Structural modifications of certain pyrrolizidine alkaloids during sample extraction and its impact on analytical results

doi:10.21203/rs.3.rs-4913802/v1

Download PDF

Research Article

Structural modifications of certain pyrrolizidine alkaloids during sample extraction and its impact on analytical results

https://doi.org/10.21203/rs.3.rs-4913802/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Pyrrolizidine alkaloids (PAs), a group of plant toxins often contaminating food or feed, are typically extracted from samples using liquid extraction. The crude extracts are then often purified using solid-phase extraction (SPE) cartridges before being analysed by LC-MS/MS. During the development of analytical methods based on strong cation exchange SPE, certain structurally related PAs showed unexpectedly low or significantly increased recoveries, suggesting transformation reactions may be at play. To investigate this hypothesis, sample preparations were conducted using PA-free milk as food matrix, water or organic solvents, into which PA reference standards were spiked before or after critical steps of the protocol.

The results revealed a significant decrease in acetylated PA N-oxides to their corresponding deacetylated compounds, as well as the formation of epoxydic PAs from PA compounds containing chlorine and hydroxyl groups in the α position. Evaporation of the alkaline SPE eluates, combined with the use of the protic solvent methanol in cases of deacetylation, was responsible for these phenomena. An alkaline ester hydrolysis mechanism was hypothesised for the deacetylation, while an internal S_N2 reaction, similar to the chlorohydrin reaction, was suggested for the formation of epoxy PA compounds. Consequently, using different sample preparation methods may inadvertently bias the determined PA patterns.

Plant toxins

transformation

strong cation exchange

deacetylation

epoxidation

solid phase extraction

Pyrrolizidine alkaloids (PAs) are a group of toxic plant secondary metabolites occurring in tribes and genera of the Boraginaceae, Asteraceae and Leguminosae families (Smith & Culvenor, 1981). Currently, more than 660 compounds have been identified, occurring in approximately 6,000 plant species worldwide as a defence mechanism against herbivorous insects (Boppré, 2011; Mattocks, 1986). In plants, PAs primarily exist in their more hydrophilic corresponding N-oxide form. It is widely accepted that unintentional co-harvesting of PA-containing plants in crop fields is the primary cause of contamination of plant-based food and feed. Teas, herbal teas, spices and culinary herbs are among the most affected product groups (Bodi et al., 2014; EFSA, 2017; Kaltner et al., 2020). The ingestion of contaminated feed has been shown to transfer PAs to animal-based food such as milk, meat or eggs (Mulder et al., 2016; Mulder et al., 2020; Taenzer et al., 2025). When ingested, only 1,2-unsaturated PAs and their corresponding N-oxides exhibit toxic potential to humans and mammals. Passing the intestine, these compounds are metabolised to pyrrole esters, capable of forming covalent adducts with DNA or cellular proteins (Fu et al., 2004). Consequently, both short-term and long-term intake of PAs can lead to fatal intoxications and are associated with pulmonary hypertension, hepatic sinusoidal obstruction syndrome, liver cirrhosis or liver cancer in humans (Edgar et al., 2015; Edgar et al., 2020; Yang et al., 2017).

For the purpose of preventive consumer protection, food authorities have established regulatory limits for PAs in frequently contaminated foodstuff. In the European Union, these limits were first implemented in 2020 and apply to a sum of 35 PAs (European Commission, 2023). Besides regulatory limits, PAs have been analysed in several matrices for research purposes for decades. As PAs are considered harmful even when ingested in small amounts over a long period of time, sensitive analytical methods are necessary to detect PA trace levels down to the ppb range. To avoid signal-decreasing interfering matrix compounds as well as possible, liquid-chromatography coupled to sensitive mass spectrometry (LC-MS) has been established as the instrumental “gold standard” in PA analytics (Casado et al., 2022; Crews et al., 2010). Prior to detecting PAs, they have to be properly extracted from the respective sample material. Commonly, acidic extraction solvents are used, followed by further purification and concentration steps such as the use of solid phase extraction (SPE) cartridges. Herein, SPE cartridges based on reversed phase or strong cation exchange (SCX) material are commonly used (Keuth et al., 2022; Kwon et al., 2021; Picron et al., 2018).

In earlier method development studies, discrepancies in recovery rates of certain, back then newly incorporated PA reference standards were noticed. Specifically, a near-total loss of 7Oacteylintermedine Noxide and 7Oacteyllycopsamine Noxide was observed, along with increased recoveries for the corresponding deacetylated PA Noxides (Kaltner et al., 2019). In a subsequent study, poor recoveries were obtained for three PAs with an incorporated chlorine atom, while three epoxydic PAs resulted in increased recovery rates (Klein et al., 2022). In both cases, it was hypothesised that unknown transformation reactions during the sample preparation procedure using SCX SPE cartridges were responsible for these findings. While these types of reactions of PAs during sample preparation have not been described in the literature before, they can have a significant impact on the reported analytical results for a given sample. Consequently, in the current study, we investigated the individual steps of the sample preparation applied in the mentioned studies by using PA reference standards spiked prior to and after potentially critical steps of the method. As the observed effects were assumed to be caused by the method itself and to avoid matrix interference, both milk (as a model food matrix) and water or extraction solvents without food matrix were used for spiking during the experiment.

The experiments and measurements were independently performed in two separate laboratories (Oberschleissheim, Lab1; Giessen, Lab2). Lab1 performed experiments using PA-free dairy milk while in Lab2 the reference standards were spiked to water or pure solvents without any sample matrix.

Chemicals and reagents

For all experiments, acetonitrile and methanol (both LC-MS grade) as well as formic acid (99–100%), ammonia solution (25%) and n-hexane (all of analytical grade) were achieved from Th. Geyer (Renningen, Germany). Ammonium carbonate (HPLC grade) was acquired from Fisher Scientific (Schwerte, Germany) and ammonium hydrogen carbonate (LC-MS grade) was obtained from Merck (Darmstadt, Germany). Ultra-pure water was obtained by water purification, either using an UltraClear™ system from Evoqua Water Technologies (Barsbuettel, Germany, Lab1), or a Milli-Q™ system provided by Merck (Darmstadt, Germany, Lab2). Eighteen reference standards, namely 7Oacetylintermedine (AcIm), 7Oacetylintermedine N-oxide (AcImN), 7Oacetyllycopsamine (AcLy), 7Oacetyllycopsamine N-oxide (AcLyN), heliotrine (Ht), heliotrine N-oxide (HtN), intermedine (Im), intermedine N-oxide (ImN), jacobine (Jb), jacobine N-oxide (JbN), jaconine (Jn), lycopsamine (Ly), lycopsamine N-oxide (LyN), merenskine (Mk), merenskine N-oxide (MkN), merepoxine (Mx), merepoxine N-oxide (MxN) and senkirkine (Sk) were purchased from Phytolab (Vestenbergsgreuth, Germany) or from Cfm Oskar Tropitzsch (Marktredwitz, Germany). Stock solutions of each compound were prepared with either methanol or acetonitrile/water (50/50,v/v) (1 mg/mL) and shared between Lab1 and Lab2 to conduct the experiments with identical reference standards. Spike solutions of the standards (10 µg/mL) were obtained by subsequent dilution of stock solutions in methanol. All reference standard solutions were stored at -20°C.

General sample preparation procedure: The following general sample preparation was performed: 3.0 mL of milk (Lab1) or 2.0 g of water (Lab2) were weighed into a 50 mL polypropylene sample tube and 30 mL formic acid (2%, aq.) and 15 mL n-hexane were added. After shaking (30 min) and centrifugation (2,600 g), 15 mL of the aqueous phase were loaded onto a Bond Elut Plexa PCX 6 mL 200 mg SPE cartridge (Agilent, Waldbronn, Germany), preconditioned with 5 mL of methanol and 5 mL of formic acid (2%, aq.). After washing with 10 mL of water and 10 mL of methanol, the analytes were eluted with 6 mL of ammoniated methanol (5%). Eluates were evaporated to dryness under nitrogen at 50°C. Therefore, Lab1 used a Turbovap II water bath apparatus (Biotage, Uppsala, Sweden) while Lab2 evaporated with a FSC400D Sample Concentrator with an aluminium heat block (Cole Parmer, St. Neots, United Kingdom). Residues were reconstituted in 500 µL (Lab1) and 1,000 µL (Lab2) methanol/water (10/90, v/v) and filtered through a 0.2 µm PVDF syringe filter (Macherey-Nagel, Düren, Germany) into a glass vial.

Instrumentation and measurements: As LC-MS instrumentation, Lab1 used a Shimadzu HPLC system (Duisburg, Germany) hyphenated to an API 4000 triple quadrupole MS from Sciex (Darmstadt, Germany). The system of Lab2 consisted of a Dionex Ultimate 3000 RS by Thermo Scientific (Dreieich, Germany) coupled to an API 3200 triple quadrupole MS from Sciex (Darmstadt, Germany). Both Lab1 and Lab2 used chromatographic settings already published by Klein et al. (2022). In brief, separation was performed on Kinetex™ EVO C18 columns 100×2.1 mm with particle size 2.6 µm (Lab1) or 1.7 µm (Lab2), respectively. The columns were protected by SecurityGuard™ ULTRA EVO C18 2.1 mm pre-columns (all from Phenomenex, Aschaffenburg, Germany). At a flow of 0.3 mL/min and an oven temperature maintained to 30°C, 10 µL were injected into the system. Eluent solvents were ammonium carbonate (Lab1) or ammonium hydrogen carbonate solution (Lab2), (10 mmol/L at pH 9.0, A) and acetonitrile (B). The used LC gradient, exact ion source parameters and selected m/z transitions are extensively described elsewhere (Klein et al., 2022). Analyte levels were calculated for each sample using calibration solutions (Lab1: 0.1, 1.0, 2.5, 5.0, 10 ng/mL; Lab2: 1.0, 2.5, 5.0, 10, 25, 50 ng/mL) of a mix of PAs in methanol/water (10/90, v/v).

Software

Mass spectrometry data acquisition, data processing and peak integration was performed using Analyst 1.6 and Multiquant 3.0 (both from Sciex, Darmstadt, Germany). Chemical structures were drawn using ChemDraw Prime V 20.1 (PerkinElmer, Waltham, MA, United States). Atomic charges and the respective figure were calculated and drawn using the browser-based tool ”Atomic Charge Calculator II” (Raček et al., 2020) in automatic mode.

Spiking experiments

For the experiments, a total of eleven setups were prepared, comprising four using milk and seven using solely water or solvents as spiked mediums. Table 1 provides a comprehensive overview of all tested sets.

Table 1

Analytes and timepoints used for the spiked milk (M) and solvent (S) experiments
Used matrix / solvent	Name	Spiked analytes	Final c in vial	Timepoint of spiking
Milk	M1	AcImN, AcLyN, HtN, Sk	5 ng/mL	Prior to sample preparation
	M2			Prior to evaporation
	M3			Spiked to pure MeOH prior to evaporation, reconstituted in milk extract
	MC			Added to evaporated residue (= control)
Water	S1	AcImN, AcLyN, HtN, Jn, MkN, Sk	50 ng/mL	Prior to sample preparation
Water	S2			Prior to SPE purification
Water	S3			Prior to evaporation
MeOH + NH₃ ^a	S4			Spiked to solvent prior to evaporation
ACN	S5			Spiked to solvent prior to evaporation
ACN + NH₃ ^a	S6			Spiked to solvent prior to evaporation
MeOH + NH₃ ^a	SC			Added to evaporated residue (= control)
AcImN, 7Oacetylintermedin N-oxide; AcLyN, 7Oacetyllycopsamine N-oxide; ACN, acetonitrile; HtN, heliotrine N-oxide; Jn, jaconine; MeOH, methanol; MkN, merenskine N-oxide; Sk, senkirkine; SPE, solid phase extraction; ^a 5% NH₃ in the final mixture

In Lab1, to examine the deacetylation process in matrix, PA-free milk was spiked with AcImN and AcLyN to a theoretical concentration of 5 ng/mL per compound in the final sample extract. HtN and Sk as control compounds were added in the same amount. The following experiments with milk were prepared in triplicates: (M1) Spiking of milk aliquots before the sample preparation procedure, (M2) spiking the SPE eluate of blank milk prior to evaporation, (M3) spiking 6 mL of pure methanol prior to evaporation and subsequently reconstituting the residue with syringe-filtered (0.2 µm, PVDF) blank milk matrix extract, and, as a control (MC), by adding 50 µL of a standard mix in methanol and 450 µL water to an evaporated blank milk residue, followed by vortex mixing and syringe filtering into a glass vial. To calculate changes in analyte levels due to the sample preparation procedure, the mean signal responses of the four spiked analytes in experiments M1, M2 and M3 were compared to their respective mean peak area (n = 3) in the control experiment MC.

In Lab2, matrix-free solvents were spiked instead. Spiking was conducted in each experiment in a way to achieve a theoretical concentration of 100 ng/mL per compound in the measurement vial. First, to check for deacetylation and epoxidation of respective PAs, AcImN and AcLyN, HtN and Sk as well as the chlorine-consisting compounds MkN and Jn, were spiked to 2.0 g of water and extracted as described above (S1). Next, water was extracted, and the six analytes were spiked after the liquid extraction step and prior to SPE purification (S2) or after SPE elution and prior to evaporation (S3), respectively. Furthermore, in three experiments, the compounds were directly spiked to different solvents, subsequently evaporated, reconstituted, and filtered (0.2 µm, PVDF). Herein, the six PAs were added to ammoniated methanol (5%, S4), pure acetonitrile (S5) or ammoniated acetonitrile (5%, S6). In a control setup, 2.0 g of water was treated according to the described extraction procedure, with the evaporated residue reconstituted in 50 µL of a PA mix in methanol and 450 µL water (SC).

To express changes in the non-spiked, but arising analytes ImN, LyN, Jb, and MxN, their mean signal responses (n = 3, Lab1) or their mean concentrations (n = 3, Lab2) were reported relative to levels of the control analyte HtN in the respective experiment.

Deacetylation

In milk samples, the deacetylation of AcImN and AcLyN during the extraction and clean-up procedure was investigated by spiking these two compounds as well as HtN and Sk at different steps of the sample preparation (MC, M1 – M3), as described above. Compared to the reference setup (MC), the mean concentrations (n = 3) of the control PAs ranged from 100 ± 4% (M3) to 105 ± 7% (M2) for HtN and from 93 ± 7% (M3) to 105 ± 8% (M1) for Sk, respectively, indicating the overall stability of the used control compounds (Fig. 1A). In contrast, this was not the case for the acetylated N-oxides. When spiked to the milk prior to any sample preparation (M1), their mean concentrations were reduced to 2.5 ± 0.7% (AcImN) or 2.3 ± 0.9% (AcLyN). Also, when spiked to SPE eluates prior to nitrogen evaporation (M2), remaining concentrations of 13 ± 2% (AcImN) or 12 ± 2% (AcLyN) were detected, respectively. Interestingly, in experiment M3, where AcImN and AcLyN were spiked into pure methanol, evaporated, and reconstituted in extract of PA-free milk, the levels of both compounds were barely reduced and 95 ± 7% (AcImN) or 93 ± 7% (AcLyN) of the initial concentration was still detectable. A corresponding increase of ImN and LyN was observed due to changes in the ratio of their mean signal response, compared to the respective mean concentration of the stable reference analyte HtN. In the control setup MC, still 1.6 ± 0.1% of ImN and 2.4 ± 0.2% of LyN were detected. Also, in experiment M3, their ratio to the concentration of HtN remained low with 6.1 ± 1.2% (ImN) and 8.2 ± 0.8% (LyN), whereas in setups M1 and M2 their ratios greatly increased to 108 ± 4% and 60 ± 5% (ImN) and to 89 ± 2% and 75 ± 6% (LyN), respectively. An increase in the amount of the corresponding tertiary amine PAs, namely AcIm, AcLy, Im, and Ly, were not detected in any of the milk setups.

The observation of decreased amounts of AcImN and AcLyN due to the sample preparation procedure was consistent with findings from solvent setups conducted at Lab2 (Fig. 1B). When spiked prior to sample preparation (S1), both analytes were detected only in small amounts, at 2.3 ± 0.4% for AcImN and 2.3 ± 0.2% for AcLyN, compared to their levels in the control setup (SC). Similar results were obtained in experiment S2, where the analytes were spiked prior to the SPE purification step (2.5 ± 0.9% for AcImN, 2.5 ± 1.1% for AcLyN). When added to the SPE eluate prior to evaporation (S3), the detected levels were 11 ± 1% for AcImN and 10 ± 2% for AcLyN, consistent with the results observed in the corresponding experiment in milk (M2). In setups S4, S5 and S6, the analytes were spiked into ammoniated methanol, ammoniated acetonitrile, or pure acetonitrile and subsequently evaporated. In ammoniated methanol (S4), the levels of AcImN and AcLyN were also decreased to 4.2 ± 1.0% (AcImN) or 4.4 ± 1.1% (AcLyN). In contrast, spiking the acetylated N-oxides to acetonitrile (S5) or ammoniated acetonitrile (S6) resulted in hardly any loss of the compounds (AcImN: 93 ± 2% or 92 ± 6%; AcLyN: 96 ± 3% or 95 ± 4%). Consistent with the findings of the milk setups, no corresponding tertiary amine PAs of the analytes (AcIm, AcLy, Im, Ly) were detected above the method’s limit of detection (LOD) in any of the solvent experiments. In sum, a strong decrease of mean levels of acetyl PA N-oxides, corresponded with significantly increased levels of non-acetylated PA N-oxides, was observed for the spiked analytes both in milk and solvent samples. This deacetylation (Fig. 2A) was solely detected in setups with AcImN and AcLyN being present in ammoniated methanol and evaporated to dryness (M1, M2, S1 – S4). In summary, the current study showed a decrease of AcImN and AcLyN ranging from 77–93% concurrent with an increase of the corresponding analytes ImN and LyN.

Epoxidation: In the current study, the potential epoxide formation in certain PAs due to the applied sample extraction procedure was investigated. Therefore, the respective analytes Jn and MkN were spiked to the solvent setups conducted at Lab2. In the solvent control setup (SC), corresponding epoxide PAs Jb and MxN were detected, but only in traces smaller than the limit of quantitation (LOQ). In contrast, when Jn and MkN underwent at least the evaporation step of the regular sample extraction procedure (S1 - S3), Jb and MxN were formed to a certain extent (Jb: 18 ± 2%, 17 ± 3% or 11 ± 1%; MxN: 61 ± 5%, 59 ± 8% or 40 ± 2%; in relation to the concentration of HtN). In setups where the analytes were directly spiked into ammoniated methanol (S4), acetonitrile (S5) or ammoniated acetonitrile (S6) and then were subsequently evaporated, the findings differed between S4 to S6. In S4, the Jn mean level was reduced to 54 ± 6% and its corresponding epoxide Jb was detected (25 ± 3%, relative to HtN). Also, MkN was almost completely lost with mean proportions of 5.6 ± 2.4%, while its corresponding epoxide PA MxN was formed to the highest extent, (87 ± 2%, relative to HtN) compared to all other solvent experiments. On the other hand, in experiment S5, Jb was hardly formed (2.6 ± 0.2%) and also MxN was detected in comparably low concentrations of 16 ± 4%, relative to the mean level of HtN. In setup S6, however, the presence of ammonia in the acetonitrile solution led to a reduced concentrations of Jn (70 ± 5%) and MkN (26 ± 10%) while Jb and MxN were formed (21 ± 4% and 75 ± 8%, relative to HtN). Thus, the latter results were similar to those of setups S1 – S4. The corresponding tertiary amine or N-oxide PAs (JnN, Mk, JbN, Mx) of the investigated analytes were not detected above the LOD in any of the solvent setups. The detected decrease of PAs with chlorine and a hydroxyl group in α position, herein Jn and MkN, correlated with a high increase of their respective epoxide compounds Jb and MxN. The formation of corresponding epoxide PAs (Fig. 2B) was observed in all solvent experiments where ammoniated solvents, regardless of whether they were protic (methanol) or aprotic (acetonitrile), were used (S1 - S4; S6). Herein, the extent of this reaction was different between the structures, ranging between 11–25% for Jn to Jb and 40–87% for MkN to MxN (Fig. 2B). Consequently, MxN was generally formed to a much higher extent than Jb. It is expected that Mk and JnN, the structurally corresponding PAs to MkN and Jn, the compounds tested herein, react accordingly.

In former method development studies, deacetylation or epoxidation of certain PAs due to sample extraction conditions had not been noticed or investigated. As comparable analytical methods did not commonly include the analytes AcImN, AcLyN, Jn, Mk, or MkN, the phenomena described in this study were likely missed. Herein, it was shown that AcImN and AcLyN deacetylate to their corresponding compounds ImN and LyN when handled in methanol under alkaline conditions (Fig. 2A). Regarding a potential mechanism of the observed deacetylation, no analogues were found in the literature. First, a nucleophilic substitution (S_N) at the C7 atom was considered. If an S_N1 mechanism was assumed, the acetylic group would have to be released spontaneously, which is unlikely, forming a carbocation intermediate stabilised by protic solvents such as methanol. A subsequent nucleophilic attack by hydroxide anions present in an alkaline environment would result in racemic mixtures of ImN and rinderine N-oxide (RrN), or LyN and echinatine N-oxide (EnN), respectively. If an S_N2 mechanism occurred, a free hydroxyl group would react via a nucleophilic backside attack, resulting in an inversion of the stereochemistry at C7, leading to RrN or EnN as the respective sole reaction product (Fig. S1 A). The analytical method used in the current study was able to detect and to chromatographically separate the latter two analytes from ImN and LyN (Fig. S1 B), but neither RrN nor EnN were detected in any of the sample extracts where deacetylation occurred. This demonstrated that neither a S_N1 nor a S_N2 mechanism at C7 took place (Fig. S1 C). Consequently, we hypothesised a nucleophilic ester hydrolysis, initiated by a methanolate ion formed from methanol under alkaline conditions. The methanolate attacks the carbonyl C atom of the acetyl side chain, forming a negatively charged intermediate. Relocation of a lone electron pair back to the oxygen-carbon bond releases methyl acetate as a leaving group, leading to the deacetylated corresponding PA N-oxides while maintaining the existing stereochemistry at the C7 atom (Fig. 3A). Considering our experimental results, we currently consider the latter mechanism as the most probable one. To prove our hypothesis, reference standards labelled with stable isotopes at positions crucial for the proposed mechanism, such as the oxygen bound to the C7 atom, could be used in comparably designed future studies, but are not commercially available so far.

It was demonstrated that AcImN and AcLyN deacetylate in methanol under alkaline conditions to ImN and LyN. In contrast, AcIm and AcLy did not undergo a deacetylation to Im and Ly, which was already known from earlier studies (Kaltner et al., 2019). Consequently, potential differences in the atomic charges of the C7 or the acetyl carbon atom were assumed to occur between the acetyl PA and its corresponding acetyl PA N-oxide. The atomic charges of AcLy and AcLyN were calculated and compared, revealing negligible charge differences of the two C7 atoms (Fig. S2). As expected, the atomic charges of the acetyl carbon atoms of AcLy or AcLyN were much more positive than of the C7 atoms, strongly supporting the proposed reaction mechanism of a nucleophilic methanolate attack at this highly electrophilic site of the molecules (Fig. 3). As the partial atomic charges did not explain the reaction taking place solely for acetylated PA N-oxides, it was hypothesised that the presence of an N-oxide may stabilise the intermediate charged molecule in a way that the deacetylation can solely occur in these molecules.

However, although this deacetylation occurs solely under specific extraction conditions (protic solvents and alkaline pH commonly present in SCX SPE-based analytical methods), it should be particularly taken into account if different extraction methods were used. While AcImN and AcLyN are not included in the PA analyte set of Commission Regulation (EU) 2023/915 (so-called PA35, 21 PAs + 14 potentially co-eluting isomers), their deacetylated forms ImN and LyN are indeed part of this analyte set. Applying a sample extraction method based on alkaline pH and protic solvents may, therefore, affect reported results and may lead to unintended variations regarding the PA profiles in samples or even to varying PA (sum) amounts in samples. For instance, if AcImN or AcLyN are considerably present in a sample, their conversion to ImN and LyN due to a SCX SPE sample preparation method using alkaline conditions might result in differing sum amounts than those obtained with another extraction technique.

In a study from Zan et al. (2023), three PA-containing plants used in traditional Chinese medicine were investigated for their PA levels and PA patterns. Herein, the authors demonstrated Arnebia guttata to contain AcLy and AcLyN. However, an unusual PA N-oxide/tertiary amine PA ratio distinctly below 1 was reported for AcLyN/AcLy in the plant samples, while for other PA compounds this ratio was, as expected, ≥ 1, giving a hint to potential deacetylation of AcLyN. Although, using alkaline methanol to elute the PAs from the SPE and subsequently evaporating the eluates and, thus, applying conditions comparable to them in the current study, a recovery of 89% was reported for AcLyN. However, these results were contrary to findings of the current study, the reason for this remains unclear.

In this study, the epoxide formation of chlorine-containing PAs under alkaline conditions were systematically investigated, demonstrating the transformation of Jn to Jb and MkN to MxN to a certain degree (Fig. 2). This transformation was attributed to conditions during sample extraction similar to those found in the literature for the so-called chlorohydrin reaction (Carrà et al., 1979). In the study by Klein et al. (2022) on the development of an analytical method to detect PAs in milk, chlorohydrin-reaction-like conditions were applied during the SPE step of sample extraction analogous to the protocol used in the current study. The authors observed decreased recoveries of jaconine (Jn) by up to 65%, while its corresponding epoxide compound, jacobine (Jb), showed increased recoveries of up to 127%. Similarly, the chlorine-containing PAs Mk and MkN showed decreased recoveries of 69% (Mk) and 13% (MkN), while their corresponding epoxides, Mx and MxN, had recoveries of up to 121% (Mx) and 175% (MxN), respectively. These observations support the theory of a chlorohydrin-like reaction mechanism transforming Jn to Jb, Mk to Mx and MkN to MxN. It is likely that this epoxide formation occurred also in other earlier studies without recognising this effect (Kaltner et al., 2019).

In a study on the transfer of pyrrolizidine alkaloids from dry tea to tea infusions, brewing with buffer solutions at pH 9 led to significant decreases in Jn, Mk and MkN, with losses of up to 99% (Kaltner et al., 2025). Conversely, only MxN, the corresponding epoxide of the latter compound, was found in increased amounts exceeding 100%, while Jb and Mx were present at comparable levels when brewing was performed with pure water. These results suggest that the latter tertiary amine PAs are generally less stable at cooking temperatures. Nevertheless, the observed loss of PAs with chlorine and hydroxyl groups in the α position at alkaline pH aligns with the findings of the current study. Although neither the PA compounds prone to epoxide formation nor their epoxide reaction products are included in the PA35 set of analytes, unintended changes in the chemical structures of PA analytes due to certain conditions during sample extraction may affect scientific findings and, thus, should not be neglected in future studies.

In the current study, potential deacetylation of AcImN and AcLyN to ImN and LyN as well as epoxidation of Jn to Jb and MkN to MxN during sample preparation were systematically investigated in spiked milk or solvent. Both effects, deacetylation and epoxide formation, were tracked back to the evaporation step of the ammoniacal methanol eluate under a stream of nitrogen at 50°C. In case of epoxidation, these conditions are comparable to those applied in the so-called chlorohydrin reaction, and a similar reaction mechanism may be assumed. For deacetylation, a potential reaction mechanism similar to the alkaline ester hydrolysis was supposed. However, it was not fully understood why deacetylation solely affected the N-oxides and not the corresponding tertiary amine PAs, which should be further investigated in future studies. In consequence, our findings demonstrated that reported results of PA amounts in a sample could differ, depending on the used extraction techniques of PAs from food matrices. Although it seems unlikely for a reported result to considerably vary in the individual PA amounts if different extraction methods were used, the described effects should nonetheless be taken into account, in particular when applying the PA35 set for regulated PAs.

FUNDING

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation, Project 455407584).

Author Contribution

Conceptualization, methodology, analysis and investigation: L.M.K., F.K.; Writing - original draft preparation, supervision, visualisation and funding: F.K.; Writing - review and editing: L.M.K.

ACKNOWLEDGMENTS

The authors thank Eszter Borsos, University of Vienna, for her helpful thoughts and fruitful discussions on the hypothetical deacetylation mechanism of acetylated PA N-oxides and especially for her patience with the constantly rearranging nitrogen atom during the atomic charge calculations.

DATA AVAILABILITY

The datasets generated and analysed during the current study are available from the corresponding author upon reasonable request.

Bodi, D., Ronczka, S., Gottschalk, C., Behr, N., Skibba, A., Wagner, M., Lahrssen-Wiederholt, M., Preiss-Weigert, A., & These, A. (2014). Determination of pyrrolizidine alkaloids in tea, herbal drugs and honey. Food Additives & Contaminants: Part A, 31(11), 1886-1895. https://doi.org/10.1080/19440049.2014.964337
Boppré, M. (2011). The ecological context of pyrrolizidine alkaloids in food, feed and forage: an overview. Food Additives & Contaminants: Part A, 28(3), 260-281. https://doi.org/10.1080/19440049.2011.555085
Carrà, S., Santacesaria, E., Morbidelli, M., Schwarz, P., & Divo, C. (1979). Synthesis of Epichlorohydrin by Elimination of Hydrogen Chloride from Chlorohydrins. 1. Kinetic Aspects of the Process. Industrial & Engineering Chemistry Process Design and Development, 18(3), 424-427. https://doi.org/10.1021/i260071a012
Casado, N., Morante-Zarcero, S., & Sierra, I. (2022). The concerning food safety issue of pyrrolizidine alkaloids: An overview. Trends in Food Science & Technology, 120, 123-139. https://doi.org/https://doi.org/10.1016/j.tifs.2022.01.007
Crews, C., Berthiller, F., & Krska, R. (2010). Update on analytical methods for toxic pyrrolizidine alkaloids. Analytical and Bioanalytical Chemistry, 396(1), 327-338. https://doi.org/10.1007/s00216-009-3092-2
Edgar, J. A., Molyneux, R. J., & Colegate, S. M. (2015). Pyrrolizidine alkaloids: Potential role in the etiology of cancers, pulmonary hypertension, congenital anomalies, and liver disease. Chemical Research in Toxicology, 28(1), 4-20. https://doi.org/10.1021/tx500403t
Edgar, J. A., Molyneux, R. J., & Colegate, S. M. (2020). Linking Dietary Exposure to 1,2-Dehydropyrrolizidine Alkaloids with Cancers and Chemotherapy-Induced Pulmonary and Hepatic Veno-Occlusive Diseases. Journal of Agricultural and Food Chemistry, 68(22), 5995-5997. https://doi.org/10.1021/acs.jafc.0c02582
European Commission (2023). Commission Regulation (EU) No 2023/915 of 25 April 2023 on maximum levels for certain contaminants in food and repealing Regulation (EC) No 1881/2006. Official Journal of the European Communities L, 119, 103-157.
European Food Safety Authority, EFSA (2017). Risks for human health related to the presence of pyrrolizidine alkaloids in honey, tea, herbal infusions and food supplements. EFSA Journal, 15(7), e04908, Article e04908. https://doi.org/10.2903/j.efsa.2017.4908
Fu, P. P., Xia, Q., Lin, G., & Chou, M. W. (2004). Pyrrolizidine Alkaloids - Genotoxicity, Metabolism Enzymes, Metabolic Activation, and Mechanisms. Drug Metabolism Reviews, 36(1), 1-55. https://doi.org/doi:10.1081/DMR-120028426
Kaltner, F., Gottschalk, C., De Vries, E., & Mulder, P. P. J. (2025). Transfer of pyrrolizidine and tropane alkaloids from tea and herbal tea to infusions. Food Chemistry. Under review.
Kaltner, F., Rychlik, M., Gareis, M., & Gottschalk, C. (2020). Occurrence and Risk Assessment of Pyrrolizidine Alkaloids in Spices and Culinary Herbs from Various Geographical Origins. Toxins, 12(3), 155. https://doi.org/10.3390/toxins12030155
Kaltner, F., Stiglbauer, B., Rychlik, M., Gareis, M., & Gottschalk, C. (2019). Development of a Sensitive Analytical Method for Determining 44 Pyrrolizidine Alkaloids in Teas and Herbal Teas via LC-ESI-MS/MS. Analytical and Bioanalytical Chemistry, 411(27), 7233-7249. https://doi.org/10.1007/s00216-019-02117-1
Keuth, O., Humpf, H. U., & Fürst, P. (2022). Determination of pyrrolizidine alkaloids in tea and honey with automated SPE clean-up and ultra-performance liquid chromatography/tandem mass spectrometry. Food Additives & Contaminants: Part A, 39(1), 149-157. https://doi.org/10.1080/19440049.2021.1982149
Klein, L. M., Gabler, A. M., Rychlik, M., Gottschalk, C., & Kaltner, F. (2022). A sensitive LC–MS/MS method for isomer separation and quantitative determination of 51 pyrrolizidine alkaloids and two tropane alkaloids in cow’s milk. Analytical and Bioanalytical Chemistry, 414(28), 8107-8124. https://doi.org/10.1007/s00216-022-04344-5
Kwon, Y., Koo, Y., & Jeong, Y. (2021). Determination of Pyrrolizidine Alkaloids in Teas Using Liquid Chromatography–Tandem Mass Spectrometry Combined with Rapid-Easy Extraction. Foods, 10(10), 2250. https://www.mdpi.com/2304-8158/10/10/2250
Mattocks, A. R. (1986). Chemistry and Toxicology of Pyrrolizidine Alkaloids. Academic Press.
Mulder, P. P. J., de Witte, S. L., Stoopen, G. M., van der Meulen, J., van Wikselaar, P. G., Gruys, E., Groot, M. J., & Hoogenboom, R. L. A. P. (2016). Transfer of pyrrolizidine alkaloids from various herbs to eggs and meat in laying hens. Food Additives & Contaminants: Part A, 33(16), 1826-1839. https://doi.org/10.1080/19440049.2016.1241430
Mulder, P. P. J., Klijnstra, M. D., Goselink, R. M. A., van Vuuren, A. M., Cone, J. W., Stoopen, G., & Hoogenboom, R. L. A. P. (2020). Transfer of pyrrolizidine alkaloids from ragwort, common groundsel and viper’s bugloss to milk from dairy cows. Food Additives & Contaminants: Part A, 37(11), 1906-1921. https://doi.org/10.1080/19440049.2020.1798028
Picron, J. F., Herman, M., Van Hoeck, E., & Goscinny, S. (2018). Analytical strategies for the determination of pyrrolizidine alkaloids in plant based food and examination of the transfer rate during the infusion process. Food Chemistry, 266, 514-523. https://doi.org/10.1016/j.foodchem.2018.06.055
Raček, T., Schindler, O., Toušek, D., Horský, V., Berka, K., Koča, J., & Svobodová, R. (2020). Atomic Charge Calculator II: web-based tool for the calculation of partial atomic charges. Nucleic Acids Research, 48(W1), W591-W596. https://doi.org/10.1093/nar/gkaa367
Schindler, O., Raček, T., Maršavelski, A., Koča, J., Berka, K., & Svobodová, R. (2021). Optimized SQE atomic charges for peptides accessible via a web application. Journal of Cheminformatics, 13(1), 45. https://doi.org/10.1186/s13321-021-00528-w
Smith, L. W., & Culvenor, C. C. J. (1981). Plant sources of hepatotoxic pyrrolizidine alkaloids. Journal of Natural Products, 44(2), 129-152. https://doi.org/10.1021/np50014a001
Taenzer, J., These, A., Knappstein, K., Lamp, J., Dänicke, S., Saltzmann, J., Gottschalk, C., Fedotenko, I., & Jira, W. (2025). Selective and low transfer of pyrrolizidine alkaloids from Jacobaea vulgaris Gaertn. into muscle and liver of dairy cattle, goat and sheep. Food Control, 167, 110766. https://doi.org/https://doi.org/10.1016/j.foodcont.2024.110766
Yang, M., Ruan, J., Gao, H., Li, N., Ma, J., Xue, J., Ye, Y., Fu, P. P.-C., Wang, J., & Lin, G. (2017). First evidence of pyrrolizidine alkaloid N-oxide-induced hepatic sinusoidal obstruction syndrome in humans. Archives of Toxicology, 91(12), 3913-3925. https://doi.org/10.1007/s00204-017-2013-y
Zan, K., Wang, Z., Hu, X.-W., Li, Y.-L., Wang, Y., Jin, H.-Y., Zuo, T.-T., & Ma, S.-C. (2023). Pyrrolizidine alkaloids and health risk of three Boraginaceae used in TCM. Frontiers in Pharmacology, 14. https://doi.org/10.3389/fphar.2023.1075010

No competing interests reported.

ManuscriptSIPAchangesKaltner.docx

Download PDF

Reviewers agreed at journal
31 Aug, 2024
Reviewers invited by journal
25 Aug, 2024
Editor assigned by journal
19 Aug, 2024
Submission checks completed at journal
19 Aug, 2024
First submitted to journal
14 Aug, 2024

You are reading this latest preprint version

Structural modifications of certain pyrrolizidine alkaloids during sample extraction and its impact on analytical results

Status:

Version 1

Abstract

Figures

INTRODUCTION

MATERIALS AND METHODS

RESULTS

DISCUSSION

CONCLUSIONS

Declarations

FUNDING

Author Contribution

ACKNOWLEDGMENTS

DATA AVAILABILITY

References

Additional Declarations

Supplementary Files

Status:

Version 1