Analysis of Endogenous Metabolites using Multifunctional Derivatization and Capillary RPLC-MS

Abstract

Heterogeneity in metabolite structure and charge state complicates their analysis in electrospray mass spectrometry (ESI-MS). Complications such as diminished signal response and quantitation can be reduced by sequential dual-stage derivatization and capillary RP LC-ESI-MS analysis. Our sequential dual-stage chemical derivatization reacts analyte primary amine and hydroxyl groups with a linear acyl chloride head containing a tertiary amine moiety. Analyte carboxylate groups are then coupled to a linear amine tag with a tertiary amine moiety. This increase in the number of tags on analytes increases analyte proton affinity and hydrophobicity. We derivatized 250 metabolite standards which on average improved signal to noise by >44-fold, with an average limit of detection of 66 nM and R² of 0.98. This system detected 107 metabolites from 18 BAECs, 111 metabolites from human urine, and 153 from human serum based on retention time, exact mass, and MS/MS matches from a derivatized standard library. As a proof of concept, aortic endothelial cells were treated with epinephrine and analyzed by the dual-stage derivatization. We observed changes in 32 metabolites with many increases related to energy metabolism, specifically in the TCA cycle. A decrease in lactate levels and corresponding increase in pyruvate levels suggest that epinephrine causes a movement away from glycolytic reliance on energy and a shift towards the more efficient TCA respiration for increasing energy.

Graphical Abstract

1. Introduction

Endogenous small molecules play important roles in biological systems including signaling, energy production, and metabolic networks.^{1, 2} High sensitivity quantitation using targeted and untargeted methods yields a deeper understanding of complex biological interactions.^{3, 4} However, existing metabolomic methods are hindered by analyte heterogeneity e.g. polarity, charge state, and abundance.⁵ Mass spectrometry (MS) has been widely used to detect metabolites, but the sensitivity and selectivity often depend on analyte ionization efficiency.^{6, 7} LC-ESI-MS is used to analyze a wide scope of compounds due to its higher loading capacity and larger mass range.^8–10 Given the wide range of metabolite polarity, no single chromatographic method yields high resolution across all metabolite classes.

Chemical derivatization is used to increase chromatographic performance and electrospray efficiency. Coupling reactions are targeted toward a specific functional group with chemical tags which confer hydrophobicity and proton affinity.^11–13 This offers increased chromatographic performance, reduced matrix effects, and uniform signal response.^{19, 20} Collectively these factors increase the signal intensity, signal-to-noise ratio (S/N), and decrease adduct formation compared to untagged analytes.^{9, 14–18}

Recently, multiple functional groups have been targeted to increase metabolite coverage by either split sample parallel derivatization (Figure 1A)¹⁹ or sequential derivatization (Figure 1B).^{13, 20, 21} Initial work in parallel derivatization targeted amines, thiols, ketones, and aldehydes from human breast cancer cells using four types of capture agents. This improved detection limits to 100–250 pmol/10⁷ cells.²² Similarly, parallel derivatization of alcohols and carboxylic acids from plasma samples improved S/N by 44–1500—fold.²³ Carbonyls, amines, and phenols from serum were derivatized in parallel using multiple dansyl tags. This system quantified 67 metabolites and increased sensitivity by 50–1500-fold.^{24, 25}

Figure 1.

(A) Split sample and parallelderivatization of metabolites. (B) Developed dual-stage sequential derivatization of metabolites.

While parallel tagging shows substantial benefits compared to untagged protocols, expansion of tagging on each analyte using sequential derivatization allows for further increasing MS signal response and reduction of adducts. Amines and phenols from human urine samples were successfully derivatized using ¹²C₂- and ¹³C₂-dansyl chloride (DnsCl).^{19, 26} This allowed the detection of metabolites with an improved S/N ratio of greater than 20. 2-hydrazinoquinoline (HQ) and 3-nitrophenylhydrazine were used to derivatize carboxylic acids, aldehydes, and ketones in urine and serum samples, detecting over 300 metabolites in a single run analysis of ~17 min.^{27, 28} Aldehyde and carboxylate metabolites of cytosine (5-cadC) were derivatized using Girard’s reagents (GirD, GirT, and GirP) with S/N by 3 fold.^{29, 30} Amines, alcohols, and carboxylic acids were derivatized using diethylaminoacetyl chloride and diethylethylenediamine, 30 amino acids and organics were detected in bovine aortic endothelial cells with an average LOD of 76 nM.^{13, 20, 21}

In this paper, we increased metabolite coverage using sequential chemical derivatization in conjunction with capillary RPLC-nESI-MS. This shows superior performance in sensitivity and metabolite coverage over the previous CE-MS and mixed-mode full bore LC-MS systems. Ultimately, capRPLC-MS allows for improved analysis of 150 analytes from human serum.^{12, 13, 20, 21} Additionally, this system was used to investigate the metabolic effects of epinephrine on mammalian endothelial cells.

2. Materials and methods

2.1. Reagents and solutions

All chemical reagents and materials were purchased from Sigma-Aldrich (St. Louis, MO, USA) and Fisher Scientific (Pittsburgh, PA, USA) unless otherwise noted. 1[Bis(dimethylamino)methylene]-1H1,2,3triazolo[4,5b]pyridinium3oxidhexafluorophosphate(HATU),hydroxybenzotriazole (HOBt),1-Chloro-N, N,2-trimethyl-1-propenylamine, 3-(diethylamino) propionic acid, anhydrous dimethylformamide (DMF), pyridine, n, n-diethylethylenediamine, propranolol, epinephrine purchased from Sigma-Aldrich. Dulbecco’s Modified Eagle’s Medium (DMEM) and fetal bovine serum (FBS) were purchased from Sigma-Aldrich. Human serum NIIST 909c was purchased from the National Institute of Standards and Technology. Normal human urine from purchased from Lee BioSolutions (Maryland Heights, MO USA). Metabolite standards (organic acids and derivatives, amino acids and derivatives, alkyl, aryl and short diamines, catechol amines, simple sugars) were obtained from the metabolite library of standards (IROA Technologies, Sea Girt, NJ). Ammonium carbonate, phosphate buffered saline (PBS), liquid chromatography mass spectrometry (LCMS) grade acetonitrile and water, formic acid and culture dishes were purchased from Fisher Scientific.

2.2. Biological sample treatment and preparation.

Bovine aortic endothelial cells (BAECs): BAECs (passage p6–8) were cultured in 10% FBS supplemented DMEM in 6 cm culture plates at 37°C and 5% CO₂. The cells were treated with 10μM of propranolol in serum reduced media (2% FBS v/v, 300 μL HEPES, 1.38 mg sodium pyruvate) for 30 minutes. Cell media was then swapped with 1μM epinephrine in serum reduced media mentioned above and incubated at 37°C and 5% CO₂ for 15 minutes. After 30 minutes the media was removed from the plates, followed by a rinse with 600 μL PBS. Cells were lysed using six hundred microliters 80:20 MeOH:H₂O (v/v). Cell plates were scraped, and the cell solution was transferred to a vial. The chilled cell solution was sonicated using a handheld Misonix wand sonicator (Farmingdale, NY, USA) for 10 × 1s bursts. Cells were centrifuged at 21,100×g for 3 min at 4°C to remove precipitated protein and cell debris.¹³ The supernatant was collected and stored at −20°C until use. Identical conditions were applied to the control (cells without drug treatment).

Human Urine:

A urine sample (100 μL) was extracted using 1ml cold 80/20 MeOH/H₂O, incubated at −20°C for 15 minutes, and then spun down using a centrifuge at 21,100×g for 5 min at 4°C. The supernatant was collected and stored at −80°C until use^{11, 31}.

Human Serum:

Serum (100 μL) NIST 909c standard was dissolved in 4 mL of 80/20 methanol/H₂O. The mixture was centrifuged at 21,100×g at a temperature of 4°C for 5 minutes. The supernatant was collected and stored at −80°C until use³².

2.3. Tag synthesis and derivatization of metabolites

Tag synthesis:

The N, N-Diethylethylenediamine tag was used without adulterations. The synthesis of 3-(diethylamino) propionyl chloride was undertaken by chlorinating the corresponding carboxylic acid. Specifically ghosez chlorinating reagent (1-Chloro-N, N,2-trimethyl-1- propenylamine (300 mM)) and 3-(Diethylamino) propionic acid (300 mM) were mixed in a 1.5 mL Eppendorf Safelock Biopur tube that contained 1 mL DMF and left to react for 2 hours at room temperature. This resulted in 3-(diethylamino) propionyl chloride which tags amine/hydroxy metabolites.

Standards:

Anhydrous pyridine (4 μL) was added to 250 metabolites standards (53 μM) in 1mL DMF. The reaction crude was stirred for 2 minutes at room temperature followed by the addition of 3-(diethylamino) propinoyl chloride (22 mM) to derivatize amines and hydroxyls. After 30 minutes, the first derivatization step was quenched by adding 71mM of N, N-Diethylethylenediamine, followed by the addition of 250 mM of 1:1 HATU/HOBt which was left to react for 70 minutes at room temperature. ^{12, 13, 21}

Biological samples:

Bovine aortic endothelial lysate (20 μL, about 66,000 cells, calculated cell numbers)¹³, human urine (10 μL), and Human serum (10 μL) were individually dried down using a vacuum centrifuge and reconstituted in 1 mL DMF. Each of the three biological samples were derivatized using our dual step derivatization described earlier and dried down for analysis.

2.4. Capillary column fabrication

Capillary (inner diameter of 50μm and outer diameter of 360 μm) was purchased from Polymicro Technologies (Phoenix, AZ, USA). Gemini 3 μm C18 stationary phase was purchased from Waters Corp (Milford, MA, USA). The capillary columns were fabricated and packed as previously described³³. Integrated frits were photopolymerized 7 cm from the end of a fused silica capillary. Nano emitter tips were pulled using a Sutter p-2000 laser-based micropipette puller, then etched with 49% hydrofluoric acid. The columns were packed in-house with Gemini 3 μm C18 stationary phase to a length of 17.5 cm after equilibration.

2.5. LC-MS method

A Thermo Vanquish LC and autosampler were connected to Q-Exactive orbitrap with a flow splitter to achieve nanoflow rates. Sample (0.0137 μL) was introduced onto the column using a pressure cell. LC flow to the split was 175 μL min⁻¹ with a split ratio of 1:2100 which produced a flow rate through the column of 125 nL min⁻¹. Mobile phase (mp) A was 5 mM ammonium carbonate buffer pH 8.5, mobile phase B was 100% acetonitrile, C was 0.1% formic acid in water, and mobile phase D was 0.1% formic acid in acetonitrile. The gradient for the methods starting with high pH buffer was as follows; 0 – 1 mins: 95% A / 5%B, 1 – 8 mins: 95% A / 5%B to 2% A / 98%B, 8–10: 2% A / 98%B then switch to low pH buffer; 10 – 20 mins: 95% C / 5% D to 2% C / 98% D, 20–25: 2% A / 98%B, 25 – 32 mins: 95% A / 5%B to re-equilibrate. A Q-Exactive orbitrap equipped with a nanospray flex ion source (Thermo Fisher Scientific, Waltham, MA) was used for all MS analysis at a spray voltage of +1.75 kV in positive ionization mode. The resolution was set to R_fwhm = 70K, AGC to 1e⁶, and capillary temperature to 200°C with a scan range of 100–600 m/z.

2.6. Data processing and statistical analysis

Thermo Scientific Xcaliber Qualbrowser was used to detect targeted peaks with a maximum mass error of 7 ppm. Peak retention times and signal to noise (S/N) were extracted for further analysis. R (version 4.0.5) was used to plot all peaks and extract signal intensities of each analyte based on the retention time and m/z within a 7 ppm mass error window. The R package ggVennDiagram{Gao, 2021 #226} was used to plot a Venn diagram representing the three biological samples. A Student’s t-test with a p-value < 0.05 indicating significance was used to plot bar charts to indicate changes in biological treatment.

3. Results and Discussion

3.1. Derivatization and capillary LC-MS

This study applied our previously developed dual-stage chemical derivatization (Figure 1B) to increase metabolite coverage to analyze 250 metabolites using reverse-phase capillary LC-MS. Our sequential derivatization yields stable tagged analytes which contain tertiary amines with ethyl side chains to increase electrospray efficiency. These ethyl side chains offer increased hydrophobicity to the tag when the tertiary amine is deprotonated. Previous work used mixed-mode ion exchange and capillary electrophoresis to separate 23 metabolite standards.¹³ To improve peak shape, loading capacity, and overall metabolite detection, reverse phase capillary LC-MS was explored. The derivatization improved chromatographic retention by rendering all tagged metabolites with a hydrophobic tag. Comparing the same 23 metabolites analyzed in our previous CE-MS work to the current capRP LC-ESI-MS system, the average peak width decreased from 9.6 ±2.0 sec to 6.2 ±2.2 sec (Supplemental T1).

Capillary RP LC-ESI-MS was used to analyze the derivatized metabolites within a 32-minute run. The pKas of the tertiary amines on the tags were ~7.8 and 8.2 respectively (determined by pH titration, data not shown). The use of ammonium carbonate, pH 8.5 allowed deprotonation and sample stacking on the C18 capillary column. This increased retention of most of the analytes away from the dead time. The four-phase gradient (Figure 2A) allowed 50 singly tagged metabolites to elute before the 10-minute mark (Figure 2B). The mobile phase was decreased to pH 3.2 at time of 10min which protonated the analytes and allowed for elution of multiply tagged analytes with improved peak shape between 10–25 minutes. Figure 2B shows the separation of 250 derivatized analyte standards at 5 μM (Supplemental T1) from metabolic pathways such as TCA, amino acids synthesis, catecholamine, carbohydrate, and neurotransmitter metabolism.

Figure 2.

(A) Four-phase gradient for stacking and eluting analytes to increase sample loading capacity on a RP capLC-MS. (B)Full Chromatogram of 250 tagged metabolite standards.

Comparison of tagged to untagged metabolites showed substantial benefits of tagging. Untagged metabolites have a diverse range of structure, polarity, and abundance which complicates analysis for RPLC. The 250 untagged metabolites had an average retention time of 2.62 min on a capRP LC-ESI-MS, which coelutes with salts that suppress ionization. The retention time was increased by 4-fold after derivatization. On average, the limit of detection was 66 nM and the average linearity was R²: 0.978. (Supplemental T1). Also, good reproducibility was observed with peak intensity RSDs of 7.48%, and peak retention time RSDs of 0.35%. The average S/N was improved by 44-fold compared to the untagged metabolites. Few of the tagged analytes (11%) showed poor peak shape and large widths (>20sec FWHM) which contribute to the increase in LOD (Supplemental T1).

3.2. Metabolites Identified in BAEC, human urine, and human serum.

To show the performance of this method for profiling complex samples, we investigated cell lysates, urine, and serum. Our sequential derivatization detected 107 metabolites from BAEC (from about 18 cells injected on column)¹³, 111 metabolites from human urine, and 153 from human serum based on retention time, exact mass, and MS/MS match from the derivatized standard library (Supplemental Table T1). Figures 3A, 3B and 3C show the full chromatogram metabolites from BAEC, human urine, and human serum respectively. Figure 3D details the number of shared metabolites detected across BAECs, human urine, and human serum. Pathways such as amino acetyl tRNA metabolism, TCA cycle and glyoxylate, and dicarboxylate were observed for overlap between the three biological samples. Arginine biosynthesis was observed for overlap between urine and serum. Ornithine metabolism was observed for overlap between BAEC and urine. Glycine, serine, and threonine metabolism and arginine/proline metabolism were observed for overlap between BAEC and serum. Histidine, proline, and tyrosine metabolism were observed for only BAEC, urine, and serum respectively. For select analytes, the low concentration of metabolites resulted in a low average S/N for BAECs and human urine compared to standards. Meanwhile, the high average S/N for serum is due to the high concentration and large number of metabolites present. Taken together, the derivatization and capRP LC-ESI-MS system showed good sensitivity and reproducibility for large metabolite coverage.

Figure 3.

Full Chromatogram of (A) 107 tagged metabolites targeted in Bovine aortic endothelial cells (BAECs). (B) 111 tagged metabolites targeted in normal urine. (C) 153 tagged metabolites targeted in human serum analysis. (D) Venn diagram of tagged metabolites targeted in BAECs, normal urine, and human serum analysis.

4. Biological treatment

Based on the strong performance of the analytical system for metabolite standards and biological matrices, we investigated the ability to quantify and discover metabolic changes induced by epinephrine administration to endothelial cells. The sympathetic nervous system is driven by adrenergic receptor activation which stimulates the body’s fight-or-flight response and is a key determinant of vascular homeostasis.³⁴ Activation of neurohumoral pathways of the adrenergic receptor by the agonist epinephrine activates eNOS signaling^35–37 and superoxide dismutase in endothelial cells.³⁸

BAECs were pre-treated with the antagonist propranolol to inhibit the beta-adrenergic receptor (Figure 4) as is typically done in short time course experiments. Propranolol was then replaced with epinephrine to activate the beta-adrenergic receptors as previously described.^{35, 39} Treatment showed 32 significant metabolite changes (Figure 5A). Of these metabolites, 28% decreased upon epinephrine treatment and 72% increased. Grouping these 32 metabolites into pathways showed that the most profound changes centered around energy metabolism and those amino acid pathways closely connected to glycolysis (Figure 5B).

Figure 4.

BAECs were cultured, treated with propranolol (10μM) in serum reduced media for 30 minutes, followed by epinephrine (1μM) in serum reduced media for 15 minutes. Metabolites were extracted, derivatized and analyzed using capRPLC-ESI(+)-MS.

Figure 5.

(A) Summary of 32 metabolites that show significant changes after treatment of BAECs with 10μM propranolol for 30 minutes followed by addition of 1μM epinephrine for 15 minutes, derivatized and analyzed using capRPLC-ESI(+)-MS. P-value 0.05. (B) Metabolic pathway analysis for some of the significantly changed metabolites after treatment (citrate/isocitrate not significantly changed). Glycine, serine, and threonine metabolism, pyruvate metabolism and TCA metabolism.

Energy metabolism, specifically the citric acid cycle TCA cycle and pyruvate generally showed increased levels. The decrease in lactate levels and a corresponding increase in pyruvate levels suggest a movement away from glycolytic reliance on energy and a shift towards the more efficient TCA respiration for increasing energy. The increase in the levels of pyruvate (a fuel source) suggests the activation of the TCA. The increase in TCA metabolites is consistent with the increased energy demands of the fight-or-flight response. In addition, endothelial cells are known readily convert glutamine/glutamate to α-ketoglutarate as an energy source. Glycine, serine, and threonine metabolism that connects to pyruvate metabolism were also observed.

5. Conclusion

Our work shows the use of sequential dual-stage chemical derivatization to tag multiple functional groups including endogenous primary amines, hydroxyls, carboxylic acids, and capillary RP LC-ESI-MS for expanded metabolite coverage. The boost in signal, S/N, and improved retention that our dual-proton affinity tags offers makes it possible to detect low concentration analytes, expanding metabolite coverage. We expanded the metabolite coverage from ~20 done previously by CE-MS to 250 using capillary RP LC-ESI-MS. This work illustrates that the tagging method is applicable across multiple complex biological matrices. This work shows the utility of using high sensitivity platforms to analyze a broad array of metabolites from ultra small sample amounts. We were able to detect 107 targeted metabolites from 18 cells. Future experiments that build on this work include using alternate separation platforms like sold phase concentration to increase the S/N and further improve coverage. In addition, single cell analysis is possible with this system depending on the overall size of the cell and its metabolite mass. These analyses could provide more insight into cell-to-cell heterogeneity and various metabolic pathways in diabetic neuropathy.