Discovery and Identification of Arsenolipids Using a Precursor-Finder Strategy and Data-Independent Mass Spectrometry.

Arsenolipids are a class of lipid-soluble arsenic species. They are present in seafoods and show high potentials of cytotoxicity and neurotoxicity. Hindered by traditional low-throughput analytical techniques, the characterization of arsenolipids is far from complete. Here, we report on a sensitive and high-throughput screening method for arsenolipids in krill oil, tuna fillets, hairtail heads, and kelp. We demonstrate the detection and identification of 23 arsenolipids, including novel arsenic-containing fatty acids (AsFAs), hydroxylated AsFAs, arsenic-containing hydrocarbons (AsHCs), hydroxylated AsHCs, thiolated trimethylarsinic acids, and arsenic-containing lysophosphatidylcholines not previously reported. The new method incorporated precursor ion scan (PIS) into data-independent acquisition. High-performance liquid chromatography (HPLC) electrospray ionization quadrupole time-of-flight mass spectrometry (ESI-qToF-MS) was used to perform the sequential window acquisition of all theoretical spectra (SWATH). Comprehensive HPLC-MS and MS/MS data were further processed using a fragment-guided chromatographic computational program Precursorfinder developed here. Precursorfinder achieved efficient peak-picking, retention time comparison, hierarchical clustering, and wavelet coherence calculations to assemble fragment features with their target precursors. The identification of arsenolipids was supported by coeluting the HPLC-MS peaks detected with the characteristic fragments of arsenolipids. Method validation using available arsenic standards and the successful identification of previously unknown arsenolipids in seafood samples demonstrated the applicability of the method for environmental research.

Introduction

Arsenolipids are lipid-soluble organic arsenic compounds.[1] In the past decade, about 70 arsenolipids have been discovered from algae, fish oil, and caviar (Supporting Information Table S1). Reported arsenolipids typically include phosphatidyl-arsenosugar (AsPL),[2] arsenic-containing fatty acids (AsFAs),[3] arsenic-containing hydrocarbons (AsHCs),[4] trimethylarsenio fatty alcohols (TMAsFOHs),[5] arsenic-containing phosphatidylcholines (AsPCs), and arsenic-containing phosphatidylethanolamines (AsPEs).[6] Recent studies have found that several arsenolipids elicited cytotoxicity comparable to that of arsenite in the human bladder and liver cells.[7] Arsenolipids also penetrated physiological barriers, accumulated in the brains of vertebrates, and exerted neurotoxicity on differentiated brain cells.[8−11] These newly recognized toxicities make it important to identity and quantify diverse arsenolipids in the environment. The chemical information of arsenolipids is particularly limited because of the difficulties in characterizing various arsenolipids in complex environmental matrices. Therefore, there is a need for a high-coverage method to identify potential arsenolipids.

Currently, identifying arsenolipids significantly relies on data-dependent acquisition (DDA) using mass spectrometry (MS). However, in DDA, only a fixed number of precursors with the most intense MS peaks are fragmented and analyzed. Arsenolipids with very low signal intensities are overlooked. Data-independent acquisition (DIA) is a tandem MS/MS method in which all the ions within a selected m/z range are fragmented, and all the produced fragments are analyzed. This full-coverage and nontargeted technique can overcome the bias in the selection of precursors.[12−18]

Most arsenolipids contain a dimethylarsinoyl group, As(O)(CH3)2, or a trimethylarsinyl group, As(CH3)3, giving rise to three characteristic fragment ions: As(CH3)2+, As(CH2)+, and As(CH3)2OH2+.[6,19] Using these characteristic arsenic fragments to conduct precursor ion scan (PIS) is a promising approach to screen arsenolipids. However, in a regular PIS, it is difficult to distinguish the target product ions from the ions of similar m/z because Q3 in triple quadrupole MS is a low mass resolution analyzer. In quadrupole time-of-flight (qToF) MS, although Q3 is replaced by ToF, the step size of Q1 is still 1 Dalton (Da). The accurate mass of the precursor cannot be determined. Therefore, for finding arsenolipids, especially those unknown identities, it is beneficial to collect both the MS and MS/MS information in high mass resolution so that the interference can be avoided and the parent ion of a specific product ion can be accurately measured.

This can be achieved by incorporating PIS into the sequential window acquisition of all theoretical mass spectra (SWATH). SWATH is a subcategory of DIA where the ions to be fragmented enter qToF MS within certain m/z windows (e.g., 25 Da).[20,21] Performing PIS in SWATH has the following unique advantages. First, this method finds precursors for fragments in a nontargeted and high-mass accuracy way. Second, SWATH has a fast scan capability because of its wide step increments, and it produces better MS/MS spectra than the full fragmentation method, thereby increasing the sensitivity of detection for both analytes and fragment ions at low concentrations.

The main challenge for developing such a method lies in the data analysis, owing to the lack of a linkage between precursors and product ions in the wide MS window in SWATH. Some software, such as DIA-Umpire[22] and MS-DIAL,[23] can deconvolute the complex MS/MS spectra in DIA. However, probably because these software use the precursor peaks as the models to group product ions, some product ions with low intensities are likely to be neglected. To facilitate high coverage of detecting and identifying arsenolipids, we report here a data-processing program Precursorfinder to link product ions with their precursors using product ions as the grouping models. Such a matching process is achieved by the extraction of the information of characteristic product ions and precursor ions from SWATH data, peak detection, and subsequent wavelet coherence calculations and hierarchical clustering. We created this “PIS in SWATH” strategy by combining SWATH with Precursorfinder. We demonstrated a successful untargeted screening of arsenolipids in krill oil, tuna fillets, heads ofTrichiurus haumela(hairtail fish), and kelp.

Materials and Methods

Reagents

Sodium arsenate (AsV, Sigma-Aldrich, St. Louis, MO), monomethylarsonic acid (MMA, Chem Service, West Chester, PA), 3-amino-4-hydroxyphenylarsonic acid (3AHPAA, Pfaltz and Bauer Inc., Waterbury, CT), N-acetyl-4-hydroxy-m-arsanilic acid (NAHAA, Pfaltz and Bauer Inc.), and 3-nitro-4-hydroxyphenylarsonic acid (Rox, Sigma-Aldrich) were used to prepare the arsenic standards solution. The concentrations of the arsenic species were determined by inductively coupled plasma mass spectrometry (ICP-MS) instrument Agilent 7500cs (Agilent Technologies, Germany). Hexane, dichloromethane (DCM), isopropanol, water, ethanol, and methanol were obtained from Thermo Fisher Scientific (Waltham, MA). Capsulated krill oil and tuna fish fillets were brought from local markets in Edmonton, Alberta, Canada. Hairtail and kelp were purchased from a local market in Chongqing, China.

Sample Preparations

Tuna fillets (0.5 g), hairtail head (0.5 g), or kelp (0.5 g) was freeze-dried and extracted twice using 6 mL of DCM and methanol mixture (2/1, v/v) in water-bath sonication for 1 h. Water was added to the combined extract to make the final ratio of DCM:methanol:water 2:1:1. The mixture was then sonicated in a water bath for another 1 h. The lipid layer of the mixture was evaporated to dryness under nitrogen and redissolved in 400 μL of ethanol/isopropanol (3/1, v/v). Krill oil (100 μL) was diluted four times in hexane. All the samples were filtered through a 0.45 μm membrane filter and subjected to high performance liquid chromatography (HPLC)–MS/MS analysis.

Analytical Conditions

A Shimazdu HPLC system (pump, LC-20 AD XR; autosampler, SIL-20A XR; Shimazdu, Columbia, MD) and a reverse-phase (RP) column (ODS-3, 100 mm × 2 mm, 3-μm particle size; Phenomenex, Torrance, CA) were used for the separation of arsenicals. LC was directly coupled to an AB Sciex TripleTOF X500R system (AB Sciex, Concord, Ontario, Canada). Mobile phase A consisted of 0.05% formic acid in water, and mobile phase B consisted of 0.05% formic acid in methanol. The flow rate was 0.12 mL/min.

When testing the arsenic standards, a 10-μL aliquot of arsenic standard solution (50 μg/L iAsV, MMA, 3AHPAA, NAHAA, and Rox) was separated using HPLC with gradient elution. The mobile phase B was kept at 2% for 4 min, linearly increased from 2 to 95% in the next 2 min, then kept at 95% for 6 min, and decreased to 2% in the next 1 min. The column was kept at 2% mobile phase B for 9 min for equilibration. The sample was analyzed with the data acquired using SWATH in the negative ionization mode. The parameters of the MS ion source were as follows: ion spray voltage, −4500 V; temperature, 550 °C; ion gas 1, 45 psi; ion gas 2, 30 psi; curtain gas, 25 psi. The MS parameters were as follows: mass range, m/z 100–300; collision energy (CE), 0 V; CE spread, 0 V; declustering potential (DP), −70 V; DP spread, 0 V; SWATH windows, m/z 100–150, 149–200, 250–300; and accumulation time, 200 ms. The MS/MS parameters were as follows: mass range, m/z 50–200; CE, −35 V; CE spread, 15 V; DP, −70 V; DP spread, 0 V; accumulation time, 100 ms.

When analyzing the arsenolipids in the diluted krill oil, 1-μL aliquot of the filtered sample was separated using HPLC with gradient elution. The mobile phase B was linearly increased from 10 to 85% in the first 20 min, kept at 85% for 15 min, decreased to 10% in the next 1 min, and kept at 10% for 9 min for equilibration. The SWATH was acquired in the positive ionization mode. The MS source parameters were as follows: ion spray voltage, 5000 V; temperature, 500 °C; ion gas 1, 45 psi; ion gas 2, 30 psi; and curtain gas, 25 psi. The MS parameters were as follows: CE, 0 V; CE spread, 0 V; DP, 80 V; and DP spread, 0 V. The MS/MS parameters were as follows: CE, 45 V; CE spread, 15 V; DP, 80 V; and DP spread, 0 V. SWATH windows were m/z 270–300, 299–325, 350–375, 374–400, 399–425, 424–450, and 485–500. SWATH covered m/z 270–500 for MS and m/z 90–125 for MS/MS. The accumulation time for MS and MS/MS was 140 and 70 ms, respectively.

A 10-μL aliquot of the lipid extract of tuna fillets was analyzed by gradient HPLC separation. The mobile phase B was linearly increased from 10 to 85% in the first 20 min, kept at 85% for 30 min, decreased to 10% in the next 1 min, and kept at 10% for 9 min for equilibration. Two sets of SWATH were acquired, with one covering the MS range of m/z 200–400 and the other covering the MS range of m/z 400–600. The size of the SWATH window was 25 Da for both acquisitions. MS/MS covered m/z 50–200. The accumulation time was 100 ms for both MS and MS/MS. Other parameters of MS were the same as those for krill oil.

A 10-μL aliquot of the lipid extract of hairtail heads or kelp was analyzed using the same chromatographic and MS conditions as those used for krill oil analysis, except that the SWATH window was 25 Da, and the mass range of MS and MS/MS was m/z 100–550.

For HPLC–ICPMS measurements, the chromatographic condition was identical to that of HPLC–ESIMS. Arsenic was monitored at m/z 75. Oxygen 20% in argon was used as the optional gas. Other instrument parameters of ICPMS are shown in SI Table S2. HPLC–ICPMS was used to compare the susceptibility to the erroneous identification of coeluted arsenolipids with our method.

Identification of Precursors in SWATH

The workflow of “PIS in SWATH” is shown in Figure .

Figure 1

Workflow of “PIS in SWATH”. The sample was extracted and subjected to HPLC–ESI qToF MS for SWATH whose data was processed by Precursorfinder. Precursorfinder first detects peaks in the XICs of characteristic product ions in each SWATH window and then compares the retention times of the product ions with those of the precursor ions. Next, hierarchical clustering and wavelet coherence calculations are used to select candidate precursor ions according to the peak shape similarity. Those candidates are finally subjected to PIS for confirmation and structure identification.

Peak Detection

First, the raw extracted ion chromatogram (XIC) of the product ions (mass window: ± 0.0005 Da) in each SWATH window was denoised by Daubechies external phase wavelet and smoothed twice by moving the average and median filter. The smoothing window size was roughly 0.3–0.4% of the number of sampling points and was self-adaptive to the signal length. A local maximum in the preprocessed XIC was identified as a peak. Then, the colocated peaks were identified as the intersection of the peaks of two characteristic product ions. The height of the colocated peaks should be greater than a threshold which can be adjusted by the user of Precursorfinder. Users can also specifically pick out the peaks of interest from the colocated peaks. The full width at half maximum (FWHM) of a peak was calculated as the width of the peak when it was attenuated to 0.5 on both sides.

Selection of Precursor Ions

Precursor ions were selected on the basis of their retention time proximity and peak shape similarity with the characteristic product ions. The retention time τp of every ion in MS1 was compared with the retention time τ0 of the characteristic product ion in each SWATH window. If |τp – τ0|≤ FWHMcharacteristic product ion, this ion was added to a pool of “possible precursor ions”. This step was to ensure that the retention times of the precursor ions selected were almost identical to those of the characteristic product ion. Then, we used hierarchical clustering and wavelet coherence calculations to find out candidate precursor ions from this pool.

Hierarchical Clustering

Hierarchical clustering is an unsupervised clustering method which uses the “distance” between two data sets for grouping. The distance here is defined as: 1-correlation of the two data sets within (τ0 ± FWHM). At a certain clustering level, the possible precursor ions that were in the same cluster with the characteristic product ion were selected as the candidate precursor ions.

Wavelet Coherence Calculations

Using continuous wavelet transformation (CWT) overcomes the alteration of the peak shape that may arise from the background noise and the overlapping of multiple peaks.[24] CWT converts the data into the scale (s) and translation (time, τ) domain:where x(t) is the original signal, and ψ* is the mother wavelet that is the Morlet wavelet here. For each possible precursor ion (Figure a), wavelet transformation was run to determine the scale (slc) and transformation (τlc) of its largest wavelet coefficient in the coefficient magnitude scalogram (Figure b). Its wavelet coherence with the XIC of a product ion is shown in a coefficient heat map (Figure c), where s is the largest scale. The coherence was calculated as the mean under the area [slc, smax] × [τlc – FWHM, τlc + FWHM] (red square, Figure c). If it was larger than a certain coherence level, this ion was selected as a candidate precursor ion.

Confirmation of Precursor Ions

Because a product ion might be linked to several coeluted candidate precursor ions, after getting the list of candidate precursor ions, the product ion scan on triple TOF MS was used to confirm which precursor ions can generate the characteristic product ions and to identify the structures of the precursors.

Precursorfinder

The data processing workflow was assembled in Precursorfinder. The source code was written in Matlab R2018a (The MathWorks, Inc., Natick, MA).

Results and Discussion

Arsenic Standards in Water

We first tested the precursor-finding ability of “PIS in SWATH” using five arsenic standards, iAsV, MMA, NAHAA, 3AHPAA, and Rox. Two product ions, AsO2– (m/z 106.9120) and AsO3– (m/z 122.9069), were used to trace these arsenicals by Precursorfinder after HPLC–SWATH. The precursors above a cluster level of 1 or a wavelet coherence of 0.7 with the two fragments are summarized in SI Table S3. In the SWATH window of m/z 100–150 (Figure S1A), the signal of AsO2– and AsO3– has two peaks. Using both hierarchical clustering and wavelet coherence calculations, MMA (m/z 138.939) and iAsV (m/z 140.919) were selected as the precursors. They had a high coherence with peak 1 (0.76) and peak 2 (0.88). In the window of m/z 200–250 (Figure S1A), 3AHPAA (m/z 231.961) was selected. In the window of m/z 250–300 (Figure S1A), NAHAA (m/z 273.972) and Rox (m/z 261.935) were selected. Their retention times also matched those of AsO2– and AsO3– (Figure S1B). It demonstrates that this method can successfully find out the five arsenicals.

Arsenolipids in Krill Oil

We then applied the “PIS in SWATH” strategy on krill oil and used characteristic fragments As(CH3)2+ (m/z 104.9685) and As(CH2)+ (m/z 102.9529) to screen arsenolipids.

There are 12 peaks in the XICs of As(CH3)2+, As(CH2)+, and As(CH3)2OH2+ (m/z 122.9791) in seven SWATH windows (Figure A). These peaks suggest the existence of arsenolipids in the krill oil. The candidate precursor ions were selected, and they are summarized in Table S4. We rounded up all the candidate precursor ions to one decimal point because Q1 had a step width of 1 Da (Table S5). Using product ion scan, the ions at m/z 294.221, 301.643, 363.188, 391.219, 437.202, 437.224, 449.203, and 499.180 were confirmed to generate As(CH3)2+ and As(CH2)+ (Figure B). The retention times of the two arsenic fragments are in good agreement with those in Figure A.

Figure 2

(A) XICs of As(CH2)2+, As(CH3)2+, and As(CH3)2OH2+ in SWATH windows obtained from the analysis of a krill oil sample. Twelve peaks were detected. (B) XICs of As(CH3)2+ and As(CH2)2+ in the PIS of the candidate precursor ions at m/z 294.221, 301.643, 363.188, 391.219, 437.202, 437.224 (fragmented together with 437.202), 449.203, and 499.180.

Among the eight precursors, AsFA362, AsFA390, AsFA436, and AsFA448 are the known arsenolipids (Table S1). Their identities were confirmed by their MS/MS spectra (Figure S2). Their elution order on the C18 column was in accordance with that found by Petursdottir et al.[25] and Amayo et al.,[6,19] who reported that AsFA362 eluted first, followed by AsFA436, AsFA390, and AsFA448.

The ions of m/z 294.221, 301.643, 437.224, and 499.180 are from potential new arsenolipids. Figure A shows the MS/MS spectrum of the ion at m/z 437.224. The accurate mass measurements of the molecular ion suggest its formula to be C20H41AsO5. In addition to As(CH3)2+ and As(CH2)2+, Figure A shows the presence of fragments [M + H – C3H6O2]+ (m/z 363.1855) and [M + H – C3H6O2 – H2O]+ (m/z 345.1750) with small mass errors compared to their theoretical mass values, indicating that two hydroxyl groups are on one or two of the 17 carbons proximate to the arsenic end. On the basis of these observations, we proposed this ion to be a derivative of AsFA and named it AsFA-OH436. This is the first time that a hydroxylated AsFA is reported.

Figure 3

MS/MS spectra and proposed structures of (A) AsFA-OH436, (B) the doubly charged species of AslysoPC601 at m/z 301.6434, (C) the singly charged species of AslysoPC601 at m/z 499.1801, and (D) AsFA-OH420.

The ions at m/z 301.643 and 499.180 had identical retention times (Figure B). After checking their isotopic patterns (Figure S3), we suspected that m/z 301.643 was the doubly charged analogue of the singly charged m/z 499.180. The inspection of the MS/MS spectra of their product ion spectra confirmed that they were from the same compound (Figure B, C), as three characteristic phosphatidylcholine fragments, C5H14NO+, C2H6O4P+, and C5H15NO4P+, and arsenic-containing fragments were observed in both spectra. This arsenolipid C25H53AsNO8P (a proposed structure in Figure B) is an arsenic-containing lysophosphatidylcholine (AslysoPC601) derived from phosphatidylcholine. The ion at m/z 499.180 was generated after C5H14ON+ was lost, which might occur during in-source fragmentation. Lysophospholipids are the degradation products of phospholipids and regulate critical biological functions and disease processes.[26] Some of them have strong surface activity that can collapse red blood cells and other cell membranes, causing hemolysis or cell necrosis. The potential toxicity of AslysoPC has not been fully studied.

Arsenolipids in Tuna Fillets

We then performed the same strategy on the lipid extract of tuna fillets. Seven out of 16 SWATH windows show the peaks of As(CH3)2+, As(CH2)+, and As(CH3)2OH2+ (Figure S4A). The candidate precursor ions were selected and are summarized in Tables S6 and S7. We found that nine precursors m/z 333.214, 363.183, 359.221, 391.210, 405.214, 419.242, 421.329, 449.205, and 529.267 could generate As(CH3)2+ and As(CH2)+ in the product ion scan (Figure S4B), and their retention times were in good agreement with those shown in Figure S4A.

AsFA362, AsFA390, AsFA418, AsFA448, AsFA528, AsHC404, AsHC332, and AsHC358 are known arsenolipids (MS/MS spectra in Figure S5). They have been previously detected in algae, capelin fish, and caplin oil (Table S1). The ion at m/z 421.190 is novel. This compound might have been overlooked before because of its low concentration and its similar chromatographic behavior to that of a known arsenolipid AsFA362 (Figure S4B). Its molecular ion C19H37AsO5+ and fragment ions C19H36AsO4+, C19H34AsO3+, and C17H36AsO2+ suggest that it is an AsFA-OH species (Figure D). On the basis of these fragments, the two hydroxyl groups are probably at the end of its structure distant from arsenic. Although AsFA-OH420 and AsFA390 have the same length of the carbon chain, AsFA-OH420 eluted earlier than AsFA390 from the chromatographic column because AsFA-OH420 has two more hydrophilic hydroxyl groups (Figure S4B).

Taleshi et al. have identified AsHC404 and AsHC332 in tuna fillets.[27] In addition to the two arsenolipids, we detected AsFA362, AsFA390, AsFA418, AsFA448, AsFA528, and AsHC358. We found AsHC332 as the predominant arsenolipid. Our results are consistent with the findings of Stiboller et al.[28] who reported AsHC332 as the major species and AsFA362 as a minor constituent of the total arsenic in the muscle of tuna.

Stiboller et al.[28] found that the intensity of the molecular ion of AsFA362 was below the threshold for being selected in a DDA MS/MS experiment. This highlights the usefulness of our method for a better analysis of some arsenolipids with low intensity.

Arsenolipids in Hairtail Heads

There are 16 peaks in the XICs of As(CH3)2+, As(CH2)+, and As(CH3)2OH2+ in seven SWATH windows (Figure S6A). The candidate precursor ions are summarized in Tables S8 and S9. The ions at m/z 152.971, 333.213, 361.245, 363.186, 391.219, 405.213, 407.229, 409.245, 419.250, 425.203, 431.250, 441.235, 447.279, 449.203, and 529.266 were confirmed to generate As(CH2)2+ and As(CH3)2+ (Figure S6B), whose retention times are in good agreement with those in shown in Figure S6A. Among them, AsHC332, AsHC360, AsFA362, AsFA390, AsFA404, AsFA418, AsFA446, AsFA448, and AsFA529 have been reported (Table S1). Their MS/MS spectra are shown in Figure S7-S9.

The ions with m/z 152.971, 407.229, 409.245, 425.203, 431.250, and 441.235 are potential new arsenolipids. The accurate protonated molecular mass suggests that the ion at m/z 152.971 is a thiolated trimethylarsinic acid (TMA) C3H9AsS (Figure A). It did not generate As(CH3)2OH2+ but generated As(CH2)2+ and As(CH3)2+ (Figure A), which matched with the absence of As(CH3)2OH2+ and the presence of As(CH2)2+ and As(CH3)2+ in the SWATH window m/z 149–175 (Figure S6A). These results demonstrated that there was no oxygen but sulfur in its structure. This is the first time a thiolated TMA is reported. Thiolated TMA152 might be produced through the thiolation of TMA, a metabolite of inorganic arsenic, or through the degradation of other thiolated arsenolipids.

Figure 4

MS/MS spectra and proposed structures of (A) thiolated TMA152, (B) AsHC406, (C) AsHC408, (D) AsFA424, ( E) AsFA430, (F) AsFA440, and (G) AsHC-OH498. The positions of the double bonds on the proposed structures have not been determined.

By complementing the information of molecular and fragment ions with accurate mass measurements, the ions at m/z 407.229 and 409.245 were suggested to be AsHC406 (C23H39AsO, Figure B) and AsHC408 (C23H41AsO, Figure C), respectively. The ions at m/z 425.203, 431.250, and 441.235 were AsFA424 (C22H37AsO3,Figure D), AsFA430 (C22H43AsO3,Figure E), and AsFA440 (C23H41AsO3,Figure F), respectively, according to the presence of [M + H]+, [M + H – H2O]+, As(CH2)2+, and As(CH3)2+ in their MS/MS spectra.

Arsenolipids in Kelp

Four out of 18 SWATH windows showed the peaks of As(CH3)2+, As(CH2)2+, and As(CH3)2OH2+ (Figure S10A). The candidate precursor ions are summarized in Table S10 and Figure S11. Only m/z 361.243, 423.187, 425.203, and 499.240 generated As(CH3)2+ and As(CH2)+ in the product ion scan (Figure S10B) whose retention times matched those shown in Figure S10A.

Except AsFA360 and AsFA422 (MS/MS spectra in Figure S11), the other two precursors have not been reported previously. AsFA424 (Figure D) was also detected in the extract of hairtail heads. Figure G shows [M + H]+, [M + H – H2O]+, and [M + H – C3H8O3]+ in the MS/MS spectrum of the ion at m/z 499.240. This ion was a hydroxylated AsHC (C25H43AsO5, AsHC-OH498) with four more hydroxyl groups compared to the regular AsHCs. The fragment [M + H – C3H8O3]+ (m/z 407.1941) indicated that there are three hydroxyl groups at the end of its structure, as proposed in Figure G.

In the present work, we detected 23 arsenolipids in four kinds of seafood. We successfully identified ten new arsenolipids, including AsFA-OH436, AsFA-OH420, AslysoPC601, thiolated TMA152, AsFA424, AsFA430, AsFA440, AsHC406, AsHC408, and AsHC-OH498. The structures of AsFA-OH, AsHC-OH, AslysoPC, and thiolated TMA are reported for the first time. The identification of the previously unknown arsenolipids points to a research need for synthesizing these arsenolipids to be used for toxicological studies. The discovery of these new arsenolipids and future research into their chemical and toxicological properties will help us gain a better understanding of the health benefits and potential risks involved in the consumption of these seafood.

The new “PIS in SWATH” can overcome the problem of the erroneous identification of arsenolipids with a similar molecular mass. One example is that AsFA-OH436 was found at m/z 437.2231 (Figure A), which was very close to the m/z value of the commonly found AsFA436 (theoretical m/z 437.2037, Figure S2C). If we use regular PIS, the low mass resolution of Q1 would be unable to differentiate them. As in Figure B, the ions of m/z 437.2231 and 437.2037 fragmented together.

“PIS in SWATH” can also overcome the problem of the erroneous identification of somewhat overlapped or coeluted arsenolipids. Figure A shows that HPLC–ICPMS was not able to differentiate the coeluting arsenolipids species. AsFA436, AsFA448, and AsFA390 were almost coeluted at 24–25 min. Nevertheless, MS1 in SWATH clearly provided the molecular detection of individual arsenolipids (Figure B).

Figure 5

Analyses of krill oil samples using reverse-phase HPLC–ICP–MS and HPLC–ESI–MS. (A) HPLC separation followed by the detection of As at m/z 75 using ICP-MS. (B) Overlaid signal of AsFA362 (m/z 363.19), AsFA390 (m/z 391.22), AsFA436 (m/z 437.20), AsFA448 (m/z 449.20), AsFA-OH436 (m/z 437.22), and AslysoPC601 (m/z 499.18) detected using ESI–MS after HPLC separation.

“PIS in SWATH” can help reduce the instrument analysis time. In regular PIS on qToF MS, one sample needs to be analyzed repeatedly with different precursor selections to achieve full MS coverage. SWATH can finish the analysis in one injection.

Precautions should be taken toward two potential issues that could affect the detection and identification of new arsenolipids. The first issue relates to the detection of new arsenolipids that may not contain a dimethyl or trimethyl arsenical moiety. Our current “PIS in SWATH” technique scans for arsenolipids that produce As(CH2)2+ and As(CH3)2+. All the arsenolipids we have reported here and those reported in the literature contain an As(CH3) moiety (n = 1–3). However, novel arsenolipids not containing an As(CH3)n moiety could be missed. The second issue relates to the potentially unstable arsenolipids, which may hydrolyze under extraction or chromatography conditions. Therefore, the actual number of arsenolipid compounds present in marine organisms could be much higher than those shown in this study and those previously reported. Further research is needed to identify and quantify other arsenolipids. The strategy described here can be refined and extended to the characterization of diverse arsenic species present in the environment and biological systems.