A case of human poisoning by grayanotoxins following honey ingestion: elucidation of the toxin profile by mass spectrometry

High-resolution mass spectrometry (HRMS) was applied for the detection of grayanotoxins (GrTx) in a contaminated honey sample. This sample was provided by a hospital due to a suspicion of intoxication after a patient had shown the typical symptoms of GrTx poisoning. Subsequent analysis proved the contamination with high amounts of GrTx and other toxins belonging to grayanane-type diterpenoids. This group of natural toxins is synthesised by the plant family Ericaceae and comprises more than 60 individual toxins, but only one compound is available as a reference standard. We applied a screening approach that easily confirms the presence or absence of GrTx without access to standards. By searching for predictable mass spectrometric fragment ions, including typical in-source fragments arising from collision-induced dissociation during electrospray ionisation, the complete toxin profile was screened and allowed the mass spectrometric identification of 15 individual GrTx. The potential of this approach is especially demonstrated by the fact that at least two of these toxins have not been previously described in the literature. A semi-quantitative estimation indicated a total toxin concentration of 358 mg kg−1. An investigation of 49 honeys from the German retail market did not reveal the presence of GrTx. Graphical Abstract

GrTx have acute toxic effects and the dose-effect relationships of some of them have been intensively investigated in mice (Hikino et al. 1976). Further data were received from studies in skeletal muscle of frogs (Deguchi & Sakai 1967), rats (Fukuda et al. 1974;Nishikawa et al. 1989;Onat et al. 1991) and brine shrimp larvae (Kinghorn et al. 1978). The harmful effects of GrTx result from their binding to sodium channels of excitable membranes and the resulting inactivation of action potential. This leads to continuous depolarisation and enhancement of the calcium cation influx (Narahashi & Seyama 1974;Teuscher & Lindequist 1994;Jansen et al. 2012;Li et al. 2013). In human cases of high ingested amounts of GrTx, symptoms began within 30 min up to 4 h of latency at the most. The most striking symptoms and signs were nausea, hypotension, bradycardia and impairment of consciousness (Gunduz et al. 2008;Jansen et al. 2012).
Due to beekeeping in regions, where GrTx producing species grow, honey can be contaminated and its ingestion can result in acute toxic effects (Carey et al. 1959;White & Riethof 1959;Scott et al. 1971;Sütlüpmar et al. 1993;Koca & Koca 2007;Gunduz et al. 2008;BfR 2010, Jansen et al. 2012. This may occur particularly if honey from the eastern part of the Black Sea region is consumed, where it is known as 'mad honey'. Since most of the local beekeepers produce honey on a small scale, final products may be sourced from an area densely covered with Ericaceae and can thus contain considerable concentrations of GrTx (Jansen et al. 2012). Only limited reports are available about GrTx concentrations in honey and case reports describing human intoxication with GrTx are predominantly focused on the description of typical intoxication symptoms (Gunduz et al. 2006(Gunduz et al. , 2008. Descriptions of the analysis of toxin profiles and concentrations in suspect samples allowing a detailed assessment of ingested amounts are rarely published. One case was reported in Germany (North Rhine-Westphalia) in 2010 concerned a consumer who became ill after consumption of honey from the Black Sea region (BfR 2010). This person developed bradycardia and became unconscious 4 h after ingestion. Subsequent investigation of the honey sample revealed a GrTx III level of 43 mg kg −1 as well as the presence of pollen originating from Castanea sativa Mill. and Rhododendron ponticum L. Further investigations of contaminated honeys from North Carolina in the United States were reported with analysis by paper electrophoresis according to the procedure of Tallent et al. (1957) and GrTx I levels (referred as acetylandromedotoxin) of 100 mg kg −1 were determined (White & Riethof 1959). Another sample from British Columbia, Canada, was investigated by TLC and levels of 3 and 7 mg kg −1 honey for GrTx II and III respectively were estimated (Scott et al. 1971). A study of 10 honey samples from the Black Sea region revealed GrTx I concentrations between 1.5 and 39.3 mg kg −1 and GrTx III concentrations between 0.3 and 35.1 mg kg −1 , while GrTx II was only detected at very low concentrations (Kaplan et al. 2014). Furthermore, a detailed report on the quantitative mass spectrometric determination of GrTx in biological samples and the description of product ion spectra is given by Holstege et al. (2001). This paper describes the case of a 56-year-old man who caused a traffic accident after eating two tablespoons of wild Turkish honey. Three hours after consumption the man lost consciousness due to development of severe bradycardia (slow heart rate), which was diagnosed during physical examination in hospital. The hospital physician suspected GrTx intoxication due to the patient's symptoms and, therefore, a sample of the honey was provided for toxin analysis using MS. Currently, only GrTx III can be unambiguously identified in suspect samples since it is the only compound commercially available as a reference standard. However, using high-resolution mass spectrometry (HRMS) it should be possible to match individual accurate mass measurement information to known structural formulae, thus enabling identification of other GrTx and grayanane toxins that might be present in the sample. The common core structure of these toxins ( Figure 1 and Table 1) results in a similar and thus predictable mass spectrometric fragmentation. All toxins exclusively exhibit successive losses of water arising from collision-induced dissociation during electrospray ionisation (Holstege et al. 2001). Therefore, toxins were easily identified by exact mass determination of several typical in-source fragments resulting from repeated losses of water. For result confirmation, all analytes detected as supposed toxins were further investigated by recording product ion spectra. As criteria for the identification of an analyte as GrTx or grayanane-type toxin, the following had to be fulfilled: the detection of exact masses of the molecular ion and several diagnostic ions as well as the exhibition of a typical fragmentation pattern. As all putative toxins were subsequently confirmed by their product ion spectra, it could be demonstrated that the screening for in-source fragments resulting from losses of water is a successful approach for GrTx analysis in the absence of standards for the complete compound class.
For the quantification of detected toxins with no available reference standards, a semi-quantitative approach was used on the assumption that the mass spectrometric response of GrTx III is comparable with all other toxins.

Sample
The wild honey sample originated from the Turkish Black Sea region. Figure 1. A-nor-B-homo-ent-kaurane skeleton of grayanane respectively andromedane covering the groups of grayanotoxins, asebotoxins, rhodojaponins, craiobiotoxins, kalmitoxins, rhodomolleins and lyoniatoxins (see Table 1). Table 1. Naturally occurring Ericaceous toxins with a grayanane-type structure summarised by their common carbon core structure which leads to identical exact masses for molecular ions or in-source fragments.

Recovery for GrTx in honey
An excellent recovery of 97% ± 6.1% (n = 5) was obtained by spiking a blank honey with 167 µg kg -1 of GrTx III.
Mass spectrometry HRMS data were acquired on an Orbitrap Exactive™ system (Thermo Fisher Scientific, Bremen, Germany). The MS was operated at 3500 V using an S-lens voltage of 140 V. Two scan events were recorded in parallel by permanent switching. Firstly, a full-scan from m/z 160 to 1000, applying a resolution of 50 000; and secondly, fragments generated by an HDC 30 V experiment were acquired from m/z 90 to 800 using a resolution of 25 000.
Tandem mass spectrometry (ESI-MS/MS) data were acquired on a TSQ Vantage (Thermo Fisher Scientific, San Jose, CA, USA). GrTx were analysed using a spray voltage of 3500 V, an S-lens voltage of 140 V and a vaporiser temperature of 300°C. Q1 and Q3 were set at unit resolution .

Results and discussion
Screening for grayanotoxins and other grayanane-type toxins Due to the geographic origin of the ingested honey and the observed patient symptoms, contamination with GrTx was considered by the treating physician. A sample of the honey was sent for toxin analysis. Numerous toxins with grayanane-type structure have been reported to cause acute toxicity (Hikino et al. 1976). Therefore, our analytical approach was not focused on a single toxin but on the whole group of toxins. Apart from GrTx III, no reference standards are available making targeted detection, for example, by MRM inapplicable. Therefore, full-scan HRMS was applied to screen the sample for grayananetype toxins. The naturally occurring grayanane-type toxins produced by the Ericaceae family are summarised in Table 1 and grouped according to their common carbon core structures (see below).
The first step was to screen the sample for the exact pseudo-molecular ion masses of all toxins listed in Table 1. However, the exact mass measurement of the pseudo-molecular ion for a toxin is an insufficient positive identification criterion in isolation: additional diagnostic ions must be identified. All the investigated toxins are polyhydroxylated compounds containing multiple saturated alkyl alcohol groups that are prone to in-source fragmentation resulting in successive losses of H 2 O (−18 amu). This is exemplified by the HRMS full-scan spectrum of GrTx III which demonstrates the strong proneness to the losses of H 2 O (Figure 2a). In order to demonstrate that the fragments observed in the full-scan spectrum originate from GrTx III a product-ion spectrum was acquired by tandem mass spectrometry. This enables the targeted fragmentation of a precisely selected specific precursor ion originating from the molecule of interest. Since no molecular ion ([C 20 H 34 O 6 + H] + or the respective adduct [C 20 H 34 O 6 + Na] + ) was apparent on the tandem mass spectrometer, the in-source fragment m/z 353 [GrTx III -H 2 O] was selected as a precursor ion for recording the product ion spectrum (Figure 2b). The congruence of both spectra confirms that the exact masses of in-source fragment ions formed by successive losses of H 2 O can be used as diagnostic ions for the identification of grayanane-type toxins using HRMS. As the core structure of these toxins only contains saturated alcoholic groups, their product ion spectra solely consist of fragments originating from multiple losses of water and lack any other fragmentation. Both facts are important identification criteria as all other analytes than grayanane-type toxins that bear other functional groups than saturated alcohols would have to exhibit fragments or neutral losses which are specific for ethers, esters, sugar moieties etc. or other substance-specific fragmentations. Further, the application of HRMS scans allows the identification (and therefore exclusion) of analytes that contain nitrogen atoms or isotopes due to the presence of even molecular ions or specific isotopic masses respectively. The typical grayanane-type fragmentation pattern shown in Figure 2 leads to a grouping of these toxins according to their mass spectrometric fragmentation behaviour (Table 1). All toxins belonging to one group have either identical sum formulas or result in the same fragment ions and carbon cores due to losses of water in the MS source (Figure 2). In order to group individual toxins and to compare the carbon core structures of all grayanane-type toxins, their sum formulas were reduced to their smallest common representative structural unit. The minimum oxygen atom count is 4 (Table 1, refer to GrTx XVIII: C 20 H 32 O 4 in group 5 and GrTx X: C 22 H 32 O 4 in group 10). Therefore, all other sum formulas were theoretically recalculated to four oxygen atoms by successive the losses of water (-H 2 O). This procedure results in carbon core sum formulas with four oxygen atoms allowing a direct prediction of which toxin will form identical fragment ions during in-source fragmentation. Consequently, each carbon core group in Table 1 summarises toxins that can be detected in the same extracted ion chromatogram (exact mass filter). For instance, the sum formula of pieristoxin G (C 20 H 32 O 8 , group 1; Table 1) can be reduced to the four-oxygen atom core by a fourfold loss of H 2 O, resulting in C 20 H 24 O 4 . Obviously, no other grayanane-type toxin exhibits this core structure and, thus, no other known toxins will be detected using this exact mass filter (see below). For the unambiguous identification of GrTx III in the honey sample, the elution profile for the exact mass ion chromatograms for GrTx III, the pseudo-molecular ion and specific fragment ions resulting from repeated losses of water were compared with those of the standard (Figure 3). When extracting these GrTx III-specific ion chromatograms (Rt = 4.5 min), additional signals at different retention times were detected (Figure 3). This can be explained by the fact that in addition to GrTx III (sum formula C 20 H 34 O 6 ), GrTx II, VI and craiobiotoxin II, V, VI, VII, VIII (sum formula of C 20 H 32 O 5 ), and GrTx VII, VIII, XIX (sum formula of C 20 H 30 O 4 ) share the common carbon core structure C 20 H 30 O 4 after successive losses of H 2 O (group 4; Table 1). Therefore, this screening approach extracts and screens concurrently all toxins belonging to the same group (Table 1) because, per group, the same diagnostic ions specific for a particular carbon core are formed. It can be concluded that the suspect honey sample contains five of the grayananetype toxins summarised in group 4 (Table 1 and Table 2), including GrTx III, which was confirmed by comparison with the reference standard.
By means of the described approach, we screened the sample according to the toxin groups listed in Table 1. Signals were suspected to originate from grayanane-type toxins, if extracted ion chromatograms representing iterative water losses showed signals at the same retention time (Figure 3). For all peaks, which were identified as potential grayanane-type toxins by HRMS, product ion spectra were recorded. Those compounds displaying fragmentation patterns similar to the GrTx III standard (Figure 2) or as described in the literature (Holstege et al. 2001) were confirmed as grayanane-type toxins. Results are listed in Table 2.
As standards are not yet available, an unambiguous identification of individual toxins detected for each group is not possible, especially if several toxins share the same molecular formula. For a putative identification the highest observed m/z precursor ion per detected toxin is listed in Table 2 together with all known toxins with matching sum formulas. For instance, for group 2: two analytes eluting at Rt = 3.3 and 4.5 (min) were detected and at both retention times the highest observed sum formula was C 20 H 31 O 6 ([M + H] + for positive mode). Assuming that C 20 H 31 O 6 is the pseudomolecular ion, one of these two signals should originate from GrTx XVII and the other from a grayanane-type toxin not yet described in the literature. For group 3, three toxins were detected: two with the sum formula C 20 H 34 O 7 Na, which can be theoretically assigned as rhodojaponin VI, rhodomollein XVIII or kalmitoxin I, and one with the sum formula C 20 H 31 O 5 , which was identified as rhodomollein XIX (since no other toxin with that corresponding formula has been described in the literature). Most of the naturally occurring grayanane-type toxins belong to groups 3 and 4 and eight of the 15 toxins detected in the honey sample are summarised therein, with the most intensive signals belonging to toxins of group 4. Interestingly, two group 5 analytes were detected, although only GrTx XVIII is reported in the literature. Presumably, a further unreported grayanane-type toxin with the corresponding molecular formula was detected in the honey sample.
Semiquantitative determination of grayanotoxins and other grayanane-type toxins to estimate toxin intake The concentration of GrTx III in the sample was determined by integrating the peak areas of selected ion chromatograms from in-source fragments obtained by HRMS ( Figure 3) and applying an external calibration via the reference standard ( Table 2). Amounts of all other toxins were estimated semiquantitatively by reference to the GrTx III standard, assuming equal response factors for all other toxins. For each unknown toxin, the concentration was estimated by comparison of the peak area for the highest MS signal with the highest signal of GrTx III for which the correlation of mass spectrometric response and concentration is known ( Table 2). The sum of individual grayanane-type toxin concentrations found in the sample was estimated at 358 mg kg −1 , while the total ingested toxin amount was estimated at 8.9 mg, calculated based on the quantity of honey consumed (Table 2) (reported by the hospital to be two tablespoons, amounting to approximately 25 g).

Investigation of honey from the German retail market
In order to assess the relevance of GrTx for human consumption, 49 honey samples of different geographic as well as botanical origin from retail markets were analysed. A full sample description is given in the Supplementary data online. No GrTx could be detected in any of the investigated samples.

Conclusions
A combination of MS tools allowed a detailed characterisation of the toxin profile in a suspect honey sample. The exact mass measurement of pseudo-molecular and fragment ions proved the presence of 15 grayanane-type toxins comprising 12 containing 20 carbon atoms and three containing 22. At least two of these have, to the best of our knowledge, not been previously described in the literature. Of the two compounds detected with molecular formulas corresponding to C 20 H 30 O 6 (group 2) and the two with C 20 H 32 O 4 (group 5), only one matching toxin has been reported in each case, namely GrTx XVII and XVIII, respectively. Grayananetype toxins containing 21, 23, 24 or 25 carbon atoms, which have been reported in the literature, were not detected. For most of the compounds detected, molecular sum formulas were assigned, corresponding to toxins already reported in the literature. GrTx III was unambiguously quantified in the sample at a concentration of 54 mg kg −1 and the semiquantitative determination of other toxins revealed an estimated total toxin concentration of 358 mg kg −1 . Coupled with the reported ingested amount, a grayanane-type toxin intake of 8.9 mg has been estimated. However, effects of the individual toxins in humans are not known. The current knowledge on relative toxicities of different toxins is limited and information is derived from only a few animal experiments. There is some evidence from mice that GrTx I and III exhibit comparable toxicities, while GrTx II is of minor toxicity (Scott et al. 1971;Hikino et al. 1976). We cannot exclude that toxins were quantified that are not relevant for risk assessment and, therefore, conclusions concerning individual toxin concentrations and the extent of their effects on the patient regarding the reported symptoms cannot be drawn. The analytical investigation of German retail honey samples did not confirm the presence of GrTx or grayanane-type toxins and it can therefore be assumed that GrTx intoxication is a potential risk limited to honey of a certain botanical origin and does not constitute a global problem. Intoxications by honey seem to be limited to regions where plants with respective producing capacities are densely growing and honey is produced on a small scale without any further dilution with honey from other production areas.