Rapid determination of Aconitum alkaloids in adulterated spice powders

Holly Lee¹, Ping Jiang¹, Jana Kinar² and Bryn Shurmer^2
1SCIEX, Canada; ²Centre for Veterinary Drug Residues, Canadian Food Inspection Agency (CFIA), Canada

Abstract

Traditional herbal medicine derived from Aconitum-based plants has been widely implicated in food poisoning incidents due to the presence of highly toxic alkaloids such as aconitine (AC). Screening for adulterated foods from contamination or improper processing requires rapid and accurate analytical methods. This technical note describes a fast extraction coupled with an LC-MS/MS method for simultaneously determining 6 Aconitum alkaloids in adulterated spice powders. The sensitivity of the QTRAP 6500+ system enabled a simple dilution approach with minimal matrix effects observed, while still achieving sub-to-low ng/g method detection limits (MDLs) of 0.55–1.5 ng/g. However, toxicological screening typically occurs only when AC poisoning is suspected. In the absence of symptoms or accounts from the victim, a method capable of providing orthogonal information would be highly useful to identify the culprit toxin. Data-dependent acquisition of enhanced product ion (EPI) scans on the QTRAP system produced MS/MS spectra to compare diagnostic fragments against reference databases. Here, quantitative and qualitative analysis of Aconitum alkaloids is simultaneously demonstrated on a single instrument platform. 

Key benefits for the analysis of Aconitum alkaloids using the QTRAP 6500+ system
 

• Rapid sample preparation and low matrix load: QTRAP 6500+ sensitivity enabled a large dilution factor and 1 µL injection volume, which reduced matrix effects to <15% and eliminated sample clean-up
  
  
• Excellent MDL performance: Recoveries of ±30% and precision of %CV <15% were achieved in sand ginger at the MDLs
  
  
• Easy method development: Automated optimization workflows in SCIEX OS software enabled users to develop and optimize method parameters without manual intervention.
  
  
• Proficiency samples showed accurate results: AC concentrations in the proficiency samples were consistent with reference values.
  
  
• Complementary compound confirmation: Positive compound identification was supported by MS/MS matching against reference spectra.

Introduction

Diester-diterpenoid alkaloids (DDAs) such as AC are the active pharmaceutical components in Aconitum or aconite herbs, which are used for their anti-inflammatory, analgesic and cardiotonic effects.¹ However, without proper processing to transform these DDAs into the less toxic monodiester alkaloids (MDAs), oral consumption can lead to poisoning where the minimum lethal dose range has been reported to be as low as 2–6 mg in humans.² Inadequate processing, contamination and erroneous substitution contributed to aconitum poisoning cases previously reported in Hong Kong, Taiwan, China and Korea.¹ AC was recently identified as the toxin responsible for 2 food poisoning incidents in Ontario and British Columbia, Canada.³ In both cases, the victims fell ill after consuming sand ginger, which was later discovered to be adulterated with aconite. Due to the severity of the symptoms and the need for a quick turnaround time, the CFIA developed a rapid procedure for quantifying AC in the suspect samples by LC-MS/MS. In this work, the method was extended to include additional alkaloids, other spice powders and spectral MS/MS acquisition for compound confirmation. 

Experimental methods

Chemicals and samples: The neat standards for aconitine (AC), hypaconitine (HA), mesaconitine (MA), benzoylaconine (BAC), benzoylmesaconine (BMA) and yunaconitine (YA) were purchased from Toronto Research Chemicals. Individual neat standards were mixed to prepare stock solutions in acetonitrile from which calibration standards (0.05–100 ng/mL) were prepared in 50:50 (v/v), methanol/0.1% formic acid in water.

Several spice powders, including sand ginger, ground ginger and five-spice powder, were purchased from local supermarkets for method development. 9 proficiency samples were provided by the CFIA and comprised of 1 blank spice powder and 8 samples with varying levels of incurred aconitine.

Sample preparation: The spice powder samples were extracted based on a method developed by the CFIA. Briefly, 1 g of sample was extracted with 10 mL of 75:25 (v/v), methanol/2% formic acid in water by shaking for 1 hour, followed by sonication for 10 minutes. After centrifuging at 4500 rpm for 10 minutes, aliquots of the supernatant were diluted with 50:50 (v/v), methanol/0.1% formic acid in water at varying dilution factors to ensure the resulting peak area response fit within the calibration curve. Water and unspiked spice powder were extracted in the same manner as reagent and procedural blanks, respectively. Sand ginger was spiked at 5 ng/g to calculate the MDLs as well as 4 different concentrations ranging from 2.5 to 5000 ng/g to assess the impact of dilution on matrix effects. Recoveries were also assessed in the other spice powders spiked at 5000 ng/g.

Chromatography: LC separation was performed on a Shimadzu Nexera LC-40 system using a Phenomenex Gemini C18 column (100 x 2 mm, 3 µm, P/N 00D-4439-B0). A flow rate of 0.5 mL/min, an injection volume of 1 µL and a column temperature of 40ºC were used. The LC conditions used are shown in Table 1.

Mass spectrometry: Analysis was performed using sMRM on the QTRAP 6500+ system in positive electrospray ionization mode. Guided optimization workflows within the SCIEX OS software were used to automate method optimization for the 6 alkaloids. The guided MRM infusion procedure was used to optimize collision energy (CE) and collision exit potential (Figure 2), while MRM flow injection analysis (FIA) was used to optimize the source and gas conditions and declustering potential (DP) (Figure 3). Table 2 lists the source and gas conditions used, while Table 3 lists the MRM transitions and optimized compound-dependent parameters.

MS/MS acquisition: Compound identification was further supported by triggering EPI scans of MS/MS spectra from a data-dependent acquisition method using sMRM as the survey scan. Only the quantifier transition was monitored for each compound in the sMRM survey scan. MS/MS spectra were acquired with dynamic fill time enabled at a scan speed of 10,000 Da/sec and with CE of 45 V and collision energy spread (CES) of ±15 V. Dynamic background subtraction (DBS) and dynamic exclusion were used to optimize the intensity threshold and frequency of triggering EPIs across each LC peak. 

Data processing: Data were acquired and processed using SCIEX OS software, version 3.3. MS/MS spectral matching was compared against the SCIEX Natural Products HR-MS/MS spectral library and reference spectra in the published literature.⁴

Linearity, dynamic range, accuracy and precision of calibration curve

Quantitation of the 6 target alkaloids was performed using a solvent-based calibration curve that spanned at least 3 orders of magnitude with r² ≥0.994. Acceptable accuracies within ±30% and %CV <25% were achieved at the lower limits of quantitation (LLOQs), which were defined by the lowest calibration level of each analyte. Accuracies within ±20% and %CV <10% were achieved at all other calibration levels (Table 4).

Quantitative performance at MDL

The MDL is defined as the lowest measured concentration of an analyte that can be reported with 99% confidence as distinguishable from the method blanks.⁵ MDLs were calculated by extracting 9 replicate sand ginger samples spiked at 5 ng/g, followed by dilution to yield in-vial concentrations at 2x the LLOQ.⁵ The MDL was calculated by multiplying the standard deviation from the 9 replicates by the t-value (2.896) at the 99% confidence level, as follows:

MDL = 𝑠 × 𝑡_{(n-1, 1-α=0.99)}

The calculated columns feature in SCIEX OS software enables the user to directly perform calculations in the results table instead of transferring the data elsewhere. The in-vial MDLs were natively calculated within the software, followed by conversion to the mass-based MDLs based on the dilution factor, extraction solvent volume and original sample mass. Apparent recoveries within ±30% of the nominally spiked concentration and %CV <15% were achieved in the MDL spikes (Table 4). Absolute recoveries of 90–104% (%CV <10%, n = 9) were also calculated based on the quotient of the peak areas in the pre- and post-spiked extracts (Table 4). Figure 4 compares the XICs of the solvent blank, LLOQ standard and the pre- and post-spiked extract, all of which exhibited good peak area responses at near-MDL levels. 

The sub- to low-ng/g MDLs (0.55–1.5 ng/g) here provided ample sensitivity to support a wide range of dilution factors. Due to the high concentrations expected in the contaminated spice powders, larger dilutions were needed to lower their responses to fit within the dynamic range. The higher dilutions also helped minimize matrix effects and enabled a single solvent-based calibration for all the spice powders tested.

Impact of dilution on matrix effects

Extensive dilution was required for the high concentrations expected in the adulterated spice powders. However, samples with low-level contamination may not require significant dilution or can be directly injected. Therefore, experiments were performed to assess the impact of matrix effects on quantitation. Post-spiked sand ginger extracts at concentrations of 2.5–5000 ng/g were diluted to a varying extent to assess matrix effects based on the following:

At concentrations of ≥5000 ng/g, a dilution factor of ≥100x minimized matrix effects to <10% (Figure 5). At concentrations <10 ng/g, lower dilutions of 2x or 5x kept the matrix effects to <15%, while matrix effects increased to 20–25% for extracts that were directly analyzed without further dilution.

Recoveries at different concentrations and in different spice powders

Due to the high levels anticipated in real-world adulteration, recoveries were also assessed at higher spiking concentrations. Excellent recoveries within ±15% of the nominally spiked concentration were achieved at the higher spiking concentrations (≥50 ng/g), while recoveries were within ±30% at the near-MDL levels (Figure 6).

Good quantitative performance was also reproducible in 3 other spice powders, as demonstrated by the tight distribution observed in the apparent and absolute recoveries in the violin plots (Figure 7.)

Method applicability to quantify and confirm the presence of Aconitum alkaloids in adulterated sand ginger

The method was applied to quantify and identify the 6 target alkaloids in the proficiency sand ginger samples. These samples were previously confirmed to be adulterated with AC and quantified for their reference concentrations. None of the alkaloids were detected above their corresponding MDLs in the blank sample. Quantitative accuracy of the method was demonstrated by the good agreement (<15% difference) between the AC concentrations measured here against the reference values (Table 5). Due to the high concentrations (mg/g) observed, a dilution factor of 10,000x was required to lower the responses to fit within the dynamic range.

In addition to comparing RTs and ion ratios between the unknowns and standards, AC identification was also supported by spectral MS/MS matching against the SCIEX Natural Products HR-MS/MS library (Figure 1 and Table 5). While ion ratios have established acceptance criteria for compound identification,⁶ false positives or false negatives may still occur especially at low concentrations or in complex matrices. As an alternative, full-scan MS/MS spectra acquired by EPI scans are rich in fragments and provide a molecular fingerprint for each compound that can be searched against a reference library or published databases.⁷

Table 6 shows the range of library hit scores obtained for the proficiency samples tested. A library hit score of N/A indicated the absence of a triggered MS/MS spectrum, which was typically observed at the lower concentrations. If the compound was not present in the library, diagnostic fragments were used instead to compare the acquired MS/MS spectra against published spectra.⁴ In addition to aconitine, HA, MA and BAC were also detected. The concentrations of HA and MA were in the range of 5–91 µg/g, while BAC was in the sub- to low-mg/g range. 

In Figure 8, the XIC peaks of HA and BAC in the standard and sample extract contained circles, each of which represented an occurrence of a triggered EPI scan. If multiple EPI scans were triggered during acquisition, the SCIEX OS software would automatically select and display the MS/MS spectrum with the highest library score, as indicated by the blue highlighted circle. This software selection can be changed by the user if they wish to manually review the other triggered MS/MS spectra. HA and BAC were identified in sample 4 based on MS/MS comparison against the library and a published list of diagnostic fragments⁴, respectively (Figure 8). The confidence in the BAC identification was further bolstered by low RT errors (<1%) and low ion ratio differences (<10%). In contrast, peak splitting was observed for HA in sample 4, which resulted in an RT shift compared to the standard and significant ion ratio differences. Based on these 2 parameters alone, HA would not be confidently identified, but the observed spectral MS/MS match against the library suggests the presence of the compound under those co-eluting peaks. Further clean-up of the extracts may be required to isolate HA from these interferences.

Conclusion
 

• The sensitivity of the QTRAP 6500+ system enabled the development of a method comprised of a rapid dilution approach and LC-MS/MS that achieved sub- to low-ng/g MDLs for the analysis of 6 Aconitum alkaloids in spice powders. 
• Minimal matrix effects (<15%) were observed due to a wide dilution range, which also removed the need for further extract clean-up and allowed the use of a single solvent-based calibration. 
• Excellent quantitative matrix performance (recovery ±30% at the MDL and ±15% at higher concentrations, %CV <15%) was achieved in different spice powders and at levels near the MDL and similar to those expected in adulterated samples. 
• Good agreement was achieved between the concentrations of aconitine measured in the proficiency samples against the reference values independently measured elsewhere. 
• MS/MS spectral data provided orthogonal information to increase the confidence in compound identification.
   

References
 

1. He, Fan. et al. Simultaneous quantification of nine aconitum alkaloids in Aconiti Lateralis Radix Praeparata and related products using UPLC-QQQ-MS/MS. Sci. Rep. 2017, 7, 13023.
   
   
2. Wang, Y. et al. Characterization of metabolites and cytochrome P450 isoforms involved in the microsomal metabolism of aconitine. J. Chrom. B. 2006, 844, 292-300.
   
   
3. Public Health Agency of Canada. Public Health Notice: Aconitine toxin found in galanga powder in Ontario with possible implications for other jurisdictions. Public Health Notice – Final update, September 27, 2022.
   
   
4. Zhang, N. et al. Qualitative and Quantitative Assessments of Aconiti Lateralis Radix Praeparata Using High-Performance Liquid Chromatography Coupled with Diode Array Detection and Hybrid Ion Trap-Time-of-Flight Mass Spectrometry. J. Chrom. Sci. 2016, 54, 888-901.
   
   
5. US Geological Survey. New Reporting Procedures Based on Long-Term Method Detection Levels and Some Considerations for Interpretations of Water-Quality Data Provided by the U.S. Geological Survey National Water Quality Laboratory. Open-File Report 99-193, 1999. 
   
   
6. 2002/657/EC: Commission Decision of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical methods and the interpretation of results (Text with EEA relevance) (notified under document number C(2002) 3044) (OJ L 221 17.08.2002, p. 8)
   
   
7. Powerful scan modes of QTRAP system technology. SCIEX technical note, 2019, RUO-MKT-02-8611-A.