Highlights.

• Field amplified stacking improves single-cell metabolite detection when using capillary electrophoresis mass spectrometry.

• Salt precipitation reduces the negative impact of inorganic salts from single-cell samples.

• A range of metabolites were quantified in individual neurons from A. californica.