Development and application of a forensic toxicological library for identification of 56 natural toxic substances by liquid chromatography–quadrupole time-of-flight mass spectrometry

Target natural toxic substances were selected on the basis of previously reported poisoning cases as follows: 31 plant toxins (coniine, lycorine, galantamine, atropine, picrotoxinin, scopolamine, picrotin, strychnine, colchicine, veratramine, cyclopamine, jervine, amygdalin, aconine, cymarin, convallatoxin, cucurbitacin E, oleandrin, benzoylmesaconine, benzoylaconine, tubocurarine, hypaconitine, mesaconitine, 14-anisoylaconine, aconitine, jesaconitine, digitoxin, digoxin, α-chaconine, α-solanine, and dioscin), 7 mushroom toxins (muscimol, ibotenic acid, muscarine, phalloidin, γ-amanitin, α-amanitin, and β-amanitin), 4 mycotoxins (aflatoxin B1, aflatoxin B2, aflatoxin G1 and aflatoxin G2), 5 marine toxins (domoic acid, tetrodotoxin, okadaic acid, dinophysistoxin-1, and brevetoxin b), and 5 frog venoms (bufotenine, resibufogenin, bufalin, cinobufagin, and batrachotoxin). In addition to the natural toxins, berberine, cinchonidine, diosgenin, and quinine were selected as target substances, which are considered to be important materials of herbal medicines. Scopolamine, aflatoxin B1, aflatoxin B2, tetrodotoxin, quinine, aflatoxin G1, aflatoxin G2, resibufogenin, bufalin, colchicine, cyclopamine, diosgenin, cinobufagin, amygdalin, benzoylmesaconine, tubocurarine, mesaconitine, 14-anisoylaconine, jesaconitine, okadaic acid, and dinophysistoxin-1 were purchased from Fujifilm Wako Pure Chemical (Osaka, Japan); muscimol, coniine, ibotenic acid, muscarine, bufotenine, cinchonidine, strychnine, cymarin, convallatoxin, cucurbitacin E, oleandrin, digitoxin, digoxin, and α-solanine from Sigma-Aldrich (St. Louis, MO, USA); picrotoxinin, picrotin, domoic acid, γ-amanitin, and β-amanitin from Abcam Biochemicals (Cambridge, UK); berberine, aconine, and benzoylaconine from Cayman Chemical (Ann Arbor, MI, USA); galantamine, atropine, and jervine from Tokyo Chemical Industry (Tokyo, Japan); α-chaconine and dioscin from Extrasynthese (Lyon, France); batrachotoxin and aconitine from Latoxan (Valence, France); phalloidin and α-amanitin from Merck Millipore (Billerica, MA, USA); lycorine, hypaconitine, brevetoxin b, and veratramine from Enzo Life Sciences (New York, NY, USA); Kishida Chemical (Osaka, Japan), LKT Laboratories (St. Paul, MN, USA), and Toronto Research Chemicals (Toronto, Canada), respectively. The stock solutions of all substances were prepared at a concentration of 10–1000 μg/mL. Muscarine, lycorine, cinchonidine, scopolamine, tetrodotoxin, quinine, berberine, resibufogenin, cinobufagin, amygdalin, hypaconitine, and α-amanitin were dissolved in distilled water (DW). Aconitine and benzoylaconine were dissolved in acetonitrile. Other toxins were dissolved in methanol solution. Stock solutions were stored at − 80 °C until analysis. Methanol, acetonitrile and DW of the HPLC grade were purchased from Kanto Chemical (Tokyo, Japan). Other common chemicals used were of the highest purity commercially available. Human whole blood was obtained from Tennessee Blood Services (Memphis, TN, USA).

Sciex Triple TOF 5600 mass spectrometer (Sciex, Framingham, MA, USA) and Shimadzu NexeraX2 LC system (Shimadzu Co., Kyoto, Japan) were used for analysis. The column used for chromatographic separation was L-column ODS (150 × 1.5 mm i.d., particle size 5.0 μm; Chemicals Evaluation and Research Institute, Sugito, Saitama, Japan). The column temperature was maintained at 40 °C, and the gradient system was used with a mobile phase (A) 10 mM ammonium formate in 5% methanol aqueous solution and (B) 10 mM ammonium formate in 95% methanol solution. Linear gradient elution was started from 100% A to 100% B over 15 min. The 100% B was held for 5 min. It was then returned to 100% A over 10 min for the next run. The autosampler was maintained at 4 °C and the injection volume was 10 µL. Electrospray ionization was used in both positive and negative modes. The optimal MS parameters were declustering potential at 80 V and information dependent acquisition (IDA) criteria set at over 50 cps. The LC–QTOF-MS system allowed the acquisition of highly sensitive full scan MS spectra with high resolution and mass accuracy. In addition, IDA can be used to collect MS/MS spectra for compound identification based on MS/MS library searching.

This LC–QTOF-MS (/MS) method had several advantages for accurate detection of natural toxic substances. For instance, the mass spectrometer used in this study, triple TOF 5600, had high throughput which enabled very fast MS/MS acquisition rates at as low as 20 ms accumulation time in IDA mode. To fully leverage the instrument speed and obtain the best depth of coverage, the IDA workflow was optimized such that software overhead is minimized. The IDA method consisted of a high-resolution TOF-MS survey scan could follow up to 50 MS/MS ions. The combined use of high-resolution MS and IDA were extremely effective for the simultaneous detection of natural toxic substances in forensic samples. The instrument gave the resolution of 35,000.

Data acquisition and processing were performed by Analyst software and Peak View incorporated with the XIC Manager application (Sciex). The XIC Manager can be used for targeted processing of high-resolution MS and MS/MS data allowing for screening and identification with the highest confidence based on retention time (RT), mass error of molecular ion, isotopic pattern, and automatic MS/MS library searching.

All target substances were analyzed to investigate their retention properties, isotopic ratios and high-resolution MS/MS spectra obtained by collision induced dissociation (CID) with the injected amount of each compound of 0.1 µg. The four spectra were acquired at the collision energy (CE) at 20, 35, and 50 eV in single enhanced product ion (EPI) mode together with collision energy spread (CES) mode, in which the CE ramp range was set to 35 ± 15 eV. The CES parameter, in conjunction with the CE, determined the collision energy applied to the precursor ion in a product ion scan. The CE is ramped from low to high energies. The selection ranges of the precursor ion and RT of each compound for acquiring the library search were 20 mDa and 4.0 min, respectively. Compound identification was based on chromatographic and mass spectrometric information, including RT error, mass error, isotope matching, and library search results. The product ion for library search could be chosen from four spectra by CID energies of (±) 20, 35, 50, and 35 ± 15 eV, automatically.

To determine the limits of detection (LODs), 5 plots with different concentrations of each substance spiked into blank blood plasma were used. The LODs were defined as the concentrations giving a signal-to-ratio of 3:1. The recovery rates were calculated by the ratio of peak area obtained from a target substance spiked into ante-extraction matrix to that obtained from the substance spiked into post-extraction matrix.

A 45-year-old male with groan was found at his home. He was taken to hospital by an ambulance, but died shortly afterward. Beside the body in the room, there were dried roots of an aconite plant. Femoral vein and right and left heart blood samples were collected at autopsy performed in our laboratory and stored at − 80 °C until analysis. The blood samples were treated and analyzed in the same way as spiked samples described above.

Registered data consisted of 56 natural toxic substances with compound name, source, formula, exact mass, polarity, exact mass of precursor ion, ion form, RT, LOD, and recovery rate (Table 1). Extracted ion chromatograms of simultaneous determination of 56 substances are shown in Fig. 1. Product ion spectra of all substances were obtained by four different CE settings (see supplementary material Fig. S1). As an example the four spectra obtained for tetrodotoxin, which is one of the important toxins in food poisoning cases in Japan, are shown in Fig. 2. The [M + H]⁺ (m/z 320.1088) was the most abundant ion at 20 eV (Fig. 2a), while it remarkably decreased at 35 eV (Fig. 2b) and became a very small peak at 50 eV (Fig. 2c), and the number and intensity of fragment ions increased instead (Fig. 2a–c). Product ion spectra obtained by CES mode showed both [M + H]⁺ and fragment ions (Fig. 2d). CES mode can collect an average MS spectrum of different CE values in one single EPI scan, resulting in a full scan spectrum with both molecular and fragment ion information that can be used in library search-based identification with increased confidence. Digoxin, α-solanine, α-chaconine, digitoxin and dioscin provided [M + HCOO]⁻, and amygdalin and cucurbitacin E showed [M + NH₄]⁺ instead of [M + H]⁺ (Table 1, supplementary material Fig. S1, nos. 36, 46, 47, 51, 57, 7, and 45, respectively); therefore, it is necessary to pay attention to this phenomenon.

Several precursor ions accompany unknown ions with strange mass defects. For example, the m/z of [M + H]⁺ of strychnine (compounds no. 20) is 335.1754, but 6 ions were observed in the range of 335–337 (336.2 could be the isotopic ion) (supplementary material Fig. S1, no. 20). This is also observed in the spectra of berberine (no. 21), aflatoxin G1 (no. 27), colchicine (no. 32), veratramine (no. 33), jervine (no. 34), aflatoxin B1 (no. 37), cyclopamine (no. 39), and Aconitum alkaloids (nos. 9, 23, 26, 29, 38, and 40–42). Unfortunately, the sources of the ions are still unknown. In future, it is necessary to analyze the assignment of these ions.

With respect to the chromatographic separation, water/methanol both containing 10 mM ammonium formate was used. The used conditions provided RTs ranging from 1.9 min to 24.3 min (Table 1). Although some hydrophilic substances like ibotenic acid, musimol, and muscarine eluted quickly, it had no problems to detect them by high-resolution MS analysis.

The present library was applied for the identification of lycorine and domoic acid spiked into blank blood plasma. Figure 3 shows the results processed using the automatic extracted ion chromatograms (XICs) and spectra by TOF-MS and TOF-MS/MS. In the XICs on the left panels of Fig. 3a and b, the peaks corresponded to the target analytes. In the TOF-MS spectra on the middle panels, the measured masses were at m/z 288.1233 for lycorine and 312.1443 for domoic acid, which matched the theoretical masses with errors of 0.9 and 0.5 ppm, respectively. In the TOF-MS/MS spectra on the right panels, the masses of the fragment ions agreed very well with those of the registered MS/MS spectra. The purity scores were 79.1 and 67.2%, respectively.

Femoral vein and right and left heart blood samples collected from a 45-year-old male at the forensic autopsy performed in our laboratory were analyzed using the present library. In all samples, aconitine, jesaconitine, hypaconitine, and mesaconitine were identified, and Fig. 4 shows representative results from the femoral vein sample. In the previous study, Niitsu et al. [22] reported the four poisons as major substances detected from blood in the cases of suicide by aconite poisoning. In the TOF-MS spectra, the measured masses matched the registered masses with mass errors of 0.5–1.0 ppm. In the TOF-MS/MS spectra, the masses of the fragment ions agree very well with the registered MS/MS spectra. The purity score was 54.6–60.3%.