Ghost Peaks of Aromatic Metabolites Induced by Corona Discharge Artifacts in LC‐ESI‐MS/MS

ABSTRACT

LC‐ESI‐MS/MS is a preferred method for detecting and identifying metabolites, including those that are unpredictable from the genome, especially in basal metazoans like Cnidaria, which diverged earlier than bilaterians and whose metabolism is poorly understood. However, the unexpected appearance of a “ghost peak” for dopamine, which exhibited the same m/z value and MS/MS product ion spectrum during an analysis of Nematostella vectensis , a model cnidarian, complicated its accurate identification. Understanding the mechanism by which “ghost peaks” appear is crucial to accurately identify the monoamine repertoire in early animals so as to avoid misassignments. Verification experiments showed that in‐source oxidation of tyramine, which produced an intense signal, was responsible for this “ghost peak.” This artifact commonly occurs among aromatic compounds with high signal intensities and appears at the same m/z as their respective in vivo oxidized metabolites. In metabolomics, spectra contain diverse signals from complex biological mixtures, making it difficult to recognize artifact peaks. To prevent misassignments, despite +16 Da differences, adequate chromatographic separation of metabolites from their respective in vivo oxidation precursors is necessary. Whereas both electrolysis and gas‐phase corona discharge can cause in‐source oxidation in ESI, corona discharge proved to be the dominant factor. Additionally, the presence of multiple oxygen atom sources was suggested by the voltage‐dependent mass shift of +16 Da to +18 Da of the “ghost peak” when using ¹⁸O‐labeled water as a solvent. Accurate metabolite identification using LC‐ESI‐MS/MS requires accounting for in‐source products that can mimic in vivo products.

1. Introduction

Liquid chromatography–electrospray–tandem mass spectrometry (LC‐ESI‐MS/MS) enhances the exploration of endogenous molecular repertoires from small samples of biological extracts with high molecular specificity [1]. Advances in modern mass spectrometry instruments have enabled high‐precision measurement of m/z values and the quantification of samples with dynamic ranges of signals over three to four orders of magnitude. However, optimization of the ESI source cannot address all components in a mixture. Suboptimal source conditions can easily generate unwanted ions, including non‐covalent clusters, for example, [nM + H]⁺, in‐source fragments, and incorporation of small adducts, for example, [M + HCOOH + H]⁺. These unwanted ions are easily detectable and can be prevented with sample dilution or electrospray parameter optimization when the sample is a pure substance; however, such ions are often overlooked amidst diverse signals when the sample is a complex mixture. During the present LC‐ESI‐MS/MS of the cnidarian model, Nematostella vectensis , a cnidarian species, a “ghost peak” for dopamine appeared, exhibiting the same m/z value and MS/MS product ions as dopamine.

Cnidaria, the sister group to all bilaterians, has a nervous system rich in neuropeptides [2]. However, the presence of other chemical neurotransmitters has been unclear due to inconsistent detection for decades using various techniques [3, 4]. Monoamines, such as serotonin and dopamine, were occasionally reported in cnidarians during the 1980s and 1990s using conventional immunohistochemistry [5, 6] and HPLC‐ED (electrochemical detection) [7, 8]. Recent advances in genome informatics, however, have shown that cnidarians lack homologous genes for bilaterian enzymes involved in the biosynthesis of serotonin and dopamine [3, 9]. Consequently, identification of these molecules in previous reports has been questioned due to their poor structural specificity. Thus, we employed LC‐ESI‐MS/MS for N. vectensis to accurately detect cnidarian monoamines; however, the substantial signal from this “ghost peak” interfered with the unambiguous identification of the relatively weak dopamine signal observed in the samples. Identifying the “ghost peak” and understanding the mechanism of its appearance are essential to avoid misassignments and to ensure accurate identification of dopamine distribution in biological tissues with imaging mass spectrometry with MALDI or desorption‐ESI, excluding the possibile presence of isobaric endogenous metabolites.

Verification experiments revealed that the “ghost peak” was due to in‐source oxidation of tyramine, resulting in a mass increase of +16 Da. Artifact ions, similar to the “ghost peak,” were identified as arising from aromatic molecules with the same m/z values as in vivo mono‐oxidized products.

In‐source oxidation in ESI has been understood as electrolysis of sample compounds at the metal electrode in solution, as demonstrated by Van Berkel et al. based on the ionization mechanism [10, 11]. Covalent additions of oxygen to protein peptides [12, 13], lactate [14], terpendole [15], and particularly catechol [16] during ESI are considered artifactitious aqueous electrochemical oxidation. Countermeasures, such as employing nonmetal capillary or bipolar counter electrodes [17, 18], as well as physical insulation using a microdialysis membrane at the tubing junction [19] have been suggested to reduce sample electrolysis. On the other hand, gas‐phase corona discharge has also been proposed as another source of analyte oxidation during ESI, as demonstrated in a study using protein model peptides [20, 21, 22]. Corona discharge occurred and oxidized protein peptides under a conventional ESI source voltage setting with 3.5 kV nebulized nitrogen gas, even without visible discharge sparking [20]. In our study, contributions of electrolysis and corona discharge oxidation to the appearance of “ghost peaks” were investigated by comparing dominant ion‐source conditions at varying concentrations and electrical settings. Simultaneously, the origin of the oxygen atoms involved in oxidation of molecules was investigated using H₂¹⁸O as the solvent.

2. Experimental Procedures

2.1. Reagents

LC‐MS‐grade ultrapure water and methanol, pure grade acetic acid, and chloroform were purchased from Fujifilm Wako and used in all extractions. 1 mM formic acid (for amino acid sequence analysis, FUJIFILM Wako, Osaka, Japan), 10 mM ammonium formate (FUJIFILM Wako, Osaka, Japan), and acetonitrile (Optima LC/MS grade, Fisher Chemical, Waltham, MA, USA) were also used. The following standard aqueous solutions were prepared: dopamine HCl from LKT Laboratories Inc. (St. Paul, MN, USA); [3‐(2‐aminoethyl)‐1,3‐dihydro‐2H‐indol‐2‐one] (or 3‐(2‐aminoethyl)indolin‐2‐one) hydrochloride (CAS#4993‐84‐4) from Chembridge Co. (CA, USA); serotonin HCl from Cosmo‐Bio Co. (Tokyo, Japan); and tryptamine HCl from Combi‐Blocks (San Diego, CA, USA). L‐tyrosine and L‐tryptophan were from FUJIFILM Wako (Osaka, Japan). Additional materials included H₂ ¹⁸O (98%‐¹⁸O), purchased from Rotem Industries Ltd. (Arava, Israel); ¹³C₆‐ring‐labeled phenylalanine, purchased from Tront Research Chemicals (Ontario, Canada); and 1,1,2,2‐d ₄‐tyramine HCl, obtained from Cambridge Isotope Laboratories Inc. (MA, USA).

2.2. Animal Sample Preparations

Laboratory‐cultured specimens of N. vectensis were subjected to extraction and mass spectrometry. Additionally, mouse brain samples were supplied for secondary use by the Animal Section at the Okinawa Institute of Science and Technology.

2.3. Sample Extraction

Adult polyps of N. vectensis (30–60 polyps) and mouse brain samples were lyophilized using an Eyela, FDS‐1000 (Tokyo Rikakikai Co. Ltd., Tokyo, Japan). Freeze‐dried samples (20 mg) were milled into powder. To each sample, 2500 pmol of ¹³C₆‐ring‐labeled phenylalanine (approximately 25 μM in the final sample solution) were added as an internal standard. Undissolved and nonpolar fractions in the water–methanol mixture were eliminated using a modified Bligh–Dyer extraction method [23]. For liquid extraction, the following procedure was used: (1) 260 μL of 0.1%(v/v) acetic acid, 670 μL of methanol, and 340 μL of chloroform were added to milled samples in glass tubes. (2) Mixtures were shaken for 1 min with a vortex mixer (Vortex Genie 2, Scientific Industry Inc., Bohemia NY, USA). (3) Ultrasonication (AS ONE, Osaka, Japan) was performed in an ice bath for 10 min.

Under acidic aqueous conditions, amines are usually extracted into the aqueous methanol fraction. Following the initial extraction steps described above, the procedure continued as follows. Three hundred thirty microliters of chloroform was added to the mixture, followed by 1 min of shaking and ultrasonication in an ice bath for 5 min. Then, 330 μL of water was added, followed by another 1 min of shaking and ultrasonication in an ice bath for 5 min. This step was repeated once more. The aqueous phase was stored at −20°C for 1 h in the dark to allow precipitation. Then, the supernatant was lyophilized to remove the solvent and stored at −20°C until LC‐ESI‐MS/MS analyses. The methanol–water soluble fraction was dissolved in 50–80 μL of water containing 20 μM caffeine, which was added as an internal standard to monitor ionization among samples. SI‐labeled d ₄‐tyramine was spiked into powdered samples. This was done to monitor oxidation products during the extraction procedure, thereby distinguishing them from in vivo oxidation products.

2.4. Standard LC‐ESI‐MS/MS for Metabolomics

Online liquid chromatography was performed with a Waters M‐Class UPLC system (Waters, Milford, MA, USA) using a hybrid column combining hydrophilic and hydrophobic separations with cation‐exchange, anion‐exchange, and ODS‐coated substrates (SW‐C18 011, 300 × 1 mm, 1.6 μm, Scherzo, Kyoto, Japan). The gradient was as follows: 0–8 min 99% Solvent A (0.1% aqueous formic acid), 1% Solvent B (16.4 mM ammonium formate in 77% acetonitrile: %B was linearly increased to 15% from 8 to 25 min, 98% at 45 min, and then held for 50 min. Then it dropped to 1% at 50.5 min at a constant flow rate of 50 μL/min in a 60‐min run). The column oven was set at 40°C, and the autosampler was set at 11°C. For single, standard substance measurements without column separation, an isocratic solvent composition of 25% B was applied. This LC method allowed effective separation and analysis of compounds in complex samples, maintaining robust performance and reproducibility over the course of 60 min.

Mass spectrometry was conducted for m/z 50–700 with 60 000 mass resolution in both positive and negative ion modes with an ESI source mounted on Q‐Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). The analyzer was set to auto‐gain control (AGC) target 3e⁶. The maximum ion trap (IT) was 100 ms. The tolerance of m/z to estimate molecular formulae was set to ±0.0005 Da. A stainless steel capillary (OPTION‐53010, Fisher Scientific, Bremen, Germany) was employed. The capillary temperature was 320°C. Spray voltage was set to 3.0 kV at the optimum setting and changed from 2.0 to 4.5 kV in 0.5 kV steps during experiments. Data reduction was conducted using Xcalibur Software (Thermo Scientific, Bremen, Germany) and MZmine 4.1.0 [24].

Collision‐induced dissociation (CID) MS/MS (MS2) data acquisition was conducted from a fixed m/z 50 with 60 000 mass resolution of profile scans in both positive and negative ion modes in a data‐independent manner. The mass isolation mass window was 1 Da. Collision energy was set to the average at 10, 20, and 45 eV, representing the average as 25 eV. Energy‐resolved (ER)‐CID‐MS/MS was conducted for m/z 177.102 (protonated serotonin), m/z 161.107 (protonated tryptamine), m/z 154.086 (protonated dopamine), and m/z 138.092 from experimental collision energy settings (ce) 10 eV to 26 eV with 2 eV steps.

2.5. Corona Discharge Experiments With Tyramine

Data were acquired in positive and negative ion modes without nitrogen gas flow using well‐established atmospheric pressure corona discharge ionization with high (needle voltage at 3.2 kV), medium (2.5 kV), and low (2.0 kV) electric field settings [25, 26] in a Q Exactive Focus quadrupole‐orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). Twenty microliters of a tyramine‐methanol solution at concentrations ranging from 20 to 2000 μM was dropped onto the heater, maintained at approximately 250°C, beneath the corona probe, resulting in the introduction of tyramine vapor into the corona discharge source in open air. The net loading amount on the heater was estimated to be 0.40 nmol in positive ion mode and 0.43 and 43 nmol in negative ion mode examinations.

2.5.1. Offline ESI Experiments With Tyramine

ESI‐MS of 10 μM and 50 μM tyramine in 0.1% aqueous formic acid was performed using a nonmetallic emitter and an Exactive Orbitrap mass spectrometer, (Thermo Fisher Scientific, Bremen, Germany). Disposable polypropylene gel loading tips (tip i.d. ~0.1 mm, Eppendorf, Hamburg, Germany) were used as electrospray emitters and liquid reservoirs. A platinum wire (diameter: 0.2 mm, Nilaco, Tokyo, Japan) was inserted into each gel loading tip to make electric contact with sample solutions [27]. Measurements were performed with the spray voltage set to 3.0 kV.

3. Results and Discussion

3.1. Dopamine Ghost Peaks Originated From In‐Source Oxidation of Tyramine

Standard LC‐ESI‐MS/MS parameters were optimized to detect as many chemical components as possible with high structural specificity in a single analytical run switching positive and negative ion modes. Molecules were detected, typically from glycine at 3 min to indole acetic acid at 40 min, with multiply charged oligopeptides > MW1000 eluting later, at 40–59 min. Signal fluctuation in LC‐ESI‐MS/MS, monitored using doped caffeine (20 μM) as an internal standard, was 7.2% on average, ranging from 2.2% to 8.2% during 12‐fold sample analyses. Recovery of ¹³C₆‐ring phenylalanine spiked into dried samples prior to the extraction ranged from 51.1% to 127%. Fluctuations in recoveries were due to uneven removal of insoluble precipitates among the samples. Ion enhancement or suppression observed by ¹³C₆‐ring phenylalanine recovery primarily caused by the matrix effect of coeluted molecules including endogenous phenylalanine, because the recovery was 98.9% for the blank sample, which lacked a biological matrix. Selected ion monitoring (SIM) with the mass selection window setting of 0.4 Da (at narrowest) enhances signal‐to‐noise ratios of targeted m/z values; however, we conducted a full mass scan to monitor any qualitative changes in ion mass caused by unexpected ion fragmentation or adduct formations that appeared at unpredicted m/z. Then, mass extracted chromatograms were described for ions. Though m/z values of biological samples are displayed to three decimal places to distinguish them from other diverse signals, m/z values for standards are presented as nominal for simplicity. The detected repertoire of metabolites and their quantities in N. vectensis will be discussed in subsequent biological studies in terms of their biological production and functions. However, this paper highlights the artifactitious ions observed in ESI.

A dopamine “ghost peak” at retention time (RT) 22 min appeared in the mass chromatogram of N. vectensis at m/z 154.086, along with protonated dopamine at RT 17 min (Figure 1A). On the other hand, the “ghost peak” was absent in mouse brain, where dopamine was detected. Identical CID‐MS/MS product ion spectra at m/z 154.086 were observed between the “ghost peak” and protonated dopamine standard (Figure S1A,B). Simultaneous detection of a peak at m/z 152.071 in negative ion mode RT 22 min with weak intensity (Figure S2) indicated that the “ghost peak” at m/z 154.086 is indeed a protonated form. Several possibilities were explored to identify the origin of the dopamine “ghost peak.” The standard, 3,4‐dihydroxytyramine, a structural analogue and isomer of dopamine, was detected at the same m/z as dopamine; however, it proved distinct from the “ghost peak” based on its MS/MS spectrum. Artificial peak splitting in LC due to inappropriate elution conditions with the hybrid column, which combines hydrophilic and ionic separation conditions, was rejected, because spiking authentic dopamine enlarged the dopamine peak, but not the “ghost peak.” Another possibility was detection of in‐source fragments of derivatized dopamine at RT 22 min. Dopamine derivatives such as sulfo‐conjugates [28], glucuronides [29], and phospho‐conjugates [30] have been reported in vertebrates. These derivatives may lose their labile conjugate groups during ionization after LC separation, resulting in detection of an ion with the same structure as dopamine, but at a different RT. However, satellite ions typical of derivatized forms, such as [dopamine ester conjugate + Na or NH₄]⁺, were absent. Similarly, negative ions corresponding to [dopamine ester conjugate −H]⁻ were not detected, even though neutral loss is suppressed in deprotonated conjugates compared to their protonated forms. Last, the RT of the dopamine “ghost peak,” which was identical to that of protonated tyramine at m/z 138.091, suggested that the “ghost peak” originated from in‐source oxidation (+16 Da) of tyramine, which was strongly detected in N. vectensis (Figure 1B). This was consistent with the absence of the “ghost peak” in the mouse brain sample, which lacks tyramine [31]. The +16 Da peak was obscured among diverse signals typical of complex biological mixtures at the elution time of tyramine at RT 22 min, and the contribution of tyramine at m/z 138.091 was not visible in the mass chromatogram of dopamine at m/z 154.086 (Figure 1A). The peak area of the dopamine “ghost peak” (tyramine +16 Da) was only 0.013 times that of tyramine. This slight conversion of tyramine into tyramine +16 Da had a negligible impact on the determination of tyramine. However, the isobaric interference of the +16 Da product at m/z 154.086 is critical for accurate estimation of a relatively weak dopamine signal in N. vectensis . Despite the mass difference of +16 Da, coelution or insufficient separation of tyramine and dopamine in the LC condition would cause serious overestimates or false positives of dopamine.

FIGURE 1.

Comparative mass chromatograms detecting (A) protonated dopamine at m/z 154.086 and (B) protonated tyramine at m/z 138.091. Black, green, and orange lines indicate standards, Nematostella vectensis , and mouse brain, respectively. Standard concentrations were 3 μM for dopamine (RT 17) and 30 μM for tyramine (RT 22).

The “ghost peak” problem is compounded by the fact that some ion masses of in‐source mono‐oxidation products can be identical to those of expected in vivo metabolites, leading to spurious assignments. Expanding the in‐source oxidation scheme to other aromatic metabolites, for example, in‐source oxidation of precursor “P(138)” produces a +16 Da product “P + O” that is detected at the same m/z (154) as the in vivo mono‐oxidation product “M(154)” (Figure 2). The molecular structure of dopamine is derived from a combination of 3‐hydroxylation and decarboxylation of tyrosine. According to identical MS/MS product ion spectra between the dopamine “ghost peak” and dopamine, in‐source oxidation of tyramine also inserts oxygen at the phenol 3‐position, similar to in vivo oxidation (hydroxylation). Tyramine is usually a trace metabolite in nervous systems of vertebrates in a well‐established pathway in which primal hydroxylation is followed by decarboxylation [31]. Therefore, there is generally no need to be concerned about the dopamine “ghost peak” caused by small amount of tyramine in mouse brain measurements. However, when analyzing a critical target in a novel biological sample, such as dopamine in an extract of N. vectensis , in which the content of compounds including tyramine is unknown, it is important to consider the possibility of artificial signals derived from extremely intense signals. To avoid misassignment in dopamine measurements, sufficient LC separation from tyramine is necessary, especially in certain insects [32, 33], microorganisms used in fermentation [34, 35], and plants [36, 37] that accumulate tyramine among metabolites derived from tyrosine.

FIGURE 2.

A schematic table presenting combinations of potential precursors “P” and their corresponding in vivo product “M” with isobaric in‐source oxidized products “P + O” for tyrosine, tyramine, and tryptophan. Numbers and formulae in parentheses represent detected masses as protonated forms in positive‐ion mode.

By analogy with the “ghost peak” for dopamine, similar in‐source oxidations were investigated for other aromatic molecules. Indeed, P (tyrosine) + O at m/z 198.076 at RT 15 min appeared in the mass chromatogram in both N. vectensis and mouse brain along with M (L‐dopa) at RT 10 min (Figure S3). The resemblance of MS/MS product ion spectra indicated structural similarity between the in‐source oxidized P (tyrosine) + O and M (L‐dopa) (Figure S1D–F). It is consistent that the p‐hydroxyphenyl structure (an analog of tyramine) is oxidized into a 3,4‐dihydroxy structure in the ESI source, as well as in vivo oxidation (Figure 2). In mouse brain, in vivo synthesis of serotonin is initiated by 5‐hydroxylation of tryptophan, catalyzed by tryptophan hydroxylase (EC 1.14.16.4) [31]. In‐source products P (tryptophan) + O also appeared in the mass chromatogram at m/z 221.0921 in both N. vectensis and mouse brain at the same RT as P (tryptophan) at m/z 205.097 (RT 32 min) (Figure S4), though M (5‐hydroxytryptophan) was only observed in mouse brain at RT 24 min. The MS/MS product ion spectrum at m/z 221.0921 of P (tryptophan) + O, RT 32 min, did not match that of the 5‐hydroxytryptophan standard. Instead, it resembled the 2‐hydroxytryptophan standard (Figure S1E–G). Therefore, the in‐source oxidized product of tryptophan was distinguished from in vivo 5‐hydroxytryptophan detected in mouse brain by their ion structures, although their m/z values are identical.

Serotonin (5‐hydroxytryptamine) was detected only in mouse brain at RT 31 min in the mass chromatogram at m/z177.1025 (Figure S5A). Instead, three peaks at RT 29, 33, and 37 min were found in N. vectensis . Based on the same retention time as tryptamine at m/z 161.107 (Figure S5B), the peak at RT 37 min was identified as P (tryptamine) + O. Though the signal intensity of the peak at RT 37 min was insufficient to obtain an MS/MS spectrum, 2‐hydroxytryptamine was the most likely candidate as an analog of in‐source 2‐hydroxytryptophan for P (tryptophan) + O (Figure S1E–G). The MS/MS product ion spectra of the two peaks at RT 29 min and RT 33 min detected in N. vectensis were quite similar, with only slight differences compared to that of serotonin (Figure S1H–J). However, a few product ions observed in single CID‐MS/MS spectra were insufficient to characterize their ion structures, so further characterization with LC‐ESI‐MS/MS was conducted.

3.2. Characterization of Detected Ions at m/z 177.102 From N. vectensis

Based on the preferential loss of NH₃ (−17 Da) in MS/MS spectra, we speculated that ions at RT 29 min and RT 33 min might be structural analogs of serotonin, such as positional isomers of the hydroxyl group (Figure S1H–J). To verify their ion structures, breakdown curves obtained by energy‐resolved MS/MS (CID) [38] were compared with authentic standards of positional isomers: 6‐, 3‐, 7‐, and 4‐hydroxytryptamine, along with 3‐(2‐aminoethyl)‐1,3‐dihydro‐2H‐indol‐2‐one abbreviated as 3‐(2‐aminoethyl)indolin‐2‐one, a tautomer of 2‐hydroxytryptamine (Figure S6). Given the strong resemblance of the breakdown curves between m/z 177 in mouse brain and serotonin (Figure S6A,H), the ion structures at RT 29 min and RT 33 min of N. vectensis were identical to that of 3‐(2‐aminoethyl)indolin‐2‐one. The same LC retention times as 3‐(2‐aminoethyl)indolin‐2‐one was observed for the RT 33 min fraction (Figure S7). On the other hand, the aqueous structure of the minor RT 29 min ion in N. vectensis was still unclear due to lack of standards matching the retention time. In‐source fragments of fragile derivatives of 3‐(2‐aminoethyl)indolin‐2‐one were candidates for this fraction; however, no corresponding precursor signals were found. As analogue structures of 3‐(2‐aminoethyl)indolin‐2‐one, several isomeric forms including 2‐hydroxyindole structure have been proposed as derivatives of endogenous 5‐methoxy N‐acetylated tryptamine (melatonin) in mammalian metabolites [39, 40, 41] (Figure S8). Semiempirical calculations indicated that one of the tricyclic structures is relatively stable and irreversible (Figure S8C) [40]. The gas phase ion structure of the in‐source oxidation product of tryptamine, which is considered an oxidized ion at the 2‐position, is involved in these structural variants (Figure S8) and is indistinguishable from signals at m/z 177.102 observed in N. vectensis .

3.3. Smaller Contribution of Electrolysis to In‐Source Oxidation

Electrospray involves electrolytic processing of molecules in addition to protic solvents [10]. In‐source covalent oxygen additions have been reported for peptides [13] and other small molecules, such as indole analogs [15], phosphatidylethanolamine (PE) [42], lactate [14] and catechol [16, 17]. Voltametric studies have shown that several amino acids having relatively lower oxidation potentials such as Cys, Met, Trp, Tyr, and His are easily electrolyzed [43, 44, 45, 46, 47, 48, 49]. Therefore, in‐source oxidation of phenolic and indole structures in ESI is not surprising. The observed structure of P (tyramine) + O with an added oxygen at the phenol 3‐position of aqueous para‐ or 4‐substituents was consistent with the proposed electrolytic conversion of tyramine in anodic voltammetry (Figure S9A) [43]. The product structure becomes dopamine, and further anodic oxidation of catechol (3,4‐dihydroxy structure) can lead to oligomerization, which is similar to that of in vivo melanin formation from L‐dopa [50, 51, 52] (Figure S9A). Furthermore, anodic oxygen addition to the indole 2‐position was also consistent with the observation of P (tryptophan) + O structure with MS/MS experiments (Figure S9B) [44]. Concerns about the abiotic oxidation of the aromatic molecules during sample preparation prior to LC‐ESI‐MS/MS were resolved by the absence of d ₄‐dopamine derived from spiked d ₄‐tyramine after sample collection (Figure S10).

The appearance of the dopamine “ghost peak” is obviously caused by ESI in‐source reaction of tyramine; however, simple electrolysis did not account for the “ghost peak” in negative ion mode, in which electrons are supplied to the sample from the cathode (Figure S2). Corona discharge is another side effect in ESI, forming local plasma, activating surrounding gases near the electrode [20, 53, 54, 55]. To identify the dominant factor in appearance of dopamine “ghost peak” in a simple system, we compared the extreme conditions of the source settings for corona discharge and electrolysis using a standard solution instead of a complex biological mixture. The dependence of P + O appearance upon the P concentration was explored for tyramine and tryptamine (Figure S11). Almost constant P (tyramine) + O to P area ratios ranging from 0.013 to 0.019 (average of 0.016) were observed in a concentration range from 10 to 100 μM with an onset emerging P + O at 10 μM (Figure S11A,C). The value of the ratio was consistent with that observed in N. vectensis samples (0.013). In contrast to tyramine, P (tryptamine) + O was not detected in that concentration range, even though the signal intensity of P (tryptamine) was comparable to that of P (tyramine) (Figure S11B,D). This indicates that tryptamine is less likely to be oxidized in‐source than tyramine; however, P (tryptamine) + O appeared in the N. vectensis sample, where the tryptamine signal was extremely high. The ratio of P (tryptamine) + O to P (tryptamine) observed in the sample was 0.004, lower than 0.016 for the tyramine case. Preferential in‐source oxidation of tyramine to tryptamine was contrary to the expected aqueous oxidation based on their oxidation potentials. The electric potentials of tyramine and tryptamine, representing p‐substituted phenol and 3‐substituted indole, were estimated to be nearly identical, with E _p  = +0.65 V in the 25 μM, pH 7 (with 0.1 M phosphate buffer) shown in differential pulse voltammetry at a glassy carbon electrode [43, 44]. Otherwise, slightly lower redox potentials in aqueous solution E ₀ for tryptamine indole (1.08 V [46] and 0.7–0.8 V [47]) than those of phenol (1.5 V [46] and 0.71–1.4 V [48]) were estimated by thermodynamic calculations.

This contradiction indicates that aqueous electrolysis was not predominantly responsible for in‐source oxidation causing the “ghost peak.” Next, dependency of P + O appearance upon the spray voltage (kV) was examined for settings ranging from 2.0 to 4.0 kV, injecting 1 μL of 1 mM tyramine and tryptamine solutions into the ESI. In contrast to the gradual increase of signals of P (tyramine) at m/z 138, a steep increase of P (tyramine) + O at m/z 154 appeared at spray voltages higher than 3.0 kV (Figure 3A). The same trend was also observed for P (tryptamine) + O at m/z 177 (Figure 3B). The steep increase in appearance of P + O in response to an increase in the spray voltage with an onset at 3.0 kV also indicated that P + O was not formed by stoichiometric charge balancing in electrolysis. This was further supported by an absence of P + O using an offline ESI with a polypropylene gel‐loading tip having an inserted Pt wire (Figure S12). The setup achieved pure electrolysis, preventing the corona discharge effect [27]. The consistent absence of P (tyramine) + O signals at m/z 154 by infusing tyrosine solutions at 10 and 50 μM concentrations for 1 min demonstrated that the ESI mechanism did not oxidize tyramine in solution. Based on the negligible contribution of electrolysis to in‐source oxidation, the influence of corona discharge was suspected. To evaluate the actual contribution of the gas‐phase corona discharge effect on in‐source oxidation, tyramine vapor was introduced into the corona discharge source by dropping tyramine methanol solution onto the heater beneath the corona probe. The net loading amount was 0.40 nmol in positive ion mode (Figure 4A) and 0.43 and 43 nmol in negative ion mode examinations (Figure 4B,C). Up to 2% of P (tyramine) + O to P (tyramine) signal effectively appeared at needle voltages higher than +3.5 kV with 0.40 nmol tyramine, accompanied by a trace signal of P (tyramine) + 2O (Figure 4A). Furthermore, a lower needle voltage of −2.0 kV, creating a low electric field, was sufficient to achieve 4.2% of P (tyramine) + O relative to P (tyramine) in negative ion mode, with a similar loading amount of P (tyramine) at 0.43 nmol as in the positive ion mode (Figure 4B). The higher electron emissions from metal needles in negative ion mode likely enhance reactive oxygen species (ROS) formation compared to positive ion mode [56, 57]. Although increasing the loaded amount of tyramine by a hundredfold to 43 nmol increased the total signal intensity in negative ion mode, the proportion of P (tyramine) + O relative to P (tyramine) did not change significantly, ranging from 3.9% at −2.0 kV to 19% at −3.2 kV (Figure 4C), compared to 4.2% at −2.0 kV to 24% at −3.2 kV for 0.43 nmol (Figure 4B).

FIGURE 3.

Absolute signal intensities corresponding to peak areas versus spray voltages. (A) P (tyramine) at m/z 138 and P (tyramine) + O (left of the vertical scale) at m/z 154 (right of the vertical scale), similarly, (B) P (tryptamine) at m/z 161 and P (tryptamine) + O at m/z 177.

FIGURE 4.

Stacked bar charts of absolute intensities of ions observed with corona discharge ion source by changing needle voltages for (A) 0.40 nmol tyramine in positive ion mode, (B) 0.43 nmol tyramine in negative ion mode, and (C) 43 nmol tyramine in negative ion mode. Gray, black, and blue indicate P (tyramine), P (tyramine) + O, and P (tyramine) + 2O, respectively.

3.4. Origin of Oxygen

Corona discharge oxidation of a sample molecule involves multiple steps in gas phase, such as formation of discharged electrons and radicals, subsequent ROS formation, and ROS attacks on molecules. Unlike aqueous electrolysis, the relative feasibility of oxidation among molecular structures is poorly predicted by their estimated redox characteristics alone, as mentioned above. In‐source oxidation of hemoglobin caused by corona discharge has been confirmed even under standard ESI settings [20]. In that report, solvent water (such as H₂O + e⁻ → H· + ·OH + e⁻) was thought to be the dominant source of ROS, rather than molecular oxygen from air. The lowest oxidation potential of water and its abundance in the system were thought to provide primary ROS in the gas phase via the discharge process [20, 56]. On the other hand, atmospheric oxygen rather acted as an electron scavenger, reducing the onset voltage of protein oxidation in the 3–4 kV source range [20]. However, our corona discharge experiments showed that P (tyramine) + O was formed without solvent water (Figure 4). In order to confirm the origin of oxygen forming P + O in the ESI, tyramine dissolved in ¹⁸O (98%) water and acetonitrile (85:15) was introduced into the ESI source by infusion injection. At the standard ESI setting at 3.0 kV spray voltage, only P (tyramine) + ¹⁶O was detected (Figure 5A). A comparable level of P (tyramine) + ¹⁸O to P (tyramine) + ¹⁶O signals appeared when the source voltage was increased to 4.5 kV (Figure 5B). This observation contradicts the previous prediction that solvent water was the major ROS source at an ESI voltage around 3.0 kV [20]. Potential oxygen sources to form P + O changed, depending on the spray voltage settings, indicating a complex process of oxygen addition to tyramine mediated by corona discharge. Analyte molecules also become potential ROS suppliers. Ions suspected to be dopamine quinone structure were also detected at the same retention time of P (tyramine) + O as another “ghost peak,” though the intensity was very weak (Figure S13). The quinone structure is a reversible intermediate to form P (tyramine) + O (Figure S9A), acting as both a scavenger and a supplier of electrons in the system. This intermediate formation could explain preferential in‐source P + O formation of tyramine to tryptamine. Chemical properties of gas‐phase molecules may also be responsible for the complex in‐source oxidation regulated by corona discharge.

FIGURE 5.

Mass chromatograms of protonated P (tyramine) at m/z 138, protonated P (tyramine) + ¹⁶O at m/z 154, and protonated P (tyramine) + ¹⁸O at m/z 156 (156.0910) for 100 μM tyramine dissolved in an 85:15 mixture of ¹⁸O (98%) water and acetonitrile in infusion injection. Spray voltage (A) at 3.0 kV and (B) at 4.5 kV. Flows were repeatedly switched ON and off. NL indicates normalized level in the relative intensity.

3.5. Possible Countermeasures Reducing the Unwanted Appearance of P + O

Mono‐oxidation frequently occurs in in vivo metabolism, regardless of whether it is enzymatically or nonenzymatically mediated. Despite their biological significance, predicting the appearance of P + O, which is isobaric with in vivo M, is difficult both qualitatively and quantitatively due to the complexity of factors related to the corona discharge oxidation effect, which is influenced not only by source conditions but also by sample characteristics. When comparing metabolites between groups, which may differ in chemical composition, it is worth focusing on signals strongly detected in that specific sample as potential precursors for forming suspicious P + O. In fact, identification of the dopamine “ghost signal” detected in N. vectensis was facilitated by comparison with the stronger signal of tyramine compared to that of mouse brain, where the dopamine ghost signal was absent (Figure 1).

Although several strategies have been proposed to reduce unwanted corona discharge oxidation in ESI, no suitable applications have been achieved in LC‐ESI‐MS/MS metabolomics so far. Reducing the spray voltage or sample concentration can suppress in‐source oxidation but also reduces intensities of analyte peaks, thereby diminishing detectability of diverse metabolites. Even under typical ESI operating conditions at atmospheric pressure, dielectric breakdown occurs at the terminus of the metal spray needle [13, 20]. According to Paschen's law, which describes factors influencing corona discharge, a high gas pressure setting reduces the threshold voltage for dielectric breakdown by reducing the mean free path of electrons, gases, and ions [58, 59, 60, 61]. However, introducing a high‐pressure source to LC‐ESI‐MS/MS also requires adjustment of other settings, such as the desolvation temperature and the fore‐vacuum system. An aged surface, caused by metal etching or accumulated insoluble materials, promotes discharge [53, 61, 62]. Although renewing the electrode made the “ghost peak” disappear (Figure S14), oxidized products formed by corona discharge reappear as the needle ages after repeated analyses, necessitating unrealistically frequent needle replacement. During initial stages of analyzing challenging metabolites in novel biological samples, it is both straightforward and practical to ensure effective LC separation of metabolites derived from in vivo mono‐oxidation or hydroxylation from their respective precursors. This simple approach helps avoid misidentification of molecular species during data analysis.

4. Conclusions

We demonstrate that corona discharge causes unexpected in‐source molecular oxidation as a side effect in both positive and negative ion ESI, especially for aromatic molecules. In‐source oxidation is regulated by multiple factors and is poorly predicted in any given analytical condition. Furthermore, artifacts are difficult to recognize because they are usually buried among diverse ions in metabolomics. To avoid misidentification of in vivo products by appearance of “ghost peaks,” here we highlight ensuring sufficient LC separations of metabolites from their respective in vivo oxidation precursors, despite +16 Da differences.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1 Comparative CID‐MS/MS spectra obtained in positive‐ion mode at a 25 eV collision energy setting, averaged 10, 20, and 45 eV, for standards and observed peak signals in biological sample analyses. Spectra were described for peaks with sufficient signal intensity to obtain substantial product ions, regardless of the samples. A) RT 21 min “ghost peak” and B) dopamine standard at m/z 154.086, C) RT 15 min “ghost peak” and D) L‐dopa standard at m/z 198.076, E) RT 32 min “ghost peak”, F) 2‐ hydroxytryptophan, and G) 5‐hydroxytryptophan standard at m/z 221.092, H) RT 29 min and I) RT 33 min of N.vectensis, and J) serotonin standard at m/z 177.102. RTs correspond to the peaks observed in each mass chromatogram at m/z values.

Figure S2. Comparative mass chromatograms at A) m/z 152.071 corresponding to deprotonated dopamine and at B) m/z 136.075 corresponding to deprotonated tyramine in negativeion mode. Black, green, and orange lines indicated standards, N. vectensis , and mouse brain, respectively. Standard concentrations were 3 μM for dopamine(A), and 30 μM for tyramine(B).

Figure S3. Comparative mass chromatograms at A) m/z 198.076 corresponding to protonated tyrosine, and at B) m/z 182.081 to protonated L‐dopa. Black, green, and orange lines indicate standard, N. vectensis , and mouse brain, respectively. Standard concentrations were 3 μM for L‐dopa(A), and 35 μM for tyrosine(B).

Figure S4. Comparative mass chromatograms at A) m/z 221.092 corresponding to protonated 5‐hydroxytryptophan, and at B) m/z 205.097 to protonated tryptophan. Black, green, and orange lines indicate standard, N. vectensis , and mouse brain, respectively. Standard concentrations were 0.6 μM for 5‐hydroxytryptophan(A) and 31 μM for tryptophan(B).

Figure S5. Comparative mass chromatograms at A) m/z 177.103 corresponding to protonated serotonin, and at B) m/z 161.107 to protonated tryptamine. Black, green, and orange lines indicate standard, N. vectensis , and mouse brain, respectively. Standard concentrations were 6 μM for serotonin(A) and 6 μM for tryptamine(B).

Figure S6. Comparative breakdown curves of MS/MS at m/z 177 obtained for Energy‐resolved CID product ion spectra for hydroxyl position isomers of tryptamine standards for experimental collision energy settings (ce) 10–40 eV with 2 eV steps. A) serotonin (5‐hydroxytryptamine), B) 6‐hydroxytryptamine, C) 3‐(2‐aminoethyl)‐1,3‐dihydro‐ 2H‐indol‐2‐one, D) 7‐hydroxytryptamine, E) 4‐hydroxytryptamine, F) RT 33 min and G) RT 29 min detected in N. vectensis , and H) RT 31 min in mouse brain. Abundant product ion peaks were picked to describe the curves.

Figure S7. Comparison of retention times in mass chromatograms at m/z 177. LC conditions differed from those in standard metabolomics to separate hydroxyl group isomers. A) N. vectensis , (*) corresponding to the peak at RT 33 min in the metabolomics setting system. B) serotonin, C)6‐hydroxytryptamine, D) 3‐(2‐aminoethyl)‐1,3‐dihydro‐2Hindol‐ 2‐one, E) 7‐hydroxytryptamine, F) 4‐hydroxytryptamine.

Figure S8. Possible molecular structures of the ion detected at m/z 177 RT 29 min having the same gas phase ion structure as that of RT 33 min, assigned to 3‐(2‐aminoethyl)‐1,3‐dihydro‐2H‐indol‐2‐one (2‐aminoethyl)indolin‐2‐one). A)2‐hydroxytryptamine, B) epoxy form, C) cyclic 3‐hydroxymelatonin, and D) cyclic 2‐hydroxymelatonin tautomerism to form the indolinone derivative.

Figure S9. Proposed electrolytic oxidation in voltammetry. A) Aqueous tyramine or p‐substituted phenol releases one electron and proton at the anode followed by a nucleophilic attack of water to form dopamine. B) Aqueous tryptamine or 3‐substituted indole with oxygen added to the C2 position.

Figure S10. Mass chromatograms of d4‐tyramine added to the dried sample and the products formed in N. vectensis before extraction, A) at m/z 138.091 for tyramine, B) at m/z 142.117* for d4‐tyramine, C) at m/z 154.086 for dopamine, and D) at expected m/z 158.112* for d4‐dopamine.

Figure S11. Logarithmic scale for peak areas vs standard concentrations in 1 nM‐100 μM observed in ESI. A) P (tyramine) at m/z 138 and P (tyramine) + O at m/z 154, B) P (tryptamine) at m/z 161 and absence of P (tryptamine) + O at m/z 177. C) Mass spectra of 100 μM standard for tyramine and D) tryptamine. The peak label threshold was 1% and * indicates signals corresponding to background.

Figure S12. Mass chromatogram at A) m/z 138 corresponding to10 and 50 μM tyramine standards and B) m/z154 corresponding to expecting P (tyramine) + O during 1‐min analyses obtained using nano ESI with a gel loader tip having an inserted Pt‐wire electrode (Spray voltage was set to 3.0 kV). C) Mass spectra described at 0.7 min for 10 μM tyramine standards involving tyramine at m/z 138 but no “ghost signal” at m/z 154.

Figure S13. Mass chromatograms of A) dopamine along with the “ghost peak” at m/z 154, B) dopamine quinone form at m/z 152, and C) tyramine at m/z 138 for tyramine observed for the mixture of 100 μM tyramine and 5 μM dopamine. Retention times were shifted from the standard setting exhibited in Figure 1, due to the slight differences in LC settings.

Figure S14. Comparative mass chromatograms of an extract of N. vectensis (green) and standard (black) acquired with a brand new electrode. A) m/z 154.086 and B) m/z 138.091 in positive‐ion mode. Standard concentrations were 3 μM for dopamine (A) and 30 μM for tyramine (B).

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.