**3. Discussion**

Phytotherapeutic and nutritional use of plants and herbal-based products has (re)gained its popularity in the last years. Due to improper usage of plants or unawareness of plant toxicity, several intoxication cases with CG-containing plants have been reported recently [7,10,11]. Ingestion of toxic plant (parts) may also be possible through the food chain. Plant misidentification and inadequate control of harvesting or processing may lead to an unintentional mix of toxic plant material with the raw plant material used for production. Since more analytical methods became available for screening of organic molecules at low levels, several contaminants have surfaced as an issue of relevance in food safety. That was the case for other plant toxins, such as tropane alkaloids and pyrrolizidine alkaloids, which were found in herbal teas, herbs and spices, cereal-based food, and herbal food supplements [32–35], sometimes at levels that can represent risk for human health [36,37]. The current study aimed at developing a reliable analytical method for detection of CGs in plant-based food products. As a complementary tool for control of poisoning incidents, the method was also validated for the urine matrix. As opposed to some other human biological fluids that are used to study exposure to contaminants, urine is easily accessible from individuals of all ages, can be obtained in larger volumes, and its collection method is noninvasive.

As a detection technique, LC-MS/MS was chosen for this work, as it allows a high-throughput simultaneous detection of structurally diverse molecules, including compounds of natural origin, at trace levels and with high selectivity and specificity. In the proposed method, the SRM mode was used to obtain increased sensitivity and specificity, the parameters that are of grea<sup>t</sup> importance in the analysis of such complex matrices as herbs and urine. The selected ionization mode was ESI, the commonly used interface in the LC-MS analysis of natural toxins. Though less frequent, atmospheric pressure chemical ionization (APCI) is also used for the ionization of small molecules. Sugergat et al. [38] compared these two ionization modes for the analysis of DIGO in human serum and observed a lower intensity of the protonated molecule and a higher degree of fragmentation, resulting in lower sensitivity of the APCI mode compared to ESI.

For some molecules, formation of alkali metal adducts can be observed in ESI-MS. This can possibly be attributed to leaching from glass recipients or the presence of impurities in the mobile phase [39]. Such adducts, as compared to the molecular ions, can be unstable and might not produce fragment ions, jeopardizing the reliability of a quantitative LC-MS/MS method. The particular issue of metal adduct formation has also been reported for CGs [18]. In this study, formation of metal adducts, not prone to fragmentation, was observed when a mobile phase with HCOONH4 and HCOOH (pH 3) was used for the LC. In order to generate a sensitive single precursor ion for the analysis in the SRM mode, Bylda et al. investigated di fferent mobile phase additives and finally selected the [M + Li]+ adduct for the quantification of CGs [18]. In the absence of intense molecular ions, some applications used [M + Na]+ or [M + K]+ adducts for quantification of CGs in a single or selected ion monitoring (SIM) mode [38,40]. However, it has been noticed that SIM produced much higher detection limits with biological samples compared to SRM, while the two modes were similar in sensitivity if a standard mixture of CGs containing no matrix was injected [21]. In the current method, under the applied LC-MS conditions (see Section 5.2) with a NH4HCO3-containing mobile phase (pH 9), formation of a protonated molecular ion [M + H]+ for DIGO, OLE, OUB, and CON, and of an [M + NH4] + adduct for DIGI was achieved. Under mild fragmentation conditions, these precursor ions underwent collision-induced dissociation in the quadrupole yielding usable intense product ions. Of these fragments, the two most abundant ions were selected for the identification and quantification of these plant toxins, thereby fulfilling the requirements of the Commission Decision 2002/657/EC, which recommends the use of four identification points for confident identification of compounds in the LC-MS/MS analysis [31]. Some of the observed abundant product ions corresponded to sequential losses of the sugar moieties and elimination of hydroxyl groups from the steroid aglycones [41].

To optimize the sensitivity and selectivity of the LC-MS/MS method, SPE was used in the sample preparation of CGs in herbs and urine, in order to reduce complexity of the matrix and to allow enrichment of the analytes. This approach has previously been used for purification of some CGs in whole blood and plasma [17,24,42]. Compared to Oasis ® HLB SPE devices, used for the clean-up of urine samples in the current method, a significant reduction of matrix e ffects for herb samples was noted with Oasis ® MAX cartridges.

Currently, no maximum levels for CGs in food are set and analytical methods should preferably achieve as low as possible LOQs. The results of validation indicated that the developed method is able to detect and quantify the target CGs at low levels. In comparison to other existing methods for CGs in biological matrices and herbal-based products (Table S1), the current method is able to reach lower or similar LOQs. During the validation, matrix e ffects were observed for herbs and urine. The matrix effect is in many occasions unavoidable in MS analysis of complex matrices. It could be caused by co-eluting compounds that interfere with the ionization process of the target analytes leading to signal suppression or enhancement. The CGs were di fferently a ffected by the matrix interferences. OUB, being the most polar CG and eluting early in the chromatographic run, su ffered from a greater matrix suppression than OLE and DIGI, which eluted at the end of the run. This can be associated with the fact that hydrophilic compounds of the biological or plant matrix also eluted early in the chromatographic run giving rise to a more pronounced matrix e ffect in this region of the chromatogram. The addition of stable isotope labeled analogues of the analyzed molecules is suggested for use as internal standards to counteract the matrix e ffects. The use of commercially available or in-house synthesized labeled internal standards have already been reported for the LC-MS analysis of a number of CGs [18,19]. Other structurally related molecules, such as methyldigoxin, digitoxigenin, and gitoxigenin, were included as internal standards in some analytical methods [21,40,43]. In the current method DIGO-D, the deuterated analogue of DIGO, was applied as an internal standard for all target CGs.

The method linearity was tested and found to be good exhibiting R<sup>2</sup> >0.997, in most cases >0.999. The mean recovery of the developed method was in agreemen<sup>t</sup> with the acceptable limit of 70–120%, except for CON in herbs. A good precision of the method was demonstrated with repeatability and within-laboratory reproducibility below 20% for both matrices. The accuracy and the precision of the developed method were in the same range as for other methods, reported in the literature, for quantification of CGs in biological matrices (Table S1). The obtained MU was below 50%, complying with the SANTE/12682/2019 guidance document [44].

After completing the validation, the developed LC-MS/MS method was applied, to assess the CG contamination of herbs and herbal/spice mixtures used for culinary purposes, available on the Belgian food market. None of the target CGs were detected in these samples indicating that there is currently no safety risk for the population with regards to contamination of culinary herbs and blends with CGs.

CGs are a large and diverse group of naturally occurring toxic compounds, and, upon necessity, the LC-MS technique allows extension of the method to other plant toxins from this class. The presented method can, on the one hand, be used in food control initiatives to ensure food safety and, on the other hand, in population-wide survey studies that combine monitoring of food contamination and analysis of human biological fluids to unravel the level of exposure to plant toxins.
